Sections 7-9
Preprint
July 31, 1999
Future Effects of Long-Term Sulfur Deposition
on Surface Water Chemistry
in the Northeast and Southern Blue Ridge Province
(Results of the Direct/Delayed F|espp|sa
by
M, R, Church, K. W. Thornton, P. W, Shaffer, R/l^
G, R, Holdren, M. G. Johnson, J. J. Lee, R. S. Turner, Dl y;£
D. A. Lammers, W. G. Campbell, C. I. Lift, C,.£X Brandt, L H.legel,
G. D. Bishop, D, C, Mortenson, S. M. Piersqn, D. D; Schmoyer
A Contribution to the
.National: Acid Precipftation
U.S. Environmental Protectron^eney ;N,
Office of Research and Devejo|iinenai^asniRgl^n^i|0
Environmental Research Laboratory,
••• •• *
CM
-------
NOTICE
The information in this document has been funded wholly (or in part) by the U.S. Environmental
Protection Agency. It has been subjected to the Agency's peer and administrative review, and it has
been approved for publication as an EPA document. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
-------
CONTENTS
SECTION PAGE
Notice ii
Tables xii
Figures xix
Plates xxvii
Contributors xxix
Acknowledgments xxxi
1 EXECUTIVE SUMMARY 1-2
1.1 INTRODUCTION . 1-2
1.1.1 Project Background 1-2
1.1.2 Primary Objectives 1-3
1.1.3 Study Regions 1-4
1.1.4 Time Frames of Concern 1-4
1.2 PROCESSES OF ACIDIFICATION 1-6
1.2.1 Sulfur Retention 1-6
1.2.2 Base Cation Supply 1-7
1.3 GENERAL APPROACH 1-7
1.3.1 Soil Survey 1-8
1.3.2 Other Regional Datasets 1-8
1.3.3 Scenarios of Atmospheric Deposition 1-10
1.3.4 Data Analysis 1-10
1.4 RESULTS 1-11
1.4.1 Retention of Atmospherically Deposited Sulfur 1-11
1.4.1.1 Current Retention 1-11
1.4.1.2 Projected Retention 1-12
1.4.2 Base Cation Supply - 1-15
1.4.2.1 Current Control 1-15
1.4.2.2 Future Effects 1-15
1.4.3 Integrated Effects on Surface Water ANC 1-16
1.4.3.1 Northeast Lakes 1-16
1.4.3.2 Southern Blue Ridge Province 1-20
1.5 SUMMARY DISCUSSION 1-23
1.6 REFERENCES 1-24
2 INTRODUCTION TO THE DIRECT/DELAYED RESPONSE PROJECT 2-1
2.1 PROJECT BACKGROUND 2-1
2.2 PRIMARY OBJECTIVES 2-2
2.3 STUDY REGIONS 2-3
2.4 TIME FRAMES OF CONCERN 2-3
2.5 PROJECT PARTICIPANTS 2-6
2.6 REPORTING 2-6
3 PROCESSES OF ACIDIFICATION 3-1
3.1 INTRODUCTION 3-1
3.2 FOCUS OF THE DIRECT/DELAYED RESPONSE PROJECT 3-3
Hi
-------
CONTENTS (continued) Page
3.3 SULFUR RETENTION PROCESSES 3-3
3.3.1 Introduction 3-3
3.3.2 Inputs 3-4
3.3.3 Controls on Sulfate Mobility within Forest/Soil Systems 3-5
3.3.3.1 Precipitation/Dissolution of Secondary Sulfate Minerals 3-7
3.3.3.2 Sulfate Reduction in Soils and Sediments 3-7
3.3.3.3 Plant Uptake 3-8
3.3.3.4 Retention as Soil Organic Sulfur 3-9
3.3.3.5 Sulfate Adsorption by Soils 3-10
3.3.4 Models of Sulfur Retention 3-14
3.3.5 Summary 3-16
3.4 BASE CATION SUPPLY PROCESSES 3-17
3.4.1 Introduction T 3-17
3.4.2 Factors Affecting Base Cation Availability 3-20
3.4.2.1 Mineral Weathering 3-21
3.4.2.2 Cation Exchange Processes 3-25
3.4.3 Modelling Cation Supply Processes 3-28
3.4.3.1 Modelling Weathering 3-28
3.4.3.2 Modelling Cation Exchange Processes 3-29
4 PROJECT APPROACH 4-1
4.1 INTRODUCTION 4-1
4.2 SOIL SURVEY 4-3
4.2.1 Watershed Selection 4-3
4.2.2 Watershed Mapping 4-3
4.2.3 Sample Class Definition 4-3
4.2.4 Soil Sampling 4-4
4.2.5 Sample Analysis 4-4
4.2.6 Database Management 4-4
4.3 OTHER REGIONAL DATASETS 4-4
4.3.1 Atmospheric Deposition 4-5
4.3.2 Runoff Depth 4-5
4.4 DATA ANALYSIS 4-6
4.4.1 Level I Analyses 4-6
4.4.2 Level II Analyses 4-6
4.4.3 Level III Analyses 4-7
4.4.4 Integration of Results 4-8
4.4.5 Use of a Geographic Information System 4-9
5 DATA SOURCES AND DESCRIPTIONS 5-1
5.1 INTRODUCTION 5-1
5.2 STUDY SITE SELECTION 5-1
5.2.1 Site Selection Procedures 5-1
5.2.2 Eastern Lake Survey Phase I Design 5-1
5.2.3 Pilot Stream Survey Design 5-2
5.2.4 DDRP Target Population 5-6
5.2.4.1 Northeast Lake Selection 5-6
5.2.4.2 Southern Blue Ridge Province Stream Selection 5-25
5.2.4.3 Final DDRP Target Populations 5-25
5.3 NSWS LAKE AND STREAM DATA 5-25
5.3.1 Lakes In the Northeast Region 5-25
5.3.1.1 Lake Hydrologic Type 5-25
5.3.1.2 Fall Index Sampling 5-30
5.3.1.3 Chemistry of DDRP Lakes 5-37
iv
-------
CONTENTS (continued) Page
5.3.2 Streams In the Southern Blue Ridge Province Region 5-37
5.3.2.1 Spring Baseflow Index Sampling 5-37
5.3.2.2 Chemistry of DDRP Stream Reaches 5-40
5.4 MAPPING PROCEDURES AND DATABASES 5-40
5.4.1 Northeast Mapping 5-42
5.4.1.1 Soils 5-43
5.4.1.2 Depth to Bedrock 5-49
5.4.1.3 Forest Cover Type 5-51
5.4.1.4 Bedrock Geology 5-51
5.4.1.5 Quality Assurance 5-52
5.4.1.6 Land Use/Wetlands 5-58
5.4.1.7 Geographic Information Systems Data Entry 5-73
5.4.2 Southern Blue Ridge Province Mapping 5-90
5.4.2.1 Soils 5-93
5.4.2.2 Depth to Bedrock 5-97
5.4.2.3 Forest Cover Type/Land use 5-98
5.4.2.4 Bedrock Geology 5-98
5.4.2.5 Drainage 5-98
5.4.2.6 Quality Assurance 5-100
5.4.2.7 land Use/Wetlands 5-105
5.4.2.8 Geographic Information Systems Data Entry 5-106
5.5 SOIL SAMPLING PROCEDURES AND DATABASES 5-111
5.5.1 Development/Description of Sampling Classes 5-111
5.5.1.1 Rationale/Need for Sampling Classes 5-111
5.5.1.2 Approach Used for Sampling Class Development 5-112
5.5.1.3 Description of Sampling Classes ; 5-113
5.5.2 Selection of Sampling Sites 5-117
5,5.2.1 Routine Samples 5-117
5.5.2.2 Samples on Special Interest Watersheds 5-122
5.5.3 Soil Sampling 5-122
5.5.3.1 Soil Sampling Procedures 5-122
5.5.3.2 Quality Assurance/Quality Control of Sampling 5-123
5.5.4 Physical and ChemicaLAnalvses 5-124
5.5.4.1 Preparation Laboratories 5-124
5.5.4.2 Analytical Laboratories : 5-126
5.5.5 Database Management 5-140
5.5.5.1 Database Structure 5-140
5.5.5.2 Database Operations 5-143
5.5.6 Data Summary 5-148
5.5.6.1 Summary of Sampling Class Data 5-148
5.5.6.2 Cumulative Distribution Functions 5-150
5.6 DEPOSITION DATA 5-150
5.6.1 Time Horizons of Interest 5-161
5.6.1.1 Current Deposition 5-161
5.6.1.2 Future Deposition 5-161
5.6.2 Temporal Resolution 5-161
5.6.2.1 Level I Analyses 5-161
5.6.2.2 Level II Analyses 5-161
5.6.2.3 Level 111 Analyses 5-163
5.6.3 Data Acquisition/Generation 5-163
5.6.3.1 Level III Analyses - Typical Year Deposition Dataset 5-164
5.6.3.2 Level I and II Analyses - Long-Term Annual Average Deposition Dataset ... 5-191
5.6.4 Deposition Datasets Used in DDRP Analyses 5-200
-------
CONTENTS (continued) Page
5.7 HYDROLOGIC DATA 5-200
5.7.1 Runoff 5-200
5.7.1.1 Data Sources 5-200
5.7.1.2 Runoff Interpolation Methods 5-203
5.7.1.3 Uncertainty Estimates 5-203
5.7.2 Derived Hvdroloaic Parameters 5-204
5.7.2.1 TOPMODEL 5-204
5.7.2.2 Soil Contact (Darcy's Law) 5-209
5.7.2.3 Mapped Hydrologic Indices 5-211
6 REGIONAL POPULATION ESTIMATION 6-1
6.1 INTRODUCTION 6-1
6.2 PROCEDURE 6-1
6.2.1 Use of Variable Probability Samples 6-1
6.2.2 Estimation Procedures for Population Means 6-2
6.2.3 Estimators of Variance 6-4
6.2.4 Estimator of Cumulative Distribution Function 6-5
6.3 UNCERTAINTY ESTIMATES 6-6
6.4 APPLICABILITY 6-8
7 WATERSHED SULFUR RETENTION 7-1
7.1 INTRODUCTION 7-1
7.2 RETENTION IN LAKES AND WETLANDS 7-2
7.2.1 Introduction 7-2
7.2.2 Approach 7-4
7.2.3 Results 7-6
7.3 WATERSHED SULFUR RETENTION 7-9
7.3.1 Methods 7-9
7.3.1.1 Input/Output Calculation 7-9
7.3.1.2 Data Sources 7-11
7.3.2 Uncertainty Estimates 7-11
7.3.2.1 Introduction 7-11
7.3.2.2 Individual Variable Uncertainties 7-12
7.3.2.3 Uncertainty Calculation - Monte Carlo Analysis 7-17
7.3.3 Internal Sources of Sulfur 7-19
7.3.3.1 Introduction/Approach 7-19
7.3.3.2 Bedrock Geology 7-20
7.3.3.3 Upper Limit Steady-State Sulfate Concentration 7-24
7.3.4 Results and Discussion 7-29
7.3.4.1 Northeast 7-31
7.3.4.2 Mid-Appalachians 7-41
7.3.4.3 Southern Blue Ridge Province 7-41
7.3.4.4 Conclusions 7-43
8 LEVEL I STATISTICAL ANALYSES 8-t
8.1 INTRODUCTION 8-1
8.1.1 Approach 8-2
8.1.2 Statistical Methods 8-7
8.2 RELATIONSHIPS BETWEEN ATMOSPHERIC DEPOSITION
AND SURFACE WATER CHEMISTRY 8-9
8.2.1 Introduction 8-9
8.2.2 Approach : 8-9
8.2.3 Results and Discussion 8-9
8.2.3.1 Northeast 8-9
8.2.3.2 Southern Blue Ridge Province 8-11
8.2.3.3 Summary 8-11
vl
-------
CONTENTS (continued) Page
8.3 DERIVED HYDROLOGIC PARAMETERS 8-13
8.3.1 Soil Contact (Darcv's Law) 8-13
8.3.1.1 Introduction 8-13
8.3.1.2 Results and Discussion 8-18
8.3.2 Geomorphic/Hydroloqic Parameters 8-21
8.3.2.1 Introduction 8-21
8.3.2.2 Results and Discussion 8-22
8.3.3 TQPMQDEL Parameters 8-37
8.3.3.1 Introduction 8-38
8.3.3.2 Results and Discussion 8-41
8.3.3.3 Summary 8-48
8.4 MAPPED BEDROCK GEOLOGY 8-48
8.4.1 DDRP Bedrock Sensitivity Scale 8-50
8.4.2 Results 8-51
8.4.2.1 Sulfate and Percent Retention 8-54
8.4.2.2 Sum of Base Cations, ANC, and pH 8-59
8.4.3 Summary 8-61
8.5 MAPPED LAND USE/VEGETATION 8-62
8.5.1 Introduction 8-62
8.5.2 Data Sources 8-63
8.5.3 Statistical Methods 8-63
8.5.4 Sulfate and Percent Sulfur Retention 8-64
8.5.4.1 Northeast 8-64
8.5.4.2 Southern Blue Ridge Province 8-73
8.5.4.3 Regional Comparisons 8-73
8.5.5 ANC. Ca plus MQ. and oH 8-75
8.5.5.1 Northeast 8-75
8.5.5.2 Southern Blue Ridge Province 8-76
8.5.5.3 Regional Comparisons 8-76
8.5.6 Summary and Conclusions 8-78
8.6 MAPPED SOILS 8-78
8.6.1 Introduction 8-78
8.6.2 Approach 8-79
8.6.3 Sulfate and Sulfur Retention 8-88
8.6.3.1 Northeast 8-88
8.6.3.2 Southern Blue Ridge Province 8-92
8.6.3.3 Regional Comparisons 8-97
8.6.4 ANC. Ca plus Mo. and pH 8-102
8.6.4.1 Northeast 8-109
8.6.4.2 .Southern Blue Ridge Province 8-111
8.6.4.3 Regional Comparisons 8-113
8.6.5 Summary and Conclusions 8-113
8.7 ANALYSES OF DEPTH TO BEDROCK 8-113
8.7.1 Introduction 8-113
8.7.2 Approach 8-113
8.7.3 Sulfate and Percent Sulfur Retention 8-115
8.7.3.1 Northeast 8-115
8.7.3.2 Southern Blue Ridge Province 8-119
8.7.3.3 Comparison of Regions 8-119
8.7.4 ANC. Ca plus Ma and oH 8-119
8.7.4.2 Southern Blue Ridge Province 8-122
8.7.4.3 Comparison of Regions 8-123
8.7.5 Summary and Conclusions 8-123
vii
-------
CONTENTS (continued) Page
8.8 INTEGRATED ANALYSIS OF ALL MAPPED VARIABLES 8-124
8.8.1 introduction 8-124
8.8.2 Approach 8-124
8.8.3 Sulfate and sulfur retention 8-125
8.8.3.1 Northeast 8-125
8.8.3.2 Southern Blue Ridge Province 8-127
8.8.3.3 Regional Comparisons 8-130
8.8.4 ANC. Ca plus Ma. and oH 8-131
8.8.4.1 Northeast 8-131
8.8.4.2 Southern Blue Ridge Province 8-134
8.8.4.3 Regional Comparisons 8-137
8.8.5 Summary and Conclusions 8-137
8.9 SOIL PHYSICAL AND CHEMICAL CHARACTERISTICS 8-138
8.9.1 Introduction 8-138
8.9.2 Approach 8-138
8.9.2.1 Statistical Methods 8-140
8.9.3 Aggregation of Soil Data 8-143
8.9.3.1 Introduction 8-143
8.9.3.2 Aggregation of Soil Data 8-144
8.9.3.3 Assessment of the DDRP Aggregation Approach 8-145
8.9.3.4 Estimation of Watershed Effect 8-148
8.9.3.5 Evaluation of Watershed Effect 8-149
8.9.4 Regional Soil Characterization 8-155
8.9.5 Sulfate and Sulfur Retention 8-157
8.9.5.1 Northeast 8-157
8.9.5.2 Southern Blue Ridge Province 8-164
8.9.6 Ca plus Ma (SOBCl ANC. and pH 8-165
8.9.6.1 Northeast 8-169
8.9.6.2 Southern Blue Ridge Province 8-170
8.9.7 Evaluation of Alternative Aggregation Schemes 8-170
8.9.8 Summary and Conclusions 8-171
8.9.8.1 Alternative Aggregation Schemes 8-171
8.9.8.2 Sulfate and Sulfur Retention 8-174
8.9.8.3 Ca plus Mg (SOBC), ANC, and pH 8-175
8.9.9 Summary Conclusions 8-175
8.10 EVALUATION OF ASSOCIATIONS BETWEEN WATERSHED ATTRIBUTES .
AND SURFACE WATER CHEMISTRY 8-176
8.10.1 Introduction 8-176
8.10.2 Approach 8-176
8.10.3 Regional Characterization of Watershed Attributes 8-177
8.10.3.1 Northeast Subregions 8-177
8.10.3.2 Northeast and Southern Blue Ridge Providence 8-182
8.10.4 Sutfate and Sulfur Retention 8-182
8.10.4.1 Northeast 8-192
8.10.4.2 Southern Blue Ridge Province 8-193
8.10.5 Ca plus Ma fSOBC). ANC. and pH 8-193
8.10.5.1 Northeast 8-193
8.10.5.2 Southern Blue Ridge Province 8-197
8.10.6 Summary and Conclusions 8-197
8.10.6.1 Sulfate and Sulfur Retention 8-197
8.10.6.2 Ca plus Mg (SOBC), ANC, and pH 8-198
8.10.7 Summary Conclusions 8-198
VIII
-------
COMTENTS (continued) Page
9 LEVEL II ANALYSES - SINGLE FACTOR RESPONSE TIME ESTIMATES 9-1
9.1 INTRODUCTION 9-1
9.2 EFFECTS OF SULFATE ADSORPTION ON WATERSHED SULFUR RESPONSE TIME . . 9-2
9.2.1 Introduction 9-2
9.2.2 Section Objectives 9-4
9.2.3 Approach 9-5
9.2.3.1 Model Description 9-5
9.2.3.2 Data Sources 9-6
9.2.3.3 Model Assumptions and Limitations 9-8
9.2.3.4 Adsorption Data 9-9
9.2.3.5 Evaluation of Aggregated Data and Model Outputs 9-14
9.2.3.6 Target Populations for Model Projections 9-17
9.2.4 Results 9-18
9.2.4.1 Comparison of Northeast and Southern Blue Ridge Province Isotherm
Variables 9-18
9.2.4.2 Model Results - Northeastern United States 9-20
9.2.4.3 Model Results - Southern Blue Ridge Province 9-35
9.2.4.4 Uncertainty Analyses and Alternative Aggregation Approaches 9-51
9.2.4.5 Summary of Results from the Southern Blue Ridge Province 9-59
9.2.5 Summary Comments on Level II Suifate Analyses 9-62
9.2.6 Conclusions 9-64
9.3 EFFECT OF CATION EXCHANGE AND WEATHERING ON SYSTEM RESPONSE 9-66
9.3.1 Introduction 9-66
9.3.1.1 Level H Hypotheses 9-67
9.3.1.2 Approach 9-71
9.3.2 Descriptions of Models 9-75
9.3.2.1 Reuss Model 9-75
9.3.2.2 Bloom-Grigal Model 9-94
9.3.3 Model Forecasts 9-103
9.3.3.1 Reuss Model 9-105
9.3.3.2 Blopm-Grigal Model 9-154
9.3.4 Comparison of the Bloom-Griaal and Reuss Models 9-185
9.3.5 Summary and Conclusions 9-196
10 LEVEL III ANALYSES - DYNAMIC WATERSHED MODELLING 10-1
10.1 INTRODUCTION 10-1
10.2 DYNAMIC WATERSHED MODELS 10-3
10.2.1 Enhanced Trickle Down (ETD) Model 10-6
10.2.2 integrated Lake-Watershed Acidification Study flLWASl Model 10-7
10.2.3 Model of Acidification of Groundwater in Catchments (MAGiCl 10-13
10.3 OPERATIONAL ASSUMPTIONS 10-14
10.4 WATERSHED PRIORITIZATiON 10-14
10.4.1 Northeast 10-16
10.4.2 Southern Blue Ridae Province 10-18
10.4.3 Effects of Prioritization on Inclusion Probabilities 10-20
10.5 MODELLING DATASETS 10-20
10.5.1 Meteorological/Deposition Data 10-21
10.5.2 DDRP Runoff Estimation 10-22
10.5.2.1 Annual Runoff 10-22
10.5.2.2 Monthly Runoff 10-22
10.5.3 Morphometrv 10-24
10.5.4 Soils 10-25
10.5.5 Surface Water Chemistry 10-25
10.5.6 Other Data 10-25
10.5.7 Chloride Imbalance 10-25
10.6 GENERAL APPROACH ' 10-28
ix
-------
CONTENTS (continued) Page
10.7 MODEL CALIBRATION 10-30
10.7.1 Special Interest Watersheds 10-30
tO.7.1.1 Northeast 10-33
10.7.1.2 Southern Blue Ridge Province 10-33
10.7.2 General Calibration Approach 10-34
- 10.7.3 Calibration of the Enhanced Trickle Down Model 10-35
10.7.4 Calibration of the Integrated Lake-Watershed Acidification Model 10-38
10.7.5 Calibration of the Model of Acidification of Groundwater in Catchments 10-42
10.7.6 Calibration/Confirmation Results 10-44
10.8 MODEL SENSITIVITY ANALYSES 10-49
10.8.1 General Approach 10-50
10.8.2 Sensitivity Results 10-51
10.9 REGIONAL PROJECTIONS REFINEMENT 10-53
10.9.1 Enhanced Trickle Down 10-53
10.9.2 Integrated Lake-Watershed Acidification Study 10-54
10.9.3 Model of Acidification of Groundwater in Catchments .• 10-54
10.9.4 DDRP Watershed Calibrations 10-56
10.9.4.1 Integrated Lake-Watershed Acidification Study 10-56
10.9.4.2 MAGIC 10-59
10.9.4.3 Southern Blue Ridge Province 10-61
10.10 MODEL PROJECTIONS 10-66
10.10.1 General Approach 10-66
10.10.2 Forecast Uncertainty 10-67
10.10.2.1 Watershed Selection 10-69
10.10.2.2 Uncertainty Estimation Approaches 10-70
10.10.2.3 Relationship Among Approaches 10-74
10.10.2.4 Confidence Intervals 10-76
10.11 POPULATION ESTIMATION AND REGIONAL FORECASTS 10-76
10.11.1 Northeast Regional Projections 10-77
10.11.1.1 Target Population Projections Using MAGIC 10-77
10.11.1.2 Target Population Projections Using MAGIC and ETD 10-91
10.11.1.3 Restricted Target Population Projections Using All Three Models .... 10-113
10.11.2 Southern Blue Rldae Province 10-141
10.11.2.1 Target Population Projections Using MAGIC 10-141
10.11.2.2 Restricted Target Population Projections Using ILWAS and MAGIC . . . 10-155
10.11.3 Regional Comparisons 10-174
10.11.3.1 Northeastern Projections of Sulfate Steady State 10-174
10.11.3.2 Southern Blue Ridge Province Projections of Sulfate Steady State .... 10-178
10.11.3.3 ANC and Base Cation Dynamics - 10-178
10.12 DISCUSSION 10-195
10.12.1 Future Projections of Changes in Acid-Base Surface Water Chemistry 10-195
10.12.2 Rate of Future Change 10-197
10.12.2.1 Northeast 10-197
10.12.1.2. Southern Blue Ridge Province 10-202
10.12.3 Uncertainties and Implications for Future Changes in
Surface Water Acid-Base Chemistry 10-204
10.12.3.1 Deposition Inputs 10-205
10.12.3.2 Watershed Processes 10-207
10.13 CONCLUSIONS FROM LEVEL 111 ANALYSES 10-210
11 SUMMARY OF RESULTS 11-1
1.1 RETENTION OF ATMOSPHERICALLY DEPOSITED SULFUR 11-1
11.1.1 Current Retention 11-1
11.1.2 Projected Retention 11-3
11.2 BASE CATION SUPPLY 11-6
11.2.1 Current Control 11-6
11.2.2 Future Effects 11-7
-------
CONTENTS (continued) Page
11.3 INTEGRATED EFFECTS ON SURFACE WATER ANC 11-8
11.3.1 Northeast Lakes 11-9
11.3.2 Southern Blue Ridae Province 11-17
11.4 SUMMARY DISCUSSION 11-26
12 REFERENCES 12-1
13 GLOSSARY 13-1
13.1 ABBREVIATIONS AND SYMBOLS 13-1
13.1.1 Abbreviations 13-3
13.1.2 Symbols 13-6
XI
-------
TABLES
TABLE PAGE
1-1 Lakes in the NE Projected to Have ANC Values <0 and <50 peq L*1 for Constant
1-2
and Decreased Sulfur Deposition 1-19
S8RP Stream Reaches Projected to Have ANC Values <0 and <50 ^eq L for
Constant and Increased Sulfur Deposition 1-22
3-1 Major Rock Forming Minerals and Their Relative Reactivities 3-22
5-1 Sampling Structure for Phase I, Region 1 (Northeast), Eastern Lake Survey 5-4
5-2 Sample Structure for the Direct/Delayed Response Project - Northeastern Sample 5-8
5-3 ANC Group. Lake identification, ELS-I Phase I ANC, Weight and Inclusion
Probabilities for the Direct/Delayed Response Project Northeast Sample Watersheds ... 5-9
.5-4 Lake Identification (ID) and Name, and State and Latitudinal/Longitudinal
Location of the Northeast Sample Watersheds, Sorted by Lake ID 5-13
5-5 Lake Identification (ID) and Name, Sorted by State - Northeast Sample Watersheds ... 5-16
5-6 Stream Identification (ID), Weight, and Inclusion Probabilities for the Southern
Blue Ridge Province Direct/Delayed Response Project Sample Watersheds 5-26
5-7 Stream Identification (ID) and Name, and State and Latitudinal/Longitudinal
Location of the Southern Blue Ridge Province Sample Watersheds, Sorted by Stream ID 5-27
5-8 Stream Identification (ID) and Name, Sorted by State - Southern Blue Ridge
Province Sample Watersheds 5-28
5-9 DDRP Reciassiflcation of Northeastern Lakes Classified as "Seepage" or "Closed"
by the NSWS 5-31
5-10 Depth-to-Bedrock Classes and Corresponding Level of Confidence 5-50
5-11 Interpretation Codes for Northeast Map Overlays - Land Use/Land Cover, Wetlands,
and Beaver Activity 5-59
5-12 Northeast Watersheds Studied for Independent Field Check of Land Use and
Wetland Photointerpretations 5-63
5-13 Chi-Square Test for General Land Use Categories 5-65
5-14 Comparison of Field Check (Matched) General Land Use Determinations with
Office Photointerpretations 5-66
5-15 Chi-Square Test for Detailed Wetland Categories 5-67
5-16 Comparison of Field Check (Matched) Detailed Wetland Determinations with
Office Photointerpretations 5-68
5-17 Comparison of Beaver Dam Number (#), Breached (B) and Unbreached
(U) Status, and Lodges (L), Identified via Field Check and Office Photointerpretation
Methods 5-70
5-18 Aggregated Land Use Data for Northeast Watersheds 5-72
5-19 Watershed No. 1E1062 Soil Mapping Units 5-87
5-20 Land Use Codes Used as Map Symbols 5-99
5-21 Percent Land Use Data for Southern Blue Ridge Province Watersheds 5-107
5-22 Laboratory Analysis of DDRP Soil Samples 5-125
5-23 Analytical Variables Measured in the DDRP Soil Survey 5-127
5-24 Data Quality Objectives for Detectabfltty and Analytical Within-Batch Precision 5-131
5-25 Detection Limits for Contract Requirements, Instrument Readings, and
System-Wide Measurement In the Northeast 5-133
5-26 Detection Limits for the Contract Requirements, Instrument Readings, and
System-wide Measurement in the Southern Blue Ridge Province 5-134
5-27 Attainment of DQO's by the analytical laboratories as determined from blind
audit samples for the Northeast. 5-136
5-28 Attainment of DQO's by the Analytical Laboratories as Determined from Blind
Audit Samples for the Southern Blue Ridge Province 5-138
5-29 Quality Assurance and Quality Control Checks Applied to Each Data Batch 5-146
5-30 Medians of Pedon-Aggregated Values of Soil Variables for the DDRP Regions
and Subregions 5-160
xii
-------
TABLES (continued) Page
5-31 Monthly Values of Leaf Area Index (LAI) Used to Apportion Annual Dry Deposition
to Monthly Values 5-176
5-32 Ratios of Coarse-to-Fine Particle Dry Deposition 5-180
5-33 Ratios of Dry Deposition to Wet Deposition for DDRP Study Sites for the
Typical Year (TY) Deposition Dataset 5-182
5-34 Deposition Datasets Used in DDRP Analyses 5-201
5-35 DDRP texture classes and saturated hydraulic conductivity (K) for the NE study systems. .5-206
5-36 SCS slope classifications 5-212
5-37 Mapped and calculated geomorphic parameters collected for the NE study sites 5-215
5-38 Mapped and calculated geomorphic parameters collected for the SBRP study sites 5-219
7-1 Summary of Computed Sulfur Retention by In-lake Reduction for Lake Systems in the
Eastern United States 7-5
7-2 Intensively Studied Sites Used in Surface Water Chemistry Uncertainty Analysis 7-13
7-3 Summary Statistics on Percent Differences Between Row-weighted Average Annual
Sulfate Concentration and the Fall/Spring Row-weighted Averages 7-18
7-4 Bedrock Geology Maps Used in the DDRP Internal Sources of Sulfur Bedrock
Geology Analyses 7-21
7-5 Potential for Sulfur Contribution by Geologic Type 7-23
7-6 Summary of Watersheds (by ELS and NSS Subregion) Dropped Due to Suspected
Internal Sources of Sulfur Identified by Steady-State Analysis 7-30
7-7 Percent Sulfur Retention - Summary Statistics by Region 7-33
7-8 Summary of Sulfur Retention Status and of Watershed Variables Contributing
to Sulfur Retention for 42 Watersheds in the Northeastern United States 7-39
8-1 Surface Water Chemistry and Percent Sulfur Retention Summary Statistics
for the Northeastern DDRP Sample of 145 Lake Watersheds 8-3
8-2 Surface Water Chemistry and Percent Sulfur Retention Summary Statistics
for the DDRP Sample of 35 SBRP Stream Watersheds 8-4
8-3 Summary Statistics for Wet and Dry Deposition on the DDRP Sample
of 145 Northeastern Lake Watersheds • - 8-5
8-4 Summary Statistics for Wet and Dry Deposition on the DDRP Sample of 35
SBRP Stream Watersheds 8-6
8-5 Results of Regressions Relating Surface Water Chemistry to Atmospheric
Deposition in the Northeast Region (n = 145) < 8-10
8-6 Results of Regressions Relating Surface Water Chemistry to Atmospheric
Deposition in the Southern Blue Ridge Province (n = 32) 8-12
8-7 Estimated Population-Weighted Summary Statistics on the Darcy's Law Estimates
of Row Rate and the Index of Row Relative to Runoff 8-15
8-8 Estimated Population-Weighted Summary Statistics for Northeastern
Geomorphic/Hydrologic Parameters 8-23
8-9 Estimated Population-Weighted Summary Statistics for Southern Blue
Ridge Province Hydrologic/Geomorphic Parameters 8-24
8-10 Mapped and Calculated Geomorphic Parameters Collected
for the Northeastern Study Sites (Same as 5-37) 8-25
8-11 Mapped and Calculated Geomorphic Parameters Collected for the
SBRP Study Sites 8-28
8-12 Stratification Based on Sulfur Deposition (Wet and Dry) 8-30
8-13 Results of Stepwise Regression Relating Surface Water Chemistry versus
Geomprphic/Hydrologic Parameters for the Entire NE 8-31
8-14 Stepwise Regression Equations for Surface Water Chemistry and
Hydrologic/Geomorphic Parameters Based on Sulfur Deposition Stratification 8-33
8-15 Results of Stepwise Regression Relating Surface Water Chemistry
and Geomorphic/Hydrologic Parameters for the SBRP 8-34
8-16 Population-Weighted Summary Statistics for ln(a/KbTanB) for the Northeast 8-39
8-17 Population-Weighted Summary Statistics for ln(a/TanB) for the Southern Blue
Ridge Province 8-40
xiii
-------
TABLES (continued) Page
8-18 Spearman's Correlation Coefficients Between ln(a/KbTanB) and Surface Water Chemistry 8-42
8-19 Pearson's Correlation Coefficients Between ln(a/TanB) and NSS Pilot Chemistry 8-47
8-20 Tabulation of the Generic Bedrock Types Used to Classify the Mapped Units
Identified on State Map Legends 8-52
8-21 Tabulation of the Generic Bedrock Types Used to Classify the Mapped Units
Identified on State Map Legends 8-53
8-22 Regional and Subregional Statistics for the Bedrock Sensitivity Code Variables 8-55
8-23 Results of Regressions of Surface Water Chemistry on Bedrock Sensitivity
Code Statistics and Deposition Estimates for Northeast 8-56
8-24 Results for SBRP of Regressions of Surface Water Chemistry on Bedrock
Sensitivity Code Statistics and Deposition Estimates 8-58
8-25 Land Use and Other Environmental Variables Related to Surface Water
Chemistry of Northeastern Lakes 8-65
8-26 Factor Loadings for First 13 Principal Components after Varimax Rotation of
the Correlation Matrix of Land Use and other Environmental Variables for
Northeastern Lakes 8-66
8-27 Interpretation of the First 13 Principal Components After Varimax Rotation of the
Correlation Matrix of Land Use and Other Environmental Variables for Northeastern Lakes 8-68
8-28 Land Use and Other Environmental Variables Related to Surface Water Chemistry of.
Southern Blue Ridge Province Streams 8-69
8-29 Composition of First 11 Principal Component Analysis (PCA) Factors After Varimax
Rotation of the Correlation Matrix of Land Use and Other Environmental Variables
Related to Surface Water Chemistry of Southern Blue Ridge Province Streams 8-70
8-30 Interpretation of the First 11 Principal Components after Varimax Rotation of
the Correlation Matrix of Land Use and Other Environmental Variables for Southern
Blue Ridge Province Streams 8-71
8-31 Results of Regressions Relating Surface Water Chemistry of Northeastern Lakes to
Land Use and Other Environmental Data '. 8-72
8-32 Results of Regressions Relating Sulfate and Percent Sulfur Retention of
Southern Blue Ridge Province Streams to Land Use Data 8-74
8-33 Results of Regressions Relating ANC, Ca plus Mg, and pH of Southern
Blue Ridge Province Streams to Land Use Data 8-77
8-34 Summary Statistics for Percent Area Distribution of the 38 Soil Sampling
Classes and the 4 Miscellaneous Land Areas on the DDRP Sample of 145 NE
Lake Watersheds 8-83
8-35 Summary Statistics for the Percent Area Distribution of the 38 Soil Sampling Classes
and the 4 Miscellaneous Land Areas in the GIS 40-ft Contour on the DDRP Sample of
145 NE Lake Watersheds 8-84
8-36 Summary Statistics for the Percent Area Distribution of the 38 Soil Sampling Classes
and the 4 Miscellaneous Land Areas in the Combined GIS Bufferson the DDRP
Sample of 145 NE Lake Watersheds 8-85
8-37 Summary Statistics for the Percent Area Distribution of the 12 Soil Sampling
Classes and the 2 Miscellaneous Land Areas on the DDRP Sample of 35 SBRP
Stream Watersheds 8-86
8-38 Summary Statistics for the Percent Area Distribution of the 12 Soil Sampling Classes
and the 2 Miscellaneous Land Areas in the 100-Meter Linear GIS Buffer on the
DDRP Sample of 35 SBRP Stream Watersheds 8-87
8-39 Lake Sulfate and Percent S Retention Regression Models Developed for NE Lakes
Using Deposition, Mapped Soils (as a Percentage of Watershed Area in Soil
Sampling Classes) and Miscellaneous Land Areas as Candidate Independent Variables . . 8-89
8-40 Regression Models of Sulfate In SBRP Streams, Developed Using Deposition,
Mapped Soils (as a Percentage of Watershed Area in Soil Sampling Classes) and
Miscellaneous Land Areas (as a Percentage of Watershed Area) as
Candidate Independent Variables 8-93
8-41 Regression Models of Percent Sulfur Retention In SBRP Stream Watersheds
Developed Using Deposition, Mapped Soils (as a Percentage of Watershed Area in Soil
Sampling Classes), and Miscellaneous Land Areas as Candidate Independent Variables . 8-96
xiv
-------
TABLES (continued) Page
8-42 Lake ANC and the Sum of Lake Calcium and Magnesium Regression
Models Developed for NE Lakes Using Deposition, Mapped Soils (as a
Percentage of Watershed Area in Soil Sampling Classes) and Miscellaneous
Land Areas as Candidate Independent Variables 8-99
8-43 Lake pH Regression Models Developed for NE Lakes Using Deposition,
Mapped Soils (as a Percentage of Watershed Area in Soil Sampling Classes) and
Miscellaneous Land Areas as Candidate Independent Variables .8-101
8-44 Regression Models of ANC In SBRP Stream Watersheds, Developed Using
Deposition, Mapped Soils (as a Percentage of Watershed Area in Soil Sampling
Classes) and Miscellaneous Land Areas as Candidate Independent Variables 8-104
8-45 Regression Models of Calcium Plus Magnesium in SBRP Streams, Developed Using
Deposition, Mapped Soils (as a Percentage of Watershed Area in Soil Sampling
Classes) and Miscellaneous Land Areas as a Candidate Independent Variables 8-106
8-46 Regression Models of SOBC in SBRP Streams, Developed Using Deposition,
Mapped Soils (as a Percentage of Watershed Area in Soil Sampling Classes) and
Miscellaneous Land Areas as Candidate Independent Variables 8-107
8-47 Regression Models of Stream pH in SBRP Streams, Developed Using
Deposition, Mapped Soils (as a Percentage of Watershed Area in Soil Sampling
Classes) and Miscellaneous Land Areas as Candidate Independent Variables 8-110
8-48 Depth-to-Bedrock Classes for the Northeast and the Southern Blue Ridge Province .... 8-114
8-49 Regional and Subregional Statistics for the Depth-to-Bedrock Classes 8-116
8-50 Results for NE of Regressions of Surface Water Chemistry on Depth-to-Bedrock Classes
and Deposition Estimates 8-118
8-51 Results for SBRP of Regressions of Surface Water Chemistry on Depth-to-Bedrock
Classes and Deposition Estimates 8-120
8-52 Regression Models of Surface Water Sulfate and Sulfur Retention in the NE Lake
Watersheds Using Deposition, Derived Hydrologic Parameters, Bedrock Geology
Reaction Classes, Depth To Bedrock, Mapped Landuse/Vegetation, and Mapped
Soils as Candidate Regressor Variables 8-126
8-53 Regression Models of Surface Water Sulfate and Sulfur Retention in the SBRP
Stream Watersheds Using Deposition, Derived Hydrologic Parameters, Bedrock
Geology Reaction Classes, Depth To Bedrock, Mapped Landuse/Vegetation, and
Mapped Soils as Candidate Regressor Variables 8-128
8-54 Regression Models of Surface Water ANC, Ca plus Mg, and pH in the NE Lake
Watersheds Using Deposition, Derived Hydrologic Parameters, Bedrock Geology
Reaction Classes, Depth To Bedrock, Mapped Landuse/Vegetation, and Mapped
Soils as Candidate Regressor Variables 8-132
8-55 Regression Models of Surface Water ANC, Ca plus Mg, and pH in the SBRP
Stream Watersheds Using Deposition, Derived Hydrologic Parameters, Bedrock
Geology Reaction Classes, Depth To Bedrock, Mapped Landuse/Vegetation, and
Mapped Soils as Candidate Regressor Variables 8-135
8-56 Standard Deviations Within and Among Northeast Sampling Classes Estimated
from B Master Horizon Data 8-147
8-57 Means and Standard Deviations of Soil Characteristics by Aggregation Method
and Region 8-150
8-58 Population Means and Standard Errors for Selected Variables, by Subregion/
Region and Aggregation (Watershed Adjusted Data) 8-153
8-59 Non-parametric Correlations Between Lake Chemistry Variables and Selected
Soil Properties for the NE DDRP Watersheds 8-158
8-60 Non-parametric Correlations Between Stream Chemistry Variables and Selected
Soil Properties for the SBRP DDRP Watersheds 8-160
8-61 Results of Stepwise Multiple Regressions for DDRP Lake and Stream Sulfate
Concentrations (SO416) Versus Soil Physical and Chemical Properties 8-162
8-62 Results of Stepwise Multiple Regressions for DDRP Watershed Sulfur Retention
(S04_NRET) Versus Soil Physical and Chemical Properties 8-163
xv
-------
TABLES (continued) Page
8-63 Results of Stepwise Multiple Regressions for DORP Lake Calcium plus Magnesium
Concentrations (CAMG16) and Stream Sum of Base Cation Concentrations (SOBC)
Versus Soil Physical and Chemical Properties 8-166
8-64 Results of Stepwise Multiple Regressions for DDRP Lake and Stream ANC
(ALKANEW and ALKA11) Versus Soil Physical and Chemical Properties 8-167
8-65 Results of Stepwise Multiple Regressions for DDRP Lake and Stream pH (PHEQ11)
Versus Soil Physical and Chemical Properties 8-168
8-66 Results of Stepwise Multiple Regressions for DDRP Lake and Stream ANC
(ALKANEW and ALKA11) Versus Unadjusted and Watershed Adjusted Soil
Physical and Chemical Properties 8-172
8-67 Results of Stepwise Multiple Regressions for DDRP Lake and Stream Sulfate
(SO416) Versus Unadjusted and Watershed Adjusted Soil Physical and Chemical
Properties 8-173
8-68 Population Means and Standard Errors for Selected Variables, by Subregion/
Region and Aggregation 8-178
8-69 Non-parametric Correlations Between Lake Chemistry Variables and Selected
Watershed Attributes for the NE DDRP Watersheds 8-183
8-70 Non-parametric Correlations Between Stream Chemistry Variables and Selected
Watershed Attributes for the SBRP DDRP Watersheds 8-187
8-71 Results of Stepwise Multiple Regressions for DDRP Lake and Stream Sulfate
Concentration (S0416) Versus Watershed Attributes 8-190
8-72 Results of Stepwise Multiple Regressions for DDRP Watershed Sulfur Retention
(SO4 NRET) Versus Watershed Attributes 8-191
8-73 Results of Stepwise Multiple Regressions for DDRP Lake Calcium Plus Magnesium
Concentrations (CAMG16) and Stream Sum of Base Cations (SOBC) Versus
Watershed Attributes 8-194
8-74 Results of Stepwise Multiple Regressions for DDRP Lake and Stream ANC
(ALKA11, ALKANEW) Versus Watershed Attributes 8-195
8-75 Results of Stepwise Multiple Regressions for DDRP Lake and Stream Air
Equilibrated pH (PHEQ11) Versus Watershed Attributes 8-196
9-1 Comparison of Summary Data for Sulfate Adsorption Isotherm Data for Soils in
the Northeastern United States and Southern Blue Ridge Province 9-19
9-2 Summary Statistics for Modelled Changes in Sulfate Concentration, Percent Sulfur
Retention, and Delta Sulfate for Northeast Watersheds Using Long-Term Average
Deposition Data 9-25
9-3 Summary Statistics for Modelled Changes in Sulfate Concentration, Percent Sulfur
Retention, and Delta Sulfate for Northeast Watersheds Using Typical Year Deposition
Data 9-26
9-4 Comparison of Measured and Modelled Base Year (1985) Sulfate Data for SBRP
Watersheds, Using Long-Term Average Deposition Data 9-38
9-5 Comparison of Modelled Rates of Increase for [SO4 ] in DDRP Watersheds in the
SBRP with Measured Rates of Increase in Watersheds in the Blue Ridge and
Adjoining Appalachians 9-41
9-6 Summary Statistics for Modelled Changes In Sulfate Concentration, Percent Sulfur
Retention, and Delta Sulfate for Watersheds in the Southern Blue Ridge Province,
Using Long-Term Average Deposition Data 9-45
9-7 Summary Statistics for Modelled Changes in Sulfate Concentration, Percent Sulfur
Retention, and Delta Sulfate for Watersheds in the Southern Blue Ridge Province 9-46
9-8 Summary Comparison of Watershed Sulfur Status and Model Forecasts
in the Northeastern United States and Southern Blue Ridge Province Using
Typical Year Deposition Data. 9-63
9-9 List of the Chemical Species and Reactions Considered Within the Reuss Model
Framework. 9-78
XVI
-------
TABLES (continued) Page
9-10 Effect of pC02 on Changes Projected to Occur in Surface Water ANC Values at
50 and 100 Years Using the Reuss Model 9-90
9-11 List of Input Data for the Bloom-Grigal Soil Acidification Model 9-104
9-12 Summary Statistics for the Population Estimates of Current ANC Conditions for
Lakes in the NE Region for Five Different Deposition or Soils Aggregation Schemes .... 9-113
9-13 Descriptive Statistics of the Population Estimates for Changes in Lake Water
ANC for Systems in the NE 9-118
9-14 Summary Statistics Comparing the Projections Regarding Changes in Surface
Water ANC Values Obtained Using Different Soils Aggregation Schemes 9-122
9-15 Summary Statistics of the Differences Between the Population Estimates for
Future ANC Projections Made Using the Constant Level and Ramped Deposition
Scenarios 9-123
9-16 Summary Statistics for the Population Estimates of Current ANC Conditions for
Stream Reaches in the SBRP for Four Different Deposition Scenarios 9-126
9-17 Descriptive Statistics of the Population Estimates for Changes in Stream Reach
ANC Values for Systems in the SBRP 9-128
9-18 Summary Statistics of the Differences Between the Population Estimates for
Future ANC Projections Made Using the Constant Level and Ramped Deposition
Scenarios for Stream Reaches in the SBRP 9-133
9-19 Summary Statistics of the Projected Changes in Soil Base Saturations in the
NE Region, Obtained Using the Different Deposition Scenarios or Soil
Aggregation Schemes 9-138
9-20 Summary Statistics of the Projected Changes in Soil pH in the NE Region,
Obtained Using the Different Deposition Scenarios or Soil Aggregation Schemes .9-139
9-21 Summary Statistics of the Projected Changes in Soil Base Saturations in the
SBRP, Obtained Using the Different Deposition Scenarios 9-147
9-22 Summary Statistics of the Projected Changes in Soil pH in the SBRP,
Obtained Using the Different Deposition Scenarios 9-148
9-23 Comparison of the Changes in Soil Base Saturation and Soil pH that Are
Projected to Occur in the NE and SBRP 9-152
9-24 Regionally Weighted Median Values of Initial Annual Deposition Inputs to the
Bloom-Grigal Model for the Northeastern Region and the Southern
Blue Ridge Province 9-156
9-25 Regionally Weighted Median Values of Annual Initial Soil Chemical Values Input
Into the Bloom-Grigal Model for the Northeastern Region and the Southern
Blue Ridge Province 9-159
9-26 Bloom-Grigal Model Regional Projections of the Change in Soil pH in the
Northeastern United States 9-163
9-27 Bloom-Grigal Model Regional Projections of the Change in Percent Base Saturation
in the Northeastern United States 9-165
9-28 Bloom-Grigal Model Regional Projections of the Change in Soil pH in the
Northeastern United States 9-170
9-29 Bloom-Grigal Model Regional Projections for the Change in Percent Base Saturation
in the Northeastern United States. 9-172
9-30 Bloom-Grigal Model Regional Projections for the Change in Soil pH in the
Southern Blue Ridge Province 9-178
9-31 Bloom-Grigal Model Regional Projections for the Change in Percent Soil Base
Saturation in the Southern Blue Ridge Province 9-180
9-32 Summary of the Bloom-Grigal Projected Changes in Soil pH and Percent Base
Saturation in the NE and SBRP Under Constant LTA Deposition 9-183
9-33 Comparison of the Results from the Reuss and Bloom-Grigal Models with Regard to
the Magnitude of Changes in Soil pH and Base Saturation Projected in Soils
of the NE 9-187
9-34 Comparison of the Results from the Reuss and Bloom-Grigal Models with
Regard to the Magnitude of Changes in Soil pH and Base Saturation Projected
in Soils of the SBRP. Results Are Shown for 50 and 100 Years 9-193
xvii
-------
TABLES (continued) Page
10-1 Major Processes Incorporated in the Dynamic Mode! Codes 10-5
10-2 Meteorological Data Required by the Dynamics Model Codes 10-8
10-3 Chemical Constituents in Wet and Dry Deposition Considered
by the MAGIC, ETD, and ILWAS Codes 10-9
10-4 Chemical Constituents Included in Soil Solutions
and Surface Water for the MAGIC, ETD, and ILWAS Codes ". 10-10
10-5 Definitions of Acid Neutralizing Capacity (ANC) Used by the MAGIC, ETD,
and ILWAS Codes 10-11
10-6 Level III Operational Assumptions 10-15
10-7 Comparison of Calibration/Confirmation RMSE for Woods Lake Among
ETD, ILWAS, and MAGIC Models, with the Standard Error of the Observations 10-45
10-8 Comparison of Calibration/Confirmation RMSE for Panther
Lake Among ETD, ILWAS, and MAGIC Models, with the Standard Error
of the Observations 10-46
10-9 Comparison of Calibration RMSE for Clear Pond Among ETD,
ILWAS, and MAGIC Models, with the Standard Error of the Observations 10-47
10-10 Percent Change in RMSE for MAGIC and ETD for a Ten Percent Change in
Parameter Values. Parameters are Ranked in Descending Order of Sensitivity
from Left to Right 10-52
10-11 Watersheds, by Priority Class, for which Calibration Criteria
Were not Achieved 10-68
10-12 Deposition Variations Used in Input Uncertainty Analyses 10-72
10-13 Target Populations for Modelling Comparisons and Population Attributes 10-78
10-14 Descriptive Statistics of Projected ANC, Sulfate, pH, Calcium Plus Magnesium,
and Percent Sulfur Retention for NE Lakes in Priority Classes A -1 Using
MAGIC for Both Current and Decreased Deposition 10-81
10-15 Change in Median ANC and Sulfate Concentrations Over a 40-Year Period as
a Function of the Initial ELS-Phase I or NSS Pilot Survey ANC Groups 10-89
10-16 Descriptive Statistics of Projected ANC, Sulfate, and Percent Sulfur Retention
for NE Lakes in Priority Classes A - E Using MAGIC and ETD for Both Current
and Decreased Deposition 10-96
10-17 Descriptive Statistics for Projected ANC, Sulfate, Percent Sulfur Retention, and
Calcium Plus Magnesium for NE Lakes in Priority Classes A and B Using ETD,
ILWAS, and MAGJC for Both Current and Decreased Deposition 10-117
10-18 Descriptive Statistics of Projected ANC, Sulfate, and Percent Sulfur Retention,
and Calcium and Magnesium for SBRP Streams in Priority Classes A - E Using
MAGIC for Both Current and Increased Deposition 10-146
10-19 Descriptive Statistics of Projected ANC, Sulfate, Percent Sulfur Retention, and
Calcium Plus Magnesium for SBRP Streams in Priority Classes A and B Using
ILWAS and MAGIC for Both Current and Increased Deposition 10-159
10-20 Effects of Critical Assumptions on Projected Rates of Change 10-206
11-1 Weighted Median Projected Change in ANC at 50 Years for Northeastern DDRP
Lakes 11-11
11-2 Lakes in the NE Projected to Have ANC Values <0 and <50 peq L"1 for
Constant and Decreased Sulfur Deposition 11-14
11-3 Weighted Median Projected Change in ANC at 50 Years for DDRP SBRP
Stream Reaches 11-19
11-4 SBRP Stream Reaches Projected to Have ANC Values <0 and <50 ^eq L"1 for
Constant and Increased Suifur Deposition 11-23
XVIII
-------
FIGURES
FIGURE PAGE
1-1 Steps of the Direct/Delayed Response Project (DDRP) approach 1-9
2-1 Activities of the Aquatic Effects Research Program within the National Acid
Precipitation Assessment Program 2-4
3-1 Diagram of sulfur cycle in forest ecosystems 3-22
3-2 Diagram of terrestrial base cation cycle 3-18
4-1 Steps of the Direct/Delayed Response Project (DDRP) approach 4-2
5-1 Northeastern subregions and ANC map classes, Eastern Lake Survey Phase I 5-3
5-2 Representation of the point frame sampling procedure for selecting NSS Stage I
reaches 5-5
5-3 DDRP site locations for Subregion 1A 5-19
5-4 DDRP site locations for Subregion 1B 5-20
5-5 DDRP site locations for Subregion 1C 5-21
5-6 DDRP site locations for Subregion 1D 5-22
5-7 DDRP site locations for Subregion 1E 5-23
5-8 DDRP stream reach study sites in the Southern Blue Ridge Province 5-29
5-9 The pH-ANC relationship for (A) lakes of the ELS Phase I sampling in the
Northeast and (B) DDRP study lakes in the Northeast 5-38
5-10 The pH-ANC relationship for samples with ANC < 400 peq L taken at the
downstream nodes of stream reaches sampled in the NSS 5-41.
5-11 Location of Northeast field check sites and other DDRP watersheds 5-62
5-12 Example of digitization log sheet 5-81
5-13 Example of attribute entry log sheet 5-82
5-14 Definition of soil sampling classes for the DDRP Soil Survey in the Northeast 5-114
5-15 Definition of soil sampling classes for the DDRP Soil Survey in the Southern
Blue Ridge Province 5-116
5-16 Selection of watersheds for sampling 5-118
5-17 Selection of starting points for sampling 5-119
5-18 Field selection of a sampling point for sampling class on a watershed 5-120
5-19 Major steps and datasets from the DDRP database 5-141
5-20 Calculation percentage of regional or subregional area in each soil sampling 5-149
5-21 Relative areas of sampling classes in the Northeast subregions 5-151
5-22 Relative areas of sampling classes in the entire Northeast and Southern
Blue Region Province 5-152
5-23 Aggregated soil van'ables for individual pedons in the Northeast 5-153
5-24 Aggregated soil variables for individual pedons in the Southern Blue
Ridge Province 5-155
5-25 Calculation of cumulative distribution function for a soil variable in a
region or subregion 5-157
5-26 Cumulative distribution functions for pedon aggregated soil variables for the
Northeast and the Southern Blue Ridge Province 5-158
5-27 Sulfur deposition scenarios for the NE and SBRP for Level II and III Analyses. 5-162
5-28 Example of average annual runoff map for 1951-80 5-202
5-29 Flow chart of Darcy's Law soil contact calculation as applied to the DDRP
study sites .- 5-213
xix
-------
FIGURES (continued) Page
7-1 Estimated percent sulfur retention by in-lake processes in drainage lakes in
ELS Region 1 (northeastern United States) 7-7
7-2 Percent sulfur retention for intensively studied sites in the United States and
Canada relative to the southern extent of the Wisconsinan glaciation 7-10
7-3 Model of flow-weighted average concentration calculations for Biscuit Brook 7-16
7-4 Row chart for the determination of internal sources of sulfur using the
steady-state sulfate concentration 7-26
7-5. Scatter plot of the Monte Carlo calculated standard deviation versus the
calculated mean [SO4*1W 7-28
7-6. Comparison of percent sulfur retention calculated using (A) modified-LTA
deposition and (B) modified-LTA deposition adjusted with a 20 percent increase
in dry deposition 7-32
7-7. Population-weighted distribution of projected percent sulfur retention (upper and
lower bounds for 90 percent confidence interval): (A) Northeast; (B) Mid-Appalachians,
and (C) Southern Blue Ridge Province 7-34
7-8. Supplemental watersheds mapped for special evaluation of sulfur retention 7-36
7-9. Population-weighted distributions of projected percent sulfur retention, with upper
and lower bounds for 90 percent confidence intervals, for additional NSS subregions:
(A) Southern Appalachian Plateau, (B) Mid-Atlantic Coastal Plain, (C) Catskills/
Poconos, and (D) Piedmont 7-42
7-10 Combination regional population-weighted distributions of projected percent sulfur
retention, with upper and lower bounds for 90 percent confidence intervals, for
the Northeast, Mid-Appalachians, and Southern Blue Ridge Province 7-44
8-1 Distribution of estimated contact rate using Darcy's Law calculation 8-16
8-2 Distribution of index of contact (yr) using Darcy's Law calculation 8-17
8-3 Scatter plot of ANC.versus contact rate calculated using Darcy's Law 8-19
8-4 Scatter plot of ANC versus index of soil contact calculated using Darcy's Law 8-20
8-5 Scatter plot of ANC versus ln(a/KbTanB) 8-43
8-6 Scatter plot of Ca plus Mg versus ln(a/KbTanB) 8-44
8-7 Scatter plot of pH versus ln(a/KbTanB) 8-45
8-8 Data and regression model development flow diagrams 8-81
8-9 Model development procedure 8-141
8-10 Histograms of unadjusted and adjusted watershed means for selected SBRP
soils variables 8-151
8-11 The mean pH ± 2 standard errors for the SBRP watersheds estimated using
the common aggregation (bars) and the watershed effects adjusted aggregation
(lines) illustrate the lack of variation among the common aggregation values 8-152
9-1 Schematic diagram of extended Langmuir isotherm fitted to data points from
laboratory soil analysis 9-12
9-2 Comparison of measured lake (NE) or stream (SBRP) sulfate concentration with
computed soil solution concentration 9-16
9-3 Historic deposition inputs and modelled output for soils in a representative
watershed in the northeastern United States 9-22
9-4 Schematic of surface water response to changes in sulfur inputs 9-23
9-5 Comparison of measured, modelled and steady-state sulfate for Northeast lake
systems in 1984 9-27
9-6 Projected changes in percent sulfur retention and sulfate concentration for soils
in northeastern lake systems at 10, 20, 50 and 100 years 9-30
9-7 Box-and-whisker plots showing changes in sulfate concentration, percent sulfur
retention, and change in sulfate concentration for soils in northeastern lake
watersheds, using long-term average deposition data 9-31
9-8 Box-and-whisker plots showing changes in sulfate concentration, percent sulfur
retention, and change in sulfate concentration for soils in northeastern lake
watersheds, using TY deposition data 9-32
xx
-------
FIGURES (continued) Page
9-9 Projected time to steady-state concentration for sulfate in northeastern takes (A)
at current deposition and (B) after end of decreasing input in ramp scenario 9-34
9-10 Historic deposition inputs and modelled output for soils In stream systems in
the Southern Blue Ridge Province 9-36
9-11 Comparison of measured, modelled, and steady-state sulfate for stream systems
in the Southern Blue Ridge Province in 1985 9-39
9-12 Comparison of forecasts based on two sulfur deposition datasets for soils in S8RP
watersheds 9-42
9-13 Projected changes in percent sulfur retention and In sulfate concentration for
stream systems in the Southern Blue Ridge Province at 0, 20, 50, 100 and 140 years. . . . 9-44
9-14 Box and whisker plots showing changes in sulfate concentration, percent sulfur
retention, and change in sulfate concentration for soils in watersheds of the
Southern Blue Ridge Province 9-47
9-15 Box and whisker plots showing changes in suifate concentration, percent sulfur
retention, and change in sulfate concentration for soils in watersheds of the
Southern Blue Ridge Province 9-48
9-16 Projected time to 95 percent of steady-state sulfur concentration of Southern Blue
Ridge Province stream systems^ 9-50
9-17 Comparison of model simulation results for DDRP Southern Blue Ridge watersheds 9-53
9-18 Projected base year sulfate concentration with upper and lower bounds for
90 percent confidence intervals for Southern Blue Ridge Province watersheds 9-54
9-19 Projected time to sulfur steady state with upper and lower bounds for 90 percent
confidence intervals in Southern Blue Ridge Province watersheds. 9-56
9-20 Effects of data aggregation on simulated watershed sulfur response for soils in
DDRP watersheds of the Southern Blue Ridge Province 9-57
9-21 Evaluation of alternate soil aggregation procedures for soils in SBRP watersheds. 9-60
9-22 Schematic diagram of the principal process involved in the cycling of base
cations in surficial environments. 9-76
9-23 Plot of the log of the activity of Al vs. soil solution pH for individual soil
samples collected for DDRP 9-83
9-24 Plot of the log of the selectivity coefficient for the calcium-aluminum exchange
reaction vs. the measured base saturation in A/E horizons in the NE 9-86
9-25 Histograms of the (unweighted for the population estimates) projected present-
day ANC values for lakes in the NE 9-87
9-26 Histograms of the (unweighted for the population estimates) projected, present-
day ANC values for lakes in the NE 9-89
9-27 Flow diagram for the one-box Bloom-Grigal soil simulation model 9-97
9-28 Cumulative distribution of projected, present-day ANC values for lakes in the
study population in the NE as projected using Reuss's cation exchange model 9-109
9-29 Scatter plot of the projected, present-day ANC values for lakes in the NE,
obtained using the Reuss model vs. observed (ELS) values 9-110
9-30 Scatter plot of the present-day lake ANC values projected using the Reuss
model in conjunction with the Watershed-Based Aggregation (WBA) soils data vs.
observed (ELS) ANC values 9-115
9-31 Cumulative distribution of the projected surface water ANC values projected for
the study population of lakes in 50 years in the NE 9-116
9-33 Schematic illustration of the titratlon-like behavior displayed by soils in response
to constant loadings of acidic deposition 9-117
9-34 Cumulative distribution of projected present-day ANC values for stream reaches
in the study population in the SBRP, as projections using Reuss's cation
exchange model 9-125
9-35 Scatter plot of the projected present-day ANC values for stream reaches in the
SBRP, obtained using the Reuss model, vs. observed (NSS) values 9-127
9-36 Cumulative distribution of projected changes (at 50 years) in surface water ANC
obtained using the Reuss model for stream reaches in the SBRP 9-130
9-37 Cumulative distribution of projected changes (at 100 years) in surface water ANC
obtained using the Reuss model for stream reaches in the SBRP 9-131
xxi
-------
FIGURES (continued) Page
9-38 Comparison of measured vs. calculated soil pH values for the 580 aggregated
master horizons in the NE 9-136
9-39 Cumulative distribution of projected (a) base saturations and (b) soil pH values for
soils in NE 9-140
9-40 Cumulative distribution of projected (a) base saturations and (b) soil pH values for
soils in the NE 9-141
9-41 Plot of the measured (ELS) ANC values for lakes in the NE vs. the estimated,
watershed-level base saturations for mineral horizons in those watersheds 9-143
9-42 Plot of the changes in surface water ANC values at (a) 20, (b) 50, and (c) 100
years as projected by the Reuss model vs. the estimated, present-day, watershed-
level base saturations for mineral horizons in those watersheds 9-144
9-43 Plot of the projected changes in soil base saturations vs. 9-145
9-44 Cumulative frequencies of changes in (a) soil base saturation and (b) soil pH
for the population of soils in the SBRP 9-149
9-45 Cumulative frequencies of changes in (a) soil base saturation and (b) soil pH for
the population of soils in the SBRP 9-150
9-46 Cumulative distributions of aggregate initial soil pH and percent base saturation
in the NE and SBRP, with and without organic horizons 9-160
9-47 Regional CDFs of the projected change in the pH of soils on NE lake watersheds
under constant and ramp down (30 percent i) deposition scenarios after 20, 50,
and 100 years of LTA, LTA-rbc, and LTA-zbc deposition 9-161
9-48 Regional CDFs of the projected change in the percent base saturation of soils
on NE lake watersheds under constant and ramp down (30 percent 4-) deposition
scenarios after 20, 50, and 100 years of LTA, LTA-rbc, and LTA-zbc deposition 9-162
9-49 Regional CDFs of the projected change in the pH of soils on NE lake
watersheds under constant and ramp down (30% •!•) deposition scenarios
after 20, 50, and 100 years of LTA, LTA-rbc, and LTA-zbc deposition 9-168
9-50 Regional CDFs of the projected change in the percent'-base saturation of soils on
NE lake watersheds under constant and ramp down (30% i) deposition scenarios
after 20, 50, and 100 years of LTA, LTA-rbc, and LTA-zbc deposition 9-169
9-51 Regional CDFs of the projected change in the pH of soils on SBRP stream
watersheds under constant and ramp up (20% t) deposition scenarios after 20,
50, 100, and 200 years of LTA, LTA-rbc, and LTA-zbc deposition 9-176
9-52 Regional CDFs of the projected change in the percent base saturation of soils
on SBRP stream watersheds under constant and ramp up (20% t) deposition
scenarios after 20, 50, 100, and 200 years of LTA, LTA-rbc, and LTA-zbc deposition. . . . 9-177
9-53 Cumulative distributions of changes in soil base saturation for the population of
watersheds in the NE 9-188
9-54 Cumulative distributions of changes in soil pH for the population of watersheds
in the NE 9-190
9-55 Scatter diagrams of the projected changes in base saturation for individual
systems (not population weighted) in the NE obtained from the Reuss and
Bloom-Grigal models 9-191
9-56 Scatter diagrams of the projected changes in soil pH for individual systems (not
population weighted) in the NE obtained from the Reuss and Bloom-Grigal models 9-192
9-57 Cumulative distributions of changes in soil base saturation for the population of
watersheds in the SBRP 9-194
9-58 Distributions of changes in soil pH for the population of watersheds in the SBRP 9-195
10-1 Modelling priority decision tree: Northeast 10-17
10-2 Modelling priority decision tree: Southern Blue Ridge Province 10-19
10-3 Decision tree used to identify watersheds with net chloride export and procedures
for determining chloride imbalance 10-26
10-4 Approach used in performing long-term projections of future changes in surface
water chemistry. 10-29
10-5 Schematic of modelling approach for making long-term projections 10-31
xxii
-------
FIGURES (continued) Page
10-6 Representation of horizontal segmentation of Woods Lake, NY, watershed for
MAGIC and ETD 10-36
10-7 Representation of vertical layers of Woods Lake Basin for ETD 10-37
10-8 Representation of horizontal segmentation of Woods Lake Basin for ILWAS 10-39
10-9 Representation of vertical layers of Woods Lake Basin for ILWAS 10-40
10-10 Representation of vertical layers of Woods Lake, NY, watershed for MAGIC 10-43
10-11 Comparison of population histograms for simulated versus observed (Eastern
Lake Survey Phase I 1984 values) ANC for ILWAS and MAGIC 10-57
10-12 Comparison of population histograms for simulated versus observed (Eastern Lake
Survey - Phase I 1984 values) sulfate concentrations for ILWAS and MAGIC,
Priority Classes A and B 10-58
10-13 Comparison of population histograms for simulated versus observed (Eastern Lake
Survey Phase I 1984 values) ANC and sulfate concentrations for MAGIC, Priority
Classes A - E 10-60
10-14 Comparison of population histograms for simulated versus observed (Eastern
Lake Survey Phasee I 1984 values ) ANC and sulfate concentrations for MAGIC,
Priority Classes A -1 10-62
10-15 Comparison of population histograms for simulated versus observed (NSS Pilot
Survey values) ANC, Priority Classes A and B using ILWAS and MAGIC 10-63
10-16 Comparison of population histograms for simulated versus observed (NSS Pilot
Survey values) sulfate concentrations, Priority Classes A and B using ILWAS and
MAGIC 10-64
10-17 Comparison of population histograms for simulated versus observed (NSS Pilot
Survey values) ANC and sulfate concentrations, Priority Classes A - E using MAGIC . . . 10-65
10-18 Comparison of projection standard errors as a function of ANC (top figure)
and sulfate (bottom figure) concentrations for the NE uncertainty analysis
watersheds using ETD and MAGIC. . . 10-75
10-19 Projections of ANC and sulfate concentrations for NE lakes, Priority Classes
A -1, using MAGIC for 20, 50, and 100 years, under current deposition and a
30 percent decrease in deposition 10-79
10-20 pH projections for NE lakes, Priority Classes A -1, using MAGIC for 20, 50,
and 100 years, under current deposition and a 30 percent decrease in deposition 10-84
10-21 Box and whisker plots of ANC distributions at 10-year intervals for NE Priority
Classes A - I using MAGIC 10-85
10-22 Box and whisker plots of sulfate distributions at 10-year intervals for NE Priority
Classes A - I using MAGIC 10-86
10-23 Box and whisker plots of pH distributions at 10-year intervals for NE Priority
Classes A - I using MAGIC 10-87
10-24 Comparison of population histograms for ANC under current levels of
deposition and a 30 percent decrease in deposition for NE lakes, Priority Classes
A -1, using MAGIC 10-90
10-25 Comparison of population histograms for sulfate concentrations at current levels
of deposition and a 30 percent decrease for NE lakes, Priority Classes A -1, using MAGIC. 10-92
10-26 Comparison of MAGIC and ETD projections of ANC for NE lakes, Priority
Classes A - E. under current and decreased deposition 10-93
10-27 Comparison of MAGIC and ETD projections of sulfate concentrations for NE lakes,
Priority Classes A - E, under current and decreased deposition. 10-94
10-28 Comparison of MAGIC and ETD projections of pH for NE lakes, Priority Classes
A - E, under current and decreased deposition 10-95
10-29 Comparisons of projected change in ANC under current and decreased
deposition for NE Priority Classes A - E, using ETD and MAGIC 10-99
10-30 Comparisons of projected change in sulfate concentrations under current and
decreased deposition for NE Priority Classes A - E, using ETD and MAGIC 10-100
10-31 Comparisons of projected change in pH under current and decreased deposition
for NE Priority Classes A - E, using ETD and MAGIC 10-101
10-32 Box and whisker plots of ANC distributions projected using ETD in 10-year
intervals for NE lakes, Priority Classes A - E 10-103
xxiii
-------
FIGURES (continued) Page
10-33 Box and whisker plots of sulfate distributions projected using ETD in 10-year
intervals for NE lakes, Priority Classes A - E 10-104
10-34 Box and whisker plots of pH projected using ETD in 10-year intervals for NE lakes,
Priority Classes A - E 10-105
10-35 Box and whisker plots of ANC distributions in 10-year intervals using MAGIC for
NE lakes, Priority Classes A - E 10-106
10-36 Box and whisker plots of sulfate distributions in 10-year intervals using MAGIC
for NE lakes, Priority Classes A - E 10-107
10-37 Box and whisker plots of pH in 10-year intervals using MAGIC for NE lakes,
Priority Classes A - E 10-108
10-38 ETD ANC distributions at year 10 and year 50 for NE lakes, Priority Classes A -
E, under current and decreased deposition 10-109
10-39 MAGIC ANC distribution at year 10 and year 50 for NE lakes, Priority Classes
A - E, under current and decreased deposition 10-110
10-40 ETD sulfate distributions at year 10 and year 50 for NE lakes, Priority Classes
A - E, under current and decreased deposition 10-111
10-41 MAGIC sulfate distributions at year 10 and year 50 for NE lakes, Priority Classes
A - E, under current and decreased deposition 10-112
10-42 Comparison of ANC projections using ETD, ILWAS, and MAGIC for NE lakes,
Priority Classes A and B, under current and decreased deposition 10-114
10-43 Comparison of sulfate projections using ETD, ILWAS, and MAGIC for NE lakes,
Priority Classes A and B, under current and decreased deposition 10-115
10-44 Comparison of pH projections using ETD, ILWAS, and MAGIC for NE lakes,
Priority Classes A and B, under current and decreased deposition 10-116
10-45 Comparison of ANC projections under current and decreased deposition for NE
lakes, Priority Classes A and B, at year 20 and year 50 using ETD, ILWAS, and MAGIC. 10-122
10-46 Comparison of sulfate projections under current and decreased deposition for NE
lakes, Priority Classes A and B, at year 20 and year 50 using ETD, ILWAS, and MAGIC. 10-123
10-47 Comparison of pH projections under current and decreased deposition for NE
lakes, Priority Classes A and B, at year 20 and year 50 using ETD, ILWAS, and MAGIC. 10-124
10-48 Box and whisker plots of ANC distributions in 10-year intervals projected using
ETD for NE lakes, Priority Classes A and B 10-126
10-49 Box and whisker plots of ANC distributions in 10-year intervals projected using
ILWAS for NE lakes, Priority Classes A and B 10-127
10-50 Box and whisker plots of ANC distributions in 10-year intervals projected using
MAGIC for NE lakes, Priority Classes A and B 10-128
10-51 Box and whisker plots of sulfate distributions in 10-year intervals projected using
ETD for NE lakes, Priority Classes A and B 10-129
10-52 Box and whisker plots of sulfate distributions in 10-year intervals projected using
ILWAS for NE lakes, Priority Classes A and B 10-130
10-53 Box and whisker plots of sulfate distributions in 10-year intervals projected using
MAGIC for NE lakes, Priority Classes A and B 10-131
10-54 Box and whisker plots of pH distributions in 10-year intervals projected using
ETD for NE lakes, Priority Classes A and B 10-132
10-55 Box and whisker plots of pH distributions in 10-year intervals projected using
ILWAS for NE lakes, Priority Classes A and B 10-133
10-56 Box and whisker plots of pH distributions in 10-year intervals projected using
MAGIC for NE lakes, Priority Classes A and B 10-134
10-57 ETD ANC population distributions at year 10 and year 50 for current and
decreased deposition 10-135
10-58 ILWAS ANC population distributions at year 10 and year 50 for current and .
decreased deposition 10-136
10-59 MAGIC ANC population distributions at year 10 and year 50 for current and
decreased deposition 10-137
10-60 ETD sulfate population distributions at year 10 and year 50 for current and
decreased deposition 10-138
xxiv
-------
FIGURES (continued) Page
10-61 ILWAS sulfate population distributions at year 10 and year 50 for current and
decreased deposition 10-139
10*62 MAGIC sulfate population distributions at year 10 and year 50 for current and
decreased deposition 10-140
10-63 MAGIC ANC and sulfate projections for SBRP streams, Priority Classes A - E,
at year 20, year 50, year 100, and year 200 under current and increased deposition. ... 10-142
10-64 MAGIC pH projections for SBRP streams, Priority Classes A - E, at year 20, year
50, year 100, and year 200 under current and increased deposition 10-143
10-65 Box and whisker plots of ANC distributions in 10-year Intervals projected using
MAGIC for SBRP streams, Priority Classes A - E, for current and increased deposition. . 10-149
10-66 Box and whisker plots of sulfate distributions in 10-year intervals projected
using MAGIC for SBRP streams, Priority Qasses A - E, for current and
increased deposition 10-150
10-67 Box and whisker plots of pH distributions in 10-year intervals projected using
MAGIC for SBRP streams, Priority Classes A - E, for current and increased
deposition 10-151
10-68 MAGIC ANC population distributions at year 10 and year 50 for current and
increased deposition, SBRP streams, Priority Classes A - E 10-153
10-69 MAGIC sulfate population distributions at year 10 and year 50 for current and
increased deposition, SBRP streams, Priority Classes A - E 10-154
10-70 Comparison of ILWAS and MAGIC projections for ANC at years 0, 20, and 50 for
SBRP streams, Priority Classes A and 8, under current and increased deposition 10-156
10-71 Comparison of ILWAS and MAGIC projections for sulfate concentration at years
0, 20, and 50 for SBRP streams, Priority Classes A and B, under current and
increased deposition 10-157
10-72 Comparison of ILWAS and MAGIC projections for pH at years 0, 20, and 50 for
SBRP streams, Priority Classes A and B, under current and increased deposition 10-158
10-73 Box and whisker plots for ANC distributions in 10-year intervals projected
using ILWAS for SBRP streams, Priority Classes A and B, for current and
increased deposition 10-164
10-74 Box and whisker plots for ANC distributions in 10-year intervals projected using
MAGIC for SBRP streams, Priority Classes A and B, for current and
Increased deposition 10-165
10-75 Box and whisker plots for sulfate distributions in 10-year intervals projected
using ILWAS for SBRP streams, Priority Classes A and B, for current and
increased deposition 10-166
10-76 Box and whisker plots for sulfate distributions in 10-year intervals projected
using MAGIC for SBRP streams, Priority Classes A and B, for current and
increased deposition 10-167
10-77 Box and whisker plots for pH distributions in 10-year intervals projected using
ILWAS for SBRP streams, Priority Classes A and B, for current and
Increased deposition 10-168
10-78 Box and whisker plots for pH distributions in 10-year intervals projected
using MAGIC for SBRP streams, Priority Classes A and B, for current and
increased deposition 10-169
10-79 ILWAS ANC population distributions at year 10 and year 50 for current and
increased deposition, SBRP Priority Class A and B streams 10-170
10-80 MAGIC ANC population distributions at year 10 and year 50 for current and
increased deposition, SBRP Priority Class A and B streams 10-171
10-81 ILWAS sulfate population distributions at year 10 and year 50 for current and
increased deposition, SBRP Priority Class A and B streams 10-172
10-82 MAGIC sulfate population distributions at year 10 and year 50 for current and
increased deposition SBRP Priority Class A and B streams 10-173
10-83 Comparison of projected sulfate versus sulfate steady-state concentrations using
ETD. ILWAS, and MAGIC for NE lakes • 10-175
XXV
-------
FIGURES (continued) Page
10-84 Comparison of projected sulfate concentrations under decreased deposition
with the current sulfate steady-state concentrations using ETD, ILWAS, and
MAGIC for NE lakes 10-176
10-85 Comparison of projected sutfate concentrations between models for NE lakes
after 50 years under current and decreased deposition 10-177
10-86 Comparison of projected sulfate versus sulfate steady-state concentrations for
SBRP streams using ILWAS and MAGIC under both current and increased deposition. . . 10-179
10-87 Comparison of projected ANC between models in NE lakes after 50 years under
current and decreased deposition 10-180
10-88 Projected changes in ANC as a function of changes in sulfate for NE lakes
using ETD, ILWAS, and MAGIC for current and decreased deposition 10-181
10-89 Comparison of pH - ANC relationship for each of the models 10-183
10-90 Comparison of projected pH values between models for NE lakes after 50 years
under current and decreased deposition 10-184
10-91 Comparison of projected changes in calcium and magnesium versus changes in
sulfate using ILWAS and MAGIC for NE lakes 10-185
10-92 Change in median ANC, calcium and magnesium, and sulfate concentrations
projected for NE lakes using MAGIC under current and decreased deposition 10-186
10-93 Comparison of the change in pH after 50 years as a function of the initial calibrated
pH for MAGIC, ETD and ILWAS on northeastern lakes 10-188
10-94 Comparisons of projected ANC and sulfate concentrations and pH between
ILWAS and MAGIC after 50 years for SBRP streams 10-189
10-95 Comparison of projected AANC and Asulfate relationships in SBRP Priority
Class A and B streams using ILWAS and MAGIC 10-190
10-96 Comparison of projected AANC and Asulfate relationships and A(calcium and
magnesium) and Asulfate relationships for SBRP Priority Class A • E streams
using MAGIC. 10-191
10-97 Comparison of projected A(calcium and magnesium) and Asulfate relationships
for SBRP Priority Class A and B streams using ILWAS and MAGIC 10-193
10-98 Change in median ANC, calcium and magnesium, and sulfate concentrations
projected for SBRP streams under current and increased deposition using MAGIC 10-194
10-99 Comparison of the change in pH after 200 years as a function of the initial
calibrated pH for MAGIC on SBRP streams, Priority Classes A - E 10-196
10-100 Comparison of projected MAGIC change in pH versus derived pH after 50
years for NE lakes 10-201
xxvi
-------
PLATES
PLATE PAGE
1-1 Direct/Delayed Response Project study regions and sites 1-5
1-2 Sulfur retention and wet sulfate deposition for National Surface Water Survey and
National Stream Survey regions in the eastern United States 1-13
1-3 Changes in sulfur retention in the Southern Blue Ridge Province as projected by
MAGIC for constant sulfur deposition 1-14
1-4 Changes in median ANC of northeastern lakes at 50 years as projected by MAGIC . . 1-18
1-5 Changes in median ANC of Southern Blue Ridge Province stream reaches at 50
years as projected by MAGIC 1-21
2-1 Direct/Delayed Response Project study regions and sites 2-5
5-1. ANC of DDRP lakes by ANC group 5-24
5-2 Final DDRP classification of lake hydrologic type - Subregion 1A 5-32
5-3 Final DDRP classification of lake hydrologic type - Subregion 1B 5-33
5-4 Final DDRP classification of lake hydrologic type - Subregion 1C 5-34
5-5 Final DDRP classification of lake hydrologic type - Subregion 1D 5-35
5-6 Final DDRP classification of lake hydrologic type - Subregion 1E 5-36
5-7 Example of watershed soil map (including pedon site location) 5-75
5-8 Example of watershed vegetation map 5-76
5-9 Example of depth-to-bedrock map 5-77
5-10 Example of watershed land use map 5-78
5-11 Example of watershed geology map 5-79
5-12 Example of 40-ft contour delineations on a 15' topographic map 5-89
5-13 Example of combination buffer: (A) stream and 30-m linear buffer for streams,
(B) wetlands and 30-m linear buffer for wetlands, (C) elevational buffer for lake,
and (D) combination of all preceding buffers 5-91
5-14 ADS and NCDC sites linked with DDRP study sites for NE Subregion 1A 5-166
5-15 ADS and NCDC sites linked with DDRP study sites for NE Subregion 1B 5-167
5-16 ADS and NCDC sites linked with DDRP study sites for NE Subregion 1C 5-168
5-17 ADS and NCDC sites linked with DDRP study sites for NE Subregion 1D 5-169
5-18 ADS and NCDC sites linked with DDRP study sites for NE Subregion 1E 5-170
5-19 ADS and NCDC sites linked with DDRP study sites for the SBRP 5-171
5-20 DDRP study sites relative to distance from Atlantic Coast (<10 km, 10-50 km, >50 km). 5-178
5-21 Pattern of typical year sulfate deposition for the DDRP NE study sites. 5-184
5-22 Pattern of typical year sulfate deposition for the DDRP study sites in Subregion 1A. ... 5-185
5-23 Pattern of typical year sulfate deposition for the DDRP study sites in Subregion 1B. ... 5-186
5-24 Pattern of typical year suifate deposition for the DDRP study sites in Subregion 1C. ... 5-187
5-25 Pattern of typical year sulfate deposition for the DDRP study sites in Subregion 1D. ... 5-188
5-26 Pattern of typical year sulfate deposition for the DDRP study sites in Subregion 1E. ... 5-189
5-27 Pattern of typical year sulfate deposition for the DDRP SBRP study sites 5-190
5-28 Pattern of LTA sulfate deposition for the DDRP NE study sites 5-193
5-29 Pattern of LTA sulfate deposition for the DDRP study sites in Subregion 1A 5-194
5-30 Pattern of LTA sulfate deposition for the DDRP study sites in Subregion 1B 5-195
5-31 Pattern of LTA sulfate deposition for the DDRP study sites in Subregion 1C 5-196
5-32 Pattern of LTA sulfate deposition for the DDRP study sites in Subregion 1D 5-197
5-33 Pattern of LTA sulfate deposition for the DDRP study sites in Subregion 1E 5-198
5-34 Pattern of LTA sulfate deposition for the DDRP SBRP study sites 5-199
xxvn
-------
PLATES (continued) Page
7-1 Median percent sulfur retention by NSWS Subregion 7-35
7-2 Regional percent sulfur retention by major land resource area (MLRA) based on
target populations (ELS and NSS sites) 7-45
11-1 Sulfur retention and wet sulfate deposition for National Surface Water Survey subregions
In the eastern United States 11-2
11-2 Changes in sulfur retention in the Southern Blue Ridge Province as
projected by MAGIC for constant sulfur deposition 11-5
11-3 Changes In median ANC of northeastern lakes at 50 years as projected by MAGIC . . . 11-10
11-4 ANCs of northeastern lakes versus time, as projected by MAGIC for
constant sulfur deposition 11-12
11-5 ANCs of northeastern lakes versus time, as projected by MAGIC for
decreased sulfur deposition. 11-13
11-6 Changes in median pH of northeastern lakes at 50 years as projected by MAGIC ... 11-16
11-7 Changes in median ANC of Southern Blue Ridge Province stream reaches at 50
years as projected by MAGIC 11-18
11-8 ANCs of Southern Blue Ridge Province stream reaches versus time, as projected
by MAGIC for constant sulfur deposition 11-21
11-9 ANCs of Southern Blue Ridge Province stream reaches versus time, as projected
by MAGIC for increased sulfur deposition 11-22
11-10 Changes in pH of SBRP stream reaches as projected by MAGIC 11-24
11-11 Changes in pH of SBRP stream reaches as projected by ILWAS 11-25
xxviii
-------
PRIMARY CONTRIBUTORS TO THE DDRP REPORT
The Direct/Delayed Response Project and this Review Draft Report represent the efforts of many
scientists, technical and support staff. The primary contributors to this report are noted here.
Section 1: Executive Summary
M. R. Church, U.S. Environmental Protection Agency
Section 2: Introduction
M. R. Church, U.S. Environmental Protection Agency
Section 3: Processes of Acidification
P. W. Shaffer, NSI Technology Services Corp.
G. R. Holdren. NSI Technology Services Corp.
M. R. Church, U.S. Environmental Protection Agency
Section 4: Project Approach
M. R. Church, U.S. Environmental Protection Agency
Section 5: Data Sources and Descriptions1
L J. Blume, U.S. Environmental Protection Agency
G. E. Byers, Lockheed Engineering and Sciences Co.
W. G. Campbell, NSI Technology Services Corp.
M. R. Church, U.S. Environmental Protection Agency
D. A. Lammers, U.S.D.A. Forest Service
J. J. Lee, U.S. Environmental Protection Agency
L H. Liegel, U.S.D.A. Forest Service
D. C. Mortenson, NSt Technology Services Corp.
C. J. Palmer, NSI Technology Services Corp.
M. L Papp, Lockheed Engineering and Sciences Co.
B. P. Rochelle, NSI Technology Services Corp.
D. D. Schmoyer, Martin Marietta Energy Systems, Inc.
K. W. Thornton, FTN & Associates, Ltd.
R. S. Turner, Oak Ridge National Laboratory
R. D. Van Remortel, Lockheed Engineering and Sciences Co.
Section 6: Regionalization of Analytical Results
D. L Stevens, Eastern Oregon State University
K. W. Thornton, FTN & Associates, Ltd.
Section 7: Watershed Sulfur Retention
B. P. Rochelle, NSI Technology Services Corp.
M. R. Church, U.S. Environmental Protection Agency
P. W. Shaffer, NSI Technology Services Corp.
G. R. Holdren, NSI Technology Services Corp.
Section 8: Level I Statistical Analyses
M. G. Johnson, NSI Technology Services Corp.
R. S. Turner, Oak Ridge National Laboratory
D. L Cassell, NSI Technology Services Corp.
D. L Stevens, Eastern Oregon State University.
M. B. Adams, Automated Systems Group, Inc.
C. C. Brandt, Oak Ridge National Laboratory
W. G. Campbell, NSI Technology Services Corp.
M. R. Church, U.S. Environmental Protection Agency
G. R. Holdren, NSI Technology Services Corp.
L H. Liegel, U.S.D.A. Forest Service
xxix
-------
Section 8: Level I Statistical Analyses (continued):
B. P. Rochelle, NSI Technology Services Corp.
P. F. Ryan, University of Tennessee
D. D. Schmoyer, Martin Marietta Energy Systems, Inc.
P. W. Shaffer, NSI Technology Services Corp.
D. A. Wolf, Martin Marietta Energy Systems, Inc.
Section 9: Level II Sfngle-Factor Time Estimates1
G. R. Holdren, NSI Technology Services Corp.
M. G. Johnson, NSI Technology Services Corp.
C. I. Liff, Utah State University
P. W. Shaffer, NSI Technology Services Corp.
Section 10: Level III Dynamic Watershed Models
K. W. Thornton, FTN & Associates, Ltd.
D. L Stevens, Eastern Oregon State University
M. R. Church, U.S. Environmental Protection Agency
C. I. Liff, Utah State University
Extramural Cooperators Providing Modelling Expertise and Support:
C. C. Brandt, Oak Ridge National Laboratory
B. J. Cosby, University of Virginia
S. A. Gherini, Tetra-Tech, Inc.
G. M. Hornberger, University of Virginia
M. Lang, Tetra-Tech, Inc.
S. Lee, University of Iowa
R. K. Munson, Tetra-Tech, Inc.
R. M. Newton, Smith College
N. P. Nikolaidts, University of Connecticut
P. F. Ryan, University of Tennessee
J. L Schnoor, University of Iowa
R. S. Turner, Oak Ridge National Laboratory
D. M. Wolock, U.S. Geological Survey
Section 11: Integration and Summary
M. R. Church, U.S. Environmental Protection Agency
P. W. Shaffer, NSI Technology Services Corp.
Contributors to this section listed alphabetically
Beginning on this line, remaining contributors listed alphabetically
XXX
-------
ACKNOWLEDGMENTS
The performance of this portion of the Direct/Delayed Response Project (DDRP) and the
preparation of this report have required the efforts of hundreds of scientists and support personnel. We
acknowledge here a few of those persons who made particularly outstanding contributions. To all the
others who helped us, but who are not named here, we also extend our sincere thanks.
William Ruckleshaus led the way in calling for the initiation of the DDRP and Lee Thomas showed
a continued and very patient interest in seeing that ft was completed properly. We thank them for their
foresight and leadership.
Courtney Rlordan and Gary Foley of the EPA Office of Research and Development (ORD) provided
much encouragement and support for the Project throughout its development and implementation. We
thank them for their appreciation of the technical complexity of the task.
Rick Linthurst, the first Director of the Aquatic Effects Research Program (AERP), played an
absolutely critical role in the development and nurturing of the Project during its early years. We greatly
appreciate his early and continuing commitment to the DDRP. Dan McKenzie, as Director of the AERP,
provided important continuing support for the Project and we thank him for his efforts in helping guide
this phase of the Project to its conclusion.
Tom Murphy, Laboratory Director for EPA's Environmental Research Laboratory-Corvallis (ERL-C),
and Ray Wilhour, Bob Lackey and Spence Peterson, Branch Chiefs for ERL-C, have all supported the
Project and its staff from the first to the last. We thank them for their support.
Dwain Winters and Brian McLean from the Office of Air and Radiation at EPA-Headquarters
provided insight and suggestions for analyses of particular relevance to questions of Agency policy.
We thank them for their interest and assistance.
xxxi
-------
Dixon Landers, Technical Director of the National Surface Water Survey, Jay Messer, Technical
Director of the Pilot Stream Survey, and Phil Kaufmann, Technical Director of the National Stream Survey
and their staffs all provided valuable help in interpreting and correctly using their surface water chemistry
data. We thank especially Tim Sullivan, Joe Eilers, Jim Blick, Mark DeHaan, Alan Heriihy and Mark Mitch.
Jim Omernik (EPA), Andy Kinney (NSI) and Andy Herstrom (NSI) provided many interesting hours
of instruction and discussion on the topics of physical geography and the proper use and application of
Geographic Information Systems. Our efforts in these technical areas have certainly profited from their
valuable advice and counsel.
Bill Fallen (ORD), Chuck Frank (EPA) and his staff, Linda Looney (EPA), and Cindy Burgeson (NSI
Technology Services Corp.) all have provided much administrative assistance to help keep the Project
moving in the right direction and at the pace required. We thank them all for their efforts and assistance.
Many landowners and state and government agencies allowed us to map and sample soils on
their properties. We thank them for permission to do so.
The cooperation of the U.S. Department of Agriculture (USDA) Soil Conservation Service (SCS)
was essential to the completion of the DDRP Soil Survey. People in the SCS state offices who were
responsible for mapping of DDRP watersheds and obtaining the soil descriptions and samples included
Ed Sautter, Roy Shook (Connecticut and Rhode Island); Gene Grice, Steve Hundley (Massachusetts); Dick
Babcock, Bob Joslin, Kenny LaFlamme (Maine); Sid Pilgrim, Henry Mount (New Hampshire); Fred Gilbert,
Keith Wheeler, Will Hanna (New York); Garland Lipscomb, George Martin (Pennsylvania); Dave Van
Houten (Vermont); Talbert Gerald, Bob Wilkes (Georgia); Horace Smith, Andy Goodwin (North Carolina);
Darwin Newton, David Lewis (Tennessee); Niles McLoda (Virginia). In addition, more than 100 soil
scientists were involved in mapping and sampling.
Regional consistency and comparability was greatly assisted by the efforts of people at the SCS
National Technical Centers, especially Oliver Rice, Ted Miller (Northeast) and Larry Ratliff (South). The
xxxii
-------
continuing support of DORP activities by Milt Meyer, Ken Hinkley, and Dick Arnold of the SCS National
Office was extremely helpful.
John Warner, former SCS Assistant State Soil Scientist for New York was the Regional
Correlator/Coordinator of the Soil Survey for both the Northeast and Mid-Appalachian Regions. Hubert
Byrd, former State Soil Scientist for North Carolina, served as RCC for the SBRP Soil Survey.
Elissa Levine and Harvey Luce (University of Connecticut), Bill Waltman and Ray Bryant (Cornell
University), Cheryl Spencer and Ivan Fernandez (University of Maine), Steve Bod/ne and Peter Veneman
(University of Massachusetts), Bill Smith and Lee Norfleet (Clemson University), and Dave LJtzke and
Marilew Battling (University of Tennessee) supervised the operation of the soils preparation laboratories
for the DDRP Soil Survey.
A large and dedicated staff at EPA's Environmental Monitoring and Systems Laboratory-Las Vegas
(EMSL-LV) played an absolutely crucial role in support of the DDRP Soil Survey. Gareth Pearson and
Bob Schonbrod provided supervisory guidance for the DDRP Soil Survey activities at EMSL-LV. Lou
Blume (EPA) served as Technical Monitor for the program and was responsible for delivery of verified
field, soil preparation laboratory, and analytical databases. Lou Blume was responsible for contracting
and management of soil preparation laboratories and analytical laboratories and for the delivery of
operations reports, quality assurance reports, methods manuals and field sampling manuals for the Soil
Survey. Mike Papp of Lockheed Engineering and Sciences Corporation (LESC) was responsible for
delivery of verified field, soil preparation and analytical databases for the Soil Survey. Rick Van Remortel
(LESC) assisted in the verification of the SBRP analytical database and in the preparation of laboratory
operations and quality assurance reports. Bill Cole (LESC) was the Task Lead for the verification of the
analytical database for the NE and assisted in the preparation of the methods manual and quality
assurance report for the NE Soil Survey. Gerry Byers (LESC) assisted in the preparation of methods
manuals and quality assurance reports for the NE and SBRP. Marilew Bartling (LESC) served as the Task
Lead for the verification of Soil Survey data for the SBRP, served as a manager of a soil preparation
laboratory for the SBRP Soil Survey and contributed to the operations and quality assurance reports for
xxxiii
-------
the SBRP. Rod Slagle (LESC) served as the DDRP soils database manager at EMSL-LV. Steve Simon
and Dan Hillman (LESC) assisted in methods development and project implementation early in the
Project. Craig Palmer of the Environmental Research Center of the University of Nevada-Las Vegas
provided invaluable technical assistance on quality assurance of soiis analytical data.
Deborah Coffey (NSI) played a critical role In ensuring the quality of the watershed and soils data
gathered for the Project. She either had a major responsibility for, or assisted in, the development of
data quality objectives, field sampling manuals, laboratory methods manuals, field operations reports, field
quality assurance reports and numerous other facets of the Soil Survey. We thank her for her unswerving
attention to detail. Jeff Kern (NSI) has also assisted in helping to assure the quality of field and
laboratory
Other scientists who made major contributions to the design of the soil survey activities included
Stan Buol (North Carolina State University), John Ferwerda (University of Maine-Orono), Maurice
Mausbach (Soil Conservation Service), Ben Hajek (Auburn University), John Reuss (Colorado State
University), Mark David (University of Illinois), and Fred Kaisaki (Soil Conservation Service).
Phil Arberg (EPA) and Dave WBJiams (LESC) of EMSL-LV were responsible for acquisition and
interpretation of aerial photography of the DDRP watersheds.
Numerous extramural cooperators assisted in this Project. Jack Cosby, George Homberger, Pat
Ryan and David Wolock (University of Virginia), Jerry Schnoor, Tom Lee, Nikolaos Nikdaidis, Konstantine
Georgakakos and Harihar Rajaram (University of Iowa), Steve Gherini, Ron Munson and Margaret Lang
(Tetra-Tech, Inc.), Carl Chen and Louis Gomez (Systech. Inc.) all assisted in watershed modelling
analyses. Bob Newton of Smith College assisted in gathering supplementary watershed data for use in
calibrating the models to the Special Interest lake/watersheds in the Adirondacks. John Reuss and Mark
Walthall of Colorado State University and Tom Voice of Michigan State University performed investigations
of processes of base cation supply and sulfate adsorption, respectively, that assisted us in interpreting
our Soil Survey data and in modelling soil responses. Warren Gebert, Bill Krug, David Graczyk and Greg
xxxiv
-------
Afford of the U.S. Geological Survey (Madison, Wisconsin) supplied runoff data and maps that were
crucial to the Project. Wayne Swank and Jack Waide of the USDA Forest Service cooperated with the
Project in allowing us to use data gathered by the Coweeta Hydrologic Laboratory. Jack Waide also
provided many insights into the workings of watersheds in the Southern Blue Ridge and in the application
of watershed simulation models. Tony Olsen, Sally Wampler and Jeanne Simpson of Battelle Pacific
Northwest Laboratories provided a great deal of Information on estimates of wet deposition to sites of
interest in the Eastern United States. Tony Olsen also assisted in editing text describing analyses of the
wet deposition data. Robin Dennis and Terry Clark of the EPA's Atmospheric and Exposure Assessment
Laboratory-Research Triangle Park and Steve Seilkop of Analytical Services, Incorporated, provided key
information on estimates of atmospheric dry deposition. Steve Lindberg of Oak Ridge National
Laboratory and Bruce Hicks and Tilden Myers of the National Oceanographic and Atmospheric
Administration provided considerable assistance-in the form of discussions and preliminary data on rates
of atmospheric dry deposition. We thank all of these cooperators for their assistance.
No project of the magnitude of the DORP can be successfully completed without the assistance
of peer reviewers. The DORP benefitted immensely from peer review comments all the way from its
inception to the completion of this report.
The following scientists served as reviewers of the initial Review Draft Report: David Grigal of the
University of Minnesota, Peter Chester, R. Skeffington and D. Brown of the Central Electricity Generating
Board (London), Jerry Elwood of Oak Ridge National Laboratory, John Melack of the University of
California - Santa Barbara, Phil Kaufmann of Utah State University, Bruce Hicks of the National
Oceanographic and Atmospheric Administration, Gary Stensland of the Illinois State Water Survey, Jack
Waide of the USDA Forest Service, David Lam of the National Water Research Institute (Burlington,
Ontario), Nils Christophersen of the Institute of Hydrology (Wallingford Oxon, Great Britain), Bill McFee
of Purdue University, Steve Norton of the University of Maine, Scott Overton of Oregon State University,
Ken Reckhow of Duke University, Dale Johnson of the Desert Research Institute (Reno, Nevada), and
Gray Henderson of the University of Missouri. We thank these scientists for their efforts in reviewing a
long and complex document. We especially thank Dave Grigal (Chairman), Jerry Elwood, John Melack
XXXV
-------
and Phil Kaufmann who served on the Overview Committee of reviewers. This report benefitted greatly
from the comments and constructive criticisms of all of these reviewers.
Numerous other scientists also served as reviewers over the years of individual aspects of the
Project or of the Project as a whole. We thank them also for helping us to improve the quality of the
work that we performed.
Dave Marmorek, Mike Jones, Tim Webb and Dave Barnard of ESSA, Ltd. provided much valuable
assistance in the planning of various phases of the DDRP. Their assistance in this planning was
invaluable.
John Berglund of InstaGraphics, Inc. prepared many of the figures that appear in this report. We
thank him for the fine job that he did.
A majority of the word processing throughout the DDRP and, especially, for this report was done
by Carol Roberts of NSI. We thank Carol for her many, many hours of diligent work and for her
forbearance in helping us In our attempts to get everything "exactly right". Significant word processing
support was also provided by Laurie Ippoliti (NSI), Amy Vickland (USDA Forest Service), Lana McDonald,
Rose Mary Hall and Deborah Pettiford of Oak Ridge National Laboratory, and Eva Bushman and Suzanne
Labbe of Action Business Services.
Penelope Kellar and Perry Suk of Kilkelly Environmental Associates performed truly amazing tasks
in editing both the Review Draft and Final Draft of this report. The job could not have been completed
on time without their efforts. Ann Hairston (NSI), Amy Vickland (USDA Forest Service), Susan Christie
(NSI) and Linda Allison (ORNL) also provided important editorial assistance.
The DDRP Technical Director sincerely thanks all of the Project staff and extramural cooperators
for their unquenchable enthusiasm and dedication to seeing that this very tough job was done correctly.
Good work gang...thank you.
xxxvi
-------
SECTION 7
WATERSHED SULFUR RETENTION
7.1 INTRODUCTION
The fate of sulfur deposited in a watershed is important in determining the response of the
associated surface water because sulfate can act as a mobile anion in the soil matrix (see Section 3).
In systems at steady state with regard to sulfur deposition (i.e., inputs = outputs or zero net retention),
the leaching rate of either basic or acidic cations by the "carrier anion' sulfate has been maximized.
Given no increase in sulfur deposition, future acidification (loss of ANC) of these systems would be
determined principally by cation leaching and the possible depletion of the soil exchange complex. In
systems below steady state (i.e., inputs > outputs), the acidifying effect of sulfate-driven cation leaching
has not been maximized. As sulfate leaching increases in these systems, soil adsorption sites are filled
on a net basis and acidification and the rate of acidification increases over time. A circumneutral surface
water draining a watershed with positive net sulfur retention will continue to acidify and might become
acidic (i.e., ANC < 0) as long as rates of sulfur deposition (inputs) exceed outputs. Thus, even if sulfur
deposition decreases, some circumneutral systems will acidify and might become acidic. Knowing the
patterns of watershed sulfur retention, therefore, is important with regard to understanding and forecasting
the potential future effects of sulfur deposition on surface water chemistry. In this section we examine
regional patterns of sulfur retention, as estimated using input/output budget analyses.
The purpose of the watershed sulfur retention component of the DDRP Level I Analyses is to
estimate the current status of annual sulfur retention in watersheds of the eastern United States, with
primary emphasis on the NE, Mid-Appalachian, and SBRP Regions. The Mid-Appalachian Region provides
important information for the interpretation of sulfur retention patterns from the NE to the SBRP. Specific
objectives of this section are to
• examine the influence of in-lake sulfur retention on watershed sulfur retention estimates;
• assess the contributions of internal sources of sulfur to (and the possible influences on) sulfur
Input/output budget calculations;
7-1
-------
* characterize current average annual input/output budgets in the NE, Mid-Appalachians, and
SBRP using (1) data from intensively studied sites and (2) estimates computed using
regionally extensive datasets;
• compare annual sulfur retention patterns within and among regions to determine possible
trends relative to water chemistry, soils, and atmospheric deposition; and
• conduct an uncertainty analysis of the sulfur retention estimates based on the associated
uncertainties of the factors used in the Input/output budget calculations.
7.2 RETENTION IN LAKES AND WETLANDS
7.2.1 Introduction
Section 3.3 described several processes that can cause sulfur to be retained within watersheds.
One of the processes considered is retention by sulfate reduction in wetlands and/or lakes. Retention in
these environments occurs principally by dissimilatory reduction, with sulfate used as an electron acceptor
and with hydrogen sulfide, organic sulfur, or metal sulfides as end products (Rudd et a)., 1986; Brezonik
et al., 1987).
The occurrence of sulfate reduction in anaerobic hypolimnetic waters in lakes has long been
recognized, but has been considered unimportant in long-term sulfur budgets because sulfides are
reoxidized during lake overturn. Recent studies in several locations have shown, however, that sulfate
reduction in (anaerobic) sediments overlaid by toxic lake waters can be a major sink for sulfur (e.g., Cook
et al., 1986; Baker et al., !986a). Reduction rates are approximately first order for sulfate concentration,
and in-lake rates are apparently limited by diffusion rates into sediments (Baker et al., I986b; Kelly et al.,
1987). Sulfides produced in lake sediments are largely retained within the sediment profile on a
permanent basis, with little reoxidation or volatilization (Rudd et al., 1986; Brezonik et al., 1987). Because
sulfate reduction is rate limited (i.e., by diffusion of sulfate) rather than capacity limited (Rudd et al., 1986),
reduction will likely continue roughly at current rates (expressed as percent retention) on a long-term
basis.
Measured and computed mass transfer coefficients for sulfate vary over a relatively narrow range
(Baker et al.. 1986b; Kelly et al., 1987), but the importance of in-lake sulfur retention on lake/watershed
7-2
-------
sulfur budgets is highly variable and is greatly influenced by hydraulic residence times of lakes. Sulfur
retention within lakes has been discussed and modelled by Baker et al. (1986b) and by Kelly et al.
(1987), who developed identical equations to predict sulfate retention in lakes:
*
% SO4 Retention = _ SO4 IUU (Equation 7-1)
kS04
where: kSO4 = sulfate mass transfer coefficient (m yr'1 )
Z =,lake depth (m)
tw = hydraulic residence time (yr)
Baker et al. (I986b) and Kelly et al. (1987) computed mass transfer coefficients using sulfur input/output
budgets from the literature and determined average constants of 0.54 and 0.46 m yr*1 , respectively.
Transfer of sulfate from the water-sediment interface to the anoxic zone of the sediments occurs
principally by diffusion. Thus, absolute transfer rates are relatively low, with the result that reduction in
sediments is a small component of lake sulfur fluxes except in lakes with long hydraulic residence times.
High sulfur retention has been reported for a diverse group of seepage lakes and other lakes with long
hydraulic residence (e.g.. Baker et al., 1988; Schindler et al., I986b; Lin and Schnoor, 1986). In contrast,
Shaffer and Church (1989) evaluated in-lake alkalinity production [to which sulfate reduction is the largest
contributor (Schindler, 1986; Brezonik et al., 1987)] and sulfur retention for regional lake populations in
subregions of the Eastern Lake Survey (ELS) (LJnthurst et al., I986a), and concluded that in-lake
processes have only a minor effect on ANC and sulfur budgets for most drainage lakes in the
northeastern United States, Upper Midwest (UMW), and Southern Blue Ridge (SBR). [For this section
only, SBR refers to lakes In ELS subregion 3A, which includes portions of the Piedmont and Ridge and
Valley provinces (LJnthurst et al., 1986a), and encompasses a larger geographic area than the stream
systems within the Pilot Stream survey region (Messer et al., I986a) of the SBRP considered elsewhere
in the DDRP.J
Dynamics of sulfur in freshwater wetlands have been studied in detail at only a few sites and
probably cannot be described effectively at regional scales by relationships such as the in-lake retention
7-3
-------
expression (Equation 7-1). Rates of sulfur reduction In wetlands can be very high (Welder and Lang,
1988) and, even small wetland areas, depending on their location within a watershed, can retain a
substantial fraction of watershed sulfur inputs (Calles, 1983; Wetder and Lang, 1988). Generalization of
wetland area - sulfur budget relationships is difficult, however, because the importance of wetland
retention on watershed sulfur budgets depends on the location of the wetland in the watershed and the
portion of watershed runoff flowing through it. Also, sulfur reactions In wetlands and wet soils can
change seasonally or in wet/dry years. Wetlands and wet soils can act as sulfur sinks (reduction of
sulfur) during wet periods when the system is anaerobic, but can become major sulfate sources due to
reoxidation of sulfldes upon drying (Bayley et ai., 1986; Nyborg, 1978).
In this section, we use the sulfur retention model of Baker et al. (1986b) with hydrologic data from
the ELS (Linthurst et al., 1986a; Kanciruk et al., I986a) to estimate sulfur retention in drainage lakes
(including reservoirs) in the northeastern United States and the Southern Blue Ridge. Because we lack
models to make direct estimates of sulfur reduction in wetlands, regression analyses are used to describe
relationships between watershed sulfur input/output budgets and wetlands for DDRP watersheds. Results
of these analyses are described in Sections 7.4 and 8.5.
7-2.2 Approach
The ELS characterized lake depth and hydraulic residence time for a statistically representative set
of lakes in selected areas of the eastern United States, including the Northeast (Linthurst et al., 1986a;
Kanciruk et al., 1986a). For these analyses, we used a subset of the ELS population comprised of all
drainage lakes and reservoirs with lake areas <2000 ha. Target populations are listed in Table 7-1.
Using ELS data with Equation 7-1 and assuming a value of 0.5 m yr'1 for kSO4 (Baker et al., 1986b;
Kelly et al., 1987), we estimated sulfur retention by in-lake reduction for drainage lakes in the northeastern
United States and for DDRP watersheds. Due to major uncertainties in defining hydrologic boundaries
for seepage and closed lakes and resulting uncertainties in hydrologic and chemical budgets, estimates
of in-lake sulfur retention were made only for drainage lakes and reservoirs. Based on the sampling
7-4
-------
Table 7-1. Summary of Computed Sulfur Retention by In-lake Reduction for Lake Systems in the
Eastern United States. Data for the Southern Blue Ridge and Upper Midwest are from Shaffer and
Church (1989)
Region Drainage Lakes* twb Computed % S retention % of lakes with > 10 percent
# % (yr) median 90 %ile maximum computed S retention
ELS Region 1
1A
1B
1C
10
1E
6288
1091
1421
1276
1071
1429
(88)
(87)
(96)
(86)
(81)
(94)
.20
.23
.25
.17
.18
.23
3.1
2.8
3.9
2.5
3.7
3.0
11.1
9.0
12.8
7.9
12.5
11.1
38.8
25.6
38.8
19.4
21.1
26.6
12.5
7.7
19,1
7.9
17.3
14.4
NE DORP lakes 137 (94) .46 4.2 12.5 25.6 18.6
SBR (ELS 3A) 250 (97) .10 1.2 4.0 5.4 <.1
UMW (ELS 2) 4404 (52) .48 5.3 13.0 19.3 23.2
* 'Drainage Lakes* indicates drainage lakes and reservoirs; # is target population, % is percentage of all lakes in the H.S target
population in each region.
b Hydrologic retention time (yr).
7-5
-------
design described by Unthurst et al. (1986a), we extrapolated results from sampled lakes to obtain target
population estimates for each region and for the five ELS subregions in the Northeast.
7.2.3 Results
Estimates of sulfur retention for drainage lakes and reservoirs in ELS Region 1 (northeastern United
States, Figure 5-1) are summarized in Table 7-1 and in Figure 7-1. Computed in-lake retention was
generally low, with a median retention in the NE of 3.1 percent and more than 10 percent sulfur retention
in only 12.5 percent of northeastern drainage systems. Maximum computed retention in northeastern
lakes was 39 percent. Retention in individual ELS subregions was comparable to the region as a whole;
retention in Subregions 1A (Adirondacks) and 1C (Central New England) was slightly lower than the
regional distribution, and retention in Subregions 1B (Poconos/Catskills) and 10 (Southern New England)
slightly higher. Because drainage lakes and reservoirs comprise 88 percent of lakes in the region and
at least 81 percent of target lakes in individual subregions, retention data summarized here represent by
far the majority of target lake systems in the region. Computed retention for DDRP lakes is generally
comparable to, but is slightly higher than, that of the regional target lake population, but generally
comparable. The fraction of drainage systems in the DDRP lakes is higher than for the ELS population
estimate, due principally to reclassification of several DDRP lakes (from closed or seepage to drainage)
based on data from DDRP watershed mapping activities (Section 5.3).
For comparison, data from Shaffer and Church (1989) for two other ELS regions are also included
in Table 7-1. Lakes in the SBR (ELS Subregion 3A) are dominated by drainage systems and reservoirs,
which have very short hydraulic residence times and are consequently projected to have very low in-
lake sulfur retention. Median computed retention in SBR lakes is only 1.2 percent, and maximum
retention is 5.4 percent. Estimated retention in lakes of the UMW (ELS Region 2) is somewhat higher
than in the NE, with median projected retention of 5.3 percent and more than 10 percent retention in
almost one-fourth of drainage systems. An important difference between the NE and UMW lies in the
relative abundance of lake hydrologic types; seepage and closed lakes account for almost half of all lakes
in the UMW and in-lake processes are probably an important sulfur sink in most of these lake systems.
7-6
-------
Percent Sulfur Retention
Figure 7-1. Estimated percent sulfur retention by in-lake processes in drainage lakes in ELS
Region 1 (northeastern United States). Retention was computed using the model of Baker et al.
(1986b).
7-7
-------
The estimates of low sulfur retention in northeastern lake systems are consistent with independent
lines of evidence regarding watershed sulfur budgets and in-lake processes. Our estimates of low sulfur
retention, consistent with sulfur input/output budgets developed by Rochelle and Church (1987) and
discussed in Section 7.3, show lake/watershed systems in the region to be, on average, very close to
steady state. Data presented here also are consistent with estimates of Shaffer et al. (1988) and Shaffer
and Church (1989), based on watershed-to-lake area ratios for ELS watersheds, which suggest that in-
fake processes (principally sulfate reduction) are a minor contributor to ANC budgets in most northeastern
lake/watershed systems.
The relative importance of in-lake suifate reduction to basin sulfur budgets In most systems is
largely determined by two factors: (1) absolute rates of sulfate reduction and (2) lake hydrologic variables
(more explicitly, the volume of water into which alkalinity is released, or the annual discharge per unit lake
area). Rates of sulfate reduction (as k^ apparently vary among lakes over a fairly narrow range (Rudd
et al., 1986; Kelly et al., 1987; Brezonik et al.. 1987) and in typical drainage lakes of the eastern United
States are probably comparable to rates measured in systems in which reduction is a major component
of sulfur budgets (Brezonik et al.. 1987; Kelly et al., 1987). Hydraulic residence times of lakes, however,
vary greatly among regions. For example, at the Experimental Lakes Area in Ontario and in many
seepage lakes (e.g., Schindler et al., 1986b; Un and Schnoor, 1986; Baker et al., 1986a), residence times
are long and sulfur budgets are greatly influenced by in-lake reduction. By contrast, residence times in
most drainage lakes of the northeastern United States are short, averaging about two months (Linthurst
et al., 1986a). The relatively minor role of in-lake reduction in drainage lakes of the northeastern United
States is a consequence of short hydraulic residence times, rather than of low inherent sulfate reduction
rates. The importance of residence time is explicit in the models of Baker et al. (1986b) and of Kelly et
al. (1987). Those authors concluded that in lakes with short hydraulic residence times (one year or less),
including most lakes in the northeastern United States, in-lake processes have little net effect on
watershed sulfur budgets.
7-8
-------
7.3 WATERSHED SULFUR RETENTION
Our first investigation of the regional patterns of sulfur retention consisted of a review of sulfur
input/output budgets at intensively studied sites (Rochelle et al., 1987). Figure 7-2 summarizes the
findings from this review. Definitive statements about sulfur retention on regional scales could not be
made because of lack of spatial coverage by the intensively studied sites and inconsistencies in data used
for budget calculations. There are trends, however, in sulfur retention from North to South in the eastern
United States, especially relative to the extent of the Wisconsinan glaciation, with higher retention in the
southern areas (Figure 7-2). The DDRP Level I sulfur retention analysis examines these apparent trends
in more detail using regionally consistent sulfur input and output data (Section 5) for the surface water
sites sampled by the Eastern Lake Survey (ELS) and National Stream Survey (NSS).
7.3.1 Methods
7.3.1.1 Input/Output Calculation
In the Level I sulfur retention analysis, we use an annual mass balance approach to estimate
percent retention. The general equation used to calculate percent sulfur retention is:
% Retention = ((Sw + Sd ) - (R * Ss )/(Sw + Sd ))*100 (Equation 7-2)
where: Inputs
Sw = wet sulfur deposition (mass length* time' )
Sd = dry sulfur deposition (mass length"2 time )
Outputs
R = runoff (length time )
Ss = surface water sulfur (mass length )
Equation 7-2 relates the total sulfur input (on a mass basis) to each watershed to the total sulfur output.
We applied this equation to each of the study watersheds examined in the Level I sulfur retention
analyses (ELS and NSS sites).
7-9
-------
7-10
-------
7.3.1.2 Data Sources
7.3.1.2.1 Inputs -
Wet sulfur deposition was estimated for each site using chemistry data from the National Trends
Network/National Acid Deposition Program (NTN/NADP) network and precipitation data from the NOAA
National Climatic Data Center (NCDC) network (Section 5.6). Briefly, wet sulfate concentrations and
precipitation were kriged to each site, and wet deposition was calculated (see Wampler and Olsen, 1987,
for a detailed description of the calculation). Dry sulfur deposition was estimated based on output from
the Regional Acid Deposition Model (RADM) (see Section 5.6).
7.3.1.2.2 Outputs -
We used estimates of annual runoff for the 30-year period of 1951-80 (see Section 5.7 for details).
For the purpose of these analyses we assumed that the vast majority of sulfur leaves the watershed in
the form of dissolved sulfate (David and Mitchell, 1985; Mitchell et al., 1986). Section 5.3 discusses the
chemistry data used in these analyses. For additional information concerning the ELS and NSS surface
water sulfate estimates, see Linthurst et al. (1986a), Messer et al. (1986a), and Kaufmann et al. (1988).
Seepage lakes and closed lakes were excluded from the analyses.
7.3.2 Uncertainty Estimates
7.3.2.1 Introduction
We used a Monte Carlo approach to evaluate the uncertainty associated with estimates of annual
average sulfur retention. (The specific Monte Carlo procedure used is very similar to that described in
Section 6.3.) The critical step in applying the Monte Carlo routine is developing error rates on each of
the input/output variables used in calculating percent sulfur retention (see Equation 7-2). We determined
an uncertainty distribution for each of these variables. The uncertainty distributions were propagated
through the retention equation to determine an estimate of the overall uncertainty of the percent sulfur
retention calculations.
7-11
-------
7.3.2.2 Individual Variable Uncertainties
7.3.2.2.1 Input variables -
Two variables are used to estimate the total sulfur input to each surface water system - wet and
dry sulfur deposition. The determination of uncertainty estimates for these variables is discussed in
Section 10.10. For the sulfur retention uncertainty analyses, we used relative standard deviation (RSD)
estimates of 0.25 for Sw and 0.50 for Sd.
7.3.2.2.2 Output variables -
7.3.2.2.2.1 Runoff-
The sulfur output from each watershed is a product of the estimated annual average surface water
sulfate concentration and the annual runoff. Rochelle et a!, (in press) determined that runoff for individual
watersheds could be estimated from the map of Krug et al. (in press) within ± 15 percent. Based on
this determination, we used an RSO of 0.15 for runoff in the sulfur retention uncertainty analysis.
7.3.2.2.2.2 Surface water sulfate concentration -
We estimated the annual average surface water sulfate concentration from the single fall index
value for the northeastern lakes (Section 5.3) or an average of three (Pilot Stream Survey) or two (NSS
Phase I) spring baseflow samples for the SBRP and Mid-Appalachian streams, respectively (Section 5.3).
As described below, we used extensive temporal data from intensively studied sites to estimate the
variability arising from using an index to represent average annual sulfate concentrations. Table 7-2 lists
the sites from which data were available and the frequency of data collection at each site.
First, we calculated flow-weighted annual averages for each year for each site and a spring and/or
fall flow-weighted average concentration. The fall and spring flow-weighted averages were calculated
using sulfate concentrations for samples collected during pen'ods that corresponded to the sampling
windows used in the ELS (mid-September to early November) (LJnthurst et al., 1986a) and NSS (March
15 to May 15) (Messer et al., I986a; Kaufmann et al., 1988). An additional criterion defining the NSS
sampling window was to sample prior to spring leaf-out. The spring samples collected by the NSS were
7-12
-------
Table 7-2. Intensively Studied Sites Used in Surface Water Chemistry Uncertainty Analysts
Site Name
Years of Sample Frequency
Study for Chemistry
Reference
Northeast
ILWAS
Woods Lake 5
Panther Lake 5
Sagamore Lake 5
RILWAS
Arbutus Lake 3
Black Pond 4
Bub/Sis 3
Darts Lake 3
Moss Lake 2
Pancake 1
Rondaxe Lake 4
Townsend Lake 1
West Lake 4
Windfall 3
dear Pond 4
Heart Lake 2
Otter Lake 2
SBRP and Mid-Appalachians
Coweeta 34 6
Coweeta 36 11
Deep Run 6
White Oak Run 5
Femow 10
Biscuit Brook 4
Shenandoah Nat Park 1
(52 streams)
weekly
weekly
weekly
monthly8
monthly
monthly
monthly
monthly
bimonthly
monthly
monthly
monthly
monthly
monthly
monthly
monthly
weekly
weekly
weekly
weekly
weekly
variable*
bimonthly
Goldstein,
pers. comm.
Driscoll,
pers. comm.
Waide,
pers. comm.
Galloway,
pers. comm.
Helvey,
pers. comm.
Lynch and
Dise, 1985
a Samples were collected between 10 to 13 times per year.
b Biscuit Brook is an episodic study site. Samples collected periodically through out each year, however, during selected events
extensive water chemistry samples were taken; often on an hourly time basis.
7-13
-------
non-event samples (i.e., baseflow). To maintain consistency we checked the weekly data used from
the intensively studied SBRP and Mid-Appalachian sites (Table 7-2) to ensure that no samples that were
unduly influenced by events were included in the spring flow-weighted average calculations.
Sulfur budgets for SBRP watersheds might be biased to some extent by their reliance on
streamwater sulfate concentration data collected during spring baseflow periods. Spring baseflow
chemistry closely approximates weighted mean annual chemistry computed from weekly grab samples
for many southeastern systems. However, data representing precipitation/snowmelt episodes were not
collected as part of the Pilot Stream Survey. For the few watersheds In the Southeast for which at least
some episodes have been characterized, there has been a consistent trend of increased sulfate
concentration during storm episodes [Deep Run and White Oak Run, VA (Hendrey et al., 1980; P.W.
Shaffer, unpublished data), Femow, WV (D. Helvey, personal communication), Walker Branch, TN
(Johnson and Henderson, 1979), Coweeta, NO (Swank and Waide, 1988), Panola Mountain, GA (N. Peters
and R. Hooper, personal communication)]. Due to the highly variable extent of episodic sulfate increases
and the extremely limited data available for the region, the episodic bias in sulfur budgets for the
Southeast cannot be quantified. In one system in the SBRP for which detailed sulfate export budgets
have been determined (Coweeta WS #2, three years of data), sulfate export calculated from flow-
proportional sampling was 19 percent higher than export calculated from weekly grab sample date (Swank
and Waide, 1988). The only other watershed in the region for which comparable analyses have been
completed is Panola Mountain, GA. Panola, located in the Piedmont near Atlanta, is physiographically
and climatically different from the DDRP watersheds in the SBRP, and is subject to extreme episodic
increases in sulfate following prolonged dry periods. Estimation of annual sulfate export from spring
baseflow samples at Panola appears to underestimate total annual sulfate export by as much as 50
percent (N. Peters and R. Hooper, unpublished data). The bias observed at Panola should be regarded
as an upper bound that might be approached by a few SBRP systems. The climate and moisture
regimes of SBRP watersheds are more similar to those at Coweeta than to the more xeric conditions at
Panola and at Walker Branch, TN (which also experiences large episodic increases in sulfate but with
uncertain effects on sulfate export budgets; Johnson and Henderson, 1979), suggesting that the 19
7-14
-------
percent bias observed at Coweeta is probably not atypical of SBRP watersheds (J. Waide, personal
communication).
Data from two of extensively studied sites, Biscuit Brook, NY, and Shenandoah National Park
(SNP), VA, required special considerations. The Biscuit Brook data were collected as part of a program
to monitor events. As a result, the dataset contained sections of very extensive temporal data (hourly)
along with more infrequent sampling through the year. For this dataset, the flow-weighted annual average
sulfate concentration was estimated by calculating the area under the hydrograph to properly weight the
influence of any particular event flow value on the overall average (Figure 7-3). We were not able to
perform a complete hydrograph separation analysis (such as described by Dunne and Leopold (1978))
due to the highly variable temporal sampling of the flow measurements. To determine the spring flow-
weighted average we used a flow of 10 cfs as the maximum flow that could be regarded as equal to
baseflow. The 10-cfs "limit" was determined after examination (simple hydrograph separation) of four
years of available data.
The SNP dataset contains bimonthly flow and water chemistry data for one year. We calculated
the flow-weighted annual average using the six flow and concentration measures. Two of the six values
fell within or were close to the March 15 to May 15 time frame used to calculate estimates of spring
baseflow sulfate concentration. The two samples were collected near the beginning and the end of the
period (March 15-19 and May 17-20, respectively). Although the March sample was barely within the
period, there was evidence that the flows were higher than the usual spring baseflow values for the SNP
area (P. Shaffer, personal communication). The May sample was well after leaf-out and the concentration
values were low compared to more extensive data available for Deep Run and White Oak Run. [These
two watersheds, located in the SNP, are included in the 52-site SNP dataset. They also have been
extensively monitored as part of the Shenandoah Watershed Acidification Survey (SWAS) (P. Shaffer,
personal communication).] Although the March sample had very high flows, as noted above, we used
it, rather than the sample from May 15, for the error analysis. The March sample sulfate concentration
7-15
-------
i : i
I I 1 I I
I I i I I I
I I I I I I
't t* t,
t = time
f = flow
c « sulfate concentration
Avg.c=
Figure 7-3. Model of flow-weighted average concentration calculations for Biscuit Brook.
7-16
-------
was closer than the May sample to the spring flow-weighted average sulfate concentrations for Deep Run
and White Oak Run.
After we calculated the two flow-weighted averages (annual average and fall average for lakes or
spring average for streams) for all years of data for the intensive study sites listed in Table 7-2, we then
calculated an estimate of the percent difference (%Diff) between the two averages, as described in
Equation 7-3. %Diff = ((lnd_Avg - Ann_Avg)/lnd_Avg)*100 (Equation 7-3)
where: lnd_Avg = flow-weighted average sulfate concentration for the
index sample time frame (spring or fall)
Ann_Avg - the flow-weighted annual average sulfate concentration
In the final step, we used the estimate of %Diff for each watershed and year to determine an
estimate of the uncertainty associated with using the fall index value or the spring baseflow estimate to
represent average annual chemistry for the sulfur retention analysis. First, we determined the distribution
of %Diff for each set of data (e.g., ILWAS, RILWAS, Femow). Next, we estimated an appropriate
uncertainty estimate to be used in the uncertainty calculations for the sulfur retention analyses using the
standard deviations around the mean %Drff for each of the intensively studied datasets (Table 7-3). The
combined dataset had a mean value only slightly above zero and was slightly skewed to higher values.
Both of these aspects can be attributed to the SNP data, which are probably somewhat high because
March data were used. The overall distribution was approximately bell shaped, with over 95 percent
inside ±2 standard deviations of zero. Therefore, a log normal distribution with an RSD of 9.4 percent
was used to describe the uncertainty.
7.3.2.3 Uncertainty Calculation - Monte Carlo Analysis
Once uncertainty estimates were obtained for each of the input/output variables, the next step was
to combine the information to obtain an overall estimate on uncertainty on the percent retention estimate.
We did this using Monte Carlo analysis. The basic strategy employed in the Monte Carlo analysis was
7-17
-------
Table 7-3. Summary Statistics on Percent Differences Between Flow-weighted Average Annual
Sulfate Concentration and the Fall/Spring Flow-weighted Averages
Study Site Na
Mean
Median Std.
Dev.
Min
Max
NE
RILWAS 12 -1.1 -3.6 10.0 -9.7 15.6
ILWAS 32 -3.0 -3.0 8.4 -36.4 8.9
SBRP & Mid-
Appalachian
Femow
Coweeta-34
Coweeta-36
Biscuit
SNP w
SWASb
10
6
11
3
52
11
-6.7
-8.5
-1.8
4.2
8.9
-3.3
-5.7
-7.5
-1.2
1.5
8.6
-2.4
7.5
6.0
10.5
5.6
5.4
5.5
-19.0
-15.9
.. -28.0
0.4
0.0
-14.7
4.5
0.0
8.7
10.6
23.3
5.0
137 1.4 1-5 9.5 -36.4 23.3
" N is a combination of the number of years of data and the number sites.
° SWAS includes White Oak Run and Deep Run and stands for the Shenandoah
Watershed Acidification Study.
7-18
-------
to randomly select for each iteration a value for each input/output variable (e.g., runoff) from the
distribution of possible values as determined by the associated uncertainty for that variable. These
randomly selected input/output variables were then used to calculate an estimate of percent sulfur
retention for that particular iteration. We randomly selected 11 watersheds from the study regions and
ran the Monte Carlo uncertainty using 10,000 iterations. (This number of iterations was chosen based
on the simplicity of the sulfur retention equation and the computer CPU required. We performed several
tests to evaluate the Influence of the number of iterations on convergence of the mean and standard
deviation and found that results were generally the same with significantly less than 10,000 iterations.)
The overall uncertainty of the percent retention estimate was determined from the variance of the percent
sulfur retention estimates calculated from the Monte Carlo iterations.
Based on the results of the Monte Carlo, we determined that a multiplicative normal distribution best
described the percent sulfur retention uncertainty. Finally, we plotted the standard deviation of the Monte
Carlo runs for each watershed against the average percent retention from the runs. Equation 7-4 presents
the results of the linear regression that describes this relationship.
Std. Dev. = 30.1 - 0.30 (Avg.) (Equation 7-4}
R2 = 0.99
Prob>F = 0.0001
MSE = 0.15
This relationship, along with estimated variance, was used to calculate a 90 percent confidence interval
about the percent sulfur retention population estimates presented in Section 7.3.4 (see Figures 7-7 and
7-10).
7.3.3 Internal Sources of Sulfur
7.3.3.1 Introduction/Approach
Sources of sulfur within a watershed can be important factors affecting the interpretation of the
annual percent sulfur retention estimates calculated using the input/output budget analyses. In the
7-19
-------
DDRP we are interested in percent sulfur retention relative to sulfur deposition. Sulfur from sources
such as acid mine drainage, natural weathering of sulfide-bearing bedrock, or sulfate-containing
sedimentary rocks can increase the surface water concentration of sulfate, thus biasing the results of the
annual input/output budget calculations. We have used two approaches to identify watersheds with
internal sources of sulfur. The first approach uses Information on bedrock geology to identify watersheds
associated with sulfur-bearing bedrock. In the second approach we determine an estimated surface water
sulfate concentration for each site that, if exceeded, indicates (at a designated probability level) that at
least some sulfate Is derived from internal sources. This calculation Is based on the determination of
theoretical steady-state sulfate concentrations. This section describes the methods used for, and the
results of, (1) the bedrock geology analyses and (2) the computation of an upper limit steady-state sulfate
concentration.
7.3.3.2 Bedrock Geology
The first step in the bedrock analysis was to identify the types of bedrock within each of the DDRP
watersheds. The DDRP subset of watersheds (NE=145, SBRP=35) within the ELS lake and NSS stream
populations was selected to test whether the approach could be used to identify systems with potential
internal sources of sulfur. Using the QIS, we overlaid watershed boundaries onto state geology maps
(Section 5.4.1.7.3.1 and 5.4.2.7.2.1) and then identified the mapped bedrock units within the boundaries
(Plate 5-11). State geology maps used in the analysis are listed in Table 7-4.
After we identified the mapped units associated with a watershed, we then assessed the potential
for each unit to contribute sulfur to surface waters. We developed a three-level stratification for classifying
each bedrock type. Mapped units with high probabilities for contributing sulfur were assigned the value
"Y°. Primarily, these units consisted of calcareous rocks or of rocks identified in the state map legends
as "sulfitic", "pyrite-bearlng", or a similar description. Bedrock units containing potentially large amounts
of sulfur, but with more limited contact with surface waters, were assigned a value of T". These units,
consisting of black and gray shales, sulfitic slates, fossil'rferous sediments (potential carbonate sources),
and "rusty weathering* metasediments, all probably contain substantial amounts of mineral sulfides.
7-20
-------
Table 7-4. Bedrock Geology Maps Used in the DORP
Internal Sources of Sulfur Bedrock Geology Analyses
State
Scale
Reference
CT
GA
MA
ME
NH
NY
NC
PA
Rl
SC
TN
VT
WV
1:250,000
1:5000,000
1:250,000
1:500,000
1:250,000
1:250,000
1:500,000
1:250,000
1:250,000
1:250,000
1:250,000
1:250,000
1:250,000
Rodgers (1985)
Pickering and Murray (1976)
Zen (1983)
Osberg et al. (1985)
Billings (1980)
Isachsen and Fisher (1970)
Brown (1985)
Miles (1980)
Quinn (1971)
Overstreet and Bell (1965)
Hardeman (1966)
Doll et al, (1961)
Cardwell et al. (1968)
7-21
-------
Because of limited permeabilities, however, the units in most cases will retain the native sulfur unless the
bedrock has been disturbed (e.g., by quarrying or mining operations). All other bedrock types were
assigned a classification of "NT, indicating a low potential for supplying sulfur to local surface waters.
Table 7-5 summarizes the classification scheme.
These are two caveats to the above classification system. First, we assigned the classifications
Independent of potential weathering rates. Although both the rapidly weathering bedrock (e.g., limestone)
and the more resistant material (e.g., sulfitic schist) are assigned the same code, the more highly
weatherable rock yields a higher flux of sulfur per unit time. Second, we assigned the classifications
based on data compiled at a state map level. This latter fact causes two potential problems. First,
because of the scales of the state geology maps, local concentrations of sulfide-bearing bodies are
frequently not delineated; therefore, potential local sulfur sources in individual watersheds are not always
identified. Second, as a result of correlation difficulties, the location of contacts between contiguous
units might not be depicted accurately on the watersheds. This could result in the mis-identification of
the presence (or absence) of sulfur-bearing units on a particular watershed.
Using the above classification scheme, we formulated and tested the hypothesis that watersheds
having large area! percentages of bedrock falling into the "Y" and "P" groups would more likely have
excess sulfur appearing in the input/output budgets (i.e., net negative retention of sulfur). Evaluation
of this hypothesis, however, indicated no significant correlation between net sulfur retention and the group
classification.
We attribute the lack of a correlation between these variables to several factors. First, DDRP
watersheds were selected and stratified based on lake ANC (Section 5.2). No systems with ANC values
greater than 400 j/eq L"1 were included in the northeastern sample population, thereby effectively
eliminating from the sample most watersheds with carbonate-bearing bedrock. In those watersheds with
carbonate-bearing bedrock, the fraction of areal coverage is generally sufficiently small to mask any
internal contributions to the sulfur budgets. As a result, because of the restrictions of our target
7-22
-------
Table 7-5. Potential for Sulfur Contribution by Geologic Type
Sulfur Contribution Geologic Type
Potential
Y Calcareous
Limestones
Dolostones
Sulfitic
Marbles
Carbonaceous
Pyrrte-bearlng
P Black/gray shales
Fossiliferous
Rusty weathering (schists)
N Alt other types (includes sandstones,
conglomerates, most metamorphics,
igneous, etc)
7-23
-------
population and, thus, sample, we do not get an evenly balanced sample of the different bedrock types.
Second, DDRP watersheds acting as large net sources for sulfur (e.g., 101-093, 1E1-123) generally are
associated with major disturbances (e.g., quarrying operations). This observation suggests local sources
for the sulfur and, thus, information not identified on state geology maps. As noted above, the
disturbances probably enhance the flux of sulfur from bedrock to the surface water, magnifying the
internal contributions to the sulfur budgets. Finally, watersheds exhibiting modest excess sulfur fluxes,
but associated with "N"-type bedrock, probably reflect unidentified local sources of sulfur. Again, the
discrepancy could result from scale problems with the state maps, or could simply reflect local
concentrations of sulfur-bearing minerals.
In summary, at the level of resolution currently available, bedrock geology does not explain a
significant portion of the high sulfate outputs found in the sulfur input/output budgets. Although in many
Instances, local sources for sulfur are bedrock-related, it is not possible to isolate those sources using
information compiled for state geology maps. More detailed investigations (outside of the scope of this
Project) would be required to isolate and identify these sources and resolve the discrepancies.
7.3.3.3 Upper Limit Steady-State Sulfate Concentration
7.3.3.3.1 Introduction -
The second approach selected to determine an estimate of the number of systems with internal
sources of sulfur was based on an estimated steady-state sulfate concentration. As discussed previously
(Section 7.1), steady state is obtained when sulfur outputs from a watershed equals inputs. The sulfate
concentration of the surface water at that point is the steady-state suifate concentration, and an estimate
can be computed from the inputs and the runoff, as noted in Equation 7-5 below:
IS°4" U =
-------
Steady-state sulfate concentration can be computed for any individual site for which we have estimated
inputs and runoff. If the observed (i.e., by ELS and NSS) sulfate concentration at a site is greater than
the computed steady-state concentration, a source of sulfur internal to the watershed is suspected.
As discussed previously (Section 7.3.2.3), each of the estimates of inputs and runoff has an
associated uncertainty. The computed steady-state sulfate concentration has an uncertainty that Is a
function of these input uncertainties. Thus, we can compute for each surface water (i.e., lake or stream
reach) an upper limit steady-state sulfate concentration that, if exceeded, serves to indicate the
occurrence (with known probability) of an internal source of sulfur.
7.3.3.3.2 Objectives -
The objectives of the steady-state sulfate concentration analysis are
(1) to apply an uncertainty analysis (of the type presented in Section 7.3.2) to determine an
estimate of the steady-state sulfate concentration and associated uncertainty, and
(2) to calculate an upper limit steady-state sulfate concentration that, if exceeded, indicates the
presence of internal sources of sulfur.
7.3.3.3.3 Calculation of steady-state sulfate -
7.3.3.3.3.1 Data -
We used the long-term annual average estimates of wet and dry sulfur (Section 7.3.1.2.1) and the
30-year average annual runoff (Section 5.7) to calculate estimates of steady-state sulfate concentration
for each watershed (Equation 7-5).
7.3.3.3.3.2 Monte Carlo analysis -
Figure 7-4 presents a flow chart of the steady-state sulfate analysis and subsequent use of the
steady-state sulfate concentration to identify internal sources of sulfur. Briefly, the first step in developing
the upper limit steady-state sulfate concentration is to determine an estimate of the uncertainty associated
with the steady-state calculation. We conducted a Monte Carlo analysis similar to the one discussed in
Section 7.3.2.3 using the parameter uncertainty estimates for Sw , Sd , and R. We performed Monte Carlo
7-25
-------
Determine uncertainty estimate for wet
and dry sulfur deposition and runoff
Run Monte Carlo simulations (10,000)
for random watersheds
1
Determine regression relationship
between Monte Carlo estimates of mean
[SO 4 ] ss & standard deviation (sd)
Use regression equation to calculate standard
deviation on steady-state calculations for
watersheds (all NSWS)
Compute (SC^Iss upper limit
- [S04]ss+2sd
Suspected internal sulfur sources;
watershed removed from analysis
No internal sulfur sources based
on steady-state analyses
Figure 7-4. Flow chart for the determination of internal sources of sulfur using the steady-state
sulfate concentration.
7-26
-------
simulations (10,000 iterations) for 34 watersheds selected at random from the NE, SBRP, and Mid-
Appalachian study sites (NSWS). The results of the Monte Carlo analysis provided an estimate of the
standard deviation around the mean steady-state sulfate using the uncertainty estimates for each of the
34 watersheds.
7.3.3.3.3.3 Development and application of the regression equation -
We then plotted the standard deviation versus the calculated mean steady-state sulfate
concentration (based on 10,000 runs) for the 34 watersheds (Figure 7-5) and determined the regression
equation with the standard deviation as the independent variable.
Est. Std. Dev. = - 4.9 + 0.339 mean [SO42~ 1^ (Equation 7-6)
R2 = 0.99, p = 0.0001
We then substituted the computed (or nominal) value of steady-state sulfate concentration for each
watershed into Equation 7-6 to calculate an associated standard deviation applicable to each site (note
that there is an individual estimate for each site). Analyses of the Monte Carlo runs for the 34 watersheds
indicate that a log normal distribution best describes the uncertainty associated with steady-state sulfate
concentration. In applying the regression equation to each watershed, we conducted a log transformation
of the prediction procedure to reflect the observed distribution of the uncertainty in steady-state sulfate
concentration.
The final step in the analysis of steady-state internal sources was to apply the calculated standard
deviation on steady-state sulfate concentration to determine an upper limit. We added twice the estimated
standard deviation (97.5 percent confidence interval) to the computed steady-state sulfate concentration
and then compared the result to the measured sulfate concentration. If the computed upper limit steady-
state sulfate concentration was equal to, or greater than, the measured sulfate concentration, then we
assumed no significant internal sources of sulfur. Conversely, if it was less than the measured sulfate
concentration, we strongly suspected that some sources of sulfur were being contributed to the surface
waters in addition to that estimated from atmospheric deposition.
7-27
-------
100-
80-
60-
40-
20-
100 200
Mean Sulf ate OieqL-1)
300
Figure 7-5. Scatter plot of the Monte Carlo calculated standard deviation versus the calculated
mean [S04 ]. (based on 10,000 runs per watershed) n=34.
7-28
-------
This analysis does not work well for sites in regions that retain large amounts of deposited sulfur
(e.g., the SBRP). In such regions, statistical outlier analyses (e.g., see Section 8) need to be performed
to identify unusually disturbed or affected sites.
A summary by region of the number of ELS and NSS watersheds removed from the average annual
percent sulfur retention analysis is given in Table 7-6. These sites were identified using the upper limit
steady-state sulfate concentration estimates. The additional sulfur is probably from some internal
weathering source (as discussed above) or possibly could be due to a very localized emissions source.
7.3.4 Results and Discussion
We calculated percent sulfur retention for sites located in the NE, Mid-Appalachians, SBRP, and
several adjacent regions. Sites identified as having internal sulfur sources through the steady-state sulfate
concentration analysis were eliminated (Table 7-6). Also, three SBRP sites sampled as part of the Pilot
Stream Survey were dropped due to outlier surface water chemistry (i.e., ANC > 1000 ^eq L'1 ).
We used a modified version of the long-term annual average (LTA) sulfur deposition in these
analyses. This modified LTA sulfur deposition does not include the 20 percent increase in dry deposition
discussed in Section 5.6 for the TV and standard LTA deposition data. The TY and standard LTA data
were only available for the primary DDRP study sites (NE-145, SBRP-35). The sulfur retention analyses
use surface water chemistry from approximately 1000 sites (lakes and stream reaches) sampled as part
of the ELS and NSS. The dry sulfur deposition data provided by AREAL-RTP (see Section 5.6) were
the only internally consistent dry sulfur deposition data for all of the ELS and NSS sites. These dry
sulfur deposition estimates were combined with the long-term wet sulfur deposition estimates to form the
modified-LTA deposition dataset. To test the overall effects of not using the 20 percent increase in these
analyses, we adjusted the modified-LTA data with a 20 percent increase in dry sulfur for the ELS
Northeast sites. This adjustment created a dataset analogous to the TY and standard LTA data. We then
calculated percent sulfur retention using the adjusted data and compared the results to the unadjustffied
7-29
-------
Table 7-6. Summary of Watersheds (by ELS and
NSS Subregion) Dropped Due to Suspected Internal
Sources of Sulfur Identified by Steady-State
Analysis
Region # of Watersheds
Eastern Lake Survey
1A 3
1B 5
1C 15
1D 12
1E 18
3A (SBRP) 2
Pilot Stream Survey 0
National Stream Survey
1D 12
2BN 9
2CN 23
2X 7
3A 2
38 7
7-30
-------
modified-LTA sulfur retention results (Figure 7-6). An inspection of Figure 7-6 indicates that there is only
a slight shift in the distribution of percent sulfur retention between the two datasets. This slight shift Is
unimportant relative to the principal conclusions drawn from these regional sulfur retention analyses.
Similarly, analyses using the TY dataset yield results very close to those computed using the modified-
LTA dataset. Thus, the latter dataset Is used for the remainder of the analyses presented in this section.
7.3.4.1 Northeast
Results from analyses using the modified-LTA deposition data Indicate that lake systems in the NE
are generally at or near zero percent net sulfur retention (Table 7-7; Figure 7-7A; Plate 7-1). Rochelle and
Church (1987) conducted a sulfur retention analysis using runoff and deposition data for the water year
prior to ELS and NSS sampling and showed similar results. Also, we examined sulfur retention patterns
in the NE for individual ELS subregions. Although lakes In Subregion 1B have the highest percent net
retention, lakes in alt subregions are, on average, very close to zero percent net retention (Table 7-7).
7.3.4.1.1 Evaluation of sulfur retention mechanisms in NE watersheds -
Although most NE watersheds are near sulfur steady state, a small number of watersheds are
characterized by high apparent sulfur retention. During development of preliminary sulfur input/output
budgets for lakes in the northeastern United States, we identified a subset of watersheds for which budget
analyses indicated significant sulfur retention. Because analyses of sulfate adsorption at that time
suggested that adsorption was likely to delay sulfur response in NE watersheds for a very limited time,
it was unclear how sulfur was being retained in this subset of watersheds. In an effort to understand
sulfur retention in these systems and In an effort to evaluate potential future sulfate increases at those
sites, we identified for additional analysis a group of 45 NE watersheds having high computed sulfur
retention. Soils, vegetation, land use, depth to bedrock, and bedrock geology were mapped on 44 of
the watersheds (permission to map was denied for one) during the fall of 1987 and spring and summer
of 1988 (Figure 7-8). Watersheds were mapped by the USDA SOS according to protocols developed for
the original NE DDRP soil survey (Section 5.4), except that mapping criteria were modified to require
discrete mapping of wetland areas 2 acres or larger, rather than the 6-acre map unit delineations used
7-31
-------
20
15
I-
•5 10.-
5-
H
W
If
-100 -90-60-70-60-50-40-30-20 -10 0 10 20 30 40 50 60 70 80 90 100
Midpoint Percent Sulfur Retention
B
•5 10
-100-90 -80 -70 -60 -SO -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90 100
Midpoint Percent Sulfur Retention
Figure 7-6. Comparison of percent sulfur retention calculated using (A) modified-LTA deposition
and (B) modified-LTA deposition adjusted with a 20 percent increase in dry deposition.
7-32
-------
Table 7-7. Percent Sulfur Retention - Summary Statistics by Region
Region
NE
ELS Rgl
1A
18
1C
10
1E
Mld-App
NSS 2Bn
NSS 2Cn
SBRP
PSS
ELS Rg. 3a
Misc."
S. App. Pla.
NSS2X
Piedmont
NSS3A
Mid Atlantic Coastal
NSS3B
Poconos/Catskills
NSS 1D
Na
5,828
1099
1285
1190
966
1288
12,580
6,478
2,031
247
4,329
7,199
Plain
9,535
2,724
Mean
-5.0
-11.9
8.9
-3.8
-11.5
-9.4
27.9
-4.0
67.5
68.0
43.4
68.4
30.5
-21.5
Median
-5.2
-13.9
7.6
-7.2
-8.9
-11.7
39.6
3.1
75.4
78.6
50.3
78.0
34.2
-29.2
Std. Dev.
27.6
22.5
25.9
27.3
29.9
26.5
43.1
31.8
23.2
32.8
37.5
24.1
38.4
31.2
Min.
-69.8
-63.5
-65.6
-65.7
-69.8
-61.9
-82.6
-82.7
-54.3
-64.4
-63.6
-9.8
-60.4
-71.2
Max.
73.3
60.6
73.3
63.3
53.9
51.0
90.6
55.2
88.0
92.9
86.0
91.9
92.7
67.1
3 Estimated target population calculated using NSWS weights (see LJnthurst et al., 1986a; Messer et al., 1986a;
Kaufmann et al., 1968) for information on weights.
b Additional regions sampled as part of NSS Phase I (see Plate 7-1).
7-33
-------
NE
to
O OA
Q.
S 0.6
O
~ 0.4
I
00
Upper Sound
Projected
Lower Bound
-100 -SO 0 SO 100
Percent Sulfur Retention
Mid-Appalachians
Ur
OL
a
0.8
a
0.4
OJ2
0.0
Upper Bound
Projected
-100 -50 0 50 100
Percent Sulfur Retention
I0r
O OS
1
0.6
(P
0.4
SBRP
Upper Bound
Projected
Lower Bound
0.0
-100 -SO 0 SO tOO
Percent Sulfur Retention
Figure 7-7. Population-weighted distribution of projected percent sulfur retention (upper and lower
bounds for 90 percent confidence interval): (A) Northeast; (B) Mid-Appalachians, and (C) Southern
Blue Ridge Province.
7-34
-------
Plate 7-1. Median percent sulfur retention by NSWS Subregion.
7-35
-------
NSWS SUBREGIONS
MEDIAN % SULFUR RETENTION
AND WET SULFATE DEPOSITION
2.25
MEDIAN PERCENT
SULFUR RETENTION
0 0 - 20
0 20 - 40
Q 40 - 60
0 60 - 80
• 80 - 100
Average Annual
Wet Sulfate m 2-7
Deposition (g m* yr~')* 3.00-
3.25
Eostern Uk« Sumy
2.50,
2.00'
-2.25
Vedion
Svkr«gr«n I Uttmtion
1A
IB
1C
ID
IE
-14
8
-7
-9
-12
2.00
Notional Strum Sumy
Median
X ktltntitn
3
40
34
50
n
n
2Cn
2Ba
IB
n
2As
'Deposition for 1980 - 1984
(A- Olsen. Personal Communication)
-------
DDRP STUDY SITES
Supplemental Watershed Mopping
• Supplemental Mapping Sites
+ Study Sites
Rgure 7-8. Supplemental watersheds mapped for special evaluation of sulfur retention.
7-36
-------
for other soils and for vegetation. After completion of mapping, soil map units were correlated to the
soil sampling classes defined for the initial DDRP NE soil survey, except for soils on parts of two
watersheds in Pennsylvania. Those soils were correlated to sample classes defined for Mid-Appalachian
soils, and the watersheds were dropped from the analysis. We have used these mapping data to assist
In an analysis of retention in these watersheds.
7.3.4.1.1.1 Approach -
Watersheds for this analysis were selected from the NE lakes sampled in Region 1 as part of the
ELS, using preliminary watershed sulfur input/output budgets developed with 1984 Water Year data.
Criteria for watershed inclusion were (1) lake type - limited to drainage lakes and reservoirs; (2)
watershed area - less than 3000 ha; and (3) watershed sulfur budgets (1984 Water Year data)
characterized by one or more of the following:
• at least 20 percent sulfur retention
* a 20 Meq L"1 or greater difference between lake sulfate and steady-state sulfate concentration
• lake sulfate concentrations at or below the tenth percentile of sulfate concentrations in the
respective ELS Subregion.
The budget/concentration criteria were not intended as independent selection criteria; rather, multiple
criteria were defined to ensure inclusion of lake systems with high apparent absolute and/or relative sulfur
retention. With few exceptions, watersheds met at least two of the sulfur budget/concentration criteria,
and most met all three.
We assessed watershed sulfur budgets using procedures and uncertainty estimates as described
in Section 7.3. Based on uncertainty analyses presented in Section 7.3, we determined that retention
should be regarded as significant if computed percent sulfur retention exceeds 37.5 percent.
7-37
-------
7.3.4.1.1.2 Results and Discussion
Table 7-8 summarizes sulfur budget status for the 42 NE watersheds considered, and also lists
computed in-lake sulfur retention and proportions of wet soils on each watershed. Using the criterion
of 37.5 percent to define significant retention, 27 of the 42 watersheds had significant (positive) sulfur
retention, if actual retention was not significant for any of the 535 lakes in the ELS sample (all regular
ELS Region 1 drainage lakes and reservoirs, excluding DDRP lakes and lakes with watersheds > 3000
ha) from which the 42 watersheds of concern were identified, an expected 13 lakes would fall above the
95 percent confidence window of 37.5 percent. Assuming that retention estimates for each watershed
are independent, if there were in fact no lakes with significant sulfur retention in the sampie population,
the probability of observing significant (computed) retention in >. 27 watersheds is 0.00057. From these
results we can conclude that although sulfur retention in many of the 42 watersheds is not statistically
significant, a small proportion of watersheds in the NE target population are characterized by significant
positive sulfur retention.
Indirect evaluation of soils data for the NE virtually rules out the possibility that significant net sulfate
adsorption is presently occurring in these watersheds. Response times for the 38 NE soil sample classes
are comparable to NE watershed responses presented in Section 9.2; none of the sample classes is
projected to be retaining sulfate or to have solution sulfate concentrations less than steady state under
current conditions (based on the historic deposition sequences used in Section 9). Therefore, it is highly
unlikely that the observed retention on these watersheds can be explained by adsorption.
Direct estimation of in-lake sulfur retention, using the model of Baker et al. (1986b) (Section 7.2),
suggests that in-lake processes are a minor sulfur sink in most NE lake systems. For many of the 42
watersheds under consideration here, however, the relative importance of in-lake reduction is apparently
much higher (Table 7-8). Computed retention for the 42 lakes ranged from 0.9 to 23.1 percent, exceeding
10 percent for 10 watersheds and 20 percent for 2 watersheds. If watershed sulfur budgets are adjusted
by computed in-lake retention, retention in 12 of 27 watersheds drops below the 37.5 percent threshold
that is used to define significant retention. In other words, for almost half of the 27 watersheds with
7-38
-------
Table 7-8. Summary of Sulfur Retention Status and of Watershed Variables Contributing to
Sulfur Retention for 42 Watersheds in the Northeastern United States
LAKE ID
1A1-019
1A1-037
1A2-001
1A2-036
1A2-038
1A2-056
1A2-057
1A3-018
1B1-004
1BHD06
1B1-OQ7
1B1-008
1B1-038
1B2-059
1B2-069
1B3-003
1B3-013
1B3-029
1B3-068
1C1-046
1C1-069
1C2-055
1C2-061
1C3-034
1D1-007
1D1-011
1D1-021
1D1-058
1D2-006
1D2-013
1D2-087
1D3-004
1D3-083
1E1-010
1E1-017
1E1-023
1E1-036
1E1-060
1E1-072
1E1-097
1E2-004
1E2-046
ws area
(ha)
94
105
225
574
91
78
232
153
106
124
146
131
501
57
168
105
125
37
120
422
274
295
368
136
165
641
180
464
138
192
148
185
21
217
418
234
148
84
197
49
198
634
WA:LA
11.4
14.2
15.2
20.8
4.3
7.1
4.3
9.6
11.8
5.0
12.4
3.5
78.2
4.0
15.5
7.3
13.4
6-6
10.2
17.5
25.9
8.9
11.7
12.3
18.2
14.3
6.8
12.0
20.2
8.6
10.4
15.0
3.0
17.3
44.9
28.2
6.8
8.3
1.2
1.7
11.0
12.8
rtn.
time (yr)
0.24
0.94
0.08
0.11
0.34
1.17
1.31
0.55
0.32
0.59
0.18
2.25
0.02
0.59
0.20
1.81
0.26
0.62
0.22
0.43
0.06
0.99
0.25
0.25
0.13
0.32
0.25
0.10
0.12
0.30
0.14
0.16
0.86
0.35
0.07
0.04
0.18
1.31
6.60
1.36
0.08
0.25
sulfate
lake
80.6
61.1
64.8
84.3
78.3
42.5
84.0
54.0
91.7
104.5
79.5
102.4
62.6
72.3
121.8
118.3
93.1
67.4
73.9
61.2
50.2
65.5
59.1
52.4
93.7
74.3
79.5
69.3
58.8
93.5
67.7
81.6
95.6
33.6
29.1
36.8
40.9
35.1
38.2
36.5
38.4
44.2
(uea L'1)
s-s
98.7
105
95.9
126
106
108
104
119
179
183
187
184
143
179
216
144
149
199
147
100
108
109
75.6
92.2
124
128
98.5
120
128
130
130
124
143
50.4
50.3
53.8
65.9
64.3
65.6
67.3
60.8
90.1
ws sulfur
itn (%)
18.3
41.9*c
32.4
33.2
25.9
60.6*
19.0
54.7*
48.8*
43.0*
57.4*
44.2*
56.2*
59.6*
43.5*
17.8
37.6*
66.2*
49.7*
39.0*
53.7*
39.8*
21.8
43.2*
24.3
42.0*
19.3
42.1*
53.9*
28.2
47.8*
34.3
33.0
33.2
42.2*
31.7-
38.0*
45.3*
41.7*
45.7*
37.0
51.0*
Intake
rtn(%)a
4.7
4.9
4.3
4.5
12.0
8.0
13.0
8.1
7.1
14.0
6.9
18.0
0.9
16.0
6.2
11.0
5.4
12.0
6.8
4.4
2.5
8.3
4.5
5.6
4.1
4.5
7.7
5.5
3.7
7.6
7.0
4.9
16.0
3.5
1.4
2.3
8.9
7.0
23.0
22.0
6.4
5.4
adj. ws
rtn (%)b
13.6
37.0
28.1
28.6
13.5
52.6*
6.5
46.5*
41.7*
29.0
50.5*
26.4
55.2*
43.2*
37.3
7.2
32.2
54.7*
42.9*
34.5
51.2*
31.5
17.3
37.6*
20.2
37.5
11.6
36.7
50.2*
20.6
40.9*
29.4
17.3
29.8
40.8*
29.4
29.1
38.3*
18.6
23.7
30.6
45.5*
% of watershed
wetland
8.1
15.9
2.3
4.4
4.5
10.7
0.0
0.0
0.0
7.7
2.7
9.9
0.6
7.0
0.0
7.2
3.3
0.0
4.3
9.7
7.9
14.2
13.3
13.6
14.7
10.6
19.2
21.6
17.2
8.8
17.6
2.5
0.0
9.6
8.4
36.5
46.7
24.4
17.0
5.4
90.4
20.0
H02 + H03
19.3
10.3
3.9
8.0
4.1
15.4
0.7
4.1
0.0
1.9
4.9
3.6
17.2
7.0
0.0
0.3
0.3
0.9
1.5
4.9
4.2
4.8
9.1
5.8
8.0
5.5
12.0
4.8
12.8
0.6
8.9
1.9
0.0
9.6
6.1
21.7
26.0
16.6
7.9
4.2
12.7
16.2
area
I25
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
38.6
51.4
17.4
34.8
9.8
46.1
3.2
8.4
15.3
0.0
11.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
* Computed using Equation 7-1.
Percent retention minus percent in-lake retention,
c Asterisks Indicate significant retention (a = 0.05).
7-39
-------
significant computed sulfur retention, in-lake reduction has a significant influence on watershed sulfur
retention status.
Wetlands and wet soils might also contribute significantly to sulfur retention in many of these
watersheds. Table 7-8 lists percentages of watershed areas covered by wetlands (SCS land use
classifications) and by soils in sample classes H02 and H03 (wetlands) and 125 (deep, very poorly to
somewhat poorly drained aquepts). These data indicate that all but 3 of the 15 watersheds with
significant sulfur retention (after adjusting for in-lake retention) have at least 10 percent coverage of
wetlands or wet soils (these area! proportions are not additive on watersheds; much of the area
designated as wetlands on SCS land use maps is also classified in soil sample classes H02 or H03);
wetland area exceeds 10 percent on 19 of the 42 watersheds with maximum coverage exceeding 90
percent. Evidence of net sulfur retention in wetlands is inferential; actual sulfur retention in these soils
depends not only on the extent but also the location of these areas within a watershed and the fraction
of watershed runoff flowing through them.
Analyses to date suggest that there is significant sulfur retention in a small proportion of NE lake
watersheds. Evaluation of Level II modelling data (Section 9) also indicates that adsorption is unlikely
to play a significant role in that retention. The most likely processes contributing to retention in this
group of watersheds are a combination of in-lake retention, which is important in those lakes having long
hydrologic retention times (Section 7.2), and reduction in wetlands/wet soils. Unlike adsorption, reduction
in lake sediments and wet soils is a rate-limited, rather than a capacity-limited process; retention by
reduction mechanisms can therefore continue at current rates indefinitely because no capacity exists to
be filled or exhausted. Reduction in lakes provides a permanent sink for sulfur, but the extent of retention
in wetlands and wet soils can change on an annual or even seasonal basis. During dry periods, soils
in wetlands and other anaerobic areas could reoxidize, resulting in oxidation of reduced sulfur and,
potentially, its release as sulfate. The role of wetlands and wet soils can consequently shift from that of
sulfate sink to source during dry periods; the potential for long-term retention in such systems is thus
dependent on watershed hydrologic conditions.
7-40
-------
7.3.4,2 Mid-Appalachians
The Mid-Appalachian Region does not present as clear a picture of percent sulfur retention as the
NE (Table 7-7; Figure 7-7B). For this study we have defined the Mid-Appalachian Region as a
combination of NSS Subregions 2Bn and 2Cn (Plate 7-1). Kaufmann et al. (1988) defined these regions
as the Valley and Ridge and Northern Appalachians, respectively. We found that percent retention was
more evenly distributed with no strong patterns of low or high percent net retention. In general, for
both subregions, percent net retention is low with average values less than 30 percent. Subregion 2Cn
has significantly lower percent retention estimates on the average than Subregion 2Bn. Subregion 2Cn
receives higher sulfur deposition than does Subregion 2Bn (Plate 7-1). Although Subregion 2Cn probably
has a high incidence of potential acid mine drainage influence, systems identified by Kaufmann et al.
(1988) as having potential internal sulfur sources were also subsequently identified in our steady-state
analysis (Section 7.3.3.3) and dropped from this analysis and presentation of results.
The Southern Appalachian Plateau and the Mid-Atlantic Coastal Plain have percent retention on the
average of 30 to 40 percent (Table 7-7; Figure 7-9). In these regions there Is a pattern toward higher
net retention, although a large amount of scatter in percent sulfur retention remains. The Catskills/
Pocono Region has a median net sulfur retention of -21.5 percent. This region is a transitional area from
the NE, where glaciated soils predominate, to the Mid-Appalachians, where older and more weathered
soils predominate.
7.3.4.3 Southern Blue Ridge Province
Median net sulfur retention for the SBRP is approximately 75 percent (Table 7-7; Figure 7-7C).
Rochelle and Church (1987), working with sulfur deposition data from Water Year 1984, found similar
results. The average percent sulfur retention for the Piedmont Region (adjacent to the SBRP) is also high
compared to the NE and Mid-Appalachians (median = 78.0, Table 7-7 and Figure 7-9).
7-41
-------
Southern Appalachian Plateau
Mid-Atlantic Coastal Plain
to
5 as
o
a.
£U
I*.
0
3
E
a 02
0.0
Upper Bound
Projected
Lower Bound
to
JO 0.8
O
Q.
0.4
&
3
-100 -50 0 50
Percent Sulfur Retention
too
0.0
Upper Bound
Projected
Lower Bound
-100 -50 0 50
Percent Sulfur Retention
100
Catskills/Poconos
Piedmont
1.0
0.8
0.6
o
1"
I
0.0
to
0.8
*: 0.4
CO
Upper Bound
Projected
Lower Bound
I
Upper Bound
Projected
Lower Bound
-100 -50 0 SO 100
Percent Sulfur Retention
0.0
-100 -50 0 50
Percent Sulfur Retention
100
Figure 7-9. Population-weighted distributions of projected percent sulfur retention, with upper and
lower bounds for 90 percent confidence intervals, for additional NSS subregions: (A) Southern
Appalachian Plateau, (B) Mid-Atlantic Coastal Plain, (C) Catskills/Poconos, and (D) Piedmont.
7-42
-------
7.3.4.4 Conclusions
When collectively examined, definite spatial trends in net sulfur retention are evident among the NE,
the Mid-Appalachian Region, and the SBRP. Percent sulfur retention generally increases from North to
South in the eastern United States (Figure 7-10; Plate 7-1). Plate 7-2 provides an additional view of the
North to South regional patterns of percent sulfur retention. Using the broad major land use resource
area (MLRA) classes (USDA, 1981) to stratify the NSWS study sites, Plate 7-2 Indicates again that the
SBRP and adjacent areas are retaining higher amounts of incoming sulfur deposition when compared to
the Mid-Appalachian Region.
Also, indications are that net retention of sulfur in the NE on the average is zero or close to zero.
Net sulfur retention in the Mid-Appalachian Region appears to be in transition between the NE and
SBRP. A simple analysis of variance indicates that, on the average, percent net retention is significantly
different among these three regions.
We attribute the spatial patterns in sulfur retention shown here to two key factors (1) soil type, and
(2) sulfur deposition. Whereas soils of the SBRP are predominately weathered Ultisols and Inceptisols
that tend to have high sulfate adsorption capacities, the NE Region is dominated by Spodosols, which
characteristically have low sulfate adsorption capacities (e.g., see discussion by Rochelle et at. (1987)).
Soils of the Mid-Appalachian Region are predominately Inceptisols and Ultisols. Given the current patterns
of wet sulfur deposition (Plate 7-1) and assuming that the Mid-Appalachian region has received elevated
levels of sulfur deposition for a considerable period of time, it is apparent that this region is in transition
toward a situation of lowered percent net sulfur retention and significantly elevated surface water sulfate
concentrations. We feel that this change is a direct consequence of elevated atmospheric sulfur
deposition. The SBRP is probably undergoing such a transition but with a lag, or "delay", in time. The
dynamics of transitions in the NE and SBRP are the subject of DDRP analyses in Section 9 and 10 of
this report. Analyses for the Mid-Appalachian Region will be examined in subsequent DDRP activities.
Relationships among sulfur deposition, edaphic characteristics, and sulfur retention in the NE and SBRP
are examined in Section 8.
7-43
-------
c
o
2
CL.
-------
Plate 7-2. Regional percent sulfur retention by major land resource area (MLRA) based on target
populations (ELS and NSS sites).
7-45
-------
MAJOR LAND RESOURCE AREAS
MEDIAN % SULFUR RETENTION
AND WET SULFATE DEPOSITION
MEDIAN PERCENT
SULFUR RETENTION
m < o
H o - 20
gg 20 - 40
40 - 60
gj 60 - 80
H 80 - 100
HiUQ* LAID RESOURCE AREAS
R * Nor theaitern Forggi ond
Fornt Region
S - Noriktro A.llotiic Slop*
DiTersified Farming Region
II - Eott ond Central Forming
OHO1 Fonit Region
P - So o tli Allude and Gulf
Slope Cash Crops, Fortit
god Lhittock Regioo
2-50
2.25
2.00'
Average Annual
Wet Sulfa-te
Deposition (g m* yr"1)*
3.25
2.00
2.25
-2.25
VLRA it DUN
REGION I RETENTM
R
S
N
P .
-n
2$
n
73
Deposition for 1980 ~ 1984
(A. Olsenr' Personal Communication)
-------
SECTION 8
LEVEL I STATISTICAL ANALYSES
8.1 INTRODUCTION
The chemistry and quality of surface waters in natural settings are the result of inputs from
deposition, terrestrial processes, and in-lake or in-stream processes. In this section we consider the
relationships between subtending surface water chemistry and inputs from deposition and the physical
and chemical attributes of the catchments. The scope of these analyses includes the DDRP sample of
northeastern lakes and streams in the Southern Blue Ridge Province (SBRP). We will not, however,
consider in-lake or in-stream processes explicitly in this analysis because data are not available for these
processes on a regional basis.
Level I analyses are designed primarily to address the first two DORP objectives (see Section 2.2):
(1) regional description of soil and watershed characteristics and (2) characterization of the relationships
between watershed attributes and surface water chemistry. These Level I analyses are of particular
importance to the DDRP because they are designed to corroborate some of the fundamental assumptions
of the DDRP on a regional basis. Previous research has generally been limited to observations from a
small sample of sites. To make model-based regional projections of future surface water chemistry, it
is important to determine whether or not the findings of previous studies on watershed and surface water
chemistry relationships can also be observed on a regional basis, if they cannot, other approaches may
need to be taken. At the same time, it is critical to know if we are overlooking important relationships
that should be included in the Level II and Level III Analyses.
The principal objective of the analyses in this section is to determine which soil and watershed
characteristics are most strongly related to surface water chemistry. Some of the questions we hope to
answer are the following: Can surface water chemistry be linked to specific watershed and soil
characteristics? Are there controls on surface water chemistry that are not yet identified? Which
deposition and/or watershed factors explain most of the observed variability in surface water sulfate
8-1
-------
concentrations? Do the characteristics of the near-stream or near-lake areas have a greater influence on
surface water chemistry than the watershed as a whole?
We realize that many of the results of these analyses may only provide further evidence to support
relationships already known to exist. However, because of the quality, consistency, and extent of the
data used in these analyses, new relationships between watershed characteristics and surface water
chemistry are likely to be identified, and at the same time previously observed relationships will be
reaffirmed.
8.1.1 Approach
The approach used In this section is an empirical, statistical evaluation of the relationships between
selected watershed attributes gathered for the ODRP sample of watersheds and the chemistry of the
surface water draining these watersheds. The principal dependent variables considered in this analysis
include surface water sulfate concentrations, percent watershed sulfur retention (% S retention), surface
water acid neutralizing capacity (ANC), the sum of surface water concentrations of calcium and
magnesium (Ca plus Mg), and surface water pH. Although there are a number of other variables that
could be considered, these are of primary interest to the DDRP. With the exception of % S retention,
each of the dependent variables is a direct measure of surface water chemistry. Percent S retention is
computed as the ratio of the difference between watershed sulfur inputs (from deposition) and surface
water sulfur concentrations to sulfur inputs (see Section 7). Percent S retention is a measure of the
amount of sulfur arriving via deposition that is retained by the watershed. A summary of the dependent
variable data from the northeastern sample of 145 lakes and S8RP sample of 35 streams is presented
in Tables 8-1 and 8-2, respectively.
The deposition data used in this section are the "long-term annual average" (LTA) deposition data.
These data have annual resolution and represent atmospheric deposition as of the early to mid-1980s.
The LTA deposition dataset is described more fully in Section 5.6.3.2 and summary statistics are given
in Tables 8-3 and 8-4 for the 145 northeastern and 35 SBRP sample watersheds, respectively.
8-2
-------
Table 8-1. Surface Water Chemistry and Percent Sulfur Retention Summary Statistics
for the Northeastern DORP Sample of 145 Lake Watersheds
Variable3
Mean Std. Dev.
Min.
Q1D Median
Q3° Max.
Sulfate
% S retent.
ANC
Ca+Mg
PH
112.6
-9.7
126.3
223.1
6.9
45.2
41.3
113.6
126.4
0.8
33.8
307.7
-53.3
35.0
4.5
82.4
-23.3
33.3
125.3
6.7
105.5
-6.5
97.3
191.8
7.2
130.8
14.9
213.0
292.6
7.5
303.6
61.1
391.6
560.3
8.0
" Units on sulfate, ANC, and Ca+Mg are yeq L"1. Sulfur retention is expressed as a percent. pH is unitless.
b Q1 is the 25th percentile, and O3 is the 75trt percentile.
8-3
-------
Table 8-2. Surface Water Chemistry and Percent Sulfur Retention Summary Statistics
for the DDRP Sample of 35 S8RP Stream Watersheds
Variable3
Mean
Std. Dev.
Min.
01'
Median
Q3b Max.
Suifate
% S relent.
ANC
Ca+Ma
SOBC*
PH
40.3
65.1
286.8
285.4
371.0
7.1
34.1
26.0
447.9
455.1
466.2
0.41
14.7
-60.5
16.2
46.0
92.8
6.4
19.8
60.1
98.8
85.8
156.0
6.9
23.6
74.9
126.5
117.2
223.4
7.0
42.2
79.1
171.1
189.4
244.7
7.2
178.6
88.6
1710.5
1841.6
1958.5
8.4
* Units on sulfate, ANC, Ca+Mg, and SOBC are peq L'1. Sulfur retention is expressed as a percent. pH is unitiess.
b Q1 is the 25th percentile, and 03 is the 75th percentile.
c SOBC = Sum of base cations (Ca+Mg+Na+K)
8-4
-------
Table 8-3. Summary Statistics for Wet and Dry Deposition on the DDRP Sample
of 145 Northeastern Lake Watersheds (units are ;ieq m )
Variable
Mean Std. Dev.
Mia
Q1S
Median
Q3a
Max.
SO4-WET
SO4-DRY
H-WET
H-DRY
CA+MG-WET
CA+MG-DRY
44900
22800
46800
24600
8200
10600
10300
10100
12000
16300
3600
3300
26700
9300
24500
1600
4800
3000
35200
16000
36000
11000
5800
8500
46100
20100
47300
23900
7300
10100
53800
26100
57800
32300
9200
13300
62300
60400
67300
77400
24100
19500
' Q1 is the 25th percentile, and 03 is the 75th percentile.
8-5
-------
Table 8-4. Summary Statistics for Wet and Dry Deposition on the DDRP Sample of 35
SBRP Stream Watersheds (units are f/eq m )
Variable
Mean Std Dev. Min.
Qla Median
Q3a
Max.
SO4-WET
SO4-DRY
H-WET
H-DRY
CA+MG-WET
CA+MG-DRY
52400
33000
45700
23300
10600
18000
4600
4300
3900
5400
800
3900
40800
20400
36300
11100
8500
6800
49800
30900
42300
19800
10200
16500
52900
33400
45200
22100
10700
19600
54900
34700
48000
24800
11100
20200
69400
42400
61100
36500
13200
22400
Q1 is the 25th percentile, and Q3 is the 75th pereentile.
8-6
-------
For this analysis we have grouped catchment attributes into six groups. The variables in these
groups serve as the independent or explanatory variables. The groups are: (1) derived hydrologic
variables (Section 8.3), (2) mapped bedrock geology (Section 8.4), (3) land use and vegetation (Section
8.5), (4) mapped soils (Section 8.6), (5) depth to bedrock (Section 8.7), and (6) measured chemical and
physical soil properties (Section 8.9). Variables in each of the groups are thought to have significant
influence on some aspect of surface water chemistry. We consider deposition with each of the attributes
to identify the key relationships between the dependent variables and each attribute. We include
deposition In each of these analyses because it is inextricably linked to surface water chemistry. Failure
to include deposition would in all likelihood result in inclusion of surrogate deposition variables in the
regression models. As a separate analysis we also consider the relationship between the deposition
variables and surface water sulfate concentrations and ANC (Section 8.2).
Because none of these attribute groups alone can fully account for the observed variability in the
dependent variables, we also consider them in combination. In Section 8.8 we combine the deposition
and the mapped variables (groups 1-5), excluding the measured chemical and physical properties; and
then in Section 8.10 we integrate deposition and ail of the watershed attributes.
Statistical Methods
In Section 8 there are tables presenting descriptive statistics of the explanatory variables, as well
as tables presenting results of regression analyses. In each case, the descriptive statistics are population-
weighted, unless otherwise noted. Population weighting provides estimates of the parameters in the target
population, rather than estimates for the DDRP sample only. None of the regression analyses in this
section is weighted. Based on the discussion in DuMouchel and Duncan (1983) and on the similarity
of the across-strata relationships among the variables, weighted regressions were deemed unnecessary.
Additionally, in the tables of regression results we have included a plus (+) or minus (-) sign to
indicate the direction of significant relationships, rather than a numeric estimate of the regression
parameter. These statistical analyses should be considered descriptive rather than predictive. Regression
8-7
-------
estimates have been excluded to discourage their use in predictive equations or naive computations of
the relative importance of the explanatory variables.
The standard statistical approach used in this section begins with a stepwise regression of the
explanatory variables on the surface water chemistry. This approach enables us to select explanatory
variables in a way that avoids serious problems with collinearity. The stepwise regression was
implemented in SAS (SAS Institute Inc., 1985,1987), using a value of 0.15 for both the significance level
for entry and the significance level for remaining in the model. Mallows' C_ statistic was used as a model
selection criterion. Significance levels for the explanatory variables are given in tables in each section.
The selected model was then run as a standard linear regression to perform residual analyses,
checking for outliers, leverage points, and problems with standard regression assumptions (Belsley et
ai., 1980). Cook's D statistic was used to identify leverage points (Madansky, 1988), i.e., observations
that might exert an extreme influence on the estimates of the regression parameters. In addition, the
effect on the regression parameters was assessed using the calculated DFBETAs (Belsley et ai., 1980).
Plots of the studentized residuals were used to check for outliers, as well as homoscedasticity
(constant variance of the residuals across the range of the dependent variable). Specific instances
where log transformations of ANC or Ca plus Mg were necessary to produce homoscedasticity are
discussed in Sections 8.3 and 8.7. If outliers or leverage points were found to be affecting the regression,
the stepwise regression and subsequent residual analyses were performed again without the problematic
observations.
Specific exceptions to this approach are discussed in the individual sections where the exceptions
occur. In Section 8.5, the standard statistical approach is applied to rotations of principal components
extracted from the original explanatory variables, rather than the variables themselves. In Section 8.8,
Mallow's Cp statistic could not be used as the model selection criterion in the SBRP, and Akaike's
information criterion was used instead.
8-8
-------
8.2 RELATIONSHIPS BETWEEN ATMOSPHERIC DEPOSITION AND SURFACE WATER CHEMISTRY
8.2.1 Introduction
Atmospheric deposition and its effects on surface water chemistry have been extensively studied
for several decades. Smith and Alexander (1986) found a strong positive correlation between sulfur
emissions and surface water sulfate concentration on a regional basis. Neary and Dillon (1988) found
that sulfur deposition had a positive relationship with lake sulfate and a negative relationship with ANC
for a sample of 1168 Canadian lakes. Sullivan et al. (1988D) found significant correlations between
median lake sulfate concentrations and wet sulfate deposition for the National Surface Water Survey
(NSWS) sites. In this section we examine such relationships for the DDRP sample of watersheds using
the wet and dry atmospheric deposition data for the Project (Section 5.6).
8.2.2 Approach
Surface water sulfate concentration and ANC are the two primary variables linked to the influence
of sulfur deposition on surface water chemistry, and hence these two variables are the focus of this
analysis. For explanatory variables, we used the LTA estimates of wet and dry deposition (discussed
in Section 8.1.2). In addition to the individual wet and dry deposition estimates, we also used total sulfate
deposition and total hydrogen deposition. In each case the total deposition value is the sum of the
appropriate wet and dry deposition values. The statistical analyses are discussed in Section 8.1.2.
8.2.3 Results and Discussion
8.2.3.1 Northeast
The statistical analyses show a significant positive relationship between lake sulfate and total suifate
deposition (Table 8-5). Residual analysis of this regression revealed two strong outliers with lake sulfate
levels much higher than predicted. These sites have quarry pits and will be discussed in Section 8.6.3.1.
Removing these two sites with apparent internal sources of sulfur increases the amount of explained
variability to 38 percent (Table 8-5).
8-9
-------
Table 8-5. Results of Regressions Relating Surface Water Chemistry to Atmospheric Deposition
in the Northeast Region (n = 145)
Water
Chemistry
Variable
Deposition
Adjusted Variable Regression Signrf.
R2 in Model Sign Level
Sulfate 0.27 0.27 total sulfate
second model (omitting two outliers)
0.38 0.38 total sulfate
***
ANC
0.18
0.16
wet sulfate
dry sulfate
***
***
significant at the 0.001 level
8-10
-------
There is a weaker relationship between ANC and deposition (Table 8-5). Wet and dry sulfate
together explain only 18 percent of the variability in ANC in the northeastern lakes. Notice that the
parameter estimates for wet and dry sulfate have opposite signs. In the stepwise regression used to
select a model, wet sulfate deposition was selected in the first step, and then dry sulfate deposition was
included as the deposition variable with the best relationship to the residuals from the first step. Residual
analysis indicates that this is an adjustment in the model to correct for areas with high deposition and
high ANC, such as Subregion 1B (the Poconos/Catskills, see Figure 5-1). The size of R2 for ANC is not
surprising, because ANC is strongly dependent upon mechanisms of ANC generation within watersheds
(see Section 3).
8.2.3.2 Southern Blue Ridge Province
In the SBRP, sulfate deposition variables were not significantly related to stream sulfate
concentration (Table 8-6). Because the stepwise regression used a 0.15 level of significance for entry
into the model, this result indicates that the relationship between deposition and surface water sulfate is
very weak.
The only deposition variable related to ANC was dry hydrogen deposition, but the parameter
estimate is positive and is not significant at the 0.05 level (Table 8-6). The fact that the relationship is
positive instead of negative suggests that dry hydrogen deposition may be acting as a surrogate for
some other factor. Dry hydrogen deposition is significantly negatively correlated with runoff, so this
result could represent a dilution effect due to increased runoff.
8.2.3.3 Summary
There is a significant relationship between surface water sulfate concentration and deposition in the
NE, but not in the SBRP. Nonparametric statistical analysis shows that median sulfur retention is not
significantly different from zero in the NE, but is significantly greater than zero in the SBRP. Rochelle and
Church (1987) support this conclusion. Thus, watersheds are approximately at steady state with respect
to sulfur deposition in the NE but not in the SBRP, as discussed in Section 7.3. Soils in the NE have
8-11
-------
Table 8-6. Results of Regressions Relating Surface Water Chemistry to Atmospheric Deposition
in the Southern Blue Ridge Province (n = 32)
Water
Chemistry
Variable
Adjusted
Deposition
Variable
in Model
Regression Signff.a
Sign Level
Sulfate
none selected
ANC
0.10
0.07
dry hydrogen
deposition
' S = significant at 0.15 level, but not at 0.05 level
8-12
-------
little remaining sulfate adsorption capacity, so the lake sulfate concentrations reflect the deposition
gradient (see Section 9.2). In the SBRP, the watersheds are still retaining sulfur to varying degrees.
i
Watershed processes, e.g., adsorption by soils, are the primary controls on stream chemistry, so a clear
relationship does not exist between deposition and stream concentration.
In neither region do the deposition estimates explain much of the variability In current ANC. This
observation does not mean that sulfur deposition is unimportant in causing long-term surface water
acidification (i.e., toss of ANC), but rather, highlights the important roles that watershed and soil factors
play In determining current surface water ANC. These relationships are explored further In Section 9 and
the remainder of Section 8.
8.3 DERIVED HYDROLOGIC PARAMETERS
Hydrologic processes are important factors affecting the response of surface waters to acidic
deposition (Chen et ai., 1984; Peters and Driscoll, 1987). The flowpaths foifowed by water moving
through the terrestrial portion of a watershed have been hypothesized as important in controlling the
chemistry of surface waters (Chen et al., 1984; Newton and April, 1982). Acidic deposition that rapidly
moves through the watershed system will have limited contact with the soil, resulting in reduced potential
for neutralization. In this part of the Level t Analyses, we test for relationships among mapped hydrologic,
empirically modelled, and physically modelled data and selected surface water chemistry for the DDRP
northeastern fake watersheds and SBRP stream watersheds. The objectives of these analyses are to
identify watershed characteristics that are related to surface water chemistry and to infer the influence of
potential flowpaths.
8.3.1 Soil Contact fDarcv'a Law!
8.3.1.1 Introduction
An estimate of the annual flow rate of water moving through the soil and an index of soil contact
time were calculated for each drainage lake watershed in the DDRP sample (n=136). Details of the
8-13
-------
calculation are presented In Section 5.7. Briefly, the estimate of soil-water flow rate and the index of
soil contact are calculated using Darcy's Law.
Q » KAS
where: Q = lateral soil flow
K = estimate of saturated hydraulic conductivity
A - cross sectional area of flow
S - hydraulic gradient
The index of soil contact Is calculated by dividing Q by the average annual runoff (R). Figure 5-29
outlines the steps involved in the Darcy's Law calculation. In this application, we have attempted to
use the Darcy's Law approach to model flow and index of contact time at watershed scales. The
resulting estimates of flow and index of contact are essentially estimates of the theoretical maximum
potential for runoff to contact soil in a watershed.
Table 8-7 and Figures 8-1 and 8-2 summarize the results of the flow rate and index of soil contact
calculations. The estimated flow rate and index of contact were less than 0.87 m yr"1 and 1.10 m yr*1,
respectively, for approximately 90 percent of the study watersheds. Of the remaining 14 watersheds
(approximately 10 percent), 11 are located in Subregion 1D (see Figure 5-1). This region encompasses
southern New England and is comprised mainly of Massachusetts, Connecticut, and Rhode Island. These
watersheds have a high proportion of sandy soils that probably resulted in the high flow rate and index
estimates. These sites also have low ANC concentrations, however, with 8 of the 11 1D watersheds
having ANC values less than 50 peq L"1. The resulting chemistry is probably a function of the high
deposition and the limited neutralizing capacity of the sandy soils found on many of the watersheds.
For the DDRP Level I Analyses, we have tested for correlations between the estimated flow rate
and index of contact time and ANC, sulfate, sulfur retention, pH and Ca plus Mg on a regional and
subregional level. We have excluded 8 watersheds with large rate and index values (discussed above)
from the general analysis because these sites represent a special situation in the NE and resulted in large
outlier estimates.
8-14
-------
Table 8-7. Estimated Population-Weighted Summary Statistics on the
Darcy's Law Estimates of Flow Rate and the Index of Flow Relative to
Runoff
Std.
Variable Mean Median Dev. Min. Max.
Rate (m yr"1 ) 0.45 0.09 2.34 0.002 18.2
Index (yr) 0.76 0.14 4.32 0.003 35.8
8-15
-------
30-|
20-
UJ
o
111
EC
U.
I
I
I
0 0.05 0.1 0.18 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.65 0.9 0.95 1
MIDPOINT
Estimated flow contact rate, meters/yr1
Figure 8-1. Distribution of estimated contact rate using Darcy's Law calculation.
8-16
-------
40 -i
a
30-
-
FREQUENCY
3 S 8
i . i
COTI
1
1
1
1
i
0 0.05 0.1 0.15 0.2
i
0.25
,
I
0.3
0.35
I
0.4
fc 1
0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.65 0.9 0.95 1
!
MIDPOINT
Row
rate divided by runoff
Figure 8-2. Distribution of index of contact (yr) using Darcy's Law calculation.
8-17
-------
8.3.1.2 Results and Discussion
Examining the DORP northeastern region, we found very little correlation between the calculated
estimates of the Darcy's Law flow rate and index of soil contact time and sulfate, percent sulfur retention,
Ca plus Mg, ANC, or pH (R2 ranging from 0.003 to 0.03). We also looked at correlations between the
Darcy's Law calculations and the surface water chemistry variables on a subregional level. The
subregions used were defined as part of the sampling strategy for the NSWS (see Section 5.7). We
determined that there was very little correlation at the subregional level. Figures 8-3 and 8-4 show
bivariate scatter plots of the rate versus ANC and the index versus ANC. Bivariate plots of the rate and
index versus the other surface water chemistry parameters are similar. Figures 8-3 and 8-4 indicate a
large amount of scatter in the chemistry relative to the rate and index values, particularly at the low
values where most of the data tend to be concentrated.
Peters and Murdoch (1985) observed a strong relationship between the Darcy's Law index of
hydrologic contact and surface water chemistry in the two systems (Woods and Panther Lakes) they
studied as part of the Integrated Lake/Watershed Acidification Study (ILWAS). Our results differed
significantly. One difference between the DDRP study and ILWAS is the heterogeneity of the systems
studied. ILWAS involved two watersheds that had similar physical characteristics such as basin area,
relief, lake area, percent forest cover, lake size, and lake volume (Murdoch et al., 1984). The major
difference between the two watersheds was depth to bedrock with the acidic system having very shallow
soils (low index contact; Woods Lake) and the circumneutrai system (high index contact; Panther Lake)
having very deep soils. As Indicated in Section 5.7, soil depth is a key factor in the Darcy's Law
calculation. These two watersheds probably represent the possible extremes in soil depth. There are
significant variations in many of the above-mentioned characteristics among the regional DDRP
watersheds. As an example, the DDRP lakes range in size from approximately 40 to 3000 ha. Another
factor that has been identified as having significant effects on surface water chemistry is sulfur deposition
(NAS, 1984). Wampler and Olsen (1984) found that wet sulfur deposition varied in the NE with a general
southwestern to northeastern decreasing gradient. The DDRP study watersheds are located across the
NE, and thus are subject to a high degree of variability in sulfur loading. The ILWAS watersheds,
8-18
-------
400-1
300-1
4""* aoo-
O
aa
•100-
10
20
Contact Rate (meter/yr1)
Figure 8-3. Scatter plot of ANC versus contact rate calculated using Darcy's Law.
8-19
-------
400-
I
300-
200-
o*
0)
o
100-fa
a
•
.100-
10
20
Index of Contact (yr)
Figure 8-4. Scatter plot of ANC versus Index of soil contact calculated using Carey's Law.
8-20
-------
however, are only a few kilometers apart and receive very similar sulfur deposition (Murdoch et al.,
1984).
A second factor affecting the Darcy's Law result is the precision of the data available for the DDRP
study watersheds. The parameters used in calculating the lateral soil flow (Q) were estimated as
watershed averages. For all three of the major parameters, hydraulic conductivity (K), soil depth (D,
used in estimating cross-sectional area), and slope (S), an area-weighted average was calculated based
on mapping data provided by the DDRP Soil Survey. By calculating areal averages some smoothing of
the data occurred, which might not have accurately reflected the values of these variables where the
main hydrologic activity in the watersheds occurs. Identifying which soils and depth-to-bedrock classes
are most important in affecting the basin hydrology is difficult without extensive field measurements.
8.3.2 Geomorphlc/Hydrolodic Parameters
8.3.2.1 Introduction
A significant amount of work has attempted to relate hydrologic characteristics with mapped
watershed geomorphic parameters for forested watersheds (Hewlett and Hibbert, 1967; Dingman, 1981;
Carlston, 1963; Lull and Sopper, 1966; Vorst and Bell, 1977; Woodruff and Hewlett, 1970). In general,
most previously reported research is at the event level or covers short time periods (i.e., days or weeks).
In this study we are using the NSWS index chemistry value (see Section 5.3, Unthurst et al., I986a;
Messer et al., I986a; Landers et al., 1988); therefore, hydrologic response should be viewed as an annual
representation. We assume that if a system can be interpreted as a quick response system based on
geomorphic/hydrologic information, then the system is, on the average annual basis, a quick response
system. As discussed in Section 8.3, quick response systems should have less soil-runoff interaction,
resulting in reduced potential for neutralization of acidic inputs.
In this part of Level I hydrologic analyses, we test for apparent relationships among mapped
watershed hydrologic and geomorphic parameters that might affect (or be related to) hydrologic response
and selected surface water chemistry variables for 144 lake watersheds in the NE and 32 stream
8-21
-------
watersheds in the SBRP. Three watersheds with ANC > 1000 p/eq L"1 were not included in SBRP
analyses. We are testing for correlations between chemistry and watershed factors on a large regional
scale in the NE and SBRP (see Section 8.1.2 for discussion of statistics). Tables 8-8 (NE) and 8-9
(SBRP) contain summary statistics of the geomorphic/hydrologic parameters used for this analysis.
Tables 8-10 (NE) and 8-11 (SBRP) contain variable names, descriptions, and units. Detailed information
on database development is included In Section 5.7 (also, see Rochelle et al., in press-a).
Because we were specifically interested in the relationships between hydrologic/geomorphic factors
and surface water chemistry, we chose not to include other independent variables (e.g., soils, deposition)
that could influence or control surface water chemistry. In particular, for the NE, deposition explains a
large proportion of variability in some of the surface water chemistry (see Section 8.2). In some cases
the removal of deposition as a variable might have resulted in some variables acting as deposition
surrogates. We will discuss those cases as appropriate.
We have, however, also performed analyses on northeastern watersheds stratified by sulfur deposition
(wet plus dry). In these analyses, we used a simple stratification procedure based on the distribution of
sulfur deposition for our study sites. We defined four classes based on the 25th and 75th percentiles and
the median value of the deposition data (Table 8-12). We did not analyze sulfur retention based on the
stratified watersheds because deposition is a component of retention.
8.3.2.2 Results and Discussion
8.3.2.2.1 SuJfate and Sulfur Retention -
8.3.2.2.1.1 Northeast -
We found negative relationships between surface water sulfate concentration and stream order,
runoff, and maximum relief for the NE (Table 8-13). Northeastern watersheds that had low order streams
(first and second order) were associated with high sulfate concentrations. Watersheds dominated by
lower order streams tend to be headwater systems that are more likeiy to be dominated by quickflow
runoff. Quickflow results in less potential for soil-runoff interaction and subsequent neutralization of acidic
8-22
-------
Table 8-8. Estimated Population-Weighted Summary Statistics for Northeastern Geomorphic/
Hydrologic Parameters
Variable3
MIN EL
RTO
VOL
RUNOFF
WS LA
AVV~
A,
H2*O WS
B LEN
B'WIDTH
MAX REL
REL~RAT
B PSRIM
PERIMRAT
TOTSTRM
PERIN
INT
STRMORDER
DDENSITY
PER DD
B SHAPE
ELONG
ROTUND
COMPACT
M PATH7
WM PATH
Mean
319.0
0.7
2.1
64.0
19.8
5.4
0.4
0.5
0.1
2.7
1.6
134.3
0.05
10.2
3.6
3.1
2.3
0.8
2.9
0.6
0.4
1.9
0.9
0.5
1.4
765.2
1701.6
Median
327.7
0.4
0.5
64.0
11.5
3.4
0.2
0.3
0.09
2.5
1.3
103.7
0.05
9.0
2.9
0.9
0.3
0.0
3.0
0.4
0.1
1.8
0.9
0.4
1.3
489.1
1433.6
Minimum
Value
1.5
0.03
0.04
49.1
2.6
0.15
0.02
0.02
0.01
0.3
0.26
10.7
0.003
1.7
0.76
0.0
0.0
0.0
0.0
0.0
0.0
0.2
0.5
0.06
1.1
48.3
59.5
Maximum
Value
791.0
5.7
57.0
77.6
110.1
30.2
4.6
6.4
0.4
9.5
5.4
604.7
0.2
31.4
9.9
32.8
29.3
11.8
4.0
3.2
1.9
5.1
2.4
1.3
3.3
3618.6
8125.4
* See Table 5-37 for variable names, variable descriptions, and units.
8-23
-------
Table 8-9. Estimated Population-Weighted Summary Statistics for Southern Blue
Ridge Province Hydrologic/Geomorphic Parameters
Variable*
B LEN
B'PERIM
B~SHAPE
B'WIDTH
COMPACT
TOT DD
ELONG
AUG EL
M PATH
MAX REL
REL RAT
ROTUND
RUNOFF
TOTSTRM
STRMORDER
WM PATH
WS~AREA
Mean
4.8
13.9
3.0
1.7
1.4
2.6
0.7
831.5
2398.3
539.1
0.1
0.7
82.0
11.3
2.04
2548.0
9.6
Median
4.5
12.5
2.9
1.4
1.4
2.3
0.7
716.3
1951.4
538.0
0.1
0.7
86.3
7.7
2.0
2091.8
7.3
Minimum
Value
1.8
4.9
1.9
0.8
1.1
0.8
0.5
448.8
888.4
132.9
0.02
0.5
38.1
0.0
0.0
888.4
1.5
Maximum
Value
10.8
31.5
5.2
3.5
1.9
5.3
0.8
1409.7
5611.4
1368.6
0.2
1.3
114.3
41.4
4.0
5862.7
30.0
* See Table 5-38 for variable names, variable descriptions, and units.
8-24
-------
Table 8-10. Mapped and Calculated Geomorphic Parameters Collected
for the Northeastern Study Sites (Same as Table 5-37)
Parameter
Description
Units
Measured
B CENT Drainage basin centroid expressed as
an X.Y coordinate
B_LEN Length of drainage basin; air-line km
distance from basin outlet to farthest
upper point in basin
B_PERIM The length of the line which defines km
the surface divide of the drainage
basin
AH Area of all open water bodies in drainage km2
basin
INT Total length of intermittent streams km
as defined from USGS topographic maps of
aerial photos
AL Area of the primary lake km2
L_CENT Primary lake centroid expressed as
X,Y coordinates
L_PERIM Perimeter of primary basin lake km
MAXJEL Elevation at approx. highest point m
MIN_EL Elevation of primary lake m
PERIN Total perennial stream length as defined km
from USGS topographic maps and aerial
photographs
SUB_BAS(n) Area of each subcatchment in the km2
drainage basin
STRMORDER Maximum stream order (Norton) of streams
in the watershed (aerial photos used to aid
in reducing coding problems between 7.5-
and 15-minute maps)
continued
8-25
-------
Table 8-10. (Continued)
Parameter
Description
Units
TOTSTRM
Calculated
B_SHAPE
B_WIDTH
COMPACT
DDENSITY
ELONG
H20_WS
MAX_REL
M_PATH
PER_DD
PERIMRAT
REL RAT
Total stream length; combination of
perennial and intermittent
Total watershed area
Basin shape ratio;
B_LEN 2/WS_AREA
Average basin width;
WS_AREA/B_LEN
Compactness ratio; ratio of perimeter
of basin to the perimeter of a circle
with equal area;
(PERIM)/(2 x (it x
Drainage density;
TOTSTRM/WS_AREA
Elongation ratio;
(4 x WS_AREA)/L_BEN
Ratio of open water bodies area to
total watershed area
H2O_AREA/WS_AREA
Maximum relief;
MAX_ELEV - MIN_ELEV
Estimate of mean flow path
Drainage density calculated from
perennial streams only;
PERIN/WS_AREA
Ratio of the lake perimeter
to the watershed perimeter;
Lake Perimeter/B_PERIM
Relief ratio;
(MAX_ELEV-MI N_ELEV)/B_LEN
km
km2
km
m
m
continued
8-26
-------
Table a-10. (Continued)
Parameter
Description
Units
ROTUND
WM_PATH
WS_LA
Additional
RT.
Rotundity ratio;
(B_LEN)2/(4 X WS_AREA)
Estimate of weighted mean flow
path
Ratio of the total watershed area to
the area of the primary lake
Lake retention time
Volume of the primary lake
Average annual runoff; interpolated
to each site from Krug et ai. (in press)
runoff map
m
yr
106m3
cm
8-27
-------
Table 8-11. Mapped and Calculated Geomorphic Parameters Collected for the
SBRP Study Sites.
Parameter
Description
Units
Measured
B_CENT
B_LEN
B_PERIM
MAX_EL
MIN_EL
SUB_BAS(n)
STRMORDER
TOTSTRM
WS AREA
Drainage basin centrokJ expressed as
an X,Y coordinate
Length of drainage basin; air-line km
distance from basin outlet to farthest
upper point in basin
The length of the line which defines km
the surface divide of the drainage
basin
Elevation at appro*, highest point m
Elevation at watershed outlet m
Area of each subcatchment in the km2
drainage basin
Maximum stream order (Norton) of streams
in the watershed (aerial photos used to aid
in reducing coding problems between 7.5-
and 15-minute maps)
Total stream length; perennial
Total watershed area
km
km2
Calculated
AVG_EL
B_SHAPE
B WIDTH
Average elevation;
(MAX_ELEV + MIN_ELEV)/2
Basin shape ratio;
B_LEN VWS_AREA
Average basin width;
WS_AREA/B_LEN
m
km
continued
8-28
-------
Table 8-11 (Continued)
Parameter
Description
Units
COMPACT Compactness ratio; ratio of perimeter
of basin to the perimeter of a circle
with equal area;
(PERIM)/(2 x (ir x
DDENSITY Drainage density;
TOTSTRM/WS_AREA
ELONG Elongation ratio;
(4 x WS_AREA)/BJ-EN
MAX REL Maximum relief; m
MAX_ELEV - MIN^ELEV
M_PATH Estimate of mean flow path m
REL RAT Relief ratio;
(MAX_ELEV-MIN_ELEV)/B_LEN
ROTUND Rotundity ratio;
(B_LEN)2/(4 x WS_AREA)
TOT_DD Estimated drainage density based on
crenulations identified on topographic map
WM_PATH Estimate of weighted mean flow path rn
Additional
R
Average annual runoff; Interpolated
to each site from Krug et al. (in press)
runoff map
cm
8-29
-------
Table 8-12. Stratification Based on
Sulfur Deposition (Wet and Dry)
Class Deposition
1 <2.46
2 2.46 <. 3.33
3 3.33 <. 3.74
4 >3.74
8-30
-------
Table 8-13. Results of Stepwise Regression Relating
Surface Water Chemistry versus Geomorphic/Hydrologic
Parameters for the Entire NE*
ANC
Ca + Mg S042" pH
B LEN
COMPACT
DDENSITY + +
ELEV
H20 WS
MAXREL - +
PERIN
RT
RUNOFF - -
STRMORDER + +
Adjusted R2 0.15b
0.11 0.29 0.20
* ANC. Ca + Mg, and pH: n « 141
SO42': n = 142
b Significant at the 0.15 Isvel
8-31
-------
inputs. Also, we found that the watersheds with lower order streams tended to be located at the higher
elevations. These systems typically receive higher sulfur deposition due to depositional patterns in the
NE, particularly Adirondack watersheds. The combination of high sulfur deposition and reduced potential
for soil interaction due to Increased percent quick runoff leads to higher surface water sulfate
concentrations. We found no significant relationships for the entire NE between sulfur retention and the
geomorphic/hydrologic parameters.
8.3.2.2.1.2 Northeast - stratified by sulfur retention -
The results of statistical analyses between the hydrologic/geomorphlc parameters and in-lake
sulfate are presented in Table 8-14. Although some individual parameters were significantly related to
sulfate for deposition classes 1 and 4, these were not consistent. None of the parameters appeared as
a significant predictor in more than one of the deposition classes.
8.3.2.2.1.3 Southern Blue Ridge Province —
We identified no significant correlations between sulfate or sulfur retention and the
hydrologic/geomorphic parameters for the SBRP (Table 8-15). A probable explanation for the lack of
significant correlations is the relative homogeneity of the SBRP watersheds in terms of both sulfur
chemistry data and the hydrologic/geomorphic parameter values.
8.3.2.2.1.4 Regional comparison -
In the NE, we identified stream order, runoff, and maximum relief as significant predictors for
surface water sulfate concentration. These findings suggest that headwater streams are associated with
high surface water sulfate concentrations due to a higher percentage of quick runoff. A higher
percentage of quickflow would result in less soil interaction and, consequently, higher surface water
sulfate. We found no significant relationships between the hydrologic/geomorphic parameters and sulfate
concentration or sulfur retention in the SBRP.
8-32
-------
Table 8-14. Stepwise Regression Equations for Surface Water Chemistry and
Hydrologic/Geomorphic Parameters Based on Sulfur Deposition Stratification
Class=1 ANC
COMPACT
DDENSITY +
H2O WS
MAXREL
PER DD
PERfMRAT +
RT
RUNOFF
Adjusted R2 0.49
Ciass=2 ANC
B PERIM +
B~WIDTH +
COMPACT
MAXREL +
STRORDER
WSAREA
Adjusted R2 0.39
Class=3 ANC
ELEV
DDENSITY +
H20 WS
MEANPATH
PER DD
VOL" +
WSAREA
Adjusted R2 0.68
Class=4 ANC
COMPACT
ELEV
ELONG
REL RAT
RUNOFF
Adjusted R2 0.36
pH Ca + Mg S042"
+ + +
0.36 0.54 0.31
pH Ca + Mg SO42'
*
0.27
pH Ca + Mg S042"
4- +
0.31 0.73
pH Ca + Mg SO42'
-
0.36
8-33
-------
Table 8-15. Results of Stepwise Regression Relating Surface Water Chemistry
and Geomorphic/Hydrologic Parameters for the SBRPa
ANC Log(Ca + Mg) S042" pH Sulfur Retention
TOT DD
REL~RAT
RUNOFF
Adjusted R2 0.39 0.15 b 0.16
* ANC. Ca + Mg, and pH: n = 32
S04 and sulfur retention: n » 31
b No variables met the 0.15 significance level for entry into the model.
8-34
-------
8.3.2.2.2 pH, ANC, and (Ca plus Mg) -
8.3.2.2.2.1 Northeast -
Runoff was the only geomorphic/hydrologic parameter that was related to Ca plus Mg. The
relationship is negative with an Increase in runoff resulting in a decrease in Ca plus Mg (R2 = 0.11). This
relationship is probably due to a dilution effect.
As was the case with Ca plus Mg, we found significant relationships between ANC and the
geomorphic/hydrologic parameters (Table 8-13, R2 = 0.15). We found that drainage density and stream
order were positively related with ANC and runoff was negatively related to ANC. As discussed above,
stream order is probably reflective of the relative position of the watershed (i.e., headwater), with higher
stream order systems tending to have smaller percentage contributions of quick runoff to total runoff.
The negative relationship with runoff might be due to dilution effects.
Drainage density, maximum relief, and stream order were positively related to pH (R2 = 0.20). The
positive relationships between pH and stream order and pH and drainage density are probably functions
of relative proportion of quickflow runoff associated with the high stream order systems. As discussed
above, the low stream order systems tended to be located at high elevations and have a greater potential
for quickflow runoff and high sulfur deposition. The high stream order systems we studied in the
northeastern typically are low elevation systems with gentler slopes and larger watershed areas. These
systems probably have a greater potential for soil interaction and subsequent neutralization of acidic
inputs. Drainage density was relatively low for most of the northeastern watersheds since these
watersheds are primarily lake watersheds. The higher drainage densities are generally found in the lower
elevation areas where stream development is more advanced.
8.3.2.2.2.2 - Northeast - stratified by sulfate deposition class -
The results of statistical analyses between the hydrologic/geomorphic parameters, stratified by
sulfate deposition class, and pH, ANC, and Ca plus Mg were presented in Table 8-14. No consistent
8-35
-------
relationships were found between the hydrologic/geomorphic parameters and pH or Ca plus Mg. None
of the parameters appeared as a significant predictor in more than one of the deposition classes.
We found a significant positive correlation between ANC and drainage density in deposition classes
1 and 3. A significant negative correlation between ANC and runoff was found in deposition classes 1
and 4. No other consistent relationships were found. These findings are consistent with those for the
entire NE and were discussed more fully in Section 8.3.2.2.2.1.
8.3.2.2.2.3 Southern Blue Ridge Province -
The results of statistical analyses between the hydrologic/geomorphic parameters and pH, ANC,
and Ca plus Mg were presented in Table 8-15. A log-transformation of Ca plus Mg was used in this
analysis to make the variance of the residuals constant. We found no significant relationships between
pH and the hydrologic/geomorphic parameters in the SBRP.
We found relationships between ANC and the hydrologic/geomorphic parameters in the SBRP.
ANC was negatively correlated with runoff and relief ratio and positively correlated with drainage density.
Drainage density was based on crenulations identified on a topographic map. The negative correlation
between ANC and runoff suggests that higher runoff results in lower ANC streams. This relationship
probably reflects a dilution effect. Relief ratio was negatively correlated with ANC. High relief ratio
watersheds tend to be headwater streams with a higher percentage of quick runoff, which would lead to
less Interaction of water with the soil matrix and, hence, lower ANC. The positive relationship between
ANC and drainage density may also be a function of relative position of the watershed within the region.
We also found limited relationships between Ca plus Mg and the hydrologic/geomorphic
parameters. As with ANC, Ca plus Mg was negatively correlated with runoff. As discussed previously,
the negative correlation between Ca plus Mg and runoff is probably due to a dilution effect.
8-36
-------
8.3.2.2.2.4 Regional comparisons -
We found similar hydrologic/geomorphic predictors for ANC and Ca plus Mg in the NE and SBRP.
Although we found significant predictors for pH in the NE, no significant correlations were found in the
SBRP. In the NE, stream order and drainage density were positively related to ANC. Lower stream order
watersheds are primarily headwater systems that have a high percentage of quickflow and, therefore,
would tend to have lower ANC. Similarly, drainage density is a measure of position within the watershed.
Streams with lower drainage densities tended to be headwater streams while lower elevation watersheds
tended to have a more developed drainage network. En the SBRP, ANC was also positively correlated
<
with drainage density. Additionally, relief ratio was negatively correlated with ANC. Similar to drainage
density and stream order In the NE, relief ratio is probably a function of the relative position of the
watershed. Watersheds with high relief ratios tend to be headwater systems and, therefore, have lower
ANC due to increased quickflow.
Runoff is a second factor that appeared to influence ANC and Ca plus Mg in both the NE and
SBRP. Significant negative relationships were found for both ANC and Ca plus Mg in both regions.
These relationships are probably due to the increased dilution of stream and lake chemistry in areas
where runoff is high.
Other significant predictors were found in the NE but not In the SBRP. These predictors included
basin perimeter for ANC, and drainage density, maximum relief, and stream order for pH. The
identification of a larger number of predictors in the NE may be a function of either the larger sample
size (141 in the NE vs. 32 in the SBRP) or the relative homogeneity of the SBRP.
8.3.3 TOPMODEL. Parameters
The hydrologic model TOPMODEL which is based on the variable source area concept, was used
to characterize flow path partitioning of the DDRP watersheds. TOPMODEL was chosen because the
model uses readily available topographic and soils information, and it predicts internal states that can be
used to partition streamflow. A more complete description of TOPMODEL is given in Section 5.7.2.1.
8-37
-------
8.3.3.1 introduction
TOPMODEL characterizes flowpath partitioning for each watershed by characterizing the spatially
aggregated distribution function of ln(a/KbTanB) in the NE and ln(a/TanB) in the SBRP where "a" is the
area drained per unit contour, TanB" Is the local slope, "K" is the hydraulic conductivity, and "b" is depth
to bedrock (Beven and Kirkby, 1979; Beven, 1986; Wolock et al., 1989). Details of the calculations are
presented in Section 5.7.2.1.1.3. Values of ln(a/KbTanB) and ln(a/TanB) have been correlated with the
likelihood of producing surface runoff. Surface runoff is defined as saturation-excess ("return") flow rather
than infiltration-excess ("Hortonlan") flow. High values of ln(a/KbTanB) or ln(a/TanB) suggest areas within
a watershed that are likely to produce surface runoff. These areas would typically be characterized as
topographically convergent, low transmissivity areas. Conversely, low ln(a/KbTanB) or ln(a/TanB) values
represent areas that have low potential for surface runoff generation (e.g., well-drained soils draining little
upslope area). The mean of ln(a/KbTanB) or ln(a/TanB) is the critical parameter for characterizing an
individual watershed {Wolock et al., 1989). In the NE, four watersheds were deleted from the analysis
due to a lack of relief as portrayed in the 1:250,000-scale digital elevation models (DEM), resulting in a
total of 141 study watersheds. In the SBRP, we eliminated three watersheds with ANC > 1000 peq L"
1 from the analyses resulting in a total of 32 watersheds.
For the NE, values of ln(a/KbTanB) are summarized in Table 8-16. Mean ln(a/KbTanB) values
ranged from -3.38 to 3.40 with a regional mean of 1.03. Subregional means were highest in subregion
1B (2.40), followed by Subregions 1E (1.48), 1A (0.91), 1C (0.77), and 1D (-0.67). For the SBRP, values
of ln(a/TanB) are summarized in Table 8-17. Mean ln(a/TanB) values ranged from 7.34 to 8.89 with a
regional mean of 7.81. Within Level I Analyses, we have tested for correlations between mean
ln(a/KbTanB) or ln(a/TanB) values and ANC, sulfate, sulfur retention, pH, and Ca plus Mg on a regional
scale in the NE and in the SBRP. We used Spearman's correlation coefficient rather than Pearson's, as
the scatter plots did not suggest a bivariate normal distribution. Spearman's correlation coefficient does
not require normality.
8-38
-------
Table 8-16. Population-Weighted Summary Statistics for
!n(a/KbTanB) for the Northeast
Region
Std.
Mean Dev. Min.
Max.
Northeast 1.03 1.08 -3.38 3.40
1A 0.91 0.86 -0.73 3.04
1B 2.40 2.42 1.34 3.40
1C 0.77 0.75 -1.20 1.71
1D -0.67 -0.20 -3.38 1.33
1E 1.48 1.63 -0.59 3.18
8-39
-------
Table 8-17. Population-Weighted Summary Statistics for
ln(a/TanB) for the Southern Blue Ridge Province
Region Mean Median Min. Max.
SBRP 7.81 7.74 7.34 8.89
8-40
-------
8.3.3.2 Results and Discussion
8.3.3.2.1 Northeast
Statistical correlations between ln(a/KbTanB) and surface water chemistry are given in Table 8-18.
We found no significant correlations between ln(a/KbTanB) and sulfur retention or sulfate concentration.
Noisy but significant positive relationships were found between values of 1n(a/KbTanB) and ANC (r =
0.28), Ca plus Mg (r = 0.31), and pH (r = 0.27). Scatter plots for these relationships are shown in
Figures 8-5, 8-6, and 8-7, respectively. The relationship between ln(a/KbTanB) and pH is particularly
noisy (Figure 8-7).
The positive correlations between values of ln(a/KbTanB) and ANC and Ca plus Mg are difficult to
explain. High watershed mean values of ln(a/KbTanB) suggest that a larger percentage of storm flow
originates from quickflow mechanisms (e.g., return flow), whereas watersheds with low values of
ln(a/KbTanB) are dominated by subsurface storm flow. A larger proportion of quickflow should result
in less overall contact of water with the soil matrix and, hence, lower ANC and Ca plus Mg.
Positive correlations between ln(a/KbTanB) and ANC and Ca plus Mg would seem to contradict
the findings reported earlier (e.g., see Section 8.3.2). As discussed in Section 8.3.2, we found stream
order to be consistent predictor variable for ANC. The relationship between stream order and ANC was
positive, thus lower ANC tended to be associated with lower order streams. These lower order streams
are generally high elevation with small drainage areas and, therefore, have a higher potential for quickflow
runoff, resulting in low ANC. The positive correlation between values of ln(a/KbTanB) and ANC suggests,
however, that more quickly responding systems result in higher ANC.
One possible explanation for the positive correlations between values of ln(a/KbTanB) with ANC
and Ca plus Mg is given in Wolock et al. (1989). In watersheds with high mean values of ln(a/KbTanB)
watersheds, less water passes through the soil matrix during high flows, as compared to low mean
ln(a/KbTanB) watersheds, which are dominated by subsurface storm flow. Throughout the hydrochemical
history of the catchments, more water has passed through those with the low mean ln(a/KbTanB) values
8-41
-------
Table 8-18. Spearman's Correlation Coefficients Between ln(a/KbTanB) and Surface Water
Chemistry
Region n ANC SO4 S Ret. pH Ca plus Mg
NE 141 0.28a 0.05 0.28 0.27* 0.31
* Significant at p = 0.10
a
8-42
-------
400-1
300-
_ 200-
O
100 -
o -
-100
B G
a
a
B g
B a
°a
B B
a e
Ba ® g
•„• -
Bfl
Q O
. -4
-2
In(aKbTanB)
Figure 8-5. Scatter plot of ANC versus ln(a/KbTanB). TOPMODEL was used to calculate values
of ln(a/KbTanB).
8-43
-------
600-1
500-
400-
cr
o>
.3,
ra
(0
O
300-
200-
100-
D Q
a f
-4
-2
In(aKbTanB)
Figure 8-6. Scatter plot of Ca plus Mg versus ln(a/KbTanB). TOPMODEL was used to calculate
values of ln(a/KbTanB).
8-44
-------
8 -
7 -
JC
Q.
5H
B B B
a q, a° a_ QQ
ra B
_ e QBE
* n « a • B
3 a
Q
a
0
Q D
-4 -2 0
In(ayKbTanB)
Rgure 8-7. Scatter plot of pH versus ln(a/KbTanB). TOPMODEL was used to calculate values of
ln(a/KbTanB).
8-45
-------
than through those with high mean ln(a/KbTanB) values, thereby consuming more of the buffering
capacity of the low mean catchments. If the buffering capacity of all catchments were initially the same
and finite, then the low mean catchments should be more depleted of buffering capacity. Low mean
catchments should, therefore, have lower ANC. Given this hydrochemical scenario, and assuming that
ANC represents subsurface flow chemistry, then catchments with high ln(a/KbTanB) values should have
high ANC.
Other factors may explain why ln(a/KbTanB) was not significantly related to surface water
chemistry. First, there are numerous sources of uncertainty in the calculations of ln(a/KbTanB). Digital
elevation models at 1:250,000 scale were used to compute values of "a" and TanB". The OEMs are
generalized to a large degree when compared to a watershed mapped at a scale of 1:24,000, which
tends to become particularly critical on smaller watersheds. Additional uncertainties are functions of the
errors associated with the DDRP Soil Survey information (e.g., error in map unit description, aggregation,
depth-to-bedrock estimates). Second, there are many controls on surface water chemistry that were not
considered within this analysis (e.g., watershed processes, sulfur deposition). For example, the physical
and chemical characteristics of soils within "low" versus "high" !n(a/KbTanB) areas are undoubtedly
different. The spatial variability of soils within a catchment, however, were not considered within these
analyses. Finally, because TOPMODEL characterizes the partitioning of storm flow through the concept
of variable source areas, it may be more suitable as an event model. Variable source areas tend to be
active only during storm events and would not be expected to contribute a significant amount of runoff
during baseflow conditions. Because NSWS surface water chemistry more accurately represents baseflow,
it may be unrealistic to expect an index of variable source areas to be correlated with surface water
chemistry.
8.3.3.2.3 Southern Blue Ridge Province -
Statistical correlations between ln(a/TanB) and surface water chemistry are shown in Table 8-19.
We found no significant correlations between ln(a/TanB) and sulfate, sulfur retention, pH, ANC, or Ca plus
8-46
-------
Table 8-19. Pearson's Correlation Coefficients Between ln(a/TanB) and NSS
Pilot Chemistry
Region n ANC S04 S Ret. pH Ca plus Mg
SBRP 32 0.28 -0.15 0.08 -0.07 0.18
8-47
-------
Mg. The possible factors responsible for lack of a significant correlation between ln(a/TanB) and surface
water chemistry are discussed more fully in Section 8.3.3.2.1.
8.3.3.3 Summary
In the Level I hydrologic analyses we attempted to relate empirically physically modelled parameters
and mapped geomorphic/hydrologic parameters to surface water chemistry. The objective of these
analyses was to use Indirect measurements of hydrology, which can be obtained relatively easily, to
describe surface water chemistry. These measurements include estimates of soil contact based on
Oarcy's Law, output parameters from the hydrologic model TOPMODEL, and mapped measurements of
geomorphology and hydrology.
• We did not determine any significant relationships between the Darcy's Law estimates and
surface water chemistry. The major factor determining this lack of relationship is the probable
large error associated with watershed soil depth, hydraulic conductivity, and slope estimates.
• Although a limited significant relationship was identified between TOPMODEL output and
surface water chemistry, this result was not necessarily explainable nor consistent with theory.
One probable explanation for the tack of correlation is that TOPMODEL is based on the
variable source area concept and is more appropriately an event level model.
• Relationships between the mapped geomorpnic/hydroiogtc parameters and surface water
chemistry were identified. The major variables that were significantly related were runoff,
stream order, and an estimate of basin shape.
These findings suggest that hydrologic/geomorphic characteristics are related to surface water
chemistry, although specific processes cannot be identified. Although we found little correlation between
Darcy's Law and TOPMODEL with surface water chemistry, we chose to include these analyses within
this report for documentation purposes. Our conclusions neither confirm nor repudiate the findings of
April and Newton (1985) and Chen et al. (1984).
8.4 MAPPED BEDROCK GEOLOGY
A parameter hypothesized to be important in controlling the composition of surface waters is
bedrock geology. Different lithologies exhibit different reactivities. Some, such as limestones or
dolostones, are highly reactive. Waters in contact with these rock types quickly attain equilibrium with
-------
the carbonate regardless of the acidity of the incident deposition. Other lithologies are, effectively,
unreactive. For example, quartzites will modify the composition of incident deposition only slightly. As
a result, waters evolving from quartzite systems tend to strongly reflect the composition of the incident
deposition.
In addition to lithology, a number of other factors contribute to the extent of interaction between
bedrock and soil and ground waters. Porosity and permeability of the bedrock, in conjunction with other
parameters (e.g., hydraulic head), control water contact times and the rates of infiltration through ground
water. Longer contact times provide greater opportunity for the water to react with the bedrock, thereby
increasing cation concentrations and ANC. Structural considerations, such as the strike of a bedrock unit
relative to the aspect of the watershed could influence water infiltration and contact times as well.
Unfortunately, quantifying these non-lithological characteristics of bedrock was not possible from the
sources used for this study. As a result, the analyses here focus on bedrock lithology as the variable of
interest for evaluating statistical relationships between bedrock and surface water chemistry.
The first step in the bedrock analysis was to identify the types of bedrock within each of the DDRP
watersheds. Using an ARC/INFO Geographic Information System (GIS) (see Section 5.4.1.7), watershed
boundaries were overlaid onto state geology maps and the bedrock units mapped within the boundaries
identified.
For the 145 watersheds located in the NE region, a total of 136 different mapped bedrock units
were identified. The large number of bedrock types relative to the number of watersheds results in
insufficient degrees of freedom for a reasonable statistical evaluation of the relationship between individual
bedrock types and surface water chemistry. Therefore, it was necessary to group the different units into
more generic classes in order to perform the analyses. This classification was accomplished in a two-
step process. The first step was to assign each mapped unit to a generic bedrock type. Then, we
assigned a relative reactivity to each of these generic rock types.
8-49
-------
8.4.1 DDRP Bedrock Sensitivity Scale
A number of studies have been undertaken to evaluate the relationship between bedrock geology
and surface water composition (Hendrey et al., 1980; Rapp et al., 1985; Shilts, 1981). These studies have
been used on regional scales to help identify areas that are potentially sensitive to the effects of acidic
deposition.
Hendrey et al. (1980) used a 4-point scale to delineate rocks of different reactivities. Highly reactive
rocks, such as limestones, dolostones, or highly fossiliferous rocks, were assigned scores of (4). As the
reactivity of the rocks decreased, the reactivity score was decreased. The reactivity scale of Shilts (1981)
was developed along similar lines, except that he used a value of (1) to designate the most reactive
lithologies. Some structural considerations were implicitly included in these rankings. For example, marine
shales are cation-rich, but because of limited permeability and the presence of pyrite, these units were
assigned reactivities of (2) on the Hendrey et al. scale and (3) on Shilts' scale. These ranking systems
have proven useful for identifying regions potentially sensitive to acidification.
Rapp et al. (1984) developed a 10-point scale to evaluate the relationships between bedrock
geology and surface water chemistry for lakes located in the Upper Midwest. On their scale, (1)
represents the most reactive bedrock types (limestones, marbles, calcareous tilts), while (10) represents
the least reactive units (e.g., quartzrtes, organic deposits). Significant correlations were found between the
amounts of nonreactive bedrock and surface water chemistry in their study area.
In attempting to use the above scales in the DDRP Level I analyses, several difficulties were
encountered. The Rapp et al. scale was developed for the Upper Midwest. As such, it does not contain
the range of lithologies encountered in DDRP and is not appropriate for use here. In working with the
other scales, a major problem has been the lack of resolving power for distinguishing the different
contributions of weathering to the range of compositions observed among lakes. The watershed sample
used in the DDRP was selected based on lake water ANC. Watersheds with surface water ANC > 400
eq L*1 were excluded from the study, and the majority of systems have surface waters with ANC < 200
8-50
-------
eq L"1. Using the Hendrey et al. scale, the majority of systems included in DDRP have bedrock
sensitivity scores of (2). Similar limitations have been encountered using the other scales in this
evaluation. For this reason, we wanted to develop a sensitivity or reactivity scale that would allow the
us to distinguish the relative ANC generating capacities of a group of lithologically different but, otherwise,
moderately unreactive rock types.
With this goal in mind, a 6-point scale was developed with the Intent of separating rocks with
slightly different reactivities into different categories. The top two categories, (5) and (6), are reserved
for the reactive and highly reactive lithologies of Hendrey et al. (1980), corresponding to their classes
(3) and (4). Within our classes (1) through (4), we attempted, then, to distinguish rock types that have
only slightly different reactivities with surface or ground waters.
Classification of individual mapped bedrock units was accomplished in a two-step process. First,
each mapped unit was classified according to a generic rock type. Table 8-20 lists the rock types
considered. Once this step was completed, a reactivity score was assigned to each of the generic rock
types. Table 8-21 summarizes the reactivities assigned. These assignments were reviewed by both
project participants and a limited number of individuals external to the project. Consensus was usually,
but not universally, attained for each of the scores. In all cases, project participants made final decisions
concerning the selection of the relative reactivity score. The decisions regarding final reactivity
assignments were made independent of any knowledge of the ANC of the surface waters associated with
the specific bedrock units.
8.4.2 Results
For the DDRP samples in the NE and SBRP, multiple estimates of the aggregated bedrock reactivity
were synthesized. In the following analyses, the variable Mean is the weighted average of the sensitivity
codes for a watershed, where the weights are the areal proportions of the watershed covered by the
bedrock type. The variable Max is the maximum sensitivity code observed on a watershed. The variable
HSup is the percent of the watershed covered by bedrock with sensitivity codes that are at least 5.0.
8-51
-------
Table 8-20. Tabulation of the Generic Bedrock Types Used to Classify the Mapped Units Identified
on State Map Legends
Symbol
Description
Symbol
Description
101 alkali feldspar granite
I02 granite
I03 quartz porphyry
I04 granite porphyry
I05 granophyre
106 pegmatite
I07 aplite
IOS syenite
I09 quartz syenite
110 alkali feldspar syenite
111 granodiorite
112 tonalite
113 monzonite
114 quartz monzonite
115 diorite
116 quartz diorite
117 alkali feldspar rhyolite
118 rhyolite
119 dacite
I20 obsidian/pumice
121 diorite porphyry
I22 andesite
I23 latite
I24 trachyte
125 phonolite
(26 gabbro
I27 anorthosite
I28 norite
(29 diabase
ISO basalt
131 charnockite
I32 ultramaflc(s)
C23 Organic deposits/peat
C22 mixed limestone/dolostone
C21 dolomites/dolostones
C20 limestones
C19 interfingering Is/dastics
M01 mixed metamorphics
M02 quartzite
M03 schist
M04 phyllite
M05 slate
M06 gneiss
M07 granitic gneiss/granofel
M08 greenstones
M09 amphibolites
M10 serpentinites
M11 chlor/amphib/epid schist
M12 marble
M13 suifldic schist
M14 calc-silicates
M15 leucocratic gneisses
M16 migmatites
M17 mixed metaclastics
M18 mixed types
C01 quartz sandstone
C02 sulfidic pelite/shale
C03 chert
C04 iron formations
COS pelite/mudstone
C06 shale
C07 argillite
COS conglomerate
C09 sandstone
C10 arenite/arkose
C11 graywacke
C12 sittstone
C13 mixed elastics
C14 calcareous shale
C15 calcareous siltstone
C16 calcareous sandstone
C17 calcareous arenite
C18 calcareous conglomerate
8-52
-------
Table 8-21. Tabulation of the Generic Bedrock Types Used to
Classify the Mapped Units Identified on State Map Legends
Reactivity Explanation Generic bedrock types
Score (from Table 8-22)
C01 C02 C03 C04 C23
1 minimally reactive M02 M13
I06 I07 I09
COS COG C07 COS C09 C13
2 slightly reactive M03 M04 M05 M07 M16 M17
101 I02 I03 I04 I05 I08
C10 C11 C12 C14
3 modestly reactive M01M06M11
111 112 117 118 119 I24 131
C15
4 moderately reactive M08 M09 M14 M15
113 114 115 116 I20 121 I22 I23
127 128 129 130
C19
5 reactive I25126132
C16 C17 C18 C20 C21 C22
6 highly reactive M12
8-53
-------
The statistical analyses on the DORP data used standard regression procedures discussed in Section
8.1.2.
As can be seen in Table 8-22, the variables Mean and Max do not differ much between subregions
in the NE, or between the NE and the SBRP. The average and maximum sensitivity codes for watersheds
are highest in Subregion 1A and lowest in Subregion 1B, but it Is unlikely that these differences are
significant. HSup. the percent of the watershed with sensitivity codes of 5 or 6, has the highest average
In Subregion 1E. This result means that more watersheds in this subregion are classified as having
significant percentages of reactive to highly reactive bedrock types. The data for the SBRP indicate that
the estimated sensitivity codes are similar to those for the northeastern subregions with lower proportions
of the highly reactive bedrock types.
The measurement error analyses show that there are highly significant relationships between
bedrock geology and surface water chemistry, particularly ANC and sum of base cations. These
relationships may be masked in analyses performed on the DORP watersheds, since measurement error
models cannot be used with the more detailed geological information available on these watersheds. The
possible masking of existing relationships should be kept in mind when reviewing these results.
8.4.2.1 Sulfate and Percent Retention
8.4.2.1.1 Northeast -
As discussed in Section 8.2, sulfate deposition appears to be the dominant source of sulfate in
northeastern surface waters. After sulfate deposition is taken into account, no bedrock geology variable
appears in the model (Table 8-23). This observation is not surprising, because the reactivity of a rock
type is not necessarily related to its sulfur-bearing potential. We would not expect most of the DDRP
bedrock types to act as internal sources for sulfur. Some of the more reactive rock types, especially
limestones and dolostones, will release sulfate by the dissolution of gypsum. However, we anticipate that
"disturbed" systems, e.g., mining operations for coal or base metals, would serve as the primary internal
sources of sulfur for surface waters. These operations expose fresh, unweathered sulfide minerals that
8-54
-------
Table 8-22. Regional and Subregional Statistics for the Bedrock Sensitivity Code Variables
Average
Median
Min.
Max.
Entire Region 1
Mean
Max
HSup
Subregion 1A
Mean
Max
HSup
Subregion 1B
Mean
Max
HSup
Subregion 1C
Mean
Max
HSup
Subregion 1D
Mean
Max
HSup
Subregion 1E
Mean
Max
HSup
Entire Region 2 (SBRP)
Mean
Max
HSup
2.6
3.3
3.6
3.t
3.7
0.8
2.1
2.4
0.02
2.7
3.8
3.4
2.3
2.9
1.3
2.7
3.5
14.5
2.2
2.7
0.3
2.5
3.0
0.0
3.0
3.5
0.0
2.0
2.0
0.0
2.8
3.5
0.0
2.3
2.3
0.0
2.0
2.5
0.0
2.0
2.0
0.0
1.0
1.0
0.0
2.35
2.5
0.0
1.0
1.0
0.0
1.0
1.0
0.0
1.8
2.0
0.0
1.0
1.0
0.0
1.0
1.0
0.0
6.0
6.0
100.0
4.0
6.0
18.8
3.0
5.0
0.4
3.6
6.0
43.8
3.0
5.0
18.6
6.0
6.0
100.0
3.4
6.0
8.2
8-55
-------
Table 8-23. Results of Regressions of Surface Water Chemistry on Bedrock Sensitivity Code
Statistics and Deposition Estimates for Northeast
Water
Chemistry
Variable
Sulfate
Percent
Sulfur
Retention
ANC
Ca+Mg
R2
0.3618
0.1370
0.0558
0.0566
Adjusted
0.3573
0.1169
0.0491
0.0432
Variabie
in Model
total sulfate
total sulfate
Mean
Max
HSup
H5up
Regression Signif.8
Sign Level
+ ***
.j. ***
**
+ S
± **
+ *
total H
PH
0.0878
0.0683
Max
Mean
HSup
**
*
* S = Not significant at 0.05 level
* - Significant at 0.05 level
** - Significant at 0.01 level
*** = Significant at 0.001 level
8-56
-------
can be oxidized and therefore have a high potential for contributing to the sulfur and hydrogen ion
budgets in these systems.
The stepwise regressions suggest both the Mean and Max variables exhibit negative correlations
with percent sulfur retention. That is, watersheds with the least reactive bedrock types tend to retain
higher percentages of sulfate than do the watersheds containing more reactive bedrock types. The
reasons for this correlation are not immediately obvious. One possible explanation might be that since
the more reactive bedrock types may act as minor internal sources for sulfur, watersheds containing
highly reactive Itthologies may have soils that carry higher ambient loads of sulfur. As a result, these soils
may allow a greater percentage of sulfur delivered to the watersheds via deposition to pass on to surface
waters. If this were a small effect in watersheds containing bedrock with sensitivity codes of 5 or 6, then
it Is possible that these internal sources might not have been targe enough to identify in the sulfate
regression analysis, and yet be of sufficient magnitude to have a measurable effect on the sulfur budgets.
An alternative explanation could be that the more acidic rock types provide conditions conducive to the
generation of oxides in soils, hence increase the sulfate adsorption capacities of the soils. Details of
this hypothesis will be addressed in the section on soil chemical properties and their relationship to
surface water chemistry (see Section 8.8.4.)
8.4.2.1.2 Southern Blue Ridge Province •
When we performed the regressions on the 32 SBRP watersheds, the only explanatory variable that
appeared was H5up. i.e., the percentage of the watershed covered by bedrock with sensitivity codes of
5 or 6 (Table 8-24). HSup was positively correlated with surface water sulfate and negatively correlated
with percent sulfur retention. On analysis of the residuals, however, these effects could be attributed
primarily to one watershed (2A08808), which has high stream sulfate and very low % S retention. When
this site and watershed 2A07827 (Table 8-24) were excluded from the regression, no regressor variables
were identified as significant.
8-57
-------
Table 8-24. Results for SBRP of Regressions of Surface Water Chemistry on Bedrock Sensitivity
Code Statistics and Deposition Estimates
Water
Chemistry
Variable
Sulfate
Model 2
Percent
Sulfur
Retention
Model 2
ANC
Ca + Mg
Model 2
Models
pH
* S = not significant
Adjusted Variable Regress. Signtf.
R2 R2 in Model Sign Level
0.2526 0.2277 H5up + **
none significant
0.2790 0.2550 HSup - **
none significant
0.0859 0.0554 H5up + S
0.2546 0.2297 H5up + . **
0.0791 0.0450 H5up + S
none significant
none significant
at 0.05 level
Watershed6
Removed
2A08808(B)
2A07827(B)
2A08808(B)
2A07827(B)
2A07827(L)
2A07813(O)
2A08808(L)
2A07826(O)
2A07833(L)
* - significant at 0.05 level
** = significant at 0.01 level
b (L) = site removed is a leverage point
(0) * site removed is an outlier
(B) = site removed is both a leverage point and an outlier
8-58
-------
8.4.2.1.3 Comparison of Regions -
In both regions, the more reactive bedrock types are associated both with higher surface water
sulfate and with lower percent sulfur retention. In this regard, the effect of bedrock on sulfur dynamics
within watershed appears to be similar across both regions. Although there is no a priori reason to
expect a relationship between the sensitivity scale and sulfur dynamics, our results do suggest that the
most reactive bedrock types also act as (minor) internal sources for sulfate, which influences the way in
which a watershed wilt respond to the effects of elevated sulfur deposition.
8.4.2.2 Sum of Base Cations, ANC, and pH
As discussed in Section 8.1, the sum of calcium and magnesium is used to represent base cations
in these analyses. This representation was necessary because of the non-local sources for sodium {e.g.,
sea salt and road salt) to the surface waters in many of the study watersheds.
8.4.2.2.1 Northeast -
The stepwise regression analyses indicate positive relationships between surface water ANC and
the regressor variable H5up. The positive relationship with the percent of watershed covered by high
bedrock sensitivity codes indicates that the bedrock is contributing significant amounts of ANC through
weathering.
In conjunction with these analyses, the regressions show that surface water pH has statistical
relationships with the variables Max. Mean and HSup. The positive relationship with Max suggests that
watersheds with higher bedrock sensitivity codes have higher pH values. The relationships with Mean
and HSup may indicate correction factors for particular watersheds with high or low bedrock sensitivity
codes.
We find a strong positive relationship between Ca plus Mg and the sensitivity code for the
watershed. This finding suggests that there is a relationship between the presumed reactivities assigned
8-59
-------
to the bedrock types and the rate of cation supply to surface waters. The higher reactivity rankings are
associated with higher weathering rates and, hence, stronger internal sources for base cations.
The stepwise regressions indicate other variables contribute to the regulation of base cation
concentrations in surface waters. In particular, there is a positive relationship between surface water
calcium plus magnesium and the total hydrogen ion deposition. If this correlation has any significance
in terms of ecological processes, then two explanations can be offered. First, the relationship may
indicate possible leaching of base cations from the soil exchange complex in excess of the mass
contributed by primary mineral weathering. The alternative explanation, especially for those watersheds
containing some carbonate bedrock (e.g., limestones), would be that the higher incident acidic deposition
allows for additional dissolution of the carbonate and hence contributes to the base cation budget.
8.4.2.2.2 Southern Blue Ridge Province -
The stepwise regression for the sum of calcium and magnesium concentration showed a positive
correlation with H5up. as shown in Table 8-24. Residual analysis indicated that watershed 2A08808 was
a strong leverage point. This site has already been discussed as an internal source of sulfur. Upon
removing this site, as well as two other watersheds, from the analysis, the stepwise regression procedure
still selected HSup. This correlation, however, was no longer significant at the 0.05 level.
The stepwise regression for ANC showed a positive relationship between this variable and HSup.
As in the calcium and magnesium model, the stepwise procedure selected HSup. but it was not significant
at the 0.05 level. This regression and the previous one suggest that the higher bedrock sensitivity
numbers are somewhat associated with increased base cation concentrations and ANC.
In the SBRP, the analyses do not show any consistent relationship between the bedrock sensitivity
numbers and the pH of the surface waters across the region. The stepwise regression for pH selected
none of the deposition or bedrock geology variables, presumably due to the lack of variability in the
deposition gradient across the region.
8-60
-------
8.4.2.2.3 Comparison of regions -
Hydrogen ion deposition appeared to be strongly related to base cation concentration in the NE,
but not in the SBRP. This is probably due to the much smaller deposition gradient in the SBRP. In the
NE, we observe an increase in base cation export from watersheds with increasing hydrogen ion
deposition, but in the SBRP, the change in deposition is smaller, so that the change in base cations is
not significant
In both the NE and SBRP, positive relationships were observed between bedrock sensitivity codes,
and ANC and base cation concentrations. In both regions, the regressions explained between 5 and 9
percent of the variability in the surface water variables, but due to a larger sample size, the regressors
for the NE were highly significant
The smaller sample size in the SBRP may also explain why no significant correlations between
bedrock lithology and pH were observed there, while such relationships are observed in the NE.
8.4.3 Summary
Results of the studies of the relationships between the relative reactivities of the different bedrock
types found within the DDRP watershed population and the associated surface water properties indicate
several pertinent factors. About two-thirds of the variability associated with the assignment of the
sensitivity numbers is attributable to measurement error. This means that our data are "noisy," and so
relationships may be obscured or minimized. Nonetheless, there are significant relationships between the
relative reactivities assigned to watersheds and associated surface water characteristics, in particular, base
cation concentrations and surface water ANC values. These relationships do not appear to be as strong
as we expected. In addition to measurement error the absence of strong relationships might be related
to the population of systems being studied. In essence, most of the watersheds Included in the study
are underlain by nonreactive bedrock types, so many of the differences observed in the surface water
chemistry might be more strongly controlled at this level by factors such as depth to bedrock or selected
soil properties. The multiple regression studies will address these issues (see Section 8.8).
8-61
-------
8.S MAPPED LAND USE/VEGETATION
8.5.1 Introduction
The effects of vegetation and land use on surface water chemistry are both general and site
specific. For example, species differences in root density, depth, and morphology affect how nutrients
cycle from the soil to forest vegetation as well as the physical and biological processes that influence soil
water infiltration and percolation. Both evergreen and deciduous vegetation strip or scavenge acidic
deposition materials from the atmosphere before they reach ground waters and surface waters. Long-
term effects of acidic deposition can be either beneficial or adverse to the nutrient status of forest soils
and to forest health; the deciding factors are local site nutrient status, ongoing silvicultural practices,
present forest species mix, and both the amount and type of atmospheric inputs received at specific
localities (Johnson et al., 1982a, and Johnson et a!., 1988).
Usually, surface waters within forested watersheds have lower turbidity and temperature and have
lower nutrient loadings than water from agricultural or urban watersheds (Simmons, 1976; Chang et al.,
1983; Comeau and Bellamy, 1986; Morgan and Good, 1988). One exception is forest land subjected to
clear-cut harvesting and/or extensive site preparation (Pritchett and Fisher, 1987). The magnitude of
water chemistry changes within or outside harvested watersheds is dependent upon clear-cut intensity
(Tiedmann et al., 1988) and ionic species (Lawrence and Driscoll, 1988).
Within these broad generalizations substantial site-to-site variation occurs because of inherent
natural spatial and temporal variability across the landscape. For example, in upland headwater forested
watersheds receiving acidic deposition, surface water sulfate can predominate in areas having minimal
vegetation and soil development; however, surface water concentrations of bases generally increase
downstream where interactions of forest species composition, soil depth, and geochemical weathering
are greater (Jeffries et al., 1988; Driscoll et al., 1987). In other situations, riparian zone vegetation reduces
chemical concentrations in soil water (SchnabeJ, 1985) and lowers suspended sediment loads leaving
agricultural watersheds (Cooper et al., 1986). Spatially, wetland position is also important: wetland
fringes bordering water bodies seem to be more effective in modifying water quality than are upland
8-62
-------
wetlands remote from major downstream lakes (Johnston et al., 1988). Finally, significant temporal
alterations in stream water chemistry have been attributed to both beaver activity (Driscoll et al., 1987b;
Naiman et al., 1986) and changing historical or recent land use patterns (Buso et al., 1985; Hunsaker et
al., 1986b).
Although northeastern lake and SBRP stream watersheds were primarily undisturbed and forested,
significant amounts of other land uses were present. It was also known that many northeastern lakes
had varying amounts of beaver activity and wetlands. The main DDRP objective in mapping land use
and forest vegetation cover types was to determine whether any land uses were consistently associated
with specific surface water chemistry variables. Section 8.5 examines those relationships that were found.
8.5.2 Data Sources
Land use and land cover data for northeastern lakes were obtained by interpretation of recent
1:12,000 color infrared (CIR) photography specifically acquired for DORP (Section 5.4.1.6). For SBRP
watersheds, SCS personnel determined land use from older (late 1970s) alternate black and white and
CIR, quad-sized National High Altitude Photography (NHAP) photos (see Section 5.4.2.7). Forest cover
types were determined during soil mapping activities (see Sections 5.4.1.3, 5.4.2.3). All land use, forest
cover, and wetland data were entered into GIS (see Sections 5.4.1.7, 5.4.2.8) so that information could
be analyzed by percent watershed area or actual hectare area in desired land use classes. Select data
on acidic deposition, precipitation, and runoff were also included.
8.5.3 Statistical Methods
Relationships between water chemistry variables and many environmental variables have been
examined via normal regression techniques for small (Osborne and Wiley, 1988) and very large (Hunsaker
et al., 1986a) data sets. Some of the problems with regression approaches are: selection of an
appropriate and parsimonious subset of regressors for the model; multicollinearity of the regressor
variables; peculiar distributions of some of the regressors, particularly when some variables have many
zero entries; and practical interpretability of results when many regressors appear in any one model.
8-63
-------
For these reasons, we used principal component analysis or PCA (Johnson and Wichern, 1982; SAS
Institute Inc., 1985, 1987) to analyze mapped land use and vegetation data. For land use data in the NE,
the correlation matrix of the 42 regressor variables in Table 8-25 was used to generate the principal
components. Thirteen principal components had eigenvalues greater than one. These factors (Table
8-26) were retained for further analysis. Together these principal components explained 81 percent of
the variability in the correlation matrix. We used a varimax rotation of the original factors to Improve the
interpretability of the factors (Table 8-27). Then we performed stepwise regressions of the surface water
chemistry variables on the rotated factors and examined the residuals for leverage points and outliers
after verifying the standard assumptions of regression analyses (see Section 8.1.2).
For SBRP watersheds, initial analysis showed three watersheds with ANC >_ 1000 ^eq L'1, due to
local carbonate bedrock rich in calcium and magnesium. We excluded these three watersheds from all
subsequent analyses. A correlation matrix of 39 regressor variables (Table 8-28) was used to generate
the principal components. Eleven principal components had eigenvalues greater than one. These factors
(Table 8-29) were retained for further analyses, because the correlation matrix was used to generate the
components. Together, these principal components explained 93 percent of the variability in the
correlation matrix. We used a varimax rotation of the original factors to improve factor interpretability
(Table 8-30). Finally, we performed regressions of the surface water chemistry variables on the rotated
factors for the SBRP watersheds, after examining residuals for leverage points and outliers (see Section
8.1.2).
8.5.4 Sulfate and Percent Sulfur Retention
8.5.4.1 Northeast
Lake sulfate was positively correlated with deposition (Section 8.2) and watershed development
but negatively correlated with beaver activity, wetland percent, and precipitation and runoff factors (Table
8-31). The adjusted R2 of 0.50 was the highest for all five water chemistry variables investigated.
Because NE watersheds have low sulfate adsorption capacity and are assumed to be at sulfur steady
state (Section 8.2), a positive correlation between surface water sulfate and sulfate deposition is expected.
8-64
-------
Table 8-25. Land Use and Other Environmental Variables Related to Surface Water Chemistry
of Northeastern Lakes
Variable
Kind
Variable
Name
Explanation of Variable Name
Photointerpretation B DAM
BT.ODGE
c~
C H
CABIN
E
E H
-------
Table 8-26. Factor Loadings for First 13 Principal Components after Varimax Rotation of the
Correlation Matrix of Land Use and other Environmental Variables for Northeastern Lakes
Environmental
Variables 1
L-H
L
P H
UV H
UC H
CABIN
H H
P~
H D
H~W
SO~4 W
S04~D
ELEV
CON
U DAM
O~DAM
IM~H20
B LODGE
B~DAM
SCS OPN
G ~
C H
C~
G H
LV~WET
W~
N
UC
N H
OW H
E H"
VfH
M~H
M
UV
E
Ul H
ur
ELS WET
H ~
SCS WET
PREClP
RUNOFF
MIX
HWD
98*
98*
94*
92*
81*
75*
41*
38*
-5
6
5
5
-3
-7
-4
-1
-1
-2
-1
10
-4
6
-1
3
2
2
0
12
-3
4
14
20
0
-1
17
-17
0
0
-2
23
-2
5
-4
2
0
2
-3
-3
0
9
6
10
-31*
4
88*
88*
86*
81*
61*
-67*
-1
-5
-8
-8
6
5
13
-6
5
2
4
3
7
9
1
-14
-10
-14
4
4
8
-10
-4
11
-4
-29*
10
27*
-14
-22
61*
Principal Components8
34 5
-1
-1
0
-2
4
-2
24
1
-1
-3
-8
-9
-7
-5
89*
88*
87*
78*
75*
-7
-9
41*
4
5
2
2
0
-6
37*
10
63*
46*
-1
-1
-8
9
-2
-3
-2
2
5
5
-1
17
-7
0
0
-1
4
14
4
20
-3
17
1
-2
23
-8
0
-4
2
9
3
-2
89*
86*
73*
73*
71*
0
1
0
9
-9
0
2
8
0
0
0
-58*
-1
0
-6
15
4
-18
-15
-8
-27
3
3
-4
3
-2
6
-1
-13
4
6
10
11
-17
15
2
-3
7
3
-5
4
5
-8
-6
1
96*
96*
-5
-6
5
-7
-1
34*
-2
-4
-9
-28*
6
-4
37*
-9
37*
11
-12
-9
-10
6
1
1
-4
-1
46*
-3
0
-8
4
6
8
6
-5
-2
8
9
-1
4
-8
9
1
-3
-1
-4
-4
-4
97*
93*
61*
1
-1
1
-1
-1
-3
-10
3
0
-1
0
-5
9
-5
-1
0
7
-1
-1
-9
13
7
37*
25*
-18
0
-10
-20
-2
-16
-4
16
-8
12
37*
-9
3
4
-3
-18
39*
-1
-1
-2
-4
10
80*
68*
61*
0
-1
•4
3
19
-6
9
3
6
-1
13
5
-2
continued
8-66
-------
Table 8-26. (Continued)
Environmental
Variables
L-H
L
P H
UV H
UC~H
CAgIN
H H
P"
H D
H~"W
$04 W
S04T5
ELEV
CON
U DAM
O~DAM
lM"H2
B LODGE
B~DAM
SCS OPN
G ~
C H
C~
G H
IJ7WET
W
N
UC
N H
OW H
E H"
WH
M~H
M~
UV
E
Ul H
Ul
ELS WET
H
SCS WET
PREClP
RUNOFF
MIX
HWD
8
0
0
-1
-1
1
1
3
-2
4
7
1
3
-11.
-4
0
-1
1
-2
-2
2
3
-4
-6
8
-4
-4
-1
-1
0
-5
7
1
99*
99*
-4
2
-1
-2
13
3
-7
0
-6
4
-1
9
-5
-5
12
23
-10
27*
4
35*
-2
-1
16
9
-39*
-13
-1
3
0
-1
-11
16
11
-8
-9
-1
-1
-1
-2
-5
14
-3
-10
-2
-2
-1
87*
-67*
2
-3
-10
27*
9
19
-27*
-4
5
Principal
10
-3
-3
11
-2
-4
-5
14
27*
3
4
11
9
-15
5
-7
-1
-2
0
0
12
-6
-1
-1
-8
1
1
1
0
4
4
7
15
-2
-1
-6
-2
86*
86*
-6
5
-2
12
-13
2
-9
Components*
11 12
-2
-2
-2
7
3
18
38*
0
-4
-9
-6
-5
-13
-25*
-12
6
-1
-4
14
6
-18
20
11
-8
14
14
0
6
-19
8
2
14
3
3
5
-15
0
-2
63*
62*
59*
12
12
10
0
1
1
8
-7
0
-14
21
16
-13
35*
27*
-13
23
34*
5
-7
6
-1
-2
-6
-10
3
-6
-10
-1
-1
2
2
-3
5
3
8
-2
-2
-5
5
8
-8
-1
19
10
74*
72*
-8
-17
13
1
1
1
2
-5
1
-20
1
-16
-5
-7
-11
-3
-14
15
0
-3
15
-2
5
10
-21
-6
6
-3
-3
-3
-8
20
2
2
4
2
3
-6
-1
3
1
-5
10
13
-2
. 0
89*
-62*
1 Printed values are multiplied by 100 and rounded to the nearest integer. Values
greater than 0.29 have been flagged by an asterisk.
8-67
-------
Table 8-27. Interpretation of the First 13 Principal Components After Varimax Rotation of the
Correlation Matrix of Land Use and Other Environmental Variables for Northeastern Lakes
Principal
Component
Rank General Interpretation of Principal Component
PC1 developed land: waste disposal, pits and quarries,
cabins, urban residential, and urban commercial
PC2 overall wet and dry deposition
PCS beaver activity, wetlands, and cropland
PC4 pasture land and cropland; less forest
PCS wetlands
PCS , barren and urban commercial land
PC7 open water, forest, and wetlands
PCS cemeteries
PC9 cabins, urban residential, and pits and quarries; less forest
PC10 urban industrial land and pits and quarries
PC11 wetlands and horticulture
PC12 precipitation and runoff
PC13 more mixed and less hardwood forest
8-68
-------
Table 8-28. Land Use and Other Environmental Variables Related to Surface Water Chemistry
of Southern Blue Ridge Province Streams
Variable
Kind
Variable
Name
Explanation of Variable Name
interpretations
Forest cover type
Other data
C
C H
E~
E H
F
F-H
Q
G H
H"
H H
L~
L H
k-
K H
M"
M H
N~
N H
0"
0 H
FT
R H
U"
U H
W~H
Z~
2_H
CON
HWD
MIX
OPEN
CAMG D
CAMG_W
H D
H~W
PRECIP
RUNOFF
SO4 W
percent area in cropland
area (ha) In cropland
percent area in grazed forest land
area (ha) in grazed forest land
percent ungrazed forest land
area (ha) in ungrazed forest land
percent area in managed or native pasture
area (ha) in managed or native pasture
percent area in horticulture
area (ha) in horticulture
percent area in waste disposal
area (ha) In waste disposal
percent area in rock outcrop
area (ha) In rock outcrop
percent area in cemeteries
area (ha) in cemeteries
percent area in pits and quarries
area (ha) in pits and quarries
percent area in miscellaneous land use
area (ha) in miscellaneous use
percent area in wetlands
area (ha) in wetlands
percent area in urban land
area (ha) in urban land
area (ha) in open water
percent area in ridge top barren land
area (ha) in ridge top barren land
percent area in conifers
percent area In hardwood forest
percent area in mixed forest
percent (dry) areas without forest or wetlands
dry CA + Mg deposition, g mz
wet Ca + Mg deposition, g m"
dry H deposition, g m
wet H deposition, g m"
precipitation in cm from National Climatic
Data Center, Asheville, NC
mean annual runoff, in inches
from Krug et al. (1985) (See Section 5.7.1)
wet sulfate deposition, g m'f
dry sulfate deposition, g m
8-69
-------
Table 8-29. Composition of First 11 Principal Component Analysis (PCA) Factors After
Varimax Rotation of the Correlation Matrix of Land Use and Other Environmental Variables
Related to Surface Water Chemistry of Southern Blue Ridge Province Streams
Environmental
Variables
1
HWET
S04WET
K
K H
CAMGWET
W H
PPT
L
M H
LTH
IvT
H
H H
OPEN
C
C H
F
U H
FT
R H
IT
MIX
CAMGDRY
HWD
CON
S04DRY
HDRY
0
O H
G H
G~
RUNOFF
2 H
Z~
F H
N~
N H
E~
E H
89*
88*
88*
87*
85*
72*
69*
9
9
9
9
-2
-1
-3
7
7
4
1
-8
-8
-6
-9
31*
6
13
-8
-5
-12
-10
-21
-12
41*
-3
-3
26
2
2
-7
-13
2
22
21
-14
-14
18
31*
10
99*
99*
99*
99*
-7
-8
18
64*
66*
-17
-4
-1
-1
-4
-9
9
-1
2
2
0
-1
0
8
17
9
-1
-1
-11
-2
-2
•4
-4
3
6
5
-4
•A
1
-2
-5
4
4
4
4
98*
98*
80*
74*
70*
-81*
9
-1
-1
16
-5
-7
-31*
-8
-1
6
-4
-3
10
14
-18
-4
-4
-17
-3
-3
-1
0
4
-10
-10
0
0
-2
5
-17
-2
-2
-2
-2
-3
•A
33*
-2
-2
-33*
98*
98*
98*
95*
-18
-28
2
-4
-17
14
-2
-2
19
15
-23
-2
-2
•4
-2
-2
-6
-8
5
13
18
-28
-27
17
-14
4
0
0
0
0
5
3
8
2
-1
-7
-9
-6
-6
-11
89*
60*
-88*
9
58*
-17
15
14
-23
14
-11
4
4
-44*
6
6
0
-14
PCA Factors
678
9
9
-15
-13
17
39*
44*
0
0
0
0
-7
-8
-14
12
19
10
-1
-7
-7
-1
3
51*
-19
84*
-59*
-83*
3
2
-5
-14
25
-3
-3
16
2
2
-11
-15
-12
-17
14
15
-19
4
-28
-1
-1
-1
-1
0
0
-8
-4
-5
7
0
-1
-1
-2
20
3
-14
-3
-2
-15
95*
94*
-7
-23
-23
-4
-4
49*
-3
-3
-5
-1
-21
-22
5
7
-26
18
-31*
5
5
5
5
-*
0
38*
-4
-4
-39*
11
12
12
5
-8
-14
-8
-8
5
13
-9
-8
87*
81*
-50*
-7
-7
7
•4
•4
2
6
9
6
3
-1
-1
-4
-5
-4
-1
-1
-1
-1
0
0
-9
-3
-3
8
-1
0
0
-3
4
12
4
-13
7
-14
-4
-4
-4
-13
4
98*
98*
49*
-2
-2
-4
-4
10
5
8
-5
-5
11
-9
2
-1
-1
-1
-1
-1
-1
-4
-3
-4
3
-2
0
0
-3
3
13
-2
3
9
-3
-3
-2
-3
-9
-7
-2
-2
-8
99*
99*
-4
-2
11
-4
-4
-6
-5
-3
-4
-10
-2
-2
-2
-2
-2
•4
1
1
2
-8
-4
-3
-3
-6
12
-9
11
1
20
27
-4
-2
-2
9
-8
-3
-3
-9
-3
-3
96*
94*
* Printed values are multiplied by 100 and rounded to the nearest integer. Values
greater than 0.29 have been flagged by an asterisk.
8-70
-------
Table 8-30. Interpretation of the First 11 Principal Components after Varimax Rotation of the
Correlation Matrix of Land Use and Other Environmental Variables for Southern Blue Ridge
Province Streams
Principal
Component
Rank
General Interpretation of Principal Component
PC1
PC2
PC3
PC4
PCS
PC6
PC7
PCS
PC9
PC10
PC11
overall wet and dry deposition and precipitation
cemeteries, cropland, and waste disposal land
cropland, horticulture, and open land
open land, urban lands, and wetlands
mixed forest and dry Ca plus Mg and S04 deposition
open water and dry Ca plus Mg deposition
miscellaneous and ungrazed forest land
open land and pasture; less precipitation
ridge top barren land and ungrazed forest
pits and quarries
grazed forest land
8-71
-------
Table 8-31. Results of Regressions Relating Surface Water Chemistry of Northeastern Lakes to
Land Use and Other Environmental Data3
Water
Chemistry
Variable R2
ANC 0.37
Ca + Mg 0.42
pH 0.32
SCS_wet
Sulfate 0.50
Percent 0.19
Sulfur
Retention
*n - 143
D S = Significant at 0.15 level
* s Rlnniffaant at OCK lounl
Factor
No.
4
12
4
12
2
4
7
11
12
2
3
4
5
7
9
12
5
Regr. Signif.b
Sign Level
+ ***
***
+ ***
***
**
.j. ***
+ *
+ *
***
+ ***
*
+ ***
**
*
.(. ***
S
^. ***
Factor Explanation
agriculture: SCS open dry, G, C
precipitation and runoff
agriculture: see above
precipitation/runoff
deposition
agriculture: see above
open water and wetlands
wetlands and horticulture: H, ELS_wet,
precipitation and runoff
deposition
beaver activity, water, wetlands
agriculture: see above
wetlands: SCS, LV, ELS
open water and wetlands
development: cabins, P, Uv, H; less forest
precipitation/runoff
wetlands (see above)
Significant at 0.01 level
= Significant at 0.001 level
8-72
-------
Increased sewage and animal or chemical waste loadings to streams from agricultural and residential
development also lead to greater overall surface water sulfate levels. Although sulfate deposition was
associated with surface water sulfate levels, amounts in surface waters were less in small watersheds
when beaver activity and wetland percentage were high. Low downstream sulfate concentrations, caused
by increased anaerobic conditions and sulfate retention behind beaver impoundments has been
documented by others (Oriscoll et a!., 1987; Goldstein et a!., 1987), especially during low-flow summer
months.
Percent sulfur retention was positively related to wetland percent (Table 8-31). Anaerobic wetland
conditions favor sulfate reduction processes that in turn foster increased sulfur retentbn.
8.5.4.2 Southern Blue Ridge Province
Singular land uses acted as either leverage points or outliers and influenced regressions relating
land use to both sulfate and percent sulfur retention. For sulfate, important land uses were pits and
quarries, open land, and pasture. Eliminating watersheds with these land uses left no significant land
use factors in regression models (Table 8-32). Evidently, the watershed with pits and quarries land use
had an internal source of sulfur. Agricultural practices on open land and pasture, including soil
amendments and animal husbandry, may result in increased sulfate loadings.
8.5.4.3 Regional Comparisons
In the NE, sulfate is strongly and positively correlated with deposition and agricultural and urban
development. Since soils in the region have little remaining sulfate adsorption capacity (Section 7,
Rochelle et al., 1989), incoming sulfur deposition or within-watershed generated sulfur quickly circulates
to surface waters after storm events. Exceptions are small watersheds with beaver activity or wetlands.
In these watersheds, sulfate reduction processes are the probable cause of decreased surface water
sulfate concentrations and increased percent sulfur retention.
8-73
-------
Table 8-32. Results of Regressions Relating Sulfate and Percent Sulfur Retention of Southern
Blue Ridge Province Streams to Land Use Data
Dependent
Variable
Sulfate
Percent
Sulfur
Retention
R2 n
0.78 32
0.80 31
30
0.76 32
30
Regress. Significant Factor/
Sign Land Uses Included
+ 10/pits & quarries
+ 8/open land & pasture
+ 10/pits & quarries
no significant factors
10/pits & quarries
no significant factors
Significance
Level*
***
S
***
__
***
Watersheds
Removed
2A07813(L)
2A08808(L)
—
2A08808(L)
2A07823(O)
* S = Significant at a 15 level
*** = Significant at 0.001 level
" (L) = Leverage point removed from regression
(O) = Outlier point removed from regression
8-74
-------
In the SBRP region, surface water sulfate and percent sulfur retention are both influenced by local
internal sulfur sources from pits and quarries and pasture land. However, when such watersheds are
eliminated from regression models, the homogeneous nature of the region stands out: upland forested
watersheds with little agricultural or urban development. Under such conditions, land use is unrelated
to either surface water sulfate or percent sulfur retention. Instead, both variables are more controlled by
high sulfate adsorption capacity of soils in the region (Section 7).
8.5.5 ANC. Ca plus Ma. and pH
8.5.5.1 Northeast
Lake ANC was positively correlated with agricultural land use and negatively correlated with
precipitation/runoff (Table 8-31); both factors in the regression were highly significant (p _< 0.001). In
watersheds having a greater percentage of agricultural, urban, or other disturbed land (Buso et al., 1985),
ANC values of surface waters are generally higher than those found in mostly-forested, small-headwater
watershed systems (Hunsaker et at., I986a; Jeffries et al., 1988). Where precipitation and runoff are high,
ANC in surface waters is reduced because of dilution effects.
Lake Ca plus Mg was positively correlated with agricultural land use but negatively related to
precipitation and runoff (Table 8-31); both factors in the regression were highly significant (p <. 0.001).
Successful farming and related activities are generally located on deeper and higher base status soils
unless low soil pH and poor fertility are offset by applying lime and fertilizers (Tisdale and Nelson, 1975).
Where acidic deposition is high, soil bases can be leached from the soli and replaced by hydrogen and
aluminum ions (Section 3); bases leached from the soil are flushed rather quickly from lakes associated
with high runoff. The positive correlation of Ca plus Mg with agriculture (Comeau and Bellamy, 1986)
but negative correlation with greater precipitation and runoff is indicative of these relationships.
Surface water pH was positively correlated with agriculture, wetlands, and horticulture but negatively
correlated with precipitation and runoff (Table 8-31); all factors were significant (p <. 0.05) in the
regression. Agricultural and lowland (cranberry) horticultural land uses could be associated with higher
8-75
-------
pH in surface waters via fertilizer inputs. Wetlands and water impounding via beaver activity also
contribute to sulfate reduction (Driscoll et al., 1987a) and an increase in pH (Section 7.2). Where
precipitation and runoff are high, lake pH will be reduced because of dilution effects.
8.5.5.2 Southern Blue Ridge Province
Ail regression models relating ANC, Ca plus Mg, and pH of SBRP streams to land use factors
were strongly influenced by leverage points (Table 8-33). Since Ca plus Mg and ANC are chemically
related surface water variables, those land uses that had potential and significant impact on one also
influenced the other variable. In all instances, the significant land uses were those which allowed within-
watershed inputs of base elements to SBRP streams. Deleting all the leverage points removed all
significant land uses from the ANC and Ca plus Mg models. For pH, removing only one leverage point
with open land and pasture left no significant land use in the regression.
As stated in Section 8.5.4.2, the SBRP region is very homogeneous in terms of forest and land
cover; overall, there is little agricultural or urban development. Where anthropogenic development or
disturbance is present, it has very marked and significant impacts on ANC, Ca plus Mg, and pH of local
streams.
8.5.5.3 Regional Comparisons
Agricultural land uses, particularly cultivated land and pasture were positively correlated with ANC,
Ca plus Mg, and pH in both the NE and SBRP. In the small SBRP region, single land uses were usually
leverage points or outliers in the overall analysis. Removing SBRP watersheds with leverage points and
outliers from the analysis produced a more homogeneous data set comprised mostly of forested
watersheds with little urban or other development. Under these conditions, land use was not readily
correlated with ANC, Ca plus Mg, or pH.
For northeastern lakes, sulfur deposition was negatively correlated with pH. Via sulfate reduction
processes under anaerobic conditions, northeastern wetlands mitigate the effects of high sulfur deposition
8-76
-------
Table 8-33. Results of Regressions Relating ANC, Ca plus Mg, and pH of Southern Blue Ridge Province
Streams to Land Use Data
Dependent
Variable
Regr.
n R Sign
Significant Factors/
Land Uses Included
Significance Watersheds
Level3 Removed"
ANC
30 0.11 +
27 0.50 +
0.06
26 0.21 +
25
8/ open land and pasture
2/cemeteries, wasteland *
3/cropland, horticulture, open land ***
4/open land, urban areas, wetland
4/open land, urban areas, wetland S
10/pits and quarries *
no significant factors ~-
2A07827(L)
2A07813(L)
2A07802(L)&
2A07826(L)&
2A07830(L)
2A08808(L)
Ca+Mg
pH
31 0.28 +
30 0.21 +
28 0.18 +
26 0.35 +
25
31 0.12 +
30
10/pits and quarries *
10/pits and quarries **
3/cropland, horticulture, open land S
10/pits and quarries
2/cemeteries, waste land **
4/open land, urban areas, wetlands *
no significant factors —
8/open land and pasture S
no significant factors —
2A08808(L)
2A07813(L)
2A07826(L)
2A07827(L)
2A07802(L)
2A07830(L)
—
2A07813(L)
—
* S = Significant at 0.15 level
* = Significant at 0.05 level
** = Significant at 0.01 level
*** = Significant at 0.001 level
b (L) = Leverage point removed from regression
(O) = Outlier point removed from regression
8-77
-------
and are associated with higher lake pH. In SBRP upland forested watersheds, there are few wetland or
riparian zones to mitigate deposition effects on stream pH. Presently, however, high sulfate adsorption
capacity of SBRP soils does help minimize deposition effects on stream water chemistry.
8.5.6 Summary and Conclusions
The major findings of this section are:
• In the NE, surface water sulfate is positively correlated with deposition and
extent of agricultural and urban development.
• In smail northeastern watersheds with beaver activity and wetlands, sulfate reduction
processes decrease surface water sulfate concentrations and increase percent sulfur
retention and pH.
• In the SBRP region, surface water sulfate and sulfur retention are influenced by local
Internal sulfur sources from pits and quarries.
• Agricultural land uses, particularly cultivated land and pasture were correlated with
ANC, Ca plus Mg, and pH in both the NE and SBRP. However, in the SBRP region,
removing outlying or influential sites produced a homogeneous dataset in which land
use was not readily correlated with ANC, Ca plus Mg, and pH.
• In both the NE and SBRP, forest cover is not directly related to surface water
chemistry; in the NE, greater developed land (and less forest) is correlated with higher
surface water ANC, Ca plus Mg, pH, and sulfate.
8.6 MAPPED SOILS
8.6.1 Introduction
Soils are an important component of terrestrial ecosystems. They are the principal source of plant
nutrients and provide a rooting medium for aboveground vegetation; they are the major site of within
watershed decomposition reactions. Soils host a plenitude of chemical reactions, including adsorption,
desorption, ion exchange, weathering, and precipitation reactions. These chemical reactions can affect
the composition and quality of soil water and consequently subtending surface and ground waters. Soil
physical properties, such as structure or architecture, the flowpath of soil water, the soil particle-size
distribution, the depth to impermeable layers, and soil bulk density, are also important. In natural
settings, the chemical and physical attributes of soils are inseparable. The objective of this analysis is
8-78
-------
to identify the relationships that exist between mapped soils and surface water chemistry on a regional
basis.
Some soils are known to attenuate some of the effects of chronic sulfur deposition principally
through sulfate adsorption and base cation supply reactions (e.g., cation exchange and mineral
weathering). These reactions are important in Isolation at the atomic level, however, as the scales
become coarser (i.e., atomic to micro, micro to meso, meso to watershed, watershed to regional) the
number of simultaneous, overlapping processes increases. At the regional scale the relationship between
soil properties and soil water chemistry involves thousands of hectares of soils and the composition of
a large number of lakes or many kilometers of stream reaches. The relatively simple set of relationships
(at the atomic level) becomes a complex set of diffuse relationships as the scale expands to the region.
Recognition of the fact that soils per se integrate a large number of physical and chemical processes is
the basis of the DDRP mapped soils analysis. In this analysis we use the proportion of different kinds
of soils in watersheds, at a well-defined but regional scale, to identify relationships that exist between
soils and the chemical composition of subtending surface waters.
8.6.2 Approach
As discussed in the Sections 4.1 and 4.2 and described in Lee et at. (1989a), a stratified random
sample of watersheds was selected and mapped in the DDRP. Mapping followed strict protocols, and
soil map units were regionally defined and correlated across the respective regions. The details of
watershed selection, map unit correlation, and mapping can be found in Sections 5.2 and 5.4.
In the NE 592 kinds of soils were identified. These soils are the components of the 338 map
units used to map the soils In the NE. In the SBRP 286 components and 176 map units were identified.
Because it was not tractable to characterize this large number of soils it became apparent that a smaller
set of soil units were needed to make regional soil characterization and sampling feasible. The result
was the development of the soil "sampling classes." Soil components considered to have similar
chemical and physical characteristics were grouped into unique classes that we termed a soil sampling
8-79
-------
class. In the NE, 38 different sampling classes were identified, and in the SBRP there were 12. Soil
sampling classes were the basis for soil sampling and analytical characterization and served as our main
link between the analytical data and the soils of the regions. They also serve as the basic units for
relating mapped soils to surface waters in this analysis.
All watersheds are not completely covered by soils. Other non-soil cover is present and can,
sometimes, extend over large areas. To completely assess the relationships between soils and surface
water chemistry, such areas that occurred on our sample of watersheds were also identified during the
mapping and were termed "miscellaneous land areas". Because these areas may influence the quantity
and quality of surface waters they are included in this analysis. In the NE these include: rock outcrop
(M01); pits, gravel (M02); rubble land (M03); and pits, quarry (M04). In the SBRP there were only two
miscellaneous land areas: rock outcrop (MRO) and quarry pits (MPQ).
An overview of how this analysis was conducted is presented in Figure 8-8. After the soil maps
were digitized, a summary of the soil map units and their extent on each watershed was obtained from
the GIS for each region. The relative proportion of each map unit component had been estimated for
the regions and entered into a mapping data file: Each map unit component had been assigned to a
sampling class and, therefore, the proportion of each sampling class in the respective watersheds could
be calculated. For example, 112 ha of map unit 134A was mapped on a particular watershed and map
unit 134A was defined by three components (a, b, and c) with the following percentages: 80 percent
component a, 15 percent component b, and 5 percent component c. Component a therefore accounts
for 89.6 ha (112 ha x 0.80) of the 112 ha of map unit 134A while b and c account for 16.8 ha and 5.6
ha, respectively. This calculation was repeated for each map unit on a watershed basis and the results
were pooled by sampling class. The proportions of the watersheds in the various soil sampling classes
and miscellaneous land areas were calculated by dividing the extent of the soil sampling class or
miscellaneous land area by the total area of the watershed.
8-80
-------
Data flow
Map watersheds
T
Digitize maps
Use CIS to calculate
soil map unit areas
Parse map unit areas into
sampling class areas
Calculate watershed
sample dass proportions
Sampling class and miscellaneous
land areas dataset
r
Deposition dataset
r
Dependent variables datasets
[ Start j
Modelling dataset
Run stepwise selection procedure
Use Mallow's Cp to select
unbiased model variables
T
Run standard regression
including residual analysis
Outliers/
leverage points?
Remove outliers and leverage points
Figure 8-8. Data and regression model development flow diagrams.
8-81
-------
With the GIS we can dissect or subdivide the watersheds to test various hypotheses. We are
particularly interested in evaluating the effect of the riparian zone (near-lake or near-stream) on surface
water chemistry. In the NE, two watershed buffer zones were considered in addition to the whole
watershed area. One is limited to the area within the first 40-ft contour interval (called the 40-ft contour
buffer zone) above the sampled lake. This buffer zone is used to delineate the near-lake soils and
wetlands. The other includes the same 40-ft contour buffer zone plus a 30-m linear buffer on either side
of any perennial (blue-line) stream and around contiguous wetlands. It also includes a 40-ft contour
buffer zone around any other lakes or ponds that are on the watershed in addition to the sampled lake.
Due to contour map distortions or errors there are only 144 watersheds in the 40-ft buffer zone dataset
and 143 in the combined buffer dataset. Because the resource of interest in the SBRP is streams,
elevational or contour buffer zones are not suitable, so linear buffer zones were used. These include the
area within 100 m of the blue-line streams on the DORP sample of watersheds in the SBRP. Because
2 of the 35 SBRP streams are not perennial, the SBRP buffer zone dataset has a sample size of 33.
Tables 8-34 through 8-38 summarize the distributions of the soil sampling classes and miscellaneous land
areas on the DDRP sample of watersheds. Table 8-34 is for whole watersheds in the NE, Table 8-35
is for the land within the 40-ft GIS contour buffer zone, and Table 8-36 is for the combined GIS buffer
zones. Table 8-37 is for the whole watersheds in the SBRP, and Table 8-38 is for the GIS 100-m linear
buffer zones.
Soils alone cannot explain all of the van'ation in surface water chemistry. Other factors such as
deposition and in-lake or in-stream processes also influence surface water chemistry. Regional data on
in-lake and in-stream processes do not exist, but deposition data do. For this analysis we include six
variables from the long-term annual average data sets. The details of how these data were compiled are
described in Section 5.6.3. The specific variables used in this analysis are discussed in Section 8.1.1.
A total of 48 independent variables are used the NE (38 soil sampling classes plus 4 miscellaneous iand
areas plus 6 deposition variables) and 20 in the SBRP (12 soil sampling classes plus 2 miscellaneous
8-82
-------
Table 8-34. Summary Statistics for Percent Area Distribution of the 38 Soil Sampling Classes
and the 4 Miscellaneous Land Areas on the DORP Sample of 145 NE Lake Watersheds
SMPLCLAS
£02
EOS
EOS
E06
H01
H02
H03
101
I02
I05
I06
I09
110
111
121
125
I29
I30
I33
I37
I38
1 40
141
I42
146
S01
802
805
809
810
811
812
813
814
815
S16
817
S18
M01
M02
M03
M04
MEAN
0.4
5.4
2.1
0.2
2.3
1.0
3.6
1.5
3.2
1.5
1.8
1.4
3.1
1.3
0.3
4.3
3.2
0.5
6.0
0.3
1.2
1.1
0.2
0.2
1.6
0.3
2.8
0.7
8.7
1.9
4.8
6.8
7.0
9.0
1.3
2.5
1.5
1.6
3.3
0.1
0.0
0.0
STDJ3EV
1.0
19.6
2.5
1.0
4.2
3.2
5.2
3.3
5.2
4.6
5.5
6.0
10.2
5.0
3.3
11.1
9.0
1.9
15.7
0.8
3.2
7.0
1.0
1.2
5.7
1.0
11.5
2.0
12.9
5.9
9.9
8.8
9.0
12.7
5.6
7.6
7.1
5.5
5.1
0.7
0.0
0.1
MIN
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Q!a
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.1
0.0
0.0
0.0
MEDIAN
0.0
0.0
1.4
0.0
1.0
0.0
1.4
0.3
0.6
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.1
0.0
0.7
3.3
3.9
1.2
0.0
0.0
0.0
0.0
1.6
0.0
0.0
0.0
Q3a
0.1
0.0
2.8
0.0
3.0
0.1
5.7
1.8
3.7
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.9
0.0
0.0
0.0
0.0
0.1
0.0
0.1
13.2
0.1
3.0
12.4
12.7
15.5
0.0
0.0
0.0
0.0
4.1
0.0
0.0
0.0
MAX
6.6
99.9
14.4
7.3
33.9
19.2
42.3
26.7
22.9
27.0
26.6
48.8
78.0
39.8
38.1
47.0
48.2
13.2
65.1
4.8
27.9
79.9
11.3
13.4
35.6
7.8
99.8
17.6
55.8
46.0
56.3
36.4
60.5
52.4
49.5
34.6
56.7
29.5
28.4
7.0
0.3
0.9
* Q1 is the 25th percentile, and Q3 is the 75th percentile.
8-83
-------
Table 8-35. Summary Statistics for the Percent Area Distribution of the 38 Soil
Sampling Classes and the 4 Miscellaneous Land Areas in the GIS 40-ft Contour on
the DORP Sample of 145 NE Lake Watersheds
SMPLCLAS
E02
E03
EOS
E06
H01
H02
H03
101
I02
I05
I06
I09
110
111
121
I25
I29
I30
I33
I37
I38
1 40
141
I42
I46
801
S02
SOS
S09
810
S11
812
813
814
815
S16
817
818
M01
M02
M03
M04
MEAN
1.6
6.3
1.0
0.2
1.2
2.0
10.1
2.6
3.7
1.0
1.1
1.3
2.4
1.1
0.3
6.2
2.0
0.3
4.8
1.0
2.3
1.2
0.3
0.7
2.6
0.9
6.6
1.1
6.0
2.7
4.4
3.4
3.3
8.0
0.6
1.6
1.1
1.0
1.6
0.1
0.0
0.0
STD_DEV
4.1
19.9
1.6
1.2
3.0
6.8
12.5
6.4
7.3
3.6
4.0
5.4
8.0
4.4
3.0
16.8
6.2
1.3
13.8
2.9
5.6
7.4
1.4
3.3
10.1
2.9
19.1
3.9
13.0
8.9
10.3
6.5
6.7
12.3
3.4
5.6
6.8
3.7
3.4
1.1
0.1
0.0
MIN
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Q1a
0.0
0.0
0.0
0.0
0.0
0.0
0.0
o.o
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
o.o
0.0
0.0
o.o
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
MEDIAN
0.0
0.0
0.2
0.0
0.0
0.0
3.0
0.5
0.6
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.2
0.0
1.1
0.0
0.0
0.0
0.0
0.2
0.0
0.0
0.0
Q3a
0.1
0.0
1.6
0.0
0.8
0.0
18.9
1.9
3.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.8
1.4
0.0
0.0
0.0
0.0
0.0
0.0
0.0
4.6
0.1
2.1
3.1
3.8
12.4
0.0
0.0
0.0
0.0
1.6
0.0
0.0
0.0
MAX
20.2
100.1
11.1
8.2
19.0
48.8
60.0
40.9
39.9
22.8
22.2
34.9
73.9
37.6
34.9
77.5
33.0
10.9
75.8
22.5
43.9
80.1
11.5
31.3
66.1
18.0
99.9
30.2
69.8
53.9
49.2
35.4
59.4
51.8
38.0
39.3
60.9
23.6
24.8
10.2
1.2
0.0
1Q1 is the 25th percentile, and Q3 is the 75th percentile.
8-84
-------
Table 8-36. Summary Statistics for the Percent Area Distribution of the 38 Soil
Sampling Classes and the 4 Miscellaneous Land Areas in the Combined CIS Buffers
on the DDRP Sample of 145 NE Lake Watersheds
SMPLCLAS
E02
E03
EOS
E06
H01
H02
H03
101
I02
I05
I06
I09
110
111
121
I25
I29
I30
I33
I37
1 38
1 40
141
I42
I46
S01
S02
SOS
S09
S10
S11
S12
S13
S14
S1S
S16
S17
S18
M01
M02
M03
M04
MEAN
1.5
6.0
1.1
0.2
1.4
2.4
9.6
2.9
4.4
1.0
1.1
1.1
2.5
1.0
0.2
6.9
1.7
0.2
4.6
0.9
2.5
1.3
0.2
0.4
2.7
0.9
5.6
1.1
6.1
2.5
4.1
3.7
3.5
8.3
0.7
1.7
1.1
1.0
1.7
0.1
0.0
0.0
STD_DEV
3.8
19.7
1.7
1.4
3.2
7.0
11.5
7.0
7.7
3.5
3.8
4.9
8.6
3.5
2.4
17.7
5.3
1.0
12.6
2.2
5.6
7.4
1.2
1.7
10.3
2.7
17.4
3.8
12.8
7.7
9.6
6.6
6.7
12.3
2.8
5.6
7.0
3.7
3.5
0.7
0.1
0.0
MIN
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Qla
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
MEDIAN
0.0
0.0
0.3
0.0
0.0
0.0
3.3
0.4
0.8
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
0.1
0.2
0.0
1.0
0.0
0.0
0.0
0.0
0.3
0.0
0.0
0.0
Q3a
0.1
0.0
1.6
0.0
1.1
0.1
17.3
1.9
4.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.7
1.7
0.0
0.0
0.0
0.0
0.2
0.0
0.0
6.1
0.0
2.2
5.0
4.7
12.4
0.0
0.0
0.0
0.0
1.6
0.0
0.0
0.0
MAX
19.4
100.1
11.5
11.2
20.4
49.0
60.0
41.2
39.9
23.0
22.0
34.9
73.9
26.4
27.9
74.4
32.7
8.1
62.5
14.2
43.5
80.1
11.5
17.1
65.6
16.4
100.2
30.2
69.8
53.9
51.9
35.4
59.4
53.1
24.4
40.6
64.9
23.6
25.1
6.7
0.8
0.6
" Q1 is the 25th percentile, and 03 is the 75th percentile.
8-85
-------
Table 8-37. Summary Statistics for the Percent Area Distribution of the 12 Soil
Sampling Classes and the 2 Miscellaneous Land Areas on the ODRP Sample
of 35 SBRP Stream Watersheds
SMPLCLAS
ACC
ACH
ACL
FL
FR
MSH
MSL
OTC
OTL
SHL
SKV
SKX
MPQ
MRO
MEAN
17.7
5.9
32.9
2.5
4.2
2.8
11.2
2.5
4.9
7.4
5.1
1.6
0.1
1.2
STD_DEV
28.3
9.7
26.3
3.1
10.2
7.3
16.4
10.5
9.2
8.7
8.2
3.3
0.5
1.9
MIN
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
6.0
0.0
0.0
0.0
Q18
0.0
0.0
1.8
0.0
0.0
0.0
0.0
0.0
0.2
0.0
Q.O
0.0
0.0
0.0
MEDIAN
0.7
0.5
30.7
0.9
0.0
0.0
0.0
0.0
1.1
5.1
0.6
0.0
0.0
0.7
Q3a
28.8
8.5
57.1
4.0
0.0
0.0
21.9
0.0
3.4
15.0
7.0
1.3
0.0
1.5
MAX
80.1
37.8
78.2
10.7
43.2
48.1
61.7
52.3
37.6
30.0
36.0
18.1
2.8
15.0
* 01 is the 25th percentile, and 03 Is the 75th percentile.
8-86
-------
Table 8-38. Summary Statistics for the Percent Area Distribution of the 12 Soil
Sampling Classes and the 2 Miscellaneous Land Areas in the 100-Meter Linear
GIS Buffer on the DDRP Sample of 35 SBRP Stream Watersheds
SMPLCLAS
ACC
ACH
ACL
FL
FR
MSH
v MSL
OTC
OTL
SHL
SKV
SKX
MPQ
MRO
MEAN
5.1
10.6
24.6
11.1
2.9
4.4
10.1
6.8
5.2
6.4
10.7
1.0
0.2
0.9
STD_DEV
10.9
14.8
20.7
12.9
7.2
10.1
14.2
19.9
6.7
8.3
14.4
2.3
1.0
1.5
MIN
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Qla
0.0
0.0
0.1
1.0
0.0
0.0
0.0
0.0
0.6
0.2
0.0
0.0
0.0
0.0
MEDIAN
0.7
0.9
25.0
8.1
0.0
0.0
0.0
0.0
2.0
2.8
3.0
0.0
0.0
0.1
Q3a
1.8
24.9
36.6
15.8
0.0
0.0
17.6
0.0
6.1
12.8
16.4
0.0
0.0
2.3
MAX
40.7
43.7
71.0
54.8
27.6
59.0
42.0
79.1
24.4
28.8
50.8
11.7
5.8
10.2
* Q1 is the 25th percentile, and 03 is the 75th pereentile.
8-87
-------
land areas plus 6 deposition variables). The dependent variables include four surface water chemical
measurements, sulfate fceq L"1), ANC (in peq L"1), Ca plus Mg (CAMQ in ^eq L"1), and pH. A fifth
variable, % S retention, is a calculated variable derived from deposition and surface water chemistry
values (See Section 7 for details on how the percent sulfur retention values were calculated).
8.6.3 Sulfate and Sulfur Retention
The retention of sulfur by terrestrial ecosystems is an important mechanism that can delay the
acidification of subtending surface waters. In the biogeochemical sulfur cycle there are two principal
soli or sediment mediated sulfur retention mechanisms, sulfate adsorption and sulfate reduction. These
mechanisms have been characterized and discussed in Sections 3.3, 7, and 9.2 in detail. Soils low in
organic matter content having significant amounts of hydrous oxides of iron and aluminum will tend to
retain sulfate via adsorption. Soils or sediments that are sufficiently wet to have anaerobic conditions
retain sulfate via sulfate reduction.
Sulfate concentrations in the DDRP SBRP sample of streams are, in general, much lower than in
the sample of lakes in the NE. However, the region as a whole has somewhat higher sulfur deposition
than the NE. The deposition in the SBRP is more uniform than in the NE. Because the SBRP is a
relatively small region with relatively uniform deposition, there is not a significant sulfate deposition
gradient as in the NE. Unlike the NE the watersheds in the SBRP are not at sulfur steady state. A
principal difference between the two regions is the soils. In the NE the soils are relatively young, having
less profile and secondary mineral development. In the SBRP most soils are relatively older and more
deeply weathered with abundant accumulations of secondary mineral phases (hydrous oxides of iron and
aluminum).
8.6.3.1 Northeast
Following the procedures described above and in Section 8.1, several regression models were
developed for the relationship between the independent variables (mapped soils and deposition) and
lake sulfate. The results are presented in Table 8-39. The first whole watershed model for lake sulfate
8-88
-------
Table 8-39. Lake Sulfate and Percent S Retention Regression Models Developed
for NE Lakes Using Deposition, Mapped Soils (as a Percentage of Watershed Area
in Soil Sampling Classes) and Miscellaneous Land Areas (as a Percentage of
Watershed Area) as Candidate Independent Variables8
Lake Sulfate Percent S Retention
Whole 40-ft Comb. Whole 40-ft Comb.
SO4-Wet + +
SO4-Dry „ + + +
Ca+Mg-Dry
E03 + H- +
H01 ...
H03 ---+++
106 +
111 +
I25 + +
I30 +
I33 ... + +
S02
S12 + +
S17 + +
S18 + +
R2 0.64 0.61 0.59 0.31 0.39 0.44
adjusted R2 0.61 0.59 0.57 0.29 0.36 0.41
n-lakesb 142 141 141 141 • 142 140
p-modelc 10 7 7 4 6 6
O/Ld 332423
* + and • refer to positive and negative parameter estimates, respectively
n-Jakes •= number of observations used to develop the regression model
p-model = number of regressor variables in model
0/L s number of outlier or leverage points omitted
8-89
-------
had an R2 of 0.73. Two observations, however, had unusually high lake sulfate concentrations and were
identified as being outlier and/or leverage points. These watersheds also had the only two occurrences
of the miscellaneous land area, M04 - quarry pits. This finding is consistent with the observation that
major watershed disturbances, such as mining, may outweigh any surface water chemistry effects due
to acidic deposition. Removing the watersheds with quarry pits and one other leverage point from the
whole watershed analysis resulted in the model R2 statistic dropping from 0.73 to 0.64. The variables in
the regression model, however, remained the same, except for M04 which is now excluded. Table 8-
39 shows the variables that were included in the lake sulfate regression models, indicated by the sign
of their respective parameter estimates. These signs indicate either positive or negative correlation to the
dependent variable. The inclusion of both wet and dry sulfate deposition in the regression models for
the NE is not surprising. In the 142 observation, whole watershed model these two variables account
for about 45 percent of the variability in lake sulfate. As discussed in Section 7, the NE is almost at
sulfur steady state {i.e., input ~ output) which explains why these two variables make such a large
contribution to the explanatory power of the model.
In addition to sulfur deposition variables, eight soil sampling classes are included in the best whole
watershed regression model for lake sulfate. Sampling classes E03, H03, and I33 were consistently
included in the lake sulfate regression models. E03 is positively correlated with lake sulfate. This
sampling class is characterized by coarse texture, poor development, and low sulfate adsorption capacity.
These soils are excessively drained and seem to be a non-interacting conduit for drainage waters.
Likewise, I06 and 111 are well drained and positively related with lake sulfate. These three soil sampling
classes, however, generally occur only in Subregion 1D. Therefore, it is also plausible that these three
sampling classes (E03, I06, and I) are surrogates for sea-salt sulfate contributions that may be
underestimated in the LTA deposition dataset. Their inclusion in the regression model, may have little
to do with their actual chemical and physical properties.
H03 was consistently negatively related to lake sulfate. The H03 soils are deep, wet, organic soils
characterized by low pH (dysic). Because of the negative correlation, these soils are thought to be an
8-90
-------
active site of sulfur retention via biological processes (e.g., sulfate reduction), and would be expected to
have a positive relationship with percent sulfur retention. The analysis indicates that sampling class 125
is positively correlated with lake sulfate concentrations and that 133 is negatively related. Fragipans occur
in the soils of both of these classes within 100 cm of the soil surface. The principal difference between
these sampling classes is that I25 soils are very poorly to somewhat poorly drained while 133 soils are
somewhat poorly drained to well drained and deep. Anaerobic conditions occur throughout the upper
100 cm of soil in the I25 class during some part of the year. Intuitively, this would suggest that this class
of soils should have a negative relationship with take sulfate concentrations rather than a positive one,
because of the potential for sulfate reduction during the anaerobic periods. Other factors such as the
landscape position of these soils and the timing and nature of the anaerobic periods may be responsible
for the observed relationship.
The S12 class of soils is well drained, moderately deep, coarse-loamy Spodosols with relatively
low base saturation and pH. Water moves rapidly through this class of soils and moves downhill at the
bedrock contact with little opportunity for sulfur retention. There is a positive relationship between S12
soils and lake sulfate. The reason for the negative relationship between the S02 class and lake sulfate
is unclear from this analysis; S02 soils may be a surrogate for another attribute.
In the NE the percent % S retention values range from -22 to +60, with a median value of -4.
In contrast, In the SBRP the range is from -60 to +89, with a median value of 75 percent. In the NE
the systems are almost at steady state with respect to sulfur. In contrast, the SBRP is effectively retaining
most of the sulfur inputs. We would expect markedly different results in the regression analysis.
In the NE (see Table 8-39) the best whole watershed model explains only 31 percent of the
variation in percent sulfur retention with a four-variable model. The best model is a six-variable model
using the combined buffer data with an R2 of 0.44. The sum of dry Ca and Mg deposition, H01, H03,
and S18 was included in all three models; H03 soils have a positive parameter estimate. The H01 soils
are thin (< 30 cm), organic soils overlaying bedrock or fragmental material that is freely drained. They
8-91
-------
are not wetland soils. They are probably active sites of organic matter decomposition and contribute
sulfur from organic matter to the surface waters. The class H03, as described earlier, includes wetland
soils, that presumably retain sulfur via a sulfate reduction mechanism. Sampling class S18 soils are
shallow and somewhat excessively drained. They are likely to be non-interacting conduits for drainage
waters. The I33 and S17 classes are positively related to percent sulfur retention in the buffer zone
models. The relationship with I33 was not as expected as was discussed with the lake sulfate results
above. The reason for inclusion of dry Ca plus Mg deposition in the models is not known, it may be
an artifact of the deposition data compilation or it may be functioning as a surrogate for another
deposition variable.
8.6.3.2 Southern Blue Ridge Province
The regression models developed for the stream sulfate concentrations are summarized in Table
8-40. The results for seven models are included: four using the soils and miscellaneous land area
distributions on the whole watersheds and three for the distributions In the 100 meter buffers.
The whole watershed model with 35 observations includes five variables that are all positively
correlated with stream sulfate. One is the miscellaneous land area, MPQ. As noted in the NE, the
occurrence of quarries on watersheds can have a significant effect on the subtending surface water
chemistry. The effect on the surface water is dependent upon the type of geological strata being mined.
In the SBRP sample population of watersheds, there was one occurrence of MPQ. The surface water
in this watershed had the highest observed value of stream sulfate concentration and the lowest % S
retention. It Is likely that the mining on this watershed has exposed sulfur-bearing materials which have
subsequently oxidized and impacted the surface water. The Anakeesta Formation (King et al., 1968),
which contains sulfur-bearing minerals is common in parts of the SBRP. Watershed disturbances, such
as road construction and landslides, may expose these materials as well.
The SHL and OTL soils, as defined in Section 5, are both well drained and have relatively low
organic matter content. SHL soils are shallow and OTL soils are deep. In general, the soils in these
8-92
-------
Table 8-40. Regression Models of Sulfate in SBRP Streams, Developed Using
Deposition, Mapped Soils (as a Percentage of Watershed Area in Soil Sampling
Classes) and Miscellaneous Land Areas (as a Percentage of Watershed Area) as
Candidate Independent Variables8
Whole Watersheds
Ca+Mg-Wet
OTC +
ACH
MSH
SHL + + +
OIL + + +
MSL + + +
MPQ +
R2 0.84 0.63 0.45 0.30
adjusted R2 0.82 0.58 0.42 0.28
n-streamsb 35 33 32 31
p-modelc 5421
O/ld none 234
100-m buffer zones
-
+
-
-
+ +
+ + +
+ + +
0.82 0.66 0.57
0.79 0.59 0.52
33 31 30
553
none 2 3
* + and - refer to positive and negative parameter estimates, respectively
n-lakes = number of observations used to develop the regression model
* p-model =• number of regressor variables in model
0/L = number of outlier or leverage points omitted
8-93
-------
sampling classes have low sulfate adsorption capacities. The OTC soils, however, only occur on three
watersheds in the SBRP ODRP sample, and in one instance they cover more than 50 percent of the
watershed. These soils, while high in secondary clay minerals, have pH values that are unfavorable for
sulfate adsorption and therefore do not retain sulfate to any significant degree. The watershed with the
high OTC coverage also had the second highest stream sulfate concentration. In addition to low sulfate
adsorption potential, it is likely that the calcareous parent material of the OTC is interbedded with sulfur-
bearing materials. In the 33 observation, whole watershed model, the watersheds with more than 50
percent OTC coverage and the MPQ site are not included. The resulting model is the same as the 35
observation model, except that it no longer has MPQ and the sign on the OTC variable is negative. The
sign reversal is probably caused by the low abundance of OTC on the two remaining OTCs. The 32
observation mode) has one OTC site remaining with 0.1 percent OTC coverage. In this mode, OTC was
not a significant explanatory variable in this model. The 31 observation model only had one significant
explanatory variable, the SHL sampling class. The soils in the SHL sampling class account for 30 percent
of the variability in stream sulfate concentrations alone. These soils are well drained, low In clay, have
moderate to rapid permeability, and are less than 50 cm deep. These four properties are characteristic
of soils with short hydrologic contact times that have little or no effect on the chemistry of drainage
waters passing through them. Surface waters in watersheds with an abundance of shallow soils will be
more susceptible to acidification than watersheds with deep, moderately well-drained soils.
The soils in the MSL sampling class were consistently selected with positive parameter estimates
In the stream sulfate regression models. This suggests that as the proportion of MSL soils on a
watershed increases, the concentration of sulfate in the subjacent stream also increases. The soils in the
MSL sampling class are, by definition, derived from metasediments and have low organic matter content.
Chemically, they have only intermediate sulfate adsorption potential. The positive parameter estimates
indicate, however, that these soils may be associated with sulfur-bearing materials.
The 33 observation, 100-m buffer model is similar to the 35 observation, whole watershed model,
but has lower R2 and adjusted R2 values. In the 31 observation model, the ACH and MSH sampling
8-94
-------
classes both have negative parameter estimates. These soils as a group have relatively higher organic
matter content in the surface layer than their ACC and MSL counterparts. This implies that, when these
sampling classes occur in the near-stream zone, sulfate is retained. The 30 observation model has only
three variables, SHL, OIL, and MSL, and explains approximately 60 percent of the variation in stream
sulfate concentrations.
The regression model results for sulfur retention in the SBRP are presented in Table 8-41. The
whole watershed model with 35 observations has an R2 of 0.86. This R2 is highly inflated by the
presence of one watershed. This watershed was the only observation with a negative net sulfur retention
(-66 percent). It is also the watershed with the quarry (MPQ), which serves as a source of sulfur.
Omitting this watershed from the analysis produces a model with only one variable and an R2 of 0.34.
The remaining variable is the MSL sampling class, which Is negatively correlated to percent sulfur
retention. As proposed In the stream sulfate discussion above, the MSL sampling class soils are
associated with sulfur-bearing parent materials that function as a source of sulfur. Omitting five more
possible outlier/influence points only decreases the model R2 from 0.34 to 0.33. The resulting model
contains only the MSL sampling class. On a regional basis, these soils explain one third of the variation
in sulfur retention and appear to be an important source of sulfur.
Three 100-m buffer, sulfur retention regression models are also presented in Table 8-41. The
model with 33 observations and R2 of 0.84 is biased by the watershed with the quarry. Omitting that
watershed results In a three-variable model that accounts for 52 percent of the variation in sulfur retention.
The variables in this model are all negatively correlated with sulfur retention. They include the OTC, SKX,
and MSL sampling classes. The OTC and MSL have low and Intermediate sulfate adsorption potentials,
respectively, and may be sources of sulfur. The SKX soils are coarse textured, excessively drained
Inceptisols formed in metasedimentary residuum. They may be a minor sulfur source, but more likely
they are non-Interacting soils with short hydrdogic contact times. Omitting a second watershed, this one
with 79 percent of the 100-m buffer zone area in the OTC sampling class, produces a two-variable model
that explains 45 percent of the variation in sulfur retention. In addition to the MSL sampling class
8-95
-------
Table 8-41. Regression Models of Percent Sulfur Retention In SBRP Stream Watersheds
Developed Using Deposition, Mapped Soils (as a Percentage of Watershed Area in Soil
Sampling Classes), and Miscellaneous Land Areas (as a Percentage of Watershed
Area) as Candidate Independent Variables*
Whole Watersheds
100-m buffer zones
OTC
SKX
MSH
MSL
MPQ
R2 0.86 0.34 0.33
adjusted R2 0.84 0.32 0.30
n-streamsb 35 34 29
p-modelc 4 1 1
0/La
none 1
6
0.84 0.52 0.45
0.81 0.47 0.41
33 32 31
432
none 1 2
* + and - refer to positive and negative parameter estimates, respectively
n-lakes = number of observations used to develop tfie regression model
c p-model = number of regressor variables in model
O/L = number of outlier or leverage points omitted
8-96
-------
(negative parameter estimate), the MSH sampling class is Included in this model. Based on the soil
sampling class definitions, the only difference between these two sampling classes is their organic matter
content and thickness of the surface layers; the MSH is high in organic matter and the MSL is low. In
general, the MSH soils are well drained; however, where they occur in the near-stream areas (within 100
m of the stream) they may be saturated with water at depths 100 cm or more below the soil surface for
a sufficient period to create anaerobic conditions that can potentially retain sulfur via sulfate reduction.
Because of the distribution of the MSH sampling class soils, they may also be a surrogate for watersheds
with high sulfate adsorption capacity soils.
8.6.3.3 Regional Comparisons
In the NE sulfur deposition explains the majority of the variability in lake sulfate concentrations.
The central tendency for watersheds in the NE is to be at sulfur steady state where sulfur input = sulfur
outputs. The capacity of these systems to retain sulfur effectively is inherently low (Section 9.2), and has
been exhausted (I.e., iow or negative sulfur retention). In the SBRP where sulfur retention is high and
the watersheds are retaining most of the sulfur inputs, sulfate deposition is not yet significantly related
to stream sulfate concentrations.
In the NE there is evidence to suggest that localized sources of sulfur deposition, not accounted
for in the LTA deposition dataset, may be contributing to higher sulfate concentrations in the near-coastal
watersheds in Subregion 1D. This additional sulfur deposition is probably derived from wind-blown sea-
salt aerosols. In both regions there were also indications that at least one of the soil sampling classes
is functioning as a sulfur source or is a surrogate for a source of sulfur.
Very poorly drained soils and acidic Histosols were positively related to sulfur retention; sulfate
reduction Is the likely mechanism. A stronger relationship was found when these soils occurred in the
near-lake or near-stream areas in the NE. These soils may be responsible for most of the sulfur retention
in the NE (See Sections 8.5 and 9.2).
8-97
-------
Shallow soils with short hydrologic contact times serve as non-interacting drainage water conduits
in both regions. It was also noted in both regions that the surface water in watersheds with a major
watershed disturbance, such as a quarry, have higher concentrations of sulfate. The extent of this effect
depends on the nature of the geologic strata being disturbed and the magnitude and location of the
disturbance,
8.6.4 ANC. Ca plus Ma. and oH
Acid neutralizing capacity (ANC) is an important measure of the potential of surface waters to
buffer the input of acidic deposition. Systems with zero or negative ANC are already acidic and with
low ANC are likely to be vulnerable to acid inputs. Systems with high ANC are strongly buffered
(capacity protected) against acid inputs and are therefore not likely to become acidic, even at current
levels of deposition, for some time, possibly centuries.
ANC is the principal indicator of surface water buffering. Related to ANC are the sum of base
cations (Ca, Mg, K, Na) and the surface water pH. In this analysis, the sum of the principal base cations,
Ca and Mg, is considered.
8.6.4.1 Northeast
In the NE region, regression models were developed that explain approximately one-half of the
variability in ANC (Table 8-42). The best whole watershed model has six variables, including wet sulfate
deposition. Sulfur deposition is negatively con-elated with ANC. The remaining variables are sampling
class variables and are all positively related to lake ANC. Soils in sampling classes 101,111,125, and I46
(very poorly drained and poorly drained Inceptisols) are among the classes with the highest base
saturation and pH values. It follows that these soils are sources of base cations and supply subtending
surface waters with base cations and buffer lake ANC. The soils in sampling class I06 are shallow, and
in general, low in pH and base saturation. Their contribution to ANC is questionable; they may be
functioning as a surrogate for another variable.
8-98
-------
Table 8-42. Lake ANC and the Sum of Lake Calcium and Magnesium Regression
Models Developed for NE Lakes Using Deposition, Mapped Soils (as a Percentage
of Watershed Area in Soil Sampling Classes) and Miscellaneous Land Areas (as a
Percentage of Watershed Area) as Candidate Independent Variables8
Lake ANC Lake Ca plus Mg
Whole 40-ft Comb. Whole 40-ft Comb.
SO4-Wet
H-Wet
E06
H01
H02
101
I05
I06
I09
111
I25
I46
S01
818
M04
R*
adjusted R2
n-lakesb
p-modelc
0/Ld
0.48
0.46
145
6
none
0.47
0.43
144
10
none
0.54
0.51
142
9
1
0.56
0.54
145
8
none
0.49
0.46
144
8
none
0.55
0.52
143
10
none
* + and - refer to positive and negative parameter estimates, respectively
n-Jakes a number of observations used to develop the regression model
° p-model » number of regressor variables in model
0/L = number of outlier or leverage points omitted
8-99
-------
The best ANC model is developed with combined buffer data, explaining 54 percent of the variation
in lake ANC with nine variables. Sulfate deposition and the 105 sampling class are the only variables in
the model negatively correlated with ANC. 105 occurs mainly in Subregion 1D and may serve as a
surrogate for sea-salt contributions of sulfur. Sampling classes 101,111,125,146, and I06 are in this model
as well as the whole watershed ANC model. This model also includes sampling classes S01 and S18
as variables, both having positive parameter estimates. The soils in both of these sampling classes have
intermediate base saturation (-20 percent) and pM (-4.6). S01 soils are deep and are widely distributed
across the region in small amounts. S18 soils are very shallow with a lithic or paralithic contact within
50 cm. The S18 sampling class occurs only in Subregion 1E.
Because lake Ca plus Mg is strongly related to ANC (r = 0.94), the regression models developed
for Ca plus Mg are similar to those for ANC and have comparable R2 values. Sampling class I05 and
wet hydrogen deposition (H-Wet) have negative parameter estimates. I05 was discussed above. H-WET
is strongly correlated with wet sulfate deposition (r = 0.92) and was substituted for the wet sulfate
deposition variable included in the ANC models. In the whole watershed model of Ca plus Mg, the
miscellaneous land area M04 (quarries) was included in the model. These watershed disturbances also
increase the amount of Ca and Mg in the subtending surface waters in addition to increasing the levels
of sulfate. In fact, the levels of Ca plus Mg and sulfate are quite simitar in these watersheds, and both
have fairly high ANCs. Ca and Mg appear to be the cations accompanying the mobile anion sulfate.
As long as the soils are not being depleted of base cations, this situation is little cause for concern.
The lake pH regression models were simpler than the ANC and Ca plus Mg models because they
have fewer variables and similar R2 values (Table 8-43). In all three models wet sulfate deposition had
negative parameter estimates while the sampling classes I09 and I25 had positive parameter estimates.
These results seem to be reasonable. Soils in the I25 sampling class have a relatively high base status,
which accounts for the positive correlation with lake pH. The I09 soils are also positively related to pH,
but are lower base status soils than the I25 soils. H03 is also included in the whole watershed model
with a positive parameter estimate. Soils in the H03 sampling class are deep, wet, organic soils
8-100
-------
Table 8-43. Lake pH Regression Models Developed for NE Lakes Using Deposition,
Mapped Soils (as a Percentage of Watershed Area in Soil Sampling Classes) and
Miscellaneous Land Areas (as a Percentage of Watershed Area) as Candidate
Independent Variables8
LakepH
Whole 40-ft Comb.
SO4-Wet
EOS
E06
H02
H03
105
109
125
R2 0.45 0.45 0.49
adjusted R2 0.44 0.42 0.46
n-lakesb 144 144 142
p-modelc 467
O/Ld 1 none 1
* + and - refer to positive and negative parameter estimates, respectively
n-iakes = number of observations used to develop the regression model
0 p-model = number of regressor variables in model
O/L = number of outlier or leverage points omitted
8-101
-------
principally located in wetlands. These soils are dysic, meaning that the pH of undried samples is less
than 4.5 (in 0.01 M CaCI2 ). Because pH is an intensity variable (i.e., concentration) the pH of the last
soil that drainage water passes through before it reaches the lake may initially control the pH of the lake
water. If there are extensive wetlands surrounding a lake including H03 soils, the H03 soils may be the
last soil that the drainage waters pass through; the pH of the lake will therefore be similar to the pH of
the H03.
The combined buffer model for lake pH had the highest R2 and adjusted R2 values, 0.49 and 0.46,
respectively, of the three regression models. As in the whole watershed model, the combined buffer
model includes wet sulfate deposition and the sampling classes I09 and I25. Additionally, it has EOS and
I05 with negative parameter estimates and E06 and H02 with positive parameter estimates. H03 was not
included as an important variable in either of the buffer models. The soils in the EOS sampling class are
poorly developed, very shallow (< 25 cm), underlain by hard bedrock, and also have one of the lowest
aggregate pH values. Because of their chemical and physical characteristics, an increasing abundance
of EOS soils will lead to lower lake pH values. The characteristics of the soils in the E06 sampling class
are the direct result of human activities. They are deep soils that lack pedogenic development due to
significant anthropogenic disturbance such as road construction. They are classed as Udorthents. In
general, they have moderate to high base saturation and moderate pH. Because of recent disturbance,
they may have abundant fresh weatherable mineral faces that supply base cations at a higher rate than
other soils in the region.
8.6.4.2 Southern Blue Ridge Province
ANC in the SBRP is generally higher than in the NE. The median ANC value for the region is
120 fjteq L"1. In the NE the median is 56 peq L'1. Because of the chemical characteristics of the soils
in the SBRP (i.e., higher sulfate adsorption capacities), these systems are not close to steady state with
respect to sulfur inputs and outputs. The soils in the region are retaining a significant proportion of the
sulfur in deposition, and at present, appear to be delaying the acidification of the surface waters in the
region. The ANC of these surface waters and the factors that control it are, therefore, very important.
8-102
-------
Regression models of the SBRp stream ANCs using the mapped soils, mapped miscellaneous
land areas, and wet and dry deposition are presented in Table 8-44. The whole watershed regression
models with 35, 34, and 33 observations all have very high R2 values, 0.92, 0.86, and 0.943, respectively.
These large R2 values are due to the presence of observations with very high ANC (>1000 jueq L"1 )
values. Two of the three sites are associated with the OTC sampling class and its calcareous parent
materials. These three systems are probably capacity protected against acidification. Omitting them from
the analysis and leaving 32 observations results in a three-variable model that explains 47 percent of the
variation in stream ANC. One of the 32 watersheds has 0.1 percent OTC. Because of the strong positive
relationship between calcareous materials and ANC, OTC has been included by the stepwise procedure
as a variable. The FL sampling class is also included with a positive parameter estimate. The soils in
this sampling class occur on flood plains. Compared to the other sampling classes in the SBRP, the soils
in the FL sampling class have the third highest base saturation and the second highest pH. Omitting a
fourth influence point results in a one-variable regression model for stream ANC that has an R2 of 0.40.
The single variable is the FL sampling class.
Unlike the NE, there is little or no indication that the DDRP sample of streams in the SBRP are
contaminated with Na from road salt or sea salt additions. Therefore, in addition to considering the
sum of stream Ca and Mg concentrations as a dependent variable, we have included an analysis of the
sum of the four principal base cations (Ca + Mg + Na + K), the sum of base cations (SOBC). In this
section and the two sections that follow, additional analyses of the relationship between these watershed
attributes and SOBC are included.
The initial 100-m buffer models of stream ANC are similar to the whole watershed models.
Dropping outlier/influence points results in a two-variable model that accounts for 92 percent of the
variation in stream ANC. The two variables are OTC and FL The OTC accounts for most of the
explanatory power. Omitting all of the high ANC sites from this analysis does not produce an unbiased
model.
8-103
-------
Table 8-44. Regression Models of ANC in SBRP Stream Watersheds, Developed Using
Deposition, Mapped Soils (as a Percentage of Watershed Area in Soil Sampling
Classes) and Miscellaneous Land Areas (as a Percentage of Watershed Area) as
Candidate Independent Variables8
Ca+Mg-Dry
OTC
SKV
SKX
FL
MSH
SHL
OTL
MSL
Whole Watersheds
100-m buffer zones
+ + +
FT 0.92 0.86 0.943 0.47 0.40
adjusted R2 0.90 0.84 0.938 0.41 0.38
n-streamsb 35 34 33 32 31
p-modelc 44331
0/Ld
none
1
0.91 0.933 0.924
0.88 0.926 0.919
33 32 31
732
none 1 2
* + and • refer to positive and negative parameter estimates, respectively
n-lakes = number of observations used to develop the regression model
° p-model = number of regressor variables in model
O/L - number of outlier or leverage points omitted
8-104
-------
The results of the Ca plus Mg analysis are presented in Table 8-45. Inclusion of all 35 observations
results in a four-variable regression model with an R2 of 0.90. However, as in the ANC analysis this
model is strongly influenced by three observations with exceptionally high values of Ca plus Mg. These
are the same three with high ANC. Omitting them from the analysis results in a two-variable model with
an R2 of 0.42. The two variables are Ft and MPQ. Analysis of the residuals and influence diagnostics
Indicate that the MPQ is an influence point. Omitting it and developing a model based upon 31
observations results in a one-variable model. The variable is the PL sampling class. This Is the identical
model developed for ANC with the same 31 observations.
The results for the 100-m Ca plus Mg model follow the pattern set by the whole watershed.
Inclusion of all 33 observations results in a model with a high R2 but with two strong Influence points.
Omitting these two observations and rerunning the analysis leads to a higher R2 model that has four
variables. At the same time another influence point is identified. Omitting this observation results in a
two-variable model with yet another influence point. This time, however, eliminating it and proceeding
with the analysis does not produce an unbiased model.
The results of the whole watershed and 100-m buffer zone regression analyses for SOBC are
presented in Table 8-46. The whole watershed model, including all 35 SBRP systems, produced a four-
parameter model with an R2 of 0.91. Included in this model is the calcareous sampling class OTC
(positively related to SOBC). It follows that the presence of significant amounts of calcareous material
can Increase the levels of base cations that may be transported to the surface water. Systems with high
SOBCs and ANCs are likely to be capacity protected against acidification. Analysis of the residuals found
that three systems, all with ANC and SOBC greater than 1000 jueq L*1 were strong influence points.
Dropping these and rerunning the analysis produces a three-parameter model with an R2 of 0.50. This
model also had significant influence points remaining. Omitting these resulted in a five-parameter model
built on the data from 28 systems. This unbiased model explains 75 percent of the observed variation
in SOBC. Five independent variables were included in this model. Ca and Mg in dry deposition and the
8-105
-------
Table 8-45. Regression Models of Calcium Plus Magnesium in
SBRP Streams, Developed Using Deposition, Mapped Soils (as
a Percentage of Watershed Area in Soil Sampling Classes) and
Miscellaneous Land Areas (as a Percentage of Watershed Area)
as a Candidate Independent Variables
Whole Watersheds
OTC +
SKV
SKX
FL + +
MSH
OTL +
MSL +
MPQ +
R2 0.90 0.42 0.25
adjusted R2 0.89 0.38 0.22
n-streamsb 35 32 31
p-modelc 4 2 1
0/Ld none 3 4
100-m buffer zones
+ + +
+
+ +
-
•f +
+
+
0.88 0.96 0.922
0.86 0.95 0.916
33 32 30
542
none 1 3
* + and - refer to positive and negative parameter estimates, respectively
n-lakes = number of observations used to develop the regression model
6 p-model = number of regrassor variables in model
O/L - number of outlier or leverage points omitted
8-106
-------
Table 8-46. Regression Models of SOBC* in SBRP Streams, Developed Using
Deposition, Mapped Soils (as a Percentage of Watershed Area in Soil Sampling
Classes) and Miscellaneous Land Areas (as a Percentage of Watershed Area) as
Candidate Independent Variables
Whole Watersheds
100-m buffer zones
Ca+Mg-Dry
FR
OTC
SKV
SKX
FL
MSL
OTL
ACL
ACC
MPQ
+
+
+ +
R2
adjusted R2
n-streamsc
p-modeld
O/Le
0.91 0.50 0.75
0.90 0.44 0.70
35 32 28
435
none 3 4
0.41 0.74 0.79
0.36 0.65 0.74
30 28 25
275
358
* SOBC = sum of base cations (Ca + Mg + Ma + K)
+ and - refer to positive and negative parameter estimates, respectively
° n-streams - number of observations used to develop the regression model
p-model = number of regressor variables in model
* O/L •= number of outlier or leverage points omitted
8-107
-------
SKX and ACL soils were included with negative parameter estimates. The FL and ACC soils were also
included but with positive parameter estimates.
The soils in both the SKX and ACL sampling classes tend to be low base status. The SKX soils
are formed in residuum and the ACL soils are formed in either residuum or alluvium. The presence of
these soils on a watershed Is indicative of highly weathered, low base status soils, and lower base status
surface waters. In contrast, the soils in the FL class are relatively high base status and are associated
with the higher base status surface waters. The ACC soils are very similar to the ACL soils and are
differentiated by their particle-size families: the ACC soils are clayey and the ACL soils are either fine-
loamy or coarse-loamy. The presence of the ACC soils may be indicative of readily weatherable primary
minerals, while the FL soils may represent hydrologic convergence zones where base cation enriched
drainage waters and sediments accumulate.
Table 8-46 includes three regression models developed for SOBC using deposition and the soils
and miscellaneous land areas within 100 m of the sampled stream. The FL sampling class is included,
with a positive parameter estimate, in the three models. The unbiased model (i.e., without outliers and/or
leverage points) had four parameters, all with positive parameter estimates and an R2 of 0.79. Included
were the FR, FL, MSL, and OTL sampling class soils. All parameter estimates were positive, indicating
that the soils in these classes are all associated with higher base status surface waters. The FR and MSL
are typically low base status soils, while FL and OTL are some of the highest base status soils in the
region. Because of their low base status and positive correlation to SOBC, the FR and MSL classes may
be surrogates for other watershed attributes that supply base cations to the streams.
The buffer zone model explains slightly more of the variability in SOBC than the whole watershed
model. This lends support to the hypothesis that the near-channel soils may have the greatest effects
on surface water chemistry for some variables.
8-108
-------
Stream pHs in the SBRP are higher than the northeastern lake pHs with a central tendency near
circumneutrality. The regression models for stream pH developed with ail 35 whole watershed
observations and with all 33 100-m buffer observations are identical in that they Include the same
variables. As with stream ANCs and the concentrations of Ca plus Mg, stream pH is strongly influenced
by the presence of calcareous soils. The results of the stream pH analysis are presented in Table 8-47.
Omitting the two highest OTC sites and one other observation with high influence diagnostics, produces
a modei with two variables that explains 38 percent of the observed variation in stream pH. The two
variables are the OTL and SHL sampling classes. After the OTC sampling class, the OTL class has the
highest base saturation in the region. The SHL dass consists of shallow Inceptisols and Ultisols with
moderate base saturation (-11%) and alone only accounts for approximately 3 percent of the variation
in stream pH. By far, the OTL sampling class soils are more important in explaining the variability in
stream pH..
Using only the soils and miscellaneous land areas within 100 m of the streams and the deposition
data results in a final model with two variables that has an R2 of 0.29. One of the variables is the
sampling class FR, and the other is the sum of the dry Ca and Mg deposition. Both have negative
parameter estimates. The negative parameter estimate on the FR is because, in the region as a whole,
the soils in the FR sampling class have the lowest pHs and base saturation. The negative relationship
with Ca and Mg in dry deposition does not seem reasonable. This variable may be a surrogate for
another independent variable.
8.6.4.3 Regional Comparisons
Because ANC, Ca plus Mg, and pH are interrelated, the resulting regression models within each
region are often similar. The median ANC and pH values in the northeastern lakes are lower than in
the SBRP streams even though the median Ca plus Mg concentrations are higher. This is a direct result
of elevated fake sulfate concentrations in the NE.
8-109
-------
Table 8-47. Regression Models of Stream pH in SBRP Streams, Developed Using
Deposition, Mapped Soils (as a Percentage of Watershed Area in Soil Sampling
Classes) and Miscellaneous Land Areas (as a Percentage of Watershed Area) as
Candidate Independent Variables8
Whole Watersheds
SO4-Wet
Ca+Mg-Dry
FR
OTC +
SKV
SHL +
OTL + +
MSL +
ACC
R2 0.67 0.42
adjusted R2 0.61 0.39
n-streamsb 35 33
p-modelc 5 2
0/Ld none 2
100-m buffer zones
-
-
-
+ + +
-
+ + +
+
-
0.68 0.38 0.39 0.47 0.29
0.63 0.34 0.31 0.45 0.23
33 31 30 29 28
5 2 3 1^2
none 2345
* +• and - refer to positive and negative parameter estimates, respectively
° n-lakes = number of observations used to develop the regression model
" p-model = number of regressor variables in model
O/L = number of outlier or leverage points omitted
8-110
-------
Results from both regions show that soils with high base saturation, especially those derived from
calcareous parent materials (SBRP), are associated with surface waters that have higher values of ANC,
base cations, and pH. Both lake and stream resources are susceptible to the effects of major watershed
disturbances (e.g., quarries). When these disturbances are present, the subtending surface waters will
have elevated base cation concentrations as well as elevated sulfate levels. In most cases the increase
in sulfate is balanced with concomitant increases in base cations. Therefore, the ANCs are not generally
negatively impacted by the water disturbance.
In the NE, poorly drained (wetland), organic soils that are acidic appear to decrease lake pH.
This is probably due in part to organic acids from these soils. In the SBRP frigid soils (PR sampling
class) are associated with lower pH surface waters. These soils are low pH and low base status. As
was observed in the northeastern sulfate analysis above, the coastal watersheds appear to have occult
sources of deposition that result in lower surface water ANCs, pHs, and Ca plus Mg.
In the NE the whole watershed regression models generally had about the same explanatory
power as the models developed using buffer zone data. In the SBRP, however, the models developed
using buffer zone data usually had more explanatory power; the only exception was for stream pH. This
suggests that in stream watersheds, the near-channel zones have a greater effect on surface water
chemistry for some variables than the rest of the watershed. To test this hypothesis definitively, however,
would require finer resolution mapping data than those obtained in the DDRP.
8.6.5 Summary and Conclusions
The difference in soils between the two regions accounts for most of the observed differences
seen in sulfate and sulfur retention. Compared to the soils of the SBRP the soils in the NE are young,
shallow, less developed, and have a lower overall capacity to retain sulfur. In contrast, the soils in the
SBRP are, in general, deep and highly weathered, with abundant secondary mineral phases that provide
considerable sulfur retention capacities.
8-111
-------
The major conclusions of this analysis are:
• In the NE, where sulfur retention on average is low, sulfur deposition explains more of the
observed variation in lake sulfate concentrations than any other independent variable.
• In the SBRP, where the majority of sulfur inputs are retained by watershed soils, sulfur
deposition is not yet significantly related to stream sulfate concentrations.
• Local sources of sulfur deposition from sea salt may be negatively affecting the surface
water chemistry in the near-coastal watersheds in Subregion 10 of the NE.
• Wetland soils or soils that are wet part of the year promote sulfur retention via sulfate
reduction reactions.
• Shallow soils with short hydrologic contact times and low sulfate adsorption capacities do
not interact sufficiently with drainage waters to affect their chemistry. Watersheds with these
types of soils in areas of high sulfur deposition are likely to be susceptible to surface water
acidification.
• Soils with high base saturation are associated with higher surface water ANC, pH, and base
cations.
• Poorly drained, acidic organic soils in the NE and frigid soils in the SBRP are associated
with lower pH surface waters. In the NE this relationship may be due in part to the result
of the organic acids in these soils. The frigid soils are low pH and low base status.
* Using only mapped soils information from the near-stream areas in the SBRP to develop
regression models generally produced models with more explanatory power than those
developed using only information from the whole watershed.
This type of analysis has been shown to be a useful tool for regionally assessing the relationships
between surface waters and soils, miscellaneous land areas, and deposition. Although these attributes
alone cannot account for all the variability in the observed data, there are some instances in which they
do account for most of it. Care must be exercised in evaluating the resulting regression models.
Inclusion of outliers and leverage points may result in models that are heavily biased. Sample populations
with small sample sizes are particularly susceptible to bias.
This analysis demonstrates the utility of soil sampling classes in characterizing the soils across
large geographic regions. It has helped us to assess the concept of soil sampling classes and may
lead to some revisions in the way classes are differentiated.
8-112
-------
8.7 ANALYSES OF DEPTH TO BEDROCK
8.7.1 Introduction
One of the Important findings of the Integrated Lake/Watershed Acidification Study (ILWAS) (Newton
and April, 1982; Goldstein et al., 1984) was that the depth of soil and surficial geological materials have
a significant effect on the quality of subtending surface waters. The ILWAS scientists found that the
difference In lake pH between Woods (pH 4.4-5.1) and Panther (pH 5-7.5) Lakes could be attributed to
the difference in the depths of surficial materials on these watersheds.
In addition to depth, the chemical and physical characteristics of the surficial materials are also
important. The latter affects the hydrologic flow path and the hydrologic contact time, which in turn
affect the length of time to react with the drainage waters. Short hydrologic contact times produce little
alteration in the chemistry of the drainage waters. The chemical characteristics of the surficial material
are also important. Materials without weatherable primary minerals will have little beneficial effect on
acidic inputs, even though they may be deep. Our objective In this analysis is to test this depth-to-
bedrock hypothesis on a regional basis.
8.7.2 Approach
A depth-to-bedrock map was prepared during the mapping phase of the ODRP by assigning a
depth-to-bedrock class to all soil map delineations. The procedure used in the NE was presented in
Section 5.4.1.2 and for the SBRP In Section 5.4.2.2. In addition to the depth-to-bedrock maps, depth
to bedrock was recorded for each soil component in the soils database for each region (Sections
5.4.1.1.2 and 5.4.2.1.2). The soils databases, therefore, provided an alternative approach to estimate
the extent of depth-to-bedrock classes on watersheds and subsequently subregions and regions. Using
these data rather than the data from the depth-to-bedrock maps provides a more precise method to
estimate the proportion of depth-to-bedrock classes.
The depth-to-bedrock classes developed for this analysis from the soils data for the NE and SBRP
are listed in Table 8-48. Note that the numbering of the classes proceeds from the rock outcrop (I) to
8-113
-------
Table 8-48. Depth-to-Bedrock Classes for the Northeast
and the Southern Blue Ridge Province
Northeast Region
Class Depth range (cm) Definition
NE 1
NE~1l
NE III
NE"IV
NE~V
NE~Vl
10 -25
25 -50
50- 100
100 - 150
150 +
Rock outcrop
Very shallow
Shallow
Moderately deep
Deep
Very deep
Southern Blue Ridge Province
Class Depth range (cm) Definition
SE I Rock outcrop
SEll 10-25 Very shallow
SElll 25-50 Shallow
SE~IV 50-100 Moderately deep
SE"V 100-200 Deep
SE~VI 200 - 500 Very deep
8-114
-------
the very deep (VI), and that depth classes V and VI indicate deeper soils for the SBRP than for the NE.
In both regions, depth class I represents rock outcrop on the watersheds.
The depth of each soil (component) is recorded in the soil component file. As described in Section
8.6.2, the mapped soils are linked to the soil component file. This file contains component-specific
information, including soil depth. Using this soil depth information we calculated the percentage of each
watershed in each of the depth categories. These percentages are used as the independent depth-to-
bedrock variables in the following analysis. The LTA sulfate and hydrogen deposition estimates, both wet
and dry, are also used as candidate explanatory variables.
The descriptive statistics on the proportion of these depth classes for both NE and SBRP are
presented in Table 8-49. In general, the NE has higher proportions of shallower soils than does the
SBRP. The proportions of deeper soils are not strictly comparable between the regions, because depth
classes V and VI are not the same across regions.
Within the NE, Subregions 1A and 1E have the highest percentages of rock outcrop, and
Subregions 1C and 1D have the lowest. Subreglon 1A has the highest percentage of very shallow and
shallow soils, while Subregions 10 and 1B have the lowest. Subregions 1D and 1C have the highest
proportions of the very deep soils, and Subregion 1A has the lowest proportion.
The statistical analyses used in the section are discussed in Section 8.1.2. Residual analysis
revealed heteroscedasticity in the residuals for ANC and base cations for both regions. We, therefore,
log-transformed these dependent variables in the analyses for depth-to-bedrock relationships.
8.7.3 Suifate and Percent Sulfur Retention
8.7.3.1 Northeast
In the NE, depth to bedrock seems to have little effect on surface water sulfate (Table 8-50).
Surface water sulfate Is dominated by wet and dry sulfate deposition. The positive correlation between
8-115
-------
Table 8-49. Regional and Subregionai Statistics for the Depth-to-Bedrock
Classes
NE
NE 1
NE'll
NE"III
NElV
NE~~V
NE~VI
Subregion 1A
NE I
NE'll
NE"III
NE'IV
NE"V
NE~VI
Subregion 1B
NE I
NE II
NElll
NElV
NE~V
NE~VI
Subregion 1C
NE I
NE'll
NElll
NE IV
NE~V
NE~VI
Average
3.34
4.24
10.81
13.22
0.45
67.79
Average
4.77
8.08
17.87
17.76
0.00
51.45
Average
3.47
1.76
7.31
20.56
0.00
66.73
Average
1.92
3.41
9.94
10.13
0.005
75.46
Median
1.6
2.3
10.3
11.2
0.0
70.0
Median
2.8
6.8
15.6
16.0
0.0
55.6
Median
0.0
0.0
3.8
14.5
0.0
67.5
Median
1.7
2.8
10.6
9.9
0.0
73.3
Minimum
0.0
0.0
0.0
0.0
0.0
2.8
Minimum
0.0
0.0
0.0
0.0
0.0
12.1
Minimum
0.0
0.0
0.3
6.7
0.0
26.1
Minimum
0.0
0.0
0.0
0.0
0.0
32.7
Maximum
28.4
42.8
60.5
56.7
64.8
100.0
Maximum
18.1
21.3
34.9
36.4
0.0
100.0
Maximum
24.0
12.7
26.6
48.2
0.0
92.8
Maximum
8.6
10.0
22.6
32.2
0.1
100.0
continued
8-116
-------
Table 8-49. (Continued)
Subregion 1D
NE 1
NE~II
NE III
NElV
NE~V
NE~VI
Subregion 1E
NE I
NE~U
NE'lll
NE IV
NE~V
NE~Vl
SBRP
SE t
SE'll
SE~III
SE IV
SE V
SE'VI
Average
2.20
0.88
4.11
5.73
0.00
86.47
Average
4.00
5.66
12.36
11.29
1.85
64.81
Average
1.2
0.0
10.1
23.7
14.6
50.2
Median
0.3
0.1
1.0
1.4
0.0
96.6
Median
1.6
2.0
10.8
7.6
0.0
70.0
Median
0.7
0.0
6.6
19.3
7.1
38.2
Minimum
0.0
0.0
0.0
0.0
0.0
33.2
Minimum
0.0
0.0
0.0
0.0
0.0
2.8
Minimum
0.0
0.0
0.0
0.0
0.0
12.6
Maximum
12.8
4.4
22.4
27.0
0.0
100.0
Maximum
28.4
42.8
60.5
56.7
64.8
100.0
Maximum
15.0
2.6
30.0
53.8
58.3
99.0
8-117
-------
Table 8-50. Results for NE of Regressions of Surface Water Chemistry on Depth-to-Bedrock
Classes and Deposition Estimates
Water
Chemistry Adjusted
Variable R2 R2
Suifate 0.2723 0.2621
Percent
Sulfur
Retention 0.1051 0.0983
Log(ANC+100) 0.2603 0.2446
Log(Ca+Mg) 0.2481 0.2211
pH 0.3203 0.3058
Variable
in Model
wet 804 dep.
dry SO4 dep.
dry SO4 dep.
wet S04 dep.
dry S04 dep.
NEJI
dry SO4 dep.
wet SO4 dep.
NE V
NE~"H
NE~VI
wet S04 dep.
dry SO4 dep.
NE II
Regression Signif.a
Sign Level
+ **
+ *
+ ***
***
+ ***
*
+ ***
***
**
**
S
***
+ ***
**
a S = significant at 0.15 level
* = significant at 0.05 level
** = significant at 0.01 level
*** ^ significant at 0.001 level
8-118
-------
percent sulfur retention and dry sulfate deposition may represent a spurious correlation due to the
formulation for percent retention.
8.7.3.2 Southern Blue Ridge Province
In the SBRP, depth to bedrock has a significant effect on sulfate dynamics. The percent of shallow
soils (SEJII) has a strong positive relationship with surface water sulfate and a strong negative
relationship with percent sulfur retention (Table 8-51). This suggests that, as the percent of shallow soils
increases and the percent of deep soils decreases, the amount of sulfate adsorption decreases. This
decrease in sulfate adsorption may be due to several factors. The deep soils have more mass and,
hence, more total sulfate adsorption capacity. The deep soils may also have higher contact times and
different flowpaths for the soil water.
8.7.3.3 Comparison of Regions
It appears that in-lake sulfate in the NE is predominantly controlled by atmospheric deposition
and not by the depth of surficial material. In the SBRP, the shallow (25 • 50 cm) category of depth to
bedrock accounts for about 32 percent of the variability in observed stream sulfate concentrations and
more than 40 percent of the variability in watershed sulfur retention estimates. These results imply that
shallow soils play an important role in regional sulfur dynamics in the SBRP and that they are often
associated with higher stream water sulfate concentrations and lower watershed sulfur retention.
8.7.4 ANC. Ca Plus Ma and oH
In this part of the depth-to-bedrock analysis we consider the relationships between the proportion
of watershed coverage in the various depth-to-bedrock categories and the non-sulfur dependent variables.
Unlike the NE, there is little or no indication that the DDRP sample of streams in the SBRP are
contaminated with Na from road salt or sea salt. Therefore, in addition to considering the sum of stream
Ca plus Mg as a dependent variable, we have included an analysis of SOBC (Ca + Mg + Na + K). Due
to the behavior of the residuals of the regressions, both ANC and base cations were log-transformed to
8-119
-------
Table 8-51. Results for SBRP of Regressions of Surface Water Chemistry on Depth-to-Bedrock
Classes and Deposition Estimates
Water
Chemistry
Variable
Adjusted
Variable
in model
Regression
Sign
Slgnif*
Level
Sulfate
0.3200 0.2966
SE
Percent
Sulfur
Retention
0.4211 0.4004
SE HI
Log(ANC)
Log(Ca+Mg)
Log(SOBC)
PH
0.3140
0.2740
0.2202
0.1494
0.2667
0.2239
0.1933
0.1210
dry H dep.
SE_V
SE V
wet SO4 dep.
SE_V
dry H dep.
+ *
*
**
S
*
+ *
* S = significant at 0.15 level
* = significant at 0.05 level
** = significant at 0.01 level
*** = significant at 0.001 level
8-120
-------
remove heteroscedastlcity. One hundred (100) was added to the ANC before transforming, in order to
avoid problems in taking the logarithm of non-positive numbers.
Wet sulfate deposition was negatively correlated with ANC, and dry sulfate deposition was positively
correlated with ANC (Table 8-50). Wet sulfate deposition was introduced in the regression model first,
and it presumably represents decreases in alkalinity of the surface waters with increasing deposition of
sulfate. The second deposition variable may be a correction to an overfilling with wet sulfate deposition,
or it may be a surrogate for some explanatory variable not included in the model. Since there are some
high ANC sites in Subregion 1B, it could also represent a geographic effect, as discussed in Section 8.2.
The very shallow soils represented by NE_11 (10-25 cm) are negatively correlated with ANC (Table
8-50). This results suggests that as the proportion of soils deeper than 25 cm increases, the capacity
for cation exchange increases and ANC of the surface waters increases. It also suggests that these
soils may have short hydrologic contact and therefore little or no effect on drainage water chemistry.
A similar behavior in the explanatory variables is seen for Ca plus Mg. Dry sulfate deposition is
positively correlated with in-lake base cations. This correlation may represent increased cation exchange
and leaching due to acidic deposition in a system at or near sulfur steady state. A second deposition
variable, wet sulfate, is introduced with a negative parameter in the model later. As discussed previously,
this may be a surrogate for some other variable or variables estimate or possibly a geographic effect.
Three of the depth-to-bedrock classes were included with negative parameter estimates. Included were
the deep (NE_V, 100 - 150 cm), the very deep (NE_VI, > 150 cm), and the very shallow (NEJI, 1 - 25
cm) depth categories. This result is contrary to the hypothesis suggested by the ILWAS project outlined
in Section 8.7.1, that the deeper the surficial geologic material (i.e., the deeper the depth to bedrock) the
higher the pH, ANC, and base cation status of the surface water. Our result implies the opposite
relationship, on a regional basis. Overall, our model accounts for only about 25 percent of the variability
in Ca plus Mg. Therefore, factors other than depth to bedrock are likely to account for most of the
variability in Ca plus Mg.
8-121
-------
The regression model for lake pH contained the same set of parameters as the model developed
for ANC. This model, however, had a higher R2 (0.32) than the one developed for ANC (0.26). Again,
as wet sulfate deposition increases, surface water pH decreases. Likewise, as the proportion of the
watershed with shallow soils increases, we can expect lower lake water pH.
8.7.4.2 Southern Blue Ridge Province
Due to the behavior of the residuals of the regressions, the dependent variables ANC, Ca plus
Mg, and SOBC were log-transformed to remove heteroscedasticity. The regression models developed
for these transformed variables are presented in Table 8-51. In each of these models, the depth class
SE_V (100 * 200 cm) was included with a negative parameter estimate, in the SOBC model it was the
only parameter included and explained 22 percent of the observed variability. Both the ANC and Ca plus
Mg models included a deposition variable. The estimate of dry hydrogen ion deposition was included
(positive parameter estimate) in the ANC model, whereas wet sulfate deposition was included (negative
parameter estimate) in the Ca plus Mg model. The reasons these variables were included in these
models are unclear.
The negative relationships between depth class SE_V (100 - 200 cm) and ANC, Ca plus Mg, and
SOBC suggest that this depth class represents surficial material that is highly weathered and therefore
deep, with little or no weatherabie minerals. In the SBRP, because the soils and surficial materials are
old and highly weathered, unweathered primary minerals may be prevalent only at the bedrockisoil
interface, In the saprolite. It is reasonable to assume that as these saproiites get farther from the soil
surface, the weathering rates (cation supply rates) may actually decrease because they are farther
removed from diurnal and other environmental Influences. Because of this, watersheds with abundant
deep, highly weathered soils, will probably be associated with lower ANC, pH, and base status surface
waters.
The regression model for stream pH only included the estimate of dry hydrogen deposition and
only accounts for about 15 percent of the observed variability in stream pH. Dry deposition has a
8-122
-------
positive parameter estimate, implying that as it increases so does stream pH. This apparent relationship
is unreasonable; therefore, dry hydrogen deposition is probably functioning as a surrogate for another
variable that is positively related to stream pH.
8.7.4.3 Comparison of Regions
In the NE, wet and dry sulfate deposition are important factors included in the depth-to-bedrock
regression models developed for ANC, Ca plus Mg, and pH. In these models, the very shallow (NEJI,
1 - 25 cm) depth-to-bedrock categories were consistently negatively related to the dependent variables.
This implies that as the proportion of the watershed in the very shallow depth categories increases (i.e.,
lower proportion of deeper material), we can expect the ANC, Ca plus Mg, and pH to decrease. This
is a reasonable result since shallower surficial materials are generally indicative of lower base cation
supply capacities.
In the SBRP the deep depth-to-bedrock category was negatively related to ANC, Ca plus Mg, and
SOBC. This result suggests that this depth class represents surficial material that Is highly weathered and
deep, with little or no weatherable minerals. As the proportion of the watershed in this deep material
increases, we can expect ANC, Ca plus Mg, and SOBC to be lower. The regression model for stream
pH did not include any depth-to-bedrock variables.
8.7.5 Summary and Conclusions
Depth to bedrock appears to have an important effect on sulfate dynamics in the SBRP, but not
in the NE. An important reason for this difference is that, in general, NE watersheds are at sulfate
steady state, whereas the SBRP sites are not. In both regions, depth to bedrock appears to be related
to cation supply dynamics but in opposite ways. In the NE the shallower surficial material is associated
with lower base status surface waters. In the SBRP the deeper material is also related to lower base
status surface waters. In the NE as the proportion of the watershed in the very shallow depth categories
increases (i.e., lower proportion of deeper material) we can expect the ANC, sum of Ca plus Mg
concentrations, and pH to decrease. In the SBRP it is hypothesized that the deep class of surficial
8-123
-------
material represents highly weathered materials with little or no weatherable minerals remaining. As the
proportion of the watershed in this class of material increases, lower base status surface waters can be
expected.
8.8 INTEGRATED ANALYSIS OF ALL MAPPED VARIABLES
8.8.1 Introduction
Each of the preceding sections has considered the relationship between an isolated set of
independent watershed variables and the chemistry of the subtending surface waters. These analyses
considered the relationships of surface water chemistry to atmospheric deposition only (Section 8.2),
derived hydrologic parameters (Section 8.3), mapped bedrock geology (Section 8.4), mapped
landuse/vegetation (Section 8.5), mapped soils (Section 8.6), and depth to bedrock (Section 8.7). None
of these attributes alone can explain all of the variability in the observed chemistry. The chemistry of
surface waters is the integrated result of many interacting factors, including those just mentioned.
In this analysis we combine the data from Sections 8.2-8.7 to develop regression models that
more fully account for the variability in the observed dependent variable data. Our objective is to identify
the most important relationships that exist between watershed physical characteristics and surface water
chemistry. In Section 8.10 we include the soil chemical and physical data. In the analysis presented in
this section, we do not consider any of the watershed buffer zone data.
x
8.8.2 Approach
The approach used in this analysis follows that described in Sections 8.1.1 and 8.1.2 with the
following exception. Because the number of explanatory variables in this analysis exceeded the number
of watersheds in the SBRP, Mallow's Cp statistic could not be used as a model selection criterion;
Akaike's information criterion was used instead. After each model was developed we performed residual
analysis on it, checking for leverage points and outliers, as well as for the standard regression
assumptions as described in Section 8.1.2.
8-124
-------
Sulfate and sulfur retention
8.8.3.1 Northeast
In the NE there is a strong relationship between wet sulfate deposition and sulfate concentration
in the lakes (Table 8-52). Because the watersheds are in general at sulfur steady state, the surface
waters tend to reflect the sulfur chemistry of atmospheric deposition. The negative relationship between
aquatic sulfate and precipitation indicates dilution effects caused by Increased rainfall and runoff.
Anthropogenic factors also strongly affect the sulfate concentrations. The miscellaneous land area
M04 (quarries) and the soil sampling class E06 (made land) both represent sources of sulfur from
anthropogenic watershed disturbances (Table 8-52). The positive correlation with Factor 9 also represents
anthropogenic sources: Factor 9 Indicates Increasing cabin count, urbanization, and quarries. The
positive con-elation between lake sulfate and Factor 4 (agricultural land, and cropland and pasture land)
results from some combination of anthropogenic amendments (e.g., lime, fertilizers) to the soil and
preference for conducting agricultural activities on fertile soils, which are likely to have higher pH and thus
reduced anion adsorption capacities. If these soils are limed or amended with phosphate, displacement
of sulfate from adsorption sites may result in increased sulfate moving into surface waters.
The correlations with the soil sample class H03 indicate that reduction of sulfate and retention of
sulfur by wetlands (Table 8-52) are also important. The positive relationship with H2O_WS (the water
bodies to watershed area ratio) suggests that in-lake sulfate reduction has a greater effect on sulfur
budgets in those watersheds with high watershed to lake area ratios and long lake hydrdogic residence
times (see Section 3.3.7.2). An alternative explanation could be that this relationship reflects less
opportunity for precipitation to contact soils and hence more control of sulfate concentration by the
deposition.
The first variable selected by the stepwise regression for sulfur retention is Factor 5, which
represents wetlands (Table 8-52). This correlation reiterates the importance of wetlands in the
8-125
-------
Table 8-52. Regression Models of Surface Water Sulfate and Sulfur Retention in the NE Lake
Watersheds Using Deposition, Derived Hydrologic Parameters, Bedrock Geology Reaction Classes,
Depth To Bedrock, Mapped Landuse/Vegetation, and Mapped Soils as Candidate Regressor
Variables
Adjusted
Variable8 R2 R2
Sulfate 0.7223 0.6962
(n « 141)
S Retention 0.4710 0.4360
(n = 129)
Variable Regression
in Model Sign
WET SULFATE DEP. +
M04 +
PRECIPITATION
FACTOR9 +
H03
FACTOR4 +
E06 +
H20 WS +
M03~
PERIN
TOTSTRM +
REL_RAT
FACTORS +
I46
NE II
FACTOR12
ATKBMEAN +
MAXREL +
H03 +
Signif.b
Level
***
***
***
***
***
**
S
S
S
S
S
S
**
***
***
***
***
*
*
number of observations included in regression model
• significant at 0.15 level
significant at 0.05 level
= significant at 0.01 level
- significant at 0.001 level
8-126
-------
biogeochemical sulfur cycle on a regional basis. The same rationale applies to the inclusion of soil
sampling class H03, a wetland soil, with a positive parameter estimate.
The very shallow (10-25 cm), NEJI, depth-to-bedrock class was included in the sulfur retention
model with a negative parameter estimate. This result Implies that as the proportion of watershed
coverage in very shallow surficial materials increases, watershed sulfur retention decreases. This is
apparently due to a concomitant decreased watershed sulfate adsorption (sulfur retention) capacity.
Alternatively, NEJI is highly correlated with the presence of both H01 and EOS on a watershed. H01 is
a relatively dry Histosol In which mineralization of organic matter and consequently sulfur production
could occur. The H01 soils are usually associated with high elevations and may be indicative of cloud
interception (he., increased deposition). Factor 12 (rainfall and runoff) was included with a negative
relationship with retention (Table 8-52), suggesting a dilution effect due to increased runoff.
8.8.3.2 Southern Blue Ridge Province
The first variable selected by the stepwise regression procedure is SEJII, in the regression model
of stream sulfate concentration (the shallow depth class) (Table 8-53). This is indicative of the overall
reduced sulfate adsorption capacity of shallower soils. The two depth classes SE_VI (very deep soils)
and SE_V (deep soils) which are brought in later in the stepwise procedure probably represent
corrections to the overfitting of SEJII in the regression. Since the amount of adsorption is not linearly
related to the proportion of shallow soils on a watershed, it is reasonable that the extents of SE_VI and
SE_V are used to correct for the regression estimate for SEJII.
Runoff has a negative relationship with stream sulfate concentration, indicating a dilution effect
from increased precipitation. The sample dass MSL has a positive relationship with stream sulfate
concentration (Table 8-53). This same sample class also has a negative relationship with sulfur retention
in the first two SBRP sulfur retention models. Considered together, these results suggest that MSL may
be related to or indicative of sulfur-bearing parent material. This relationship was also noted and
discussed In Section 8.6. Factor 8 (open land and pasture) has a positive relationship with stream sulfate
8-127
-------
Table 8-53. Regression Models of Surface Water Sulfate and Sulfur Retention in the SBRP Stream
Watersheds Using Deposition, Derived Hydrologic Parameters, Bedrock Geology Reaction Classes,
Depth To Bedrock, Mapped Landuse/Vegetation, and Mapped Soils as Candidate Regressor
Variables
Variable4
Adjusted
R*
Variable
in Model
Regression
Sign
Signrf.
Level
Sulfate
(n = 31)
0.8496 0.7744
S Retention
(n = 32)
0.9297 0.9052
S Retention
(n - 31)
0.6893 0.5948
S Retention
(n=30)
0.5835 0.5168
SE III +
RUftOFF
MSL +
SE VI +
FACTORS +
ACC
REL RAT
SE \7 +
MAX
SKX
MPQ
MSL
SE III
SE~VI
DRY SULFATE DEP.+
H5UP +
DRY H DEP.
FL
MSL
SE III
SE~Vl
DRY SULFATE DEP. +
H5UP +
DRY H DEP.
FL
SEJII
SE VI
SE~V
ACC +
***
S
***
***
*
S
**
**
***
***
***
S
**
**
*
S
***
***
S
**
**
*
S
***
**
S
S
* n » number of observations included in regression model
B S = significant at 0.15 level
* = significant at 0.05 level
** * significant at 0.01 level
*** = significant at 0.001 level
8-128
-------
concentrations, suggesting that anthropogenic additions or the activities of livestock are impacting stream
sulfate concentrations as discussed in the previous section.
The sample class ACC has a negative relationship with stream sulfate concentration. Soils in this
sample are derived from acid crystalline parent materials. They are clayey and have high sulfate
adsorption capacities. Thus, we expect the extent of these soils to be negatively related to in-stream
sulfate. MAX, the maximum bedrock sensitivity number on a watershed, is negatively related to sulfate
In the subtending waters (Table 8-53). In the SBRP this relationship is expected because deeper soils
are associated with more extensively weathered parent materials, which In turn results in increased
amounts of iron and aluminum oxides, the principal sites of sulfate adsorption. More weatherable bedrock
produces more of the deeper, finer textured soils abundant in iron and aluminum.
In the three successive models for sulfur retention, we see explanatory variables similar to those
selected for the stream sulfate concentration regression models. The first model for sulfur retention was
a model developed with 32 SBRP watersheds. The residual analysis identified one watershed as a strong
leverage point due to its unusual negative sulfur retention and the singular presence of MPQ (quarry) on
the site. In the second model another site was identified as an outlier and was also excluded. This
watershed also appears to have an internal source of sulfur. The variables that appear in the first two
models and not in the third model are probably site-specific or are included due to correlations with other
variables.
The miscellaneous land class MPQ (miscellaneous pits and quam'es) is negatively correlated with
sulfur retention in the first model (Table 8-53), indicating an internal source of sulfur, as previously
discussed. The sample classes MSL and PL are also negatively related to percent sulfur retention in the
first two models (Table 8-53). This result may indicate that one or both of these sample classes
occasionally has weatherable sulfur-bearing parent material. The soils in the FL sampling class have low
sulfur retention (adsorption) capacities.
8-129
-------
The depth-to-bedrock classes SEJM and SE_VI are negatively related to sulfur retention In all three
models, and SE_V Is also negatively related to retention in the final model (Table 8-53). As discussed
previously, SEJll may be indicative of the lower capacity of shallower soils to adsorb sulfate, and the
signs for SE_V and SE_VI suggest that they appear in the regressions as nonlinearity corrections for
overestimating the regression parameter for SEJll.
The bedrock geology variable HSuo is positively related to sulfur retention in the first two models,
but not in the third (Table 8-53). This result indicates, as does the negative relationship between sulfate
concentration and MAX. that more weatherable bedrock geologies tend to produce deeper soils with
higher sulfate adsorption capacities. HSup does not appear in the 30-observation regression model.
The first two models suggest some possible indicators of internal sources of sulfate (e.g., MSL,
MPQ), and the final model indicates the importance of soil depth and soil type. The sample class ACC
is positively related to sulfur retention, indicating that very clayey soils, derived from acid crystalline parent
materials, are strong adsorbers of sulfate.
8.8.3.3 Regional Comparisons
in the NE, the surface water sulfate concentrations are strongly affected by sulfur deposition. In
the SBRP, however, the watersheds are not at sulfur steady state, and hence do not mirror trends in
deposition as readily as the northeastern sites do. In both regions, watershed disturbance and agricultural
practices may outweigh the effects of deposition on surface water chemistry. Some soils have distinct
relationships with stream sulfate concentrations and watershed sulfur retention and may be indicative of
internal watershed sulfur sources. The northeastern lakes display more obvious effects of wetlands than
do streams in the SBRP, where extensive wetlands are relatively uncommon. Effects due to soil depth
and bedrock geology are more pronounced in the SBRP. In the NE, sulfur retention seems to be
primarily controlled by extent and type of wetlands. In the SBRP, sulfur retention is controlled by the soil
mass (i.e., oxyanion adsorption capacity) available to adsorb sulfate and the extent of types of soils that
adsorb more strongly.
8-130
-------
8-8-4 ANC. Ca Plus MQ. and oH
8.8.4.1 Northeast
The regression model for ANC in northeastern watersheds indicates that surface water ANC is
primarily driven by watershed-specific variables. The first variable in the model is Factor 4, the extent
of pasture and cropland in the watershed, which has a positive relationship with surface water ANC
(Table 8-54). This relationship may represent an increase in base cations through the use of soil
amendments (i.e., lime), and it may also be indicative of the selection of high base status soils for
agricultural activities. The soil sampling class I46 was also included with a positive relationship with
ANC (Table 8-54), which is expected, because the soils in this sampling class are high base status.
Their typical base saturation is over 75 percent, and their average pH is the highest of all northeastern
soil sampling classes. Factor 12 (precipitation and runoff) has a negative relationship with surface water
ANC (Table 8-54), indicating a dilution effect in the surface waters produced by increased runoff.
MA& the highest value of the bedrock sensitivity code on a watershed, has a positive relationship
with ANC (Table 8-54). The higher bedrock sensitivity numbers are associated with lithologies, such as
carbonates, that can buffer soils and surface water against changes in ANC.
The first variable in the regression model of Ca plus Mg is Factor 4 (pasture and cropland), as
in the model for ANC (Table 8-54). Again, this result probably indicates preference for agricultural
development on higher base status soils and the introduction of soil amendments. Factor 12
(precipitation and runoff) again has a negative relationship with surface water ANC and Ca plus Mg,
indicating a chemical dilution.
The soil sampling classes 146 and 125 are positively related to Ca plus Mg (Table 8-54). Both of
these sampling classes have soils with high base status and high pH. Soil-water flowing through these
soils would be expected to have more exchange of acid cations for base cations and to contribute base
cations to the surface waters. The miscellaneous land area M04 is also positively related to surface water
Ca plus Mg (Table 8-54). M04 (pits, quarries) has been shown previously to be related to increases in
8-131
-------
Table 8-54. Regression Models of Surface Water ANC, Ca plus Mg, and pH in the NE Lake
Watersheds Using Deposition, Derived Hydrologic Parameters, Bedrock Geology Reaction Classes,
Depth To Bedrock, Mapped Landuse/Vegetation, and Mapped Soils as Candidate Regressor
Variables
Variable4
ANC
(n = 138)
Adjusted
R2 R*
0.4860 0.4666
Variable
in Model
FACTOR4
I46
FACTOR12
ATKBMEAN
MAX
Regression Signrt.b
Sign Level
+ ***
+ ***
**
.j. **
+ *
Ca+Mg 0.5877 0.5662 FACTOR4 + ***
(n = 142) FACTOR12 - ***
I46 + ***
M04 + ***
I25 + ***
M01 + ***
H01 - *
pH 0.4621 0.4383 FACTOR4 + ***
(n - 143) FACTOR12 - *
WET SULFATE DEP. - ***
H03 - ***
FACTOR1 + **
DRY SULFATE DEP. + *
n = number of observations included in regression model
S = significant at 0.15 level
* = significant at 0.05 level
** = significant at 0.01 level
*** * significant at 0.001 level
8-132
-------
lake and stream sulfate concentrations (Section 8.6 and 8.8.3). Apparently base cations accompany
sulfate to the surface waters.
The sample class H01 is negatively related to Ca plus Mg (Table 8-54). The soils that make up
this class are thin mantles of organic material overlaying rock outcrop or rock fragments. These soils
generally occur on steep mountain slopes at high elevations. Precipitation failing on these soils flows
rapidly from these areas to drainage ways that feed directly into the surface waters. The lower mass
available for cation exchange and the reduced soil contact could account for the lower base cations in
the surface waters.
Surface water pH is positively related to Factor 4 (pasture and cropland) and negatively related
to Factor 12 (precipitation and runoff), as are ANC and Ca plus Mg (Table 8-54). As discussed
previously. Factor 4 probably reflects preference for agricultural development on higher base status soils
and introduction of soil amendments. Factor 12 again indicates the dilution effect of increased
precipitation. Wet sulfate deposition has a negative relationship with surface water pH (Table 8-54),
indicating the ability of increased sulfate deposition to lower the pH of subtending surface waters.
The soil sampling class H03 has a negative relationship with lake pH (Table 8-54). The soils in
this soil sampling class are dysic (low pH) wetland soils. By definition, they have low pH, typically less
than 4.5. These soils may be contributing organic acids and thus affecting the pH of the surface waters.
Also, these soils can be the last soil that the drainage waters pass through before reaching the lake.
Because of their position in the watershed, these soils may therefore have a significant ultimate effect on
surface water chemistry.
The model developed for lake pH include Factor 1 with a positive parameter estimate. Factor 1
represents developed land. This factor incorporates effects due to waste disposal sites, pits and quarries,
cabins, urban commercial land, and urban residential land. The positive relationship with pH (Table 8-
54) may indicate that the relationship is primarily driven by the base cation influx associated with some
8-133
-------
anthropogenic disturbance, such as pits or quarries. The positive relationship between surface water
pH and dry sulfur deposition suggests that this explanatory variable is a surrogate for some other factor.
8.8.4.2 Southern Blue Ridge Province
The three SBRP sites with high ANC are excluded in order to make the analysis more comparable
with the results from the NE and to adhere to the DORP design. If these sites were included the squared
correlations for ANC and Ca plus Mg would have been over ninety percent, and the squared correlation
for pH would have been over seventy percent. These increases in explained variability are all due to the
presence of highly weatherable bedrock with large amounts of carbonates and calcareous soils on these
watersheds.
The model for surface water ANC shows a positive relationship with the sample class FL and a
negative relationship with runoff (Table 8-55). The sample class FL is composed of flooded soils that
are near the stream channels and have fairly high base status and pH compared to the other soils in
the region. These soils may be the last soil that the drainage waters pass through before reaching the
surface waters. In the near-stream channel position, these soils may have a significant effect on stream
chemistry. The negative relationship with runoff suggests chemical dilution.
Two regression models were developed for Ca plus Mg in SBRP streams. The first was developed
with 32 observations. The second is based upon 29 observations after residuals analysis of the first
model identified one outlier and two leverage points. In both models, the sample class FL has a positive
relationship with stream base cations (Table 8-55), as it does with ANC. Runoff is negatively related to
stream base cations (Table 8-55). Both of these have been discussed previously.
In the first Ca plus Mg model, H5up is positively related to base cations. The higher values of
the sensitivity scale are associated with carbonate bedrocks that weather more easily and contribute
base cations to drainage waters. In both models for Ca plus Mg, there is a positive relationship with
Factor 3 (Table 8-55). Factor 3 represents larger proportions of cropland, land under horticulture, and
8-134
-------
Table 8-55. Regression Models of Surface Water ANC, Ca plus Mg, and pH in the SBRP Stream
Watersheds Using Deposition, Derived Hydrologic Parameters, Bedrock Geology Reaction Classes,
Depth To Bedrock, Mapped Landuse/Vegetation, and Mapped Soils as Candidate Regressor
Variables
Adjusted
Variable* R2 R2
ANC 0.4531 0.4154
(n = 32)
Ca+Mg 0.6714 0.6227
(n = 32)
Ca+Mg 0.7101 0.6471
(n = 29)
Variable Regression
In Model Sign
FL +
RUNOFF
FL +
RUNOFF
H5UP +
FACTORS +
FL +
FACTORS +
FACTOR4 +
DDENSITY
WET SULFATE DEP. -
Signif.b
Level
***
**
**
**
***
*
***
**
S
*
S
SOBC 0.9285 0.8927
(n = 25)
FACTORS
SE VI
OTt
WET SULFATE DEP.
WET H DEP.
ACC
MSL
SKV
***
**
*
S
PH
(n=32)
0.4312 0.3470
RUNOFF
FR
STRMORD
DRY H DEP.
*
*
*
* n - number of observations included in regression model
6 S = significant at 0.15 level
* = significant at 0.05 level
** = significant at 0.01 level
*** = significant at 0.001 level
8-135
-------
open land as defined by the SCS. As in the NE, this result may reflect the impact of soil amendments
on surface water chemistry, and it may indicate that agriculture is conducted on fertile, high base status,
flood plain soils. In the second model, Factor 4 (open land, urban development, and wetlands) is
positively related to stream base cations. The relationship with SCS open land and development
indicates anthropogenic sources of base cations to the surface waters. DDENSITY, a measure of
drainage density, is negatively related to stream base cation concentration. Higher drainage densities
usually indicate a faster runoff response and, hence, lower soil interaction. With less soil contact, the
base cation supply would tend to be lower. In the second model, wet sulfate deposition is negatively
related to base cation concentration. This may be a surrogate for increased precipitation and thus might
represent a dilution effect.
Unlike the NE, there is little or no indication that the DDRP sample of streams in the SBRP is
contaminated with sodium (Na) from road-salt or sea-salt additions. Therefore, in addition to considering
the stream Ca plus Mg as a dependent variable, we have included an analysis of the sum of the four
principal base cations, Ca + Mg + Na + K, (SOBC). The regression model developed for SOBC
explains about 92 percent of the observed variation in SOBC and contains three highly significant
variables with positive parameter estimates. These are Factor 3 (larger proportions of cropland,
horticultural activities, or open land), SE_VI (very deep depth-to-bedrock category, 200 - 500 cm), and
the OTL soil sampling class. As mentioned previously, Factor 3 is indicative of the preference of high
base status soils for agricultural purposes, which tend to be located near streams in the flood plain. In
conjunction with this, soil amendments may result in increased surface water base cation concentrations.
The very deep depth-to-bedrock category (SE_VI) is synonymous with near channel, flood plain soils.
These zones are also where base cation enriched drainage waters and sediments accumulate. The soils
in the OTL soil sampling class are generally very high base status soils, and are therefore associated with
higher base status surface waters.
In the model for surface water pH, there is a negative relationship with runoff (Table 8-55). The
soil sampling class FR is negatively related to stream pH (Table 8-55). The sampling class FR is
8-136
-------
composed of the frigid soils, which have the lowest pH and base status soils in the region. When they
occur, the drainage water pH would be expected to be reduced. Also, the proportion of these soils is
negatively correlated with the proportion of Fl_ the flooded soils with high base status and high pH.
STRMORD, the maximum Morton stream order on the watershed, is positively related to stream
pH (Table 8-55). Larger values of stream order tend to be associated with larger watersheds. These
sites have longer flow paths and more soil contact, which would elevate the pH of the drainage waters.
Dry hydrogen deposition is positively related to stream pH in the regression model (Table 8-55). The sign
of the relationship indicates that dry hydrogen deposition is probably a surrogate for another variable.
8.8.4.3 Regional Comparisons
In both regions, watershed-specific factors appear to be more important than atmospheric
deposition on the base status and pH of surface water. The effects of bedrock lithology and presence
of agricultural land appear across both regions. The base status of soils and their contact time also
affect the surface water ANC.
8.8.5 Summary and Conclusions
The specific conclusions of these analyses are:
The effect of deposition on surface water chemistry is much more distinct in the NE than
in the SBRP.
Major watershed disturbances, such as quarries and urbanization, result in increased surface
water sulfate concentrations. They also produce higher base status surface waters.
Land use, especially near-lake or near-stream agricultural activities (e.g. lime and fertilizer
amendments) may outweigh the effects of deposition on surface water chemistry.
In the NE, wetland soils are associated with sulfur retention.
Shallow soils are negatively related to sulfur retention in both the NE and SBRP. This is
probably caused by their decreased capacity to adsorb sulfate.
In the SBRP, easily weathered parent materials produce abundant iron and aluminum
oxyhydroxides. Soils formed in these types of parent materials are usually deep and have
large sulfate adsorption capacities.
In the SBRP, the very deep depth-to-bedrock category of surficial material is synonymous
with near channel, flood plain soils. These zones are also where base cation enriched
8-137
-------
drainage waters and sediments accumulate. These zones are therefore associated with
higher ANC surface waters.
8.0 SOIL PHYSICAL AND CHEMICAL CHARACTERISTICS
8.9.1 Introduction
This section evaluates the relationships between surface water chemistry and the soil physical and
chemical characteristics that were measured by the analytical laboratories as part of the DDRP soil survey.
Section 2 outlined the hypothesized basis for control of surface water chemistry by soil chemical
characteristics, i.e., sulfate retention and base cation supply. Section 3 discussed the influences of these
soil chemical characteristics in greater detail. This section uses an empirical approach to evaluate
whether the hypothesized mechanisms of soil chemical influence on surface water chemistry are
supported by relationships between measured soil chemical and physical data and water chemistry data.
Relationships between soil characteristics and surface water chemistry are evaluated in this section
using bivariate correlations and multiple regressions. The dependent variables are discussed in Section
8.1.2.
8.9.2 Approach
The candidate independent or explanatory variables considered in this section are those that were
measured at the soil analytical laboratories on soil samples taken during the DDRP soil survey. A
complete list of the measured physical and chemical characteristics was given in Table 5-22. Summary
statistics for the subset of those variables that were used in this section are given in Section 8.9.4.
The soil samples analyzed by the DDRP were'from individual subhorizons of pedons sampled
randomly from areas of occurrence of predefined sampling classes as described in Section 5.5. As
many as seven and as few as zero pedons were sampled on each watershed. In order for the data to
be used in the empirical analyses, they were aggregated through the sampling class framework and single
values calculated for each watershed according to the mass and the area of occurrence (from the
8-138
-------
mapped data, Section 5.4) of each sampling class on each watershed. This procedure is described in
more detail in Section 8.9.3.
There are questions about how data should be aggregated from single points in a heterogeneous
watershed or landscape to represent the entire study area. The related issues have been discussed in
detail by Johnson et al. (I988b). For these Level I analyses the aggregation should yield a value that
Is representative of the soils that influence the chemistry of water draining into the lake or stream as
measured by the index sample. The index sample (defined in Section 5.3) represents water that has
passed through the watersheds over different time periods and along different flow paths. For example,
some portion of the water in northeastern lakes passed slowly through the deeper soils of the watersheds
and entered the lakes or streams draining into the lakes as baseflow; another portion flowed rapidly
through shallow soils as quickflow draining directly into the takes or streams because the deeper soils
were saturated. Thus, under some hydrologic conditions, characteristics of the deeper soils on much or
all of each watershed might be relevant; under spring runoff, frozen, or storm conditions, the attributes
of the shallow soils or soils closest to the lakes or streams might be more important. Since the SBRP
stream samples were collected during baseflow conditions, the influence of shallow hydrologic flowpaths
should be relatively less important than in the NE, and characteristics of deeper soils over most of the
watersheds should be most relevant.
The soils data have been aggregated two different ways to evaluate whether the characteristics
of soils over the entire watershed or soils closest to the lake or stream are more closely associated with
the surface water chemistry. The first aggregation results in watershed values, weighted by area of
occurrence of each sampling class, representing all of the soils on the watershed. The second
aggregation results in watershed values representing the area of occurrence of each sampling class
within mapped buffer zones around the lakes and streams. The development of the buffer zones is
discussed in Section 5.4.1.7.5. The aggregation procedures are discussed in more detail in Section
8.9.3.
8-139
-------
The concept of capacity and intensity variables needs to be considered in these Level I analyses.
Capacity variables include the pool of exchangeable calcium, cation exchange capacity, or sulfate
adsorption capacity, for example. They represent either pools of available ions that may be exchanged
for other ions in solution or sinks that may remove ions from solution. The size of these pools or sinks
determines how long a process such as base cation leaching or sulfate adsorption can occur before the
pool or sink is depleted and other processes begin to occur. Intensity variables, such as pH, base
saturation, and equilibrium soil solution sulfate concentration, represent concentrations of ions that are
readily exchangeable and that quickly reach equilibrium with water in contact with the soil. In the
absence of in-stream or in-lake changes and deposition directly to the stream or lake, surface water
should reflect the values of the intensity variables of the soil with which it was last in contact For the
correlation and regression analyses presented here, both capacity and intensity variables were selected
as candidate variables to evaluate the importance of each in relationships with the index chemistry
variables.
8.9.2.1 Statistical Methods
A multiple linear regression modelling approach was used to estimate the value of a response or
dependent variable as a linear function of a set of predictor variables. Figure 8-9 illustrates the steps
used to develop the regression models. This section provides a brief summary of the modelling
approach.
The DDRP database contains information on 145 lake watersheds in the NE and 35 stream
watersheds in the SBRP. Prior to regression analysis, the distributions of the selected dependent (i.e.,
surface water chemistry) variables were examined for obvious outliers. Based on this examination, two
northeastern watersheds with high lake sulfate concentrations were dropped. In the SBRP three
watersheds were eliminated due to high stream alkalinity and an additional watershed was removed
because of high sulfate. Each of the watersheds deleted due to high sulfate concentrations had open
pits or quarries on a small portion of the watershed. The three SBRP watersheds had ANC > 1200
peq L"1 probably due to the presence of carbonate bedrock.
8-140
-------
f Enter V
Eliminate outlier watersheds
Select variables based on
hypothesized relationships and
correlation analysis
Candidate explanatory variables
Perform collinearity diagnostics
t
Reduced candidate
explanatory variables
Perform multiple linear regression
analysis and model evaluation
Figure 8-9. Model development procedure.
8-141
-------
Candidate explanatory variables were chosen in a two-stage procedure. First, explanatory variables
common to both the NE and SBRP were selected to facilitate comparison of the regression models for
the two regions. This selection was based on hypothesized relationships and nonparametric correlations
between the dependent and predictor variables. At least one soil base cation, pH, sulfate, aluminum, and
particle size variable was included in the initial set of predictors. Ammonium chloride-extracted base
cations were selected over the ammonium acetate extractions to be consistent with the Level II and III
Analyses. Concentrations were used instead of pools because of the collinearity introduced into the pool
estimates when multiplying the concentrations by a common mass.
in building and interpreting multiple regression models it is important to recognize that few
independent, i.e., explanatory, variables In watersheds are statistically independent. Soil pH, base
saturation, and exchangeable calcium are usually correlated with each other, as are total carbon,
extractable aluminum, extractable sulfate, and sulfate isotherm parameters. Candidate variables were
selected from the list in Table 5-22 to eliminate highly correlated variables (those having |r| > 0.95).
The second step In variable selection used the collinearity diagnostics from the regression
procedure (REG) in SAS (SAS Institute, Inc., 1988) to identify highly collinear predictor variables in the
initial set. When a predictor variable is nearly a linear combination of other predictor variables, parameter
estimates for these variable coefficients are unstable and have high uncertainty (Draper and Smith, 1981).
The collinearity diagnostics available in the REG procedure test for near-linear dependencies among sets
of predictors. The intercept was not included in the analysis because zero values for the predictor
variables were generally not within the range of the data (Freund and Littell, 1986). The diagnostics were
applied iteratively to the initial set of predictors. At each step, the maximum condition number was
examined and if it exceeded 30, one of the identified collinear variables was dropped. Preference was
given to keeping a collinear variable that (1) was more mechanistic, i.e., potentially causal, than other
collinear variables; (2) was considered a more reliable measure; and (3) was the only remaining variable
of its type (e.g., hydrologic, deposition, vegetation) (Hunsaker et al., 1986a). Stepwise regression was
then performed, as described in Section 8.1.2.
8-142
-------
8.9.3 Aggregation of Soil Data
8.9.3.1 Introduction
Previous regional analyses of relationships between watershed characteristics and surface water
chemistry in areas with varying levels of acidic deposition have generally been data limited so aggregation
within watersheds was not an issue (e.g., Rapp et al., 1985; Nair 1984; Eilers et al., 1983). Hunsaker et
ai. (1986a), however, used more intensive soils data and maps for the Adirondacks and found that
different aggregation procedures resulted in different associations between soil characteristics and surface
water chemistry.
There are no universally accepted or generalized procedures for aggregating watershed components
to obtain a weighted watershed average or characteristic value. Therefore, there is a variety of
aggregation procedures that might satisfy the objective of the Level I Analyses.
One issue considered in aggregating data for modelling relationships between soil chemistry and
surface water chemistry was the distinction between intensity and capacity variables. Water chemistry
at any point in time is controlled by intensity variables such as soil pH, base saturation, or aluminum
solubility. The effect of intensity variables on water chemistry is dependent on the relative cross-sectional
area of the soil through which water flows just prior to emerging as surface water. Therefore, aggregation
of intensity variables should give greater weight to that portion of the soil last encountered by the water.
Because of the difficulty in quantifying lateral versus vertical flow through watershed soils, we have not
succeeded in defining a credible aggregation scheme for intensity variables, and the method for capacity
variables (below) was used.
Changes in water chemistry over time are dependent on capacity variables such as soil cation
exchange capacity, amounts of weatherable minerals present, or amount of soluble aluminum present.
Unlike intensity variables, the effect of capacity variables is proportional to the mass of soil which the
water contacts before emerging as surface water. Consequently, the capacity variables were aggregated,
weighting by the mass of soil contacted by the water.
8-143
-------
Because aquatic chemistry represents the integrated response of an entire watershed, one
aggregation approach was to define the watershed-level quantity as a weighted combination of the
sampling classes that occur on the entire watershed. This weighting scheme used the percentage
composition of the watershed in terms of the sampling classes (i.e., each sampling class was weighted
by its area) fraction on the watershed).
Another aggregation approach was to consider only those soils in the immediate vicinity of the
lake or stream. Physical and chemical characteristics of the soils in these zones might exhibit a much
stronger relationship with water chemistry than the aggregation of ail watershed soils. For the NE a
combined buffer around the lake and streams draining into the lake was delineated; in the SBRP the area
within a 100-m zone along each side of the stream was determined (see Section 5.4.1.2 for details on
development of the buffers). In both regions the sampling class composition of the buffers was calculated
and the areal fractions were used as weights in calculating aggregated soil chemistry data for the buffers.
It should be noted that there are a number of approaches in addition to the two described here
that could be used to obtain an aggregated watershed estimate. Possible approaches include weighting
by hydrologic group, bedrock type, or vegetation type. However, given the sample design used for the
DORP, the aggregation approaches used for these Level I Analyses all involve weighting by the area of
the sampling classes on all or part of the watershed.
8.9.3.2 Aggregation of Soil Data
Extensive discussion among the DORP investigators resulted in the formulation of a common
aggregation approach that appeared to be applicable for each level of analysis (Johnson et al., 1988b)
This approach was to
(1) weight each horizon by its mass per unit area [thickness x bulk density x (1 - coarse
fragments)] to obtain a mass-weighted average for each pedon
(2) weight the pedon values by their mass per unit area to obtain a sampling class
weighted average
8-144
-------
(3) weight the sampling class value by the product of mass per unit area and area!
proportion of the sampling class on the watershed to obtain a watershed-weighted
average
Mass weighting was necessary for capacity variables {e.g., cation exchange capacity, sulfate adsorption
capacity) because these variables represent the amount of soil potentially available to react with acidic
deposition. Mass weighting was also used for aggregation of intensity variables (e.g., pH, base
saturation) because a more appropriate method was not obvious.
Coefficients for sulfate isotherms describe the partitioning of sulfate between adsorbed and
dissolved phases within the soil. Because the coefficients are derived from a function fitted to a set of
observations, the techniques used to obtain watershed estimates for these coefficients differ from the
aggregation methods described above.
The procedure involved fitting the extended Langmuir equation to isotherms for individual samples
using a nonlinear least squares routine. Estimates of net adsorbed sulfate at a set of reference points
were obtained for each sample using the fitted function, and these estimates were mass weighted to
sampling class. An isotherm was fit to the sampling class values, and net adsorbed sulfate was
estimated at the set of reference points. The net adsorbed sulfate values generated using the sampling
class isotherm coefficients were aggregated for each watershed, using the product of the sampling class
mass and the area! fraction of the sampling class on the watershed as a weight. Finally, an isotherm
was fit to the watershed estimates and the coefficients were derived from the fitted function.
8.9.3.3 Assessment of the DDRP Aggregation Approach
There are several assumptions inherent in the sampling class approach to soil characterization
described in Section 5.5.1. One important assumption is that soil components within a sampling class
are sufficiently similar so that any sample from a particular class can be used to characterize that class.
A consequence is that there may be a significant sample location effect that could inflate the estimate
of the sampling class variance. The following two sections describe procedures for evaluating the
8-145
-------
occurrence and importance of watershed effects. Additional discussion of these results can be found in
Turner et al., in review.
The sampling class definitions grouped soils having similar taxonomy or physical properties with
the assumption that chemistry of soils in a sampling class would also be similar. Comparison of the
variance within sampling classes to the variance between sampling classes, as estimated by a variance
components analysis, revealed that for most soil variables the within-class variance was equal to or
greater than the between-class variance (Table 8-56). Subsequent aggregation to watersheds resulted
in very little variance among watersheds, i.e., the watershed values for most chemical parameters were
very similar for most watersheds. The significance of this result depends on the spatial scale of the
variation. If the observed within-sampling-class variance occurred on the scale of meters (i.e., as if all
pedons for a sampling class were sampled on the same watershed), then the sampling class aggregation
scheme has accomplished a desirable smoothing of the data and it would appear that soils in the DDRP
regions are fairly uniform, especially in the S6RP. If, on the other hand, the observed within-sampling-
class variation occurred on the scale of kilometers, then aggregating through sampling class to watershed
has averaged out real watershed-to-watershed differences. Under this assumption, the uniformity of the
watershed estimates indicates that they are biased toward the regional mean.
The DDRP sampling design was not intended to directly answer the question of the scale of
variation. DDRP soil sampling was statistically designed to characterize sampling classes, not
watersheds. Given the available data we can, however, ask whether there is a watershed effect, i.e., do
the values for a specific variable from all pedons sampled on a watershed tend to be above or below
their respective sampling class means? Analyses described below revealed significant watershed effects
for most variables in both the NE and SBRP.
8-146
-------
Table 8-56. Standard Deviations Within and Among Northeast Sampling
Classes Estimated from B Master Horizon Data.
Variable"
SAND
CLAY
FRAG
AC KCL
CA CL
SBC CL
BS CL
CEC CL
AC BACL
PH~01M
PH H20
C TOT
N TOT
STOT
AL KCL
AL'PYP
AL~CD
SO4 H20
. SO4JPO4
SO4~B2b
SO4 XINb
SO4~SLPb
Within
Replicate
5.437
1.557
6.747
0.155
0.213
0.155
0.121
0.095
0.156
0.139
0.144
0.270
0.310
0.152
0.162
0.142
0.135
0.118
0.173
•
Within
Sampling
Classes
12.861
4.032
12.221
0.359
0.552
0.402
0.354
0.228
0.295
0.369
0.416
0.327
0.305
0.266
0.423
0.265
0.259
0.200
0.338
0.247
0.187
0.228
0.307
Among
Sampling
Classes
15.642
4.752
9.184
0.326
0.538
0.409
0.346
0.214
0.310
0.272
0.370 *
0.375
0.305
0.195
0.367
0.274
0.286
0.219
0.381
0.212
0.055 *
0.228
0.297
Percent Variation
Explained by ,
Sampling Class
60
58
36
45
49
51
49
47
52
35
44
57
50
43
43
52
55
55
56
42
8
50
48
* Variable labels and units are found in Table 8-59. All variables
except SAND, CLAY, and FRAG are Iog10.
b Within replicate estimates not available.
* Within variation significantly larger than among variation (p = 0.05).
8-147
-------
8.9.3.4. Estimation of Watershed Effect
A weighted, unbalanced analysis of variance model that partitions the variability of a given soil
parameter into a sampling class effect, a watershed effect, and a residual error was used to assess the
watershed-specific effect on each variable. The statistical model used in this analysis was:
y,j - 3j + b, + 0^ (Equation 8-1)
where y,j is the pedon value for a given soil parameter from sampling class I on watershed j, a; and bj
are estimates of the sampling class and watershed effects, and e(j is the residual error.
Horizon data were aggregated to the pedon prior to watershed effect analysis in order to avoid
the occurrence of missing values which would result from using only subhorizon or master horizon data,
since not alt pedons sampled had all horizons. Weighted pedon averages for capacity and intensity
variables were calculated using the aggregation approach described in Section 8.9.3.3.
It should be noted that this model does not contain a term for the sampling class by watershed
interaction. Since only one pedon was sampled in a sampling class on a watershed, there were not
enough data to estimate the interaction term. Furthermore, only a small percentage of the possible
sampling class by watershed combinations was actually sampled in each region. Also, the model does
not contain an Intercept in order to avoid the difficulties encountered in using an intercept model with
unbalanced data (Searle, 1987).
For the Northeast, there were 38 a, effects, one for each sampling class. The OTC sampling class
was not included in the SBRP analysis, because the watersheds which contained OTC were outliers with
respect to stream alkalinity and were dropped from the analysis (see Section 8.9.2). Therefore, the SBRP
model contained 11 sampling class effects. The parameterization of the model required that the number
of watershed effects, bj, be one less than the number of sampled watersheds. This parameterization
ensured that the model was full rank and that the estimates of sampling class and watershed effects were
8-148
-------
unbiased. For the Northeast this resulted in 135 b= terms, since only 136 of the 145 watersheds were
sampled. In the SBRP, three watersheds were dropped due to extreme values for stream ANC and the
model for this region contained 31 watershed effects.
The analyses were conducted using the SAS REG regression procedure (SAS Institute, Inc., 1988).
Binary Indicator variables (0 or 1) were used to classify each pedon into the appropriate sampling class
and watershed. The sampling dass estimates obtained from the regression model were aggregated to
watershed, weighting by the areal fraction of sampling class on the watershed and, for capacity variables,
the product of the areal fraction and the sampling class mass. The resulting unadjusted watershed
estimate was modified by adding the estimate of the watershed effect to give an adjusted watershed
value. The adjusted watershed values were then used as explanatory variables in the analyses described
in Sections 8.9.4 to 8.9.6.
8.9.3.5 Evaluation of Watershed Effect
There was a significant watershed effect for most variables (Table 8-57), and therefore the
watershed effect adjustment was applied uniformly to all of the data. The watershed effect adjustment
had little effect on the means of the watershed estimates but the variance was generally much greater
for the adjusted values (Table 8-57). This result was expected given the large within-sampling-class
variance. Variance that had been averaged out in the sampling class aggregation was reintroduced as
a watershed effect. The variability in the distributions of the adjusted values was more like our
expectations of the variability of natural systems (Figure 8-10). Figure 8-11 illustrates the difference in
the watershed means and standard errors for pH in 0.01 M CaCI2. Note that the adjusted watershed
means are more variable from watershed to watershed than the unadjusted means. However, the
uncertainty of the adjusted means is higher than that of the unadjusted means. The actual variance
probably lies between these two estimates.
Because the watershed effect was significant, the watershed-effect-adjusted soil chemistry was
used in the following Level I regression analyses. The large uncertainty of the adjusted estimates limits
8-149
-------
Table 8-57. Means and Standard Deviations of Soil Characteristics by
Aggregation Method and Region
Variable
Variable
NE
Unadjusted
Mean Std, Dev.
Mean
Adjusted
StdDev.
SAND*
CLAY*
SOILDEN *
CA CL
MG CL*
BS~CL*
CEC CL *
PH 01M *
AL PYP
C TOT*
SO4 H2O
S04 P04
65.5
5,17
1.27
1.92
0.45
20.3
6.40
4.02
0.29
4.00
9.66
29.0
13.2
3.82
0.17
1.52
0.37
9.72
2.93
0.13
0.12
2.70
3.82
10.7
65.0
5.21
1.27
2.10
0.39
19.6
7.11
4.28
0.29
4.08
9.57
28.9
17.9
6.33
0.22
5.81
1.09
20.9
6.75
0.42
0.19
4.72
8.22
18.9
SBRP
Unadjusted
Mean Std. Dev.
Mean
Adjusted
Std.Dev.
SAND *
CLAY*
SOILDEN *
CA CL
MG CL*
BS CL*
CEC CL *
PH 01M *
AL PYP
CTOT
SO4 H2O
SO4>04
54.7
18.3
1.31
0.26
0.23
11.03
6.81
4.32
0.25
0.93
8.82
84.9
2.13
2.56
0.07
0.06
0.03
1.98
0.86
0.06
0.09
0.44
1.28
10.6
55.1
16.7
1.27
0.37
0.24
12.3
7.30
4.37
0.29
1.21
9.42
82.8
12.1
5.69
0.13
0.40
0.15
9.04
3.19
0.20
0.21
1.13
4.61
36.0
* Watershed effect significant at p < 0.01.
8-150
-------
Unadjusted
total Clay (percent wt.)
Adjusted
2.6
7.6
12.6
17.6
22.5
27.5
32.6
37.5
42.5
47.5
6 10 18 20 25 OS 10 15 20
Frequency Frequency
Base Saturation, NH^CI (percent)
Unadjusted Adjusted
2.
r.
12.
17.
22.
27.
32.
37.
42.
47.
2.6
7.6
12.6
17.6
22.5
27.5
32.5
97.5
42.5
47.5
5 10 16 20 26 05 10 15 20
Frequency Frequency
Cation Exchange Capacity, NH
-------
5,5-
5.0
O
0
O
_ 4.5
o
3.6
"II
2222222222222222222222222222222
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
0000000000000000000000000000000
7777777777777777777778888888888
7788888888888888888888888888999
222222333380000011000
00000
13256
1111
23671
36789034522345601 146
Watershed
Figure 8-11. The mean pH ± 2 standard errors for the SBRP watersheds estimated using the
common aggregation (bars) and the watershed effects adjusted aggregation (lines) illustrate the
lack of variation among the common aggregation values.
8-152
-------
Table 8-58. Population Means and Standard Errors for Selected Variables, by Subregion/Region
and Aggregation (Watershed Adjusted Data)
Mean
± Standard Error by DDRP
Subregion/Region
NE Subregion*
Variable/
Aggregation
b 1A
IS
tc
10
IE
Overall
NE
SBflP
Soil Physical Properties
SAND
ws c
BUF
SILT
WS
BUF
ClAY
WS
BUF
FRAG
WS
BUF
THKA
WS
BUF
SOILDEN
WS
BUF
69.80 £ 3.03
73.87 ±2.46
27.S5 ± 2.38
23.55 t 1.03
2.69 ± 1.01
2.68 ± 0.82
19.88 £ 3.03
20.95 ± 245
70.14 ± 5.88
65.19 x 4.44
1.21 £ 0.03
1.21 £ 0.03
49.52 ± 1.77
45.47 ± 1.78
41.87 ± 1.38
41.84 ± 1.39
1285 ± 0.59
1280 ± 0.59
25.41 £ 1.88
24.32 * 1.88
90,92 ± 3.95
101.35 ± 4.14
1.50 ± 0.02
1.50 * 0.02
68.19 £ 2.82
• 71.11 £ 3.10
28.30 ± 2.22
24.58 £ 2.44
3.68 ± 0.95
4.33 ± 1.04
25.28 £ 262
27.39 £ 283
77.18 X 5.08
81.84 ± 5.25
1.18 x 0.03
1.08 ± 0.04
78.15 ± 219
79.42 ± 259
19.72 ± 1.73
18.81 £ 204
208 X 0.73
1.75 X 0.87
19.10 £ 225
18.47 ± 259
108.33 ± 4.99
111.11 X 9,68
1.41 £ 0.03
1.38 X 0.03
53.56 X 218
54.70 ±284
37.84 ± 1.72
35.28 ±208
8,89 X 0.73
9.80 ± 0.89
21.47 ± 205
2292 x 272
78.58 s. 3.87
81.03 ± 4.74
1.23 X 0.03
1.15 £ 0.03
6217 X 1.32
64.04 X 1.51
31.67 ± 1.04
29.37 X 1.19
6.15 £ 0.44
8.53 X 0.51
2238 X 1.27
23.08 X 1.47
8242 X 2.53
90.19 £ 269
1.29 X 0.02
1.25 X 0.02
50.35 x 1.48
5219 X 1.4S
30.04 x 0.95
30.69 X 0.93
19.61 ± 1.03
16.92 ± 1.01
19.40 x 1.89
25.32 X 1.71
10231 X 7.73
103.79 X 7.05
1.30 £ 0.02
1.25 ± 0.02
Soft Chemical Properties
CACL
" ws
BUF
MQ CL
*WS
BUF
K CL
WS
BUF
NACL
~ WS
BUF
BSCt
WS
BUF
CECCL
WS
BUF
AC BAG).
WS
8UF
PH 01M
WS
BUF
PH H20
WS
BUF
AL AO
" WS
BUF
4.87 ± 2.10
4.95 ± 1.72
0.47 x 0.32
0.89 £ 0.28
0.06 £ 0.01
0.06 £ 0.01
0.03 * 0.02
0.08 * 0.02
18.44 ± 3.94
23.30 ± 3.19
9.89 * 1.80
10.82 ± 1.31
19.82 ± 2.88
19.72 £ 2.36
4.22 ±0.09
4.38 x 0.07
4.80 ± 0.10
4,97 X 0.08
0.68 x 0.08
0.68 x 0.05
2.28 ± 1.15
3.31 ± 1.15
0.74 £ 0.17
0.93 £ 0.17
0.10 ± 0.01
0.10 ± 0.01
0.09 ± 0.01
0.08 * 0.01
29.10 ± 3.22
35.17 ± 3.30
8.93 ± 0.88
9.59 £ 0.8B
10.68 ± 1.98
11.19 £ 1.97
4.37 X 0.08
4.47 X 0.07
9.01 ± 0.07
5.14 ± 0.07
0.23 ± 0.04
0.21 x 0.03
1.36 £ 200
240 ± 1.98
0.20 X 0.30
0.58 * 0.30
0.08 ± 0.01
0.10 ± 0.01
0.09 ± 0.02
0.10 ± 0.02
22.29 ± 3.77
30.21 X 3.90
6.27 » 1.52
S.44 * 1.50
18.95 ± 273
24.57 ± 271
4.40 ± 0.08
4.50 ± 0.08
4.98 * 0.09
5.08 X 0.09
0.54 x 0.06
0.39 X 0.06
298 X 1.34
3.94 ± 1.55
0.26 ± 0.20
0.40 ± 0.23
0.05 ± 0.01
0.05 ± 0.01
0.03 £ 0.01
0.04 ± 0.01
17.49 X 3.73
20.13 X 4.24
3.98 X 1.03
9.08 t 1.19
10.04 ± 1.83
124S ± 212
4.35 ± 0.08
4.38 ± 0.09
4.83 * 0.09
466 X 0.10
0.24 x 0.04
0.23 ± 0.05
204 * 1.35
2.82 ± 1.52
0.53 ± 0.20
0.93 £ 0.23
0.13 £ 0.01
0.14 ± O.O1
0.07 £ 0.01
0.12 £ 0.02
25.39 ± 2.95
30.49 X 3.58
7.92 ± 1.03
10.84 £ 1.15
17.87 X 1.85
23.06 ± 207
4.43 ± 0.06
4.53 ± 0.08
5.01 X a07
5.10 X 0.08
0.43 X 0.04
0.36 X 0.05
2.85 ± 0.87
3.44 t 0.85
0.44 ± 0.13
0.73 ± 0.13
0.09 ± 0.01
0.09 £ 0.01
0.04 ± 0.01
0.08 ± 0.01
22.80 ± 1.91
2B.29 x 211
7.60 X 0.66
9.43 X 0.65
16.03 ± 1.20
18.94 X 1.18
4.38 X 0.04
4.46 ± 0.04
4.93 X 0.04
5.04 x 0.05
0.48 X 0.03
0.38 X 0.03
0.30 ± 0.07
0.41 X 0.07
0.27 £ 0.03
0.21 £ 0.03
0.15 * 0.01
0.14 £ 0.01
0.02 £ 0.00
0.03 £ 0.00
11.83 X 1.35
1227 £ 1.23
7.17 £ 0.48
7.86 £ 0.45
10.60 ± 0.61
11.90 £ 0.80
4.34 X 0.05
4.38 ± 0.05
5.08 X 0.08
5.12 ± 0.05
0.26 £ 0.03
0.32 £ 0.03
continued
8-153
-------
Table 8-58. (Continued)
Variable/
Aggregation
AL.CO
WS
BUF-
AL.PYP
WS
BUF
AU=OT
WS
SUF
UMEPOT
ws
8UF
C_TOT
WS
BUF
SO4H2O
WS
SUF
S04_PO4
"ws
BUF
SO4_EMX '
WS
SO4_B2 «
WS
WS
SO4_SLP 6
WS
Mean
± Standard Error by DDRP Subregion/Region
NE Subregion
1A
0.47 * 0.05
0.37 ± 0.04
0.43 ± 0.04
0.38 £ 0.03
7.46 £ 0.22
7.71 ±0.17
2.60 ± 0.08
2.76 ± 0.07
5.21 ± 1.39
5.62 ± 1.11
7.77 * 4.10
827 ±3.35
28.57 ± 6.06
24.80 * 4.98
1
30S2.2*.
870.8 ± .
120.93 * .
I
3.15 ±.
IS
0.20 ± 0.03
a 19 £ 0.03
0.18 £ O.O2
O.18 ± 0.02
7.82 ± 0.15
8.11 ±0.16
2.88 ±0.08
2.80 f 0.08
1.80 ± 0.74
2.S6 ± 0.74
13.84 ± 2.25
13.82 * 2.24
23.20 ± 3.32
21.75 ± 3,31
2165.3 ± .
952.7 ± .
289,84 ± .
1.74 * .
1C
0.31 ± 0.05
0.25 ±0.05
0.27 ± 0.03
0.23 ± 0.03
7.54 ± 0.19
7.83 ± 0.20
2.71 ± 0.08
2.86 ± 0.08
4.68 t 1.28
7.07 ± 1.27
8.35 ± 3.90
9.99 ± 3.88
28,90 ± 5.76
31.98 ± 5.70
2623.7 ± .
891.1 ±.
125.49 ± .
Z61 ± .
1D
0.15 ± O.03
0.15 ± 0.04
0.19 ± 0.02
0.18 ± 0.03
8.19 ± 0.19
8.32 ± 0.21
2.73 ± 0.08
2.79 ± 0.09
3.19 ± 0.86
4.38 ± 1.00
11.88 t 2.65
13.07 ± 3.08
40.53 ± a88
41.63 ± 4.48
1750.4 t .
1116.8 ±.
255.83 ± .
1.31 ± .
1E
0.26 * 0.03
0.23 ± 0.04
0.24 ± 0.02
0.22 ± 0.03
7.49 ± 0.15
7.83 £ 0.18
2.74 ± 0.06
2.86 ± 0.07
3.84 ± 0.87
9.87 ± 0.87
7.24 ± 2.64
8.88 ± £95
25.71 ± 3,89
30.28 ± 4.37
2338.1 ±.
872.2 ± .
133.58 1 .
2.38 ± .
Overall
NE
0.29 t 0.02
0.24 ± 0.02
0.27 ± 0.01
0.24 ± 0.01
7.88 ± 0.10
7.93 ± 0.10
2.88 ± 0.04
2.82 ± 0.04
3.86 ± 0.53
9.26 ± 0.59
9.49 ± 1.70
10.51 ± 1.88
28.24 ± 2.52
29.54 ± 2.45
2438.5 ± .
925.6 ± .
175.22 * .
2.33 ± .
S8RP
0.49 ± 0.03
0.42 ± 0.03
0.26 t 0.03
0.32 ± 0.03
.
7.22 1 0.19
7.41 ±0.18
2.51 ± 0.05
2.58 ± 0.05
1.16 ± 0.20
1.42 ± 0.20
9.68 £ 0.83
10.28 ± 0.62
87.91 £ 7.12
82.10 ± 7.02
5382.4 ±.
175.4 ± ,
38.59 ±.
32.7S £ .
Surface Water Chemistry
S0416
ws
SO4.NRET
WS
CAMQ18
WS
SOBC
WS
CMJO
WS
ALKA e
WS
PHEQ11
WS
115.46 ± 3.54
-0.12 ± 0.04
183.63 ± 2O.eO
230.82 ± 24.75
1.00 ± 0.12
80.81 ± 18.29
6.50 ± 0.18
155.16 £ 8.52
0.11 ± 0.05
327.00 ± 32.82
441.98 ± 44,41
1.63 £ 0.17
191.83 ± 31.27
7.15 ± 0.19
92.33 ± 4.99
-0.00 ± 0.07
191.85 £ 19.46
292,30 £ 24.59
1.O8 ± O.13
117.04 ± 19.33
7.01 ±0.13
128.28 ± 8.48
•0.13 ± 0,05
206.01 £ 28.23
507.77 £ 42.57
1.03 ± 0,18
84.60 ± 23.07
8.73 ± 0.21
73.90 ± 4.21
•0.13 ± 0.07
205.03 ± 19.68
292.49 £ 19.68
1.23 ± 0.13
140.02 i 18.60
7.18 ± 0.09
109.54 ± 3.58
•0.05 ± 0.03
220.22 ± 11.47
337.53 £ 16.12
1.19 ± a07
129.60 ± 10.45
8.93 ± 0.07
28.69 ± 4.00
0.72 ± 0.03
121.55 £ 15.17
204.63 £ 22.40
0.68 ± 0.11
128.79 t 16.26
7.26 ± 0.05
* 1A is the Adirondacks, 1B is the Poconos/Catskills, 1C is Central New England, 10 is Southern New England, and IE is Maine.
" Variable labels and units are found in Table 8-60.
c For each variable, WS refers to the entire watershed and BUF refers to the buffer zone.
d Error estimates were unavailable.
• ALKANEW in the NE, ALKA11 in the SBRP
8-154
-------
the predictive power of the soil variables in the regression analyses. Future surveys should be designed
to reduce this uncertainty.
8.9.4 Regional Soil Characterization
Soil physical and chemical properties were expected to vary between the NE and SBRP and
among the subregions of the NE. In this section, soils are characterized using data for measured soil
variables regionalized to the target populations. Means and standard errors for these variables are
presented for each of the northeastern subregions, for the NE as a whole, and for the SBRP in Table
8-58. The regionalized means are averages of the adjusted watershed values weighted by the inverses
of the watershed inclusion probabilities. The standard error of the regionalized mean is the weighted
standard error calculated from the adjusted watershed values weighted by the inverses of the watershed
Inclusion probabilities. Values were calculated for the whole watershed, for the combined buffer zone in
the NE, and for a 100-m buffer zone in the SBRP (see Section 5.4.1.2 for detailed description of the buffer
zones). For base cations, only values from the 1.0 N NH4 Cl extraction were used in these analyses, as
these are the values of interest to the modelling efforts. Data obtained using the 1.0 N NH4 Cl and 1.0
N NH4 OAc extractions were found to be highly correlated, so similarities may be inferred. Values in
Table 8-58, as well as the cumulative distribution frequencies shown in Section 5.5.6, can be used to
characterize the DDRP soils.
Watersheds from the five subregions of the NE differ in the primary soil properties that were
hypothesized to affect surface water chemistry (Sections 2 and 3, Church and Turner, 1986). For the
whole-watershed aggregation, base saturation (BS_CL) ranges from 17 to 30 percent, with the
Adirondacks (1A) and Southern New England (10) soils having the lowest mean base saturation. Cation
exchange capacity (CEC_CL) is lowest in Southern New England, with a mean approximately half that
of the northeastern regional mean. The highest levels of water-extractable sulfate (SO4_H2O) are found
in the two more southern subregions (Poconos/Catskills, 1B, and Southern New England); phosphate-
extractable sulfate is highest in the Southern New England soils. Sulfate isotherms also differ among the
subregions. Sulfate adsorption capacity (SO4_EMX) is highest for soils in the Adirondack Subregion and
8-155
-------
lowest in Southern New England. The Southern New England soils are also characterized by the highest
half-saturation constant (S04_B2) and the second largest equilibrium soil solution concentration
(SO4_XIN). Thus, the adsorption curve for the Southern New England soils is flatter and lower than that
for the other subregions. Soils in the Poconos/Catskills Subregion have similar isotherm parameters to
the Southern New England soils, except for a significantly higher sulfate adsorption capacity. Sulfate
isotherms for soils from the three northern subregions are distinct from those of the southern subregions.
Other soil properties also vary among the subregions. Exchangeable acidity (AC_BACL) is relatively
high in the Adirondacks soils, which are also characterized by the highest sum of base cations and the
highest cation exchange capacity. In general, extractable aluminum (AL_AO, AL_CD, AL_PYP) also is
highest in the Adirondacks soils. Soils of the Poconos/Catskills Subregion are finer-textured relative to
the other subregions and have a higher mean bulk density (SOILDEN). Soils in the Southern New
England watersheds have higher sand content. The relatively low mean CEC may be related to the
higher sand content of these soils. Soils in the Maine (1E) Subregion are similar to the Adirondacks and
Central New England (1C) soils, with relatively high levels of exchangeable acidity and total carbon.
Spodosols represent a large proportion of the soils in these three northern subregions, which may partially
explain these observations. Soil pH varies relatively little among the five subregions.
Comparing regional watershed means for the NE and SBRP, a few differences are notable. Soils
in the NE are characterized by higher concentrations of bases and base saturation, higher acidity, and
much lower clay content and phosphate-extractable sulfate. Carbon content of northeastern soils is also
higher than SBRP soils. Sulfate isotherm parameters also differ significantly between the two regions, with
the SBRP exhibiting significantly higher maximum adsorption capacity and significantly lower equilibrium
soil solution sulfate concentrations than the northeastern soils.
Mean values for soils within the buffer zones are similar to the whole-watershed means for most
variables. Differences exist for the base cation variables (CA_CL, MG_CI_ K_CU NA_CL, BS_CL), where
levels are higher in the buffers relative to the whole watershed. As these buffers represent areas of
8-156
-------
convergent flow (variable hydrologic source areas, riparian zones), this is as expected. Soils in the buffer
zones have higher total carbon content and slightly higher water-extractable sulfate levels. In the Central
New England and Maine Subregions, the mean extractable acidity of the buffer soils is higher relative to
the whole watershed mean. Differences between buffers and whole watershed values are generally larger
in the NE than in the SBRP.
8.9.5 Sulfate and Sulfur Retention
This section and Section 8.9.6 discuss the statistical relationships between measured soil physical
and chemical properties and water chemistry for the DDRP watersheds. These relationships are also
evaluated in terms of potential cause-effect controls on water chemistry. Tables 8*59 and 8-60 show the
nonparametric Spearman correlations between selected soil properties and each of the water chemistry
variables considered. Results of stepwise multiple regressions for sulfate and sulfur retention are given
in Tables 8-61 and 8-62.
8.9.5.1 Northeast
8.9.5.1.1 Whole watershed aggregation -
The coefficients of determination, or R2, range from 0 to 0.56 for sulfate and from 0.12 to 0.64
for sulfur retention in the northeastern subregions. Bivariate correlations between soil properties and
sulfate or sulfur retention are generally not high. The strongest correlation is between lake sulfate
concentration and the zero net adsorption concentration (or equilibrium soil solution sulfate concentration)
from the sulfate isotherms. This relationship makes sense mechanistically, i.e., since northeastern
watersheds are generally near steady state with respect to sulfur deposition, soil and lake sulfate
concentrations both tend to reflect deposition. The highest correlation with sulfur retention is a negative
one with extractable aluminum. Soils in the NE appear to be approaching a new equilibrium with lower
sulfate deposition; soils rich in extractable aluminum have a large adsorbed sulfate pool that is now
desorbing, resulting in an inverse relationship between extractable aluminum and sulfate retention.. The
correlations of sulfate and sulfur retention with soil pH and lime potential also fit this scenario. As would
8-157
-------
Table 8-59. Non-parametric Correlations Between Lake Chemistry Variables and Selected Soil
Properties for the NE DDRP Watersheds
Spearman Correlation Coefficients Significant at 0.05 / Significance Level / N
Variable
Units S0416" SO4 NRET
CAMG16
= 143
ALKANEW
PHEQ11
SoB Physical Properties
SAND
Sand, total
SILT
Silt, total
CLAY
Clay, total
FRAG
Fragments > 2mm diameter
THKA
Thickness adjusted for FRAG
SOILOEN
Bulk density
percent
percent
percent
percent
0.22660
cm 0.0065
0.18434 0.16098
g/ce 0.0275 0.0548
-0.34086
0.0001
0.31267
0.0001
0.30097
0.0003
0.20719
0.0130
•0.33012
0.0001
0.31272
0.0001
0.25367
0.0022
0.18355
0.0282
•0.31760
0.0001
0.30011
0.0003
0.24658
0.0030
0.18659
0.0257
Soil Chemical Properties " , ; .
CA_CL
Exchangeable calcium (NH4 Cl)
MQ_CL
Exchangeable magnesium (NH4 Cl)
K_CL
Exchangeable potassium (NH4 Cl)
NA_CL
Exchangeable sodium (NH4 Cl)
SBC CL
Surnof base cations (NH4 Cl)
BS_CLM
Base saturation
CEC.CL
Cation exchange capacity
meqftOOg
meq/100g
meq/100g
meq/100g
•0.20633
meq/100g 0.0134
percent
meq/IOOg
0.22554
0.0068
0.24922
0.0027
0.33975
0.0001
0.17816
0.0333
0.35370
0.0001
0.20727
0.0130
0.22041
0.0082
0.35981
0.0001
0.19769
0.0179
0.32607
0.0001
0.21764
0.0090
0.21303
0.0106
0.34094
0.0001
0.20094
0.0161
0.31892
0.0001
a SO416 is the lake sutfate concentration, SO4_NRET is watershed sulfur retention, CAMG16 is the lake sum of base cation
concentration, ALKA11 is the lake acid neutralizing capacity, and PHEQ11 is the air-equilibrated stream pH.
continued
8-158
-------
Table 8-59. (Continued)
Spearman Correlation Coefficients Significant at 0.05 / Significance Level / N = 143
Variable Units SO416 SO4 NRET CAMG16 ALKANEW
AC_BACL
Acidity, total exchangeable
PH 01 M
pH (0.01 M CaCB )
PH_H20
pH (deionized water)
AL_AO
Aluminum, add oxalate extr.
AL.CD
Aluminum, citrate drthionrte extr.
AL_PYP
Aluminum, pyrophosphate extr.
ALPCT
Aluminum potential
-------
Table 8-60. Non-parametric Correlations Between Stream Chemistry Variables and Selected Soil
Properties for the SBRP DDRP Watersheds
Variable
Spearman Correlation Coefficients Significant at 0.05 / Significance Level / N = 31
Units S0416* SO4 NRET SOBC ALKA11
PHEQ11
Soil Physical Properties
SAND
Sand, total
SILT
SiK, total
CLAY
Clay, total
FRAQ
Fragments > 2mm diameter
THKA
Thickness adjusted for FRAQ
SOILDEN
Bulk density
percent
0.42863
percent 0.0161
percent
percent
-0.37742
cm 0.0363
g/ce
-0.40161 -0.45121
0.0251 0.0108
Soft Chemical Properties
CA_CL
Exchangeable calcium (NH4 Cl)
MQ_CL
Exchangeable magnesium (NH4 Cl)
K_CL
Exchangeable potassium (NH4 CO
NA CL
Exchangeable sodium (NH4 Cl)
SBC.CL
Sum of base cations (NH4 Cl)
BS_CLM
Base saturation
CEC_CL
Cation exchange capacity (NH4 Cl)
0.58790 -0.40766
meq/100g 0.0005 0.0228
meq/100g
meq/100g
meq/IOOg
0.48226
meq/100g 0.0060
percent
meq/100g
0.51532 0.41331 0.51734
0.0030 0.0208 0.0029
0.49597 0.44315 0.35363
0.0045 0.0125 0.0510
0.54597 0.52823 0.50444
0.0015 0.0023 0.0038
a SO416 is the stream sulfate concentration, SO4_NRET ts watershed sulfur retention, SOBC is the stream sum of base cation
concentration, ALKA11 is the stream acid neutralizing capacity, and PHEQ11 is the air-equilibrated stream pH.
continued
8-160
-------
Table 8-60. (Continued)
Variable
Spearman Correlation Coefficients Significant at 0.05 / Significance Level / N = 31
Units SO416 SO4 NRET SOBC ALKA11
PHEQ11
AC_BACL
Total exchangeable acidity
PH 01 M
pH~(0.01M CaCI2)
PH.H20
pH (deionized water)
AL_AO
Aluminum, acid oxaJate extr.
AL_CO
Aluminum, citrate dftiionite extr.
AL_PYP
Aluminum, pyrophosphate extr.
ALPOT
Aluminum potential (pH - */3 pAI)
UMEPOT
Ume potential (pH •
C_TOT
Carbon, total
SO4.H2O
Sulfata, water extraetable
SO4_PO4
Sulfate, phosphate extraetable
SO4_EMX
Adsorption asymptote
SO4_B2
Half saturation constant
SO4JCIN
Zero net adsorption concentration
SO4_SLP
Zero net adsorption, slope
meq/100g
percent
percent
percent
-0.35524
0.0499
-0.48589
0.0056
0.36573
0.0430
-0.38790
0.0311
percent
0.43427
mg S/kg 0.0146
mgS/k
-0.49395
ueq/kg 0.0047
ueq/L
ueq/L
L/kg
0.46976
0.0077
-0.44032
0.0132
0.41129
0.0215
-0.36169
0.0456
0.37298
0.0388
-0.36653
0.0426
-0.48710
0.0055
0.46169
0.0089
-0.41774
0.0194
0.39194
0.0292
-0.44476
0.0122
-0.42702
0.0166
0.44758
0.0116
8-161
-------
Table 8-61. Results of Stepwise Multiple Regressions for DDRP Lake and Stream Sulfate
Concentrations (SO416) Versus Soil Physical and Chemical Properties
Variable"
SAND
CLAY
FRAG
THKA
SOILDEN
CA CL
MG~CL
SBC CL
BS CL
CEC CL
AC BACL
PH~01M
AL~AO
AL~CD
AL~PYP
ALPOT
C TOT
SO4 H2O
SO4~PO4
SO4~EMX
SO4~B2
SO4~XIN
SO4~SLP
Whole Watersheds Buffer Zone
Subregion* Region Region
1AC 1B 1C 10 1E NE SBRP NE SBRP
5
3
1 1
1 2{-) 2 2(-)
3
1 4
2 63
3 4(-) 4(-)
3(-) 1(-) 3 3
2 2(-) 2 2(-)
1 3 1
1
R2
0.35 0.56 0.52 0.46 None 0.27 0.66 0.43 0.62
Selected
• 1A is the Adirondacks, 1B is the Poconos/Catskilla, 1C is Central New England, 10 is Southern New England,
b and 1E is Maine.
b Variable labels and units are found in Table 8-60.
c Numbers indicate order of entry into stepwise model. (-) indicates a negative parameter estimate.
8-162
-------
Table 8-62. Results of Stepwise Multiple Regressions for DORP Watershed Sulfur Retention
(SO4_NRET) Versus Soil Physical and Chemical Properties
Variable"
SAND
CLAY
FRAG
THKA
SOILDEN
CA CL
MG CL
SBC CL
BS CL
CEC CL
AC BACL
PH~01M
AL~AO
AL CD
AL PYP
ALPOT
C TOT
SO4 H2O
SO4~PO4
S04~EMX
SO4 B2
SO4~XIN
SO4 SLP
Whole Watersheds Buffer Zone
Subregion* Region Region
1AB 1B 1C 1D 1E NE SBRP NE SBRP
5
4
20 2(-) 2(-)
3 3
2 4
1 6
2
10 K-) 10 10 10
20 3(-) 40
40
52 2 2
1
1 1
30 3(->
3(-)
R2
0.12 0.41 0.64 0.30 0.37 0.22 0.44 0.16 0.44
* 1A is the Adirondack^ 1B is the Poconos/Catskills, 1C is Central New England, 10 is Southern New England,
. and 1E Is Maine.
" Variable labels and units are found in Table 3-60.
e Numbers indicate order of entry into stepwise model. (-) indicates a negative parameter estimate.
8-163
-------
be expected, the correlations for sulfate and sulfur retention tend to be opposite in sign. A similar pattern
of relationships is apparent in the multiple regression results, with some variation among the subregions.
In the Poconos/Catskills (1B) and Southern New England (10), sulfate is positively correlated with soil
pH; In Central New England (1C) and in the region as a whole, however, the relationship with pH is
negative. In the Adirondacks (1A) and the Poconos/Catskills, the two subregions with highest sulfate
deposition, sulfate is correlated with the half-saturation concentration, a sulfate isotherm intensity factor
that is highly correlated with the concentration at zero net adsorption. This relationship is consistent with
the bivariate correlations. In the Poconos/Catskills, Central New England, and Southern New England,
sulfate is correlated with extractable aluminum. This Is consistent with the hypothesis that previously
adsorbed sulfate may be desorbing from these soils. This scenario also is supported by the sulfur
retention regressions; in most subregions and the NE overall, the greater the extractable aluminum in the
soil, the lower the net retention.
8.9.5.1.2 Combined buffer aggregation -
Van'ables selected by the stepwise regressions for the northeastern watersheds aggregated for the
combined buffers around the lakes and streams were the same as or similar to those selected for the
whole watershed aggregation. The R2 for sulfate improved significantly, with fewer variables in the
model, but the R2 for the sulfur retention model is lower. Buffer zone models were not run for the
northeastern subregions. These results alone do not allow a conclusion to be drawn regarding the
relative merits of each aggregation for these analyses.
8.9.5.2 Southern Blue Ridge Province
8.9.5.2.1 Whole watershed aggregation -
Exchangeable magnesium and maximum sulfate adsorption capacity are most strongly related to
sulfate and sulfur retention in the SBRP in the multiple linear regressions (Tables 8-61 and 8-62). R2 for
sulfate is 0.66, and 0.44 for sulfur retention. Higher exchangeable magnesium (and calcium, in the
bivariate correlations) in the soil is correlated with higher sulfur in the water and lower sulfate retention
by the soil; higher sulfate adsorption capacity is correlated with lower sulfate in the water and higher
8-164
-------
retention in the soil. Higher water-extractable sulfate in the soil is correlated with higher water suifate.
Higher base saturation soils are correlated with greater sulfur retention in the soil (though this may be
spurious because there is no bivariate correlation between these variables). In this region where the soils
have not yet reached equilibrium with atmospheric sulfur deposition, the adsorption capacity of the soil
Is a good explanatory variable of both retention and concentration in the drainage water. Water-
extractable sulfate is a readily mobilized pod of sulfate, acting in the SBRP as the soil intensity variable
associated with sulfate in the water. The reason for the strong relationship between exchangeable
magnesium (and calcium) and sulfate is possibly due to higher base status soils generally having higher
pH and hence lower sulfate adsorption, although there is no correlation between base saturation or pH
and sulfate. Another possibility would be a sulfur-rich bedrock source that is weathering both bases and
sulfur. This is supported by the correlation between the low organic meta-sedimentary MSI sampling
class soils and stream sulfate (Section 8.6.3.2).
8.9.5.2.2 100-m buffer aggregation -
There is virtually no difference in the models selected for the 100-m buffer aggregation from those
for the whole watershed in the SBRP. This suggests that estimated chemistry for the soils at the stream
sides is not more strongly associated with spring baseflow chemistry than those in the whole watershed.
Stream chemistry measured during stormflow, a time when the near-stream soils would be expected to
be more hydrologically active, might be more strongly associated with 100-m buffer soil chemistry.
8.9.6 Ca Plus Ma fSOBCl. ANC. and oH
Results of stepwise multiple regression for Ca plus Mg concentrations (sum of base cations in the
SBRP), ANC, and pH are given in Tables 8-63 through 8-65. This section summarizes the results and
discusses potential cause-effect controls on surface water chemistry. The dependent water chemistry
variables are all highly correlated with each other and therefore have very similar associations with soil
physical and chemical properties. Of the four, pH is the most dissimilar because of its nonlinear
relationship with ANC (Figure 5-9).
8-165
-------
Table 8-63. Results of Stepwise Multiple Regressions for DDRP Lake Calcium plus Magnesium
Concentrations (CAMG16) and Stream Sum of Base Cation Concentrations (SOBC) Versus Soil
Physical and Chemical Properties
Variable53
Whole Watersheds
Subregiorf
Region
Buffer Zone
Region
1Ae 1B 1C 10 1E NE SBRP NE SBRP
SAND
CLAY
FRAG
THKA
SOILDEN
CACL
MG CL
SBC_CL
BS_CL
CEC_CL
AC BACL
PH~01M
ALjAO
ALCD
ALJPYP
ALPOT
C_TOT
SO4 H2O
SO4~PO4
SO4~EMX
SO4~~B2
SO4~XIN
SO4 SLP
6
2
1
20
1
3
1
5
4
2
2
1
4(-)
0.71 0.57 0.49 0.81 0.59 0.40 0.44 0.38 0.48
* 1A is the Adirondacks, 1B Is the Poconos/Catskilla, 1C is Central New England, 10 is Southern New England,
h and 1E Is Maine.
b Variable labels and units are found in Table 8-60.
e Numbers indicate order of entry into stepwise model. (-) indicates a negative parameter estimate.
8*166
-------
Table 8-64. Results of Stepwise Multiple Regressions for DDRP Lake and Stream ANC
(ALKANEW and ALKA11) Versus Soil Physical and Chemical Properties
Variable"
SAND
CLAY
FRAG
THKA
SOILDEN
CA CL
MG CL
SBCCL
BS CL
CEC CL
AC SACL
PH'OIM
AL~AO
AL"CD
AL'PYP
ALPOT
C TOT
S04 H2O
S04 P04
SO4~EMX
SO4 B2
SO4~XIN
SO4~SLP
Whole Watersheds Buffer Zone
Subregion* Region Region
1AC 1B 1C 1D 1E NE SBRP NE SBRP
3<-)
1
3(-) 3(-) 5(-)
3
1 1 1
2 5
4
211 2 2
10
4 5(-)
2
4(-)
1(-)
3W
5(.) 3(-)
4(-)
1 2(-) 3
IT 0.75 0.62 0.53 0.83 0.47 0.43 0.44 0.36 0.47
* 1A is the AdirortdacKs, 1B is the Poconos/Catskilla, 1C is Central New England, 10 is Southern New England,
. and 1E is Maine.
6 Variable labels and units are found in Table 8-60.
c Numbers indicate order of entry into stepwise model. (•} indicates a negative parameter estimate.
8-167
-------
Table 8-65. Results of Stepwise Multiple Regressions for DDRP Lake and Stream pH
(PHEQ11) Versus Soil Physical and Chemical Properties
Variable"
SAND
CLAY
FRAG
THKA
SOILDEN
CACL
MG CL
SBC CL
BS CL
CEC CL
AC BACL
PH~01M
AL~AO
AL~CD
AL PYP
ALPOT
C TOT
SO4 H2O
SO4~PO4
SO4~EMX
SO4~B2
SO4~XIN
SO4~SLP
Whole Watersheds Buffer Zone
Subregion* Region Region
1AC 1B 1C 1D 1E NE SBRP NE SBRP
2(-)
20
4 2
3(-) 2
1 1
3(-)
11 1 1 1
6(-) 2 2(-)
4
3 . 3
5
10
3 3(-)
2(-) 1(->
4 2(-)
2 3{-)
Rz
0.33 0.84 0.64 0.46 0.71 0.31 0.45 0.30 0.48
* 1A is the Adirondacks, 1B is the Poconos/Catskills, 1C is Central New England, 1D is Southern New England,
. and 16 is Maine.
6 Variable labels and units are found in Table 8-60.
0 Numbers indicate order of entry into stepwise model. (-) indicates a negative parameter estimate.
8-168
-------
8.9.6.1 Northeast
8.9.6.1.1 Whole watershed aggregation -
The R2 values range from 0.31 to 0.84 for the northeastern subregions. Soil pH is most commonly
and most strongly associated with the water chemistry in most northeastern subregions and in the region
overall. Exchangeable calcium, base saturation, and the sum of soil base cations are highly correlated
with each other and are also commonly associated with the water chemistry. In the Poconos/Catskills
Subregion, cation exchange capacity was selected by the stepwise regressions for ANC and Ca plus Mg;
this region has the highest mean base saturation, in Southern New England, cation exchange capacity
was selected for lake pH, but with a negative sign; that subregion has the lowest base saturation, i.e.,
it has a greater proportion of acidic cations on its exchange sites. The bivariate correlations (Table 8-
59) also match the pattern seen in the multiple regressions; heavier, clay-rich soils high in exchangeable
bases, pH, and base saturation are strongly correlated with higher bases, ANC, and pH in the water
These relationships lend support to the hypothesis that exchangeable bases in soils are important controls
on the base cation supply to, and ANC of, surface waters.
A group of soil suifate-related variables also is correlated with base cations, ANC, and pH of the
northeastern DDRP lakes. The variables include intensity and capacity isotherm parameters, water and
phosphate extractable suifate, the different forms of extractable aluminum, and possibly exchangeable
magnesium. The variables appear in different combinations and with different signs in the regressions
for the different subregions. The bivariate correlations (Table 8-59) show strong positive correlations with
suifate concentration at zero net adsorption; i.e., high equilibrium suifate concentration (which is correlated
with suifate deposition) is associated with high base cation supply for the region overall. The suifate
isotherm variables SO4JEMX and SO4_SLP are correlated with low base cation supply, ANC, and pH.
In the regressions for the northeastern region as a whole, most of the suifate-related chemical parameters
are replaced by soil texture variables. The sandier soils are associated with lower base cations, ANC,
and pH in surface water; the soils with higher day content are associated with higher base cations, ANC,
and pH in the water. Further work is needed to detail possible mechanisms and subregional differences
in these relationships.
8-169
-------
8.9.6.1.2 Combined buffer aggregation -
The variables selected by the stepwise regressions for the buffer zone aggregation are more similar
to the variables selected for the subregion models than to those selected by the models for the whole
region. However, they nave slightly lower R2 values than the whole-watershed models, and much lower
R2 values than the subregion models. Buffer zone models were not run for the subregions. These limited
results suggest that the buffer zone aggregation does not help in explaining variability in the surface water
chemistry.
8.9.6.2 Southern Blue Ridge Province
8.9.6.2.1 Whole watershed aggregation -
Base saturation is most highly associated with these SBRP stream chemistry variables. Other
associated variables for pH include aluminum potential and the isotherm half-saturation constant R2
values range from 0.44 to 0.45, slightly higher than for the NE. The bivariate. correlations are consistent
with the multiple regression results. As for the NE, these relationships support the hypotheses that
exchangeable bases and mobile sulfate are important regulators of surface water chemistry.
8.9.6.2.2 100-m buffer aggregation -
The models selected for the buffer zone aggregation are very similar to those for the entire
watershed aggregation.
8.9.7 Evaluation of Alternative Aggregation Schemes
In order to examine the effect of the different aggregation schemes on the Level I Analysis results,
we ran several regressions using soil chemistry variables from the unadjusted aggregation scheme. The
results of these regressions are compared With the results from the watershed adjusted data in Tables
8-66 and 8-67. Prior to regression analysis, a collinearity analysis was conducted. Variables dropped
as a result of this analysis are marked by X's in the tables. The remaining variables were used in
stepwise regressions with ANC and sulfate as the response variables.
8-170
-------
Examination of Tables 8-66 and 8-67 shows that many more candidate explanatory variables had
to be dropped from the unadjusted data than from the adjusted data. There were fewer instances of
multi-collinearity when using the watershed aggregation. Second, the regression models based on the
adjusted data generally explained more variance in the response variables than did the models based on
the unadjusted data. The only exception to this result is SO416 in the SBRP. The adjustment for
watershed effect generally appears to increase the explanatory power of the soil chemistry variables.
8.9.8 Summary and Conclusions
8.9.8.1 Alternative Aggregation Schemes
The DORP soil sampling and common aggregation scheme (unadjusted data) probably characterizes
regional and subregional means of soil properties well. The common aggregation scheme appears to
have limitations, however, in characterizing the regional distribution of soil properties or the soil properties
of Individual watersheds. The common aggregation scheme biases individual watershed values toward
the regional mean value. An alternative aggregation approach that uses a regression model to adjust for
watershed effects appears to adjust the problem of bias toward the regional mean but adds additional
uncertainty to the estimates of watershed soil chemistry.
The common aggregation scheme was used for most Level II and III modeling because it was
the only data available at the time. The correlations and regressions conducted here used the watershed-
effects-adjusted data because they have the most explanatory power for surface water chemistry.
Additional field work would be needed to assess which aggregation scheme most closely mimics reality.
The characteristics of each aggregation scheme must be kept in mind when interpreting the results of
the models.
Although the buffer zone and whole-watershed aggregation schemes do result in slightly different
values for some of the soil physical and chemical variables, most differences are probably not significant.
The buffer zone aggregation does not result in improved regression relationships for either the NE or
SBRP, thus the advantage of using one aggregation scheme over the other for explaining index chemistry
8-171
-------
Table 8-66. Results of Stepwise Multiple Regressions for DDRP Lake and Stream ANC (ALKANEW
and ALKA11) Versus Unadjusted and Watershed Adjusted Soil Physical and Chemical Properties
Variable"
NEb
SBRP"
Unadjusted Adjusted
Unadjusted
Adjusted
SAND
CLAY
FRAG
THKA
SOIIDEN
CACL
MG CL
SBC CL
BS CL
CECCL
AC BACL
PH~01M
AL AO
AL'CD
AL PYP
ALPOT
CTOT
SO4 H2O
SO4 PO4
SO4~EMX
SO4 B2
SO4 XIN
SO4 SLP
X 3(-)
1
4
5(.)
1
2(-)
X
2
X
X
X
X
X
X
X
X
X
X
X
1
X
X
X
X
X
X
X
X
2
1
X
X
0.33
0.43
0.29
0.44
* Variable labels and units are found in Table 8-60.
X*s indicate variables dropped In collinearity analysis. Numbers indicate order of entry into stepwise model. (-) Indicates
a negative parameter estimate.
8-172
-------
Table 8-67. Results of Stepwise Multiple Regressions for DORP Lake and Stream Sulfate (S0416)
Versus Unadjusted and Watershed Adjusted Soil Physical and Chemical Properties
NEb SBRP"
Variable* Unadjusted Adjusted Unadjusted Adjusted
SAND
CLAY
FRAG
THKA
SOILDEN
CACL
MG CL
SBC CL
BS CL
CEC CL
AC BACL
PH 01 M
AL'AO
AL~CD
AL'PYP
ALPOT
C TOT
SO4 H2O
SO4 PO4
SO4 EMX
SO4~B2
SO4 XIN
SO4~SLP
X
4(") 5
s(-)
3
X
2(.)
X
4
X 6
X
X
3(-)
2
1
1
X
X
X
X
X
X
X 1
X X
X
X
X
X
X
3
X
K-) 2(-)
X
2 X
R2 0.47 0.27 0.32 0.62
* Variable labels and units are found In Table 8-60.
X'a indicate variables dropped in collinearity analysis. Numbers indicate order of entry into stepwise model. (-) indicates
a negative parameter estimate.
8-173
-------
is unclear. The buffer zone aggregation was hypothesized to be more representative because it implicitly
weighted watershed values to take into account convergent flow and last hydrologic contact with the soil.
However, from our analyses, the importance of these characteristics appear to be minor for explaining
index water chemistry. This may be due to insufficient characterization of the buffer zones. Only soil
mapping units greater than 6-10 acres were mapped; the effective buffer zones may be much smaller in
size. A more thorough soil characterization and evaluation of watershed hydrology is necessary before
the importance of buffer zones in controlling stream chemistry can be determined.
8.9.8.2 Sulfate and Sulfur Retention
The regression analyses indicated that the sulfate isotherm parameters are strongly related to
surface water sulfate. In the NE the important parameters are the equilibrium sulfate concentration and
the half saturation constant This is consistent with the hypothesis that northeastern soils are near steady
state with respect to sulfate adsorption. In the SBRP the adsorption asymptote and the extractable sulfate
are important explanatory variables for stream sulfate concentration. These variables indicate soils that
are actively adsorbing sulfate. Too much emphasis should not be placed on which particular isotherm
parameters are selected in the regressions, since the isotherm parameters are themselves strongly
correlated. It is significant, however, that the isotherm parameters are selected in both the NE and SBRP.
Even in a region near steady state, the sulfate isotherm parameters yield information about concentrations
of sulfate in the surface waters.
Variables relating to soil acidity and base status are also important but do not enter the regression
models for the regions and subreglons in a consistent manner. The relationship of surface water sulfate
concentration and soil pH varies among the subregions of the NE and is not statistically significant for
the SBRP watersheds. In general in the NE, high concentrations of sulfate in surface waters are
associated with low pH and high extractable aluminum concentrations in the soils. In the SBRP, high
sulfate concentrations are associated with high concentrations of base cations in the soils. The fact that
the two regions are approaching soil sulfate equilibrium from different directions (declining deposition and
8-174
-------
desorption in the NE versus adsorption in the SBRP) may be responsible for the variability observed in
the soil chemical relationships.
In general, the same soil variables that are important In predicting sulfate concentration are
important in the regressions for sulfur net retention, but the coefficients of the variables have opposite
signs. Values of sulfur retention are significantly higher in the SBRP then in the NE. This is consistent
with the lower observed equilibrium sulfate concentration (Table 8-58).
8.9.8.3 Ca plus Mg (SOBC), ANC, and pH
Soil pH, exchangeable base cations, and texture are most strongly related to lake ANC, pH, and
base cation concentrations in the NE. Soil base saturation has the strongest relationships in the SBRP.
The sulfate isotherm parameters are more common as explanatory variables in the NE than the SBRP.
This is consistent with the mobile anion hypothesis. The drainage water sulfate concentration, and
therefore the sulfate isotherm parameters, is relatively less important in controlling ANC in the SBRP than
in the NE.
Mean concentrations of Ca plus Mg are significantly higher in the NE than the SBRP, as expected
since soils in the SBRP are older and more highly weathered. Northeastern soils also have a higher base
saturation on average than those in the SBRP. The regressions and measured soil and surface water
attributes support the hypothesis that soil base cation availability has a stronger effect on surface water
ANC than other soil chemical properties.
8.9.9 Summary Conclusions
• Soil variables important in explaining surface water sulfate and watershed sulfur
retention include soil sulfate concentration and adsorption capacity, extractable
aluminum, soil pH, and texture.
* Soil variables important in explaining ANC, pH, and Ca plus Mg (sum of base
cations) in surface water include soil base saturation, pH, soil sulfate
concentration, and texture.
8-17S
-------
• Using a multiple regression modelling approach, measured soil chemical and
physical properties alone can account for one quarter to three quarters of the
variance in ANC, sulfate, and base cations in the lake and stream waters of the
DORP regions and subregions.
• The DDRP soils data aggregation scheme using soil sampling classes masks a
significant watershed effect. The aggregation scheme probably accurately
characterizes regional mean values, but it draws all data toward the mean, and
may affect the distribution of modelling results. Those results also will be drawn
toward the mean, underestimating the possible response of watersheds having
soil characteristics at the sensitive end of the distribution.
• Aggregating soils by buffer zones near the lakes and streams does not generally
result in better correlations with index values of surface water chemistry.
Stronger associations would likely be observed between buffer zone soil
characteristics and stormflow chemistry, when those soils are more hydrologically
active.
8.10 EVALUATION OF ASSOCIATIONS BETWEEN WATERSHED ATTRIBUTES AND SURFACE
WATER CHEMISTRY
8.10.1 Introduction
This section evaluates the relationships between surface water chemistry and all of the watershed
attributes measured in the DDRP. Many watershed characteristics have been shown to explain a
significant portion of the variance in surface water chemistry when considered individually or in groups
of related variables (see Sections 8.2 through 8.9). The analyses in this section are designed to integrate
and evaluate the various watershed attributes in explaining the variability in surface water chemistry. The
results are important in assessing whether the DDRP Levels II and Hi modelling efforts are considering
the variables most important in controlling surface water chemistry.
8.10.2 Approach
The candidate explanatory variables considered in this section include soil physical and chemical
properties, climate and deposition, geology, hydrology, physiography, vegetation, and land use
characteristics. Two basic categories of watershed attributes were used: average attribute values for a
watershed and areal proportions of a watershed meeting specified criteria. Average attributes for each
watershed include means for depth to bedrock, soil base saturation, soil permeability, deposition,
precipitation, and runoff values, among others. Mean watershed attributes were calculated by averaging
8-176
-------
the values associated with mapped areas on a watershed and weighting by the area! fraction of the
mapped area. An overview of the procedure for aggregating soil variables is given in Section 8.9.3, and
a description for the other watershed attributes can be found in Turner et al. (1989).
Although average values provide an integrated estimate of an attribute at the watershed level, such
values do not provide much information about the distribution of an attribute on a watershed.
Furthermore, mean values cannot be calculated for many attributes such as vegetation cover type or
geomorphic position. Therefore a second category of attributes was developed in order to estimate the
proportion of watersheds meeting specified criteria. Watershed proportions were derived from the
mapped data by summing the areal percentages of those mapping units on each watershed that satisfy
the specified criterion.
Summary statistics for the subset of watershed attributes that were used in these analyses are
given in Tables 8-58 and 8-68. Data derived from field mapping activities are described in Section
5.4.1.3, and the land use/wetland data obtained from photointerpretation are explained in Section 5.4.1.6.
Deposition, precipitation, and runoff data were obtained as described in Sections 5.6 and 5.7. The
regression modelling approach described in Section 8.9.2.2 was also used here.
8.10.3 Regional Characterization of Watershed Attributes
Characteristics of the sampled watersheds differ among the five subregions and between the two
study regions. The characteristics can be grouped into four categories: climate/deposition variables,
geologic parameters, hydrologic/physiographic descriptors, and land use/vegetation variables. Means
and standard errors for these means are presented for each of the northeastern subregions, for the NE
as a whole, and for the SBRP in Table 8-68.
8.10.3.1 Northeast Subregions
Long-term atmospheric deposition of ions varies among the five subregions, despite approximately
equal precipitation amounts. Sodium and chloride deposition are highest in the Southern New England
8-177
-------
I
Table 8-68. Population Means and Standard Errors for Selected Variables, by Subregion/Region
and Aggregation
Mean ± Standard Error by DDRP Subregion/Region
Variable/
A0Qiv0atfon
NE Subregion*
»> M
ID
1C
10
IE
Overall
NE
SBW
Deposition/Climate
CA..LTO c
WS
M3 LTD
~W8
HA LTD
VMS
KUD
" WS
CM LTD
WS
NH4LTD
WS
HUTD
WS
S04LTD
WS
NO3 LTD
"ws
O.J.TO
WS
PBECL
WS
RNOFT
WS
TMP AVQ
WS
CCMSTO
WS
0.15 ± 0.01
0.07 ± 0.00
0.07 * 0.01
0.03 ±0.00
0.22 ± 0.01
0.25 t 0.01
0.77 x 0.01
0.73 * 0.02
0.96 * 0.01
0.09 i 0.00
108.30 * 1.27
70.28 x 1.BO
4.84 ± 0,14
287.33 X 8.25
0.10 * 0.01
0.06 ± 0.01
0.10 ± 0.01
O.02 X 0.00
0.16 ± 0.01
0.21 ± 0.01
1.14 ± 0.02
0.87 * 0.02
0.80 ± 0.01
0.10 i 0.01
111.32 ± 1.28
53.89 X 1.68
8.29 x 0.27
128.S7 ± 1 1.81
0.08 ± 0,01
0.08 * 0.01
0.14 * 0.02
0.02 x 0.00
0.16 s 0.01
0.1S ± 0.01
0.68 ± 0.01
0,81 ± 0.02
0.41 ± 0.01
0.08 X 0.01
108.09 X 1.39
63.00 * 1.78
6.49 X 0.28
100.82 ± 6.87
0.10 t 0.01
0.19 ± 0.01
0.49 * 0,07
O.OZ iO.OO
0.28 ±0.02
0.14 * 0.01
0.72 I 0.02
0.71 £ 0.02
0.43 t a02
0.4B c 0.09
117.78 * 1.31
6Z03 * 1.68
8.18 * 0.28
22.54 ±9,03
O.OS * 0.01
0.10 1 0.01
0,28 * 0.03
O.O2 A 0.00
0.17 4 0.01
0.11 * 0.01
0.42 t 0.01
0.48 * 0.02
0.28 * 0.01
0.18 * 0.02
110.30 X 1.61
69.69 ± 1.19
5,82 * 0.22
89.80 * 13.74
0.10 i 0.00
0.08 * 0.00
0.20 ± 0.01
0.02 ± 0.00
0.18 ± 0.00
0.17 ± 0.00
0.72 x 0.01
0.68 ±0.01
0.45 ± 0.01
0.18 ± 0.01
110.89 £ 0.65
84.71 ± 0.88
6.68 ± 0.17
128.63 ± 9.58
0.20 £ 0.02
0.09 ± 0.01
0.14 ± 0.01
0.03 x 0.00
0.29 ±002
0.22 X 0.01
0.68 * 0.03
0.84 £ 0.03
0.43 * 0.01
0.12 * 0,01
145.40 t 1.19
82.08 ± 3.75
13.19 ± 0.28
. * .
Qoology ,
QEOSiN
WS
QEOMAX
WS
SEOQT4
WS
Z71 ± 0.13
3. S3 _f 0.23
0.82 ± 0.81
2.32 * 0.1S
2,70 t 0.28
4.26*3.55
2.70 i 0.14
3.9S l 0.27
3.98 * 1.86
2.13 * 0.08
S.B7 ± 0,27
1.44 ± 1.18
2.80 i 0.24
3.74 ± 0.34
20.33 * 8.21
2.SS ± 0.08
3.40 ± 0.14
687 X 1.80
2.13 X 0.12
Z8S ± 0.29
.22 ± 0.32
Photography
EL.MIN
WS
MAXRB.
WS
SLP
WS
BUF
530.33 ± 19.74
155.43 x 19.82
20.02 x O.S1
12.17 ± 0.69
388.32 ± 24.88
86.37 ± 18.88
8.88 X 0.89
7.65 X 0.84
294.87 £26.02
220.87 425.11
14.38 s 0,03
8.19 s O.A2
88.09*22.25
4Z38 ± 8.16
10.98 ± 1.02
9.67 * 1.13
160.08 * 28.60
124,83 * 15.41
13,01 X 1.35
9.32 X 1.04
301.85 ± 17.98
134.80 ± 9.73
13.88 ± 0.57
9.68 a 0.45
5B8.4B ± 37.48
537.01 x 93.21
37.75 x 3.40
34.84 ± Z8B
* 1A is the Adirondacks, 1B is the Poeonos/Catskifls, tC is Central New England, 1D is Southern New England,
and 1E is Maine.
Variable labels and units are found in Table 8-70.
e For each variable, WS refers to the entire watershed and 8UF refers to the buffer zone.
continued
8-178
-------
Table 8-68. (Continued)
Mean
± Standard Error by DDRP Subregion/Region
NE Subregion
Variable/
AggmgXion
ATNMEAN
WS
ATKBMEAN
WS
GMPFTN
WS
BUF
LOW
WS
BUF
1A
7.90 ± 0.12
0.81 ± 0.13
20.31 ± X10
48.29 ±4.89
5.85 ± 1.10
17.1S ± 3.30
18
8.47 ± 0.14
2.40 ±0.11
19.87 ± 3.08
40.15 ± 5.44
8.61 ± 1.49
19.02 £ 4.14
1C
8.31 ± 0.09
0.77 ± 0.11
28.28 ± 4.00
52.05 ± 4.88
8.88 ± 1.88
19.34 ±3.11
10
8.40 ± 0.14
-0.82 * 0.37
24.00 ± 5.21
38.38 ± 6.32
10.54 ± 2.19
17.40 ± 3.85
IE
8.29 ±0.16
1.47 ± 0.19
31.99 ± 3.98
52.94 ±4.88
5.38 ± 0.94
15.82 ± 2.28
Overall
NE
8.26 ± 0.08
1.03 ± 0.12
24.92 ± 1.79
46.58 ± 2.41
6.72 a 0.67
17.68 * 1.46
QnDp
oonr
7,81 ± 0.09
. ± .
3.17 ± 1.40
9.20 ± 3.34
Z33 ± 0.71
11.74 ± aso
Hydrology
HYD SLW
~WS
BUF
DRNSLW
"ws
BUF
PERM
WS
BUF
PRMSLW
WS
BUF
DEPTH
WS
BUF
BRDSHL
WS
BUF
IPOSHL
"ws
BUF
AREA.TER
WS
AREAH20
WS
WALA
WS
VOL
WS
DOENSfTY
WS
STRORDER
WS
FOREST
WS
BUF
CULTW
WS
BUF
PASTURE
WS
BUF
DISTURB
WS
BUF
WETLAND
WS
BUF
83.79 ± 3.31
70.48 ± 4.48
12.06 ± 1.79
27.41 ± 3.29
8.87 * 1.07
7.79 ± 1.6B
33.57 ± 3.71
39.08 ± 5.48
3.08 ±0.78
4.73 ±0.93
48.48 ± 3.98
25.88 ± 4.48
32.41 * Z85
19.33 ± 2.88
358.54 ± 87.42
48.31 ± 12.78
19.38 ± 4,99
Z(S ± 1.33
0.48 ± 0.11
1.92 ± 0.08
98.13 ± 0.88
. ± .
0.00 ± 0.00
. ± .
0.11 ± 0.08
. ± .
0.41 ± 0.21
. ± .
3.35 ± 0.70
. ± .
90.87 ± 1.78
92.87 ±2.08
42.82 ± 4.99
81.80 ± 4.43
Z36 ± 0.52
1.81 ± 0.44
81.57 ± 5.94
70.80 * 6.00
1.74 ± 0.30
2.23 ±0.35
33.15 ± 4.57
18.84 * 3,08
48.30 ± 4.18
52.03 ± 5.17
329.22 ± 78.84
28.87 ± 8.12
18.64 * 5.75
0.87 ± 0.25
0.69 t 0.18
£89*0.07
75.88 ± 4.71
. ± .
1.08 ± 0.68
. * .
11.62 * 4,24
. ± ,
8.48 ± 1.61
. * .
4.92 ± 1.11
. ± .
69.45 ± 3.66
68.BS * 4.59
23.31 ± 3.34
43.28 ± 3.95
9.22 ± 0.71
8.19 ± 1.70
46.41 ± 3.50
42.53 ± 4.50
4.68 ± 0.87
7.18 ± 1.19
29.39 ±2.32
1224 ± 2.31
22.83 ± 3.71
15.40 ± 3.03
32.87 ± 6.80
33.25 ± 6.69
11.75 ±2.28
19.40 X 3.46
21.74 ± 3.62
22.63 ± 3.34
13.33 ± 5.44
10.34 ± 3.98
15.44 ± 2.12
17.19 ± 1.97
13.22 ± 4.30
8.18 "± 121
8.98 ± 2.80
&13 ± 2.03
671.77 ± 128.58 190.82 ± 34.91
43.78 ± 7.11
28.75 ±4.85
1.39 ± 0.58
0.88 ± 0.18
3.00 ± 0.00
Urtd
91.41 ± 1.48
. t .
0.44 ± 0.17
. ± .
2.87 ± 0.72
. ± .
0,57 ± 0.15
. ± .
4,92 ± 1.23
. ± .
31.87 ± 7.98
8.68 ± 1.47
0.74 * 0.19
0.32 ± 0.12
3.00 ± 0.00
Use/Vegetation
75.69 ± 3.90
. ± .
2.55 ± 1.41
. * .
230 ± 1.42
. ± .
13.99 ± 3.54
. ± .
5.51 ± 1.26
. ± .
73.20 ± 4.95
73.58 ± 5.23
28.00 t 3.59
43.75 ± 4.69
5.81 * 1.21
6.59 ± 1.58
43.71 ± 4.06
44.82 ± 4.86
3.17 ± 0.82
4.55 ± 0.81
33.55 ± 4.83
21.09 ± 4.48
33.21 ± 4.82
28.75 ± 4.21
881.50 ± 138.68
95.61 ± 23.33
19.26 ± 2.29
4.18 ± 1.81
a71 ± 0.14
3.68 ± 0.09
89.37 ± 1.78
. ± .
3.38 ± 1.00
. ± .
t.72 ± 0.65
. ± .
0.80 ± 0,23
. ± .
4.73 ± ago
. ± .
67.01 ± 2.40
69.76 ± Z69
24.04 ± 1.81
40.09 * 2.28
7.52 * 0.87
8.58 ± 1.00
41.18 ± 2.38
42.99 ± 2.76
4,98 ± 0.98
6.52 * 0,89
31.91 ± 2,09
17.90 ± 1.81
29.73 ± 2.00
24.56 ± 2,14
478.84 ± 51.34
52.00 * 7.39
18.82 * 2.02
2.03 * 0.98
0.83 * 0.07
2.92 t 0.06
68,79 ± 1.36
. ± .
1.48 f 0.38
. ± .
3.51 t 0.87
. ± .
3.59 * 0.74
. ± .
4.63 t 0.47
. * .
5.32 ± 1.76
4.85 ± 1.36
1.69 ± 0.60
7.87 ± 2.2B
5.08 ± 0.45
5.16 ± 0.34
0.22 ± 0.16
0.69 ± 0.43
1.59 ± 0.13
£01 ± 0.15
9.43 ± 2.26
8.89 ± £03
. ± .
. ± .
966.91 ± 213.37
0.58 ± 0.39
. * .
. ± .
1.03 ± 0.17
2.23 ± 0.21
90.20 ± 100
84.44 ± 4.51
1.77 ± ZOS
3.19 ± 2.73
6.84 ± 1.93
10.50 ± 3.29
1.10 ± 0.68
1.57 ± 0.69
0.03 ± 0.07
0.31 * 0.55
8-179
-------
Table 8-68. (Continued)
Mean ± Standard Error by DDRP Subregion/Region
Variable/
Aggregation
VGTCNF
WS
euF
WSTDCD
WS
BUF
VQTDRT
WS
BUF
V«T_WEr
WS
BUF
NE Subregion
1A
16.00 £ Z51
31.48 £ 4.52
72.801 3.49
44.13 ± 4.94
0.48 ± 0.27
0.48 ± 0.37
1.60 ± 0.57
7.88 ± 2.45
IS
6.18 ± 1.48
11.55 ± 3.05
73.87 ± 4.95
68.25 ± 4.44
18.96 * 4.88
15.44 ± 3.91
2.X ± 1.18
8.29 ± 2.87
1C
19.19 £ 3.17
30.33 £ 5.18
45.21 ± 5.74
28.50 ±5.42
4.28 ± 1.15
4.30 ± 1.58
Z.4B ±0.89
a57 ± 2.30
10
27.28 ± 6.48
29.31 * 8.28
44,34 ± 7.89
42.93 ± 7.87
9.80 t 3.80
8.82 ± 1.84
4.27 £ 1.28
6.71 £ 1.74
1H
32.81 £ 4.93
43.69 £ 6.01
33.21 £ 4.92
24.31 £ 4.68
7.84 ± 1.82
4.04 ± 1.13
2.20 ± 0.59
7.08 * 1.71
Overall
ME
20.56 ± 1.97
30.19 ± 254
53.29 ± 2.82
39.81 ± 2.78
7.48 ± 1.22
5.84 * 0.96
2.50 ± 0.37
7.37 £ 1.01
SBRP
4,71 * 1.55
4,71 ± 204
38,75 ± 7.69
37.07 ± 5,71
9.49 ± 3.00
14.94 ± 4.53
0.03 £ 0.05
0.12 ± 0.19
8-180
-------
subregion, probably due to sea-salt deposition. The highest levels of calcium and magnesium deposition
are found in the Adirondacks and Southern New England, and deposition of hydrogen ions and sulfate
is highest in the Poconos/Catskills Subregion.
Although differences in bedrock sensitivity (GEO_SEN), expressed using the DDRP weatherability
index (see Section 8.4), are not large, in the Maine Subregion nearly 20 percent of the watershed area
is of sensitivity class > 4 (GEO_GT4). This is larger than for any other subregion.
Elevation of the sample lakes (EL_M1N) is highest in the Adirondacks and lowest in Southern New
England. Maximum relief (MAXREL) is highest in the Central New England Subregion and lowest in
Southern New England. The percentage of watershed area in foot or toe slope (GMP_FTN) is highest
in Maine. The highest percentage of land with poor drainage and permeability characteristics (HYD_SLW,
DRN_SLW, PERM_SLW) is consistently found in the Poconos/Catskills Subregion and the lowest in the
Southern New England Subregion. Soils in Southern New England have greatest mean depth to bedrock
(DEPTH), relatively few shallow (< 50 cm) impermeable layers (IPD_SHL), and are generally coarser-
textured than soils on watersheds in other subregions (see Section 8.9.4).
Vegetation and land use characteristics vary somewhat among the subregions. No appreciable
area of cultivated land occurs in the Adirondacks watersheds. The Southern New England watersheds
have the largest proportion of urban or disturbed land. Watersheds of the Southern New England and
Maine Subregions contain the most coniferous cover, and deciduous forest coverage is greatest in the
Adirondacks and Poconos/Catskills watersheds. The percent of open, dry vegetation (open, non-forested
land that is not wetlands) is greatest in the Poconos/Catskills Subregion; the Adirondacks Subregion has
almost no area of open, dry vegetation. The open, dry vegetation class often indicates pasture or
abandoned farm land.
8-181
-------
8.10.3.2 Northeast and Southern Blue Ridge Providence
Current atmospheric deposition is higher in the SBRP than in the NE. The mean annual
temperature in the SBRP is nearly twice that of the NE. Northeastern watersheds generally contain
more weatherable bedrock than the SBRP. Elevation, maximum relief, and slope are all higher in the
SBRP. However, the minimum elevation in the SBRP is approximately equal to the minimum of the
Adirondacks subregion. The northeastern watersheds include a higher percentage of "wet" or poorly
drained soils (based on HYD_SLW, DRN_SLW, and PERM_SLW). Despite a greater mean depth of
bedrock, the NE has a higher percentage of watershed area overlying shallow (<50 cm) bedrock.
Watersheds are generally larger in the SBRP and contain larger percentages of area in pasture;
watersheds in the NE have a greater percentage of wetlands and more coniferous vegetation.
The buffer zones are characterized by lower slopes, a higher percentage of foot and toe slopes,
and a slightly higher percentage of land with "slow" drainage. For the Adirondacks and Poconos/
Catskills Subregions, the buffers also included more soils with hydrdogic group C or D and permeability
of .< 3. In general, soils in the buffer zones are slightly deeper, with a lower percentage of area with
shallow bedrock relative to the whole watershed. The buffer zones contain a higher percentage of
lowlands and more coniferous vegetation in the NE. In the NE the percent open, dry vegetation is slightly
less in the buffer zones than in the whole watersheds, although the percent open, wet vegetation is
higher. In the SBRP the percent area with open, dry vegetation is greater in the buffer zones than in the
whole watersheds.
8.10.4 Sulfate and Sulfur Retention
This section and Section 8.10.5 discuss the statistical relationships between selected watershed
attributes and the water chemistry for the DDRP regions. These relationships are evaluated in terms of
potential cause-effect controls on surface water chemistry and identification of any important controlling
factors which are not accounted for. Tables 8-69 and 8-70 show the nonparametric Spearman
correlations between selected watershed attributes and each of the water chemistry variables considered.
Correlations with soil properties were shown in Tables 8-59 and 8-60. Results of the stepwise multiple
regressions for sulfate and sulfur retention are given in Tables 8-71 and 8-72.
8-182
-------
Table 8-69. Non-parametric Correlations Between Lake Chemistry Variables and Selected Watershed
Attributes for the NE DORP Watersheds
Spearman Correlation Coefficients Significant at 0.05 / Significance Level / N = 143
Variable Units SO416* SO4 NRET CAMG16 ALKANEW
PHEQ11
Deposition/Climate
CA_LTD
Calcium deposition, long-term
MQ_LTD
Magnesium deposition, long-term
NAJ.TD
Sodium deposition,long-term
KJLTD
Potassium deposition, long-term
CM_LTD
Calcium* magnesium, long-term
NH4J.TD
Ammonium deposition, long-term
H_LTD
Hydrogen ion deposition, long-term
SO4J.TD
Sutfate deposition, long-term
NO3_LTD
Nitrate deposition, long-term
CL_LTD
Chloride deposition, long-term
PRECJ.
Precipitation, long-term
RNOFJT
Runoff, long-term
TMP_AVG
Avg. Temp., long-term
COASTO
Distance to coast
keq/ha
keq/ha
keq/ha
keq/ha
keq/ha
keq/ha
keq/ha
keq/ha
keq/ha
keq/ha
cm
cm
C
km
0.41920
0.0001
-0.24646
0.0030
0.17295
0.0389
0.49008
0.0001
0.55209
0.0001
0.57387
0.0001
0.53753
0.0001
' -0.32362
0.0001
0.24862
0.0028
0.21602
0.0096
•0.16242
0.0526
0.21190
0.0111
0.22209
0.0077
0.17202
0.0399
•0.30330
0.0002
0.29674
0.0003
-0.34929
0.0001
-0.29001 -0.22988
0.0004 0.0057
-0.26888 -0.26041
0.0012 0.0017
•0.33078 -0.41435
0.0001 0.0001
-0.22532
0.0068
•0.20025
0.0165
-0.26048
0.0017
-0.32169 -0.30643
0.0001 0.0002
-0.24870
0.0027
-0.34727
0.0001
-0.21387
0.0103
-0.24346
0.0034
-0.40138
0.0001
-0.23133
0.0054
-0.20518
0.0140
-0.26870
0.0012
-0.30034
0.0003
a SO416 is the lake sulfate concentration, SO4_NRET is watershed sulfur retention, CAMQ16 is the lake calcium
concentration, ALKANEW is the lake acid neutralizing capacity, and PHEQ11 is the air-equilibrated lake pH.
+ magnesium
continued
8-183
-------
Table 8-69. (Continued)
Spearman Correlation Coefficients Significant at 0.05 / Significance Level / N = 143
Variable
Units S0416 S04 NRET CAMG16
ALKANEW
PHEQ11
Geology
GEO_SEN
Geological weather, index, mean
G£O_MAX
Geological weather, index, max.
GEO_GT4
Geological weather. Index > 4
-0.17501 0.22226
0.0366 0.0076
-0.18158 0.27023
0.0300 0.0011
-0.21270 0.26025
percent 0.0108 0.0017
0.22715
0.0064
0.31602
0.0001
0.30655
0.0002
0.21625
0.0095
0.31118
0.0002
0.30405
0.0002
Physiography
ELAVQ
Elevation, average
MAXREL
Relief, maximum
SLP
Slope, mean
ATNMEAN
ln(a/tan 0), mean
ATKBMEAN
ln(a/kbtan 0), mean
GMP_FTN
Footeiope, toeslope, flood plain
LOW
Lowlands
0.27059
m 0.001 1
-0.22345
m 0.0075
-0.18579
percent 0.0263
0.27666
0.0008
0.29688
0.0004
•0.27776
percent 0.0008
percent
0,27479
0.0011
0.19420
0.0201
-0.16687
0.0464
0.26179
0.0019
0.18822
0.0244
'Hydrology
HYD_SLW
Hydrologic group C or D
DRN.SLW
Drainage c!ass<= 3
PERM
Permeability, mean
PRM_SLW
Permeability class <= 3
0.29712
percent 0.0003
0.16782 0.36531
percent 0.0451 0.0001
-0.34268
cm/hr 0.0001
0.32347
percent 0.0001
0.24800
0.0028
0.34550
0.0001
-0.33047
0.0001
0.39821
0.0001
0.22654
0.0065
0.33196
0.0001
-0.31449
0.0001
0.38357
0.0001
continued
8-184
-------
Table 8-69. (Continued)
Spearman Correlation Coefficients Significant at 0.05 / Significance Level / N » 143
Variable
DEPTH
Bedrock depth, mean
BRDJ.T2
Bedrock class <= 2 (100 cm)
BRD.SHL
Bedrock < = 50 cm
IPD.SHL
Impermeable layer <= 50 cm
AREA_TER
Area, Terrestrial
AREAJH20
Area, water
WALA
Watershed area/lake area
VOL
Lake volume
DDENSfTY
Drainage density
STRORDER
Stream order, maximum
Units
m
percent
percent
percent
ha
ha
m3
SO416 SO4 NRET
-0.18515
0.0268
-0.18181
0.0298
-0.18022
0.0313
-0.17159
0.0404
-0.18537
0.0267
0.18521
0.0273
-0.37781 0.16568
0.0001 0.0488
CAMG16
0.27970
0.0007
0.21169
0.0111
0.22011
0.0085
ALKANEW
0.21578
0.0096
0.26421
0.0014
0.20137
0.0159
0.22157
0.0081
0.35154
0.0001
PHEQ11
0.20136
0.0159
0.26093
0.0016
0.22224
0.0076
0.17854
0.0335
0.21157
0.0115
0.35029
0.0001
land Use/Vegetation
FOREST
Forested land
CULTW
Cultivated land
PASTURE
Pasture/grazed land
DISTURB
Disturbed land
WETLAND
Wetland
VGT_CNF
Vegetation, coniferous
VGTJDCD
Vegetation, deciduous
percent
percent
percent
percent
percent
percent
percent
•0.23145
0.0056
0.19619
0.0193
0.26987
0.0012
0.26558
0.0014
-0.28421
. 0.0006
0.33794
0.0001
4.32069
0.0001
0.20787
0.0131
0.39902
0.0001
0.33590
0.0001
-0.22830
0.0063
0.24694
0.0030
0.44050
0.0001
0.28701
0.0005
-0.20495
0.0144
0.24764
0.0030
0.42920
0.0001
0.28373
0.0006
continued
8-185
-------
Table 8-69. (Continued)
Spearman Correlation Coefficients Significant at 0.05 / Significance Level / N = 143
Variable Unite SO416 SO4_NRET CAMQ16 ALKANEW PHEQ11
VGTJ3RY 0.19488 . 0.38888 0.43434 0.41717
Vegetation, dry open percent 0.0197 0.0001 0.0001 0.0001
VQT_WET -0.19436 0.18205
Vegetation, wet open percent 0.0200 0.0295
8-186
-------
Table 8-70. Non-parametric Correlations Between Stream Chemistry Variables and Selected
Watershed Attributes for the SBRP DDRP Watersheds
Variable
Spearman Correlation Coefficients Significant at 0.05 / Significance Level / N = 31
Units SO416" SO4 NRET SOBC ALKA11
PHEQ11
Deposition/Climate
CAJ.TD
Calcium deposition, long-term
MGJ.TD
Magnesium deposition, long-term
NA.LTD
Sodium deposition, long-term
KJ.TD
Potassium deposition, long-term
CMJ.TD
Calcium + magnesium deposition
NH4_LTD
Ammonium deposition, long-term
H_LTD
Hydrogen ion deposition, long-term
SO4_LTD
Sulfate deposition, long-term
NO3_LTD
Nitrate deposition, long-term
CL_LTD
Chloride deposition, long-term
PRECJ.
Precipitation, long-term
RNOFJT
Runoff, long-term
TMP_AVQ
Avg. Temp., long-term
keq/ha
keq/ha
keq/ha
keq/ha
keq/ha
-0.47863
keq/ha 0.0065
0.40484
keq/ha O.OZ39
keq/ha
-0.38306
keq/ha 0.0334
-0.36371
keq/ha 0.0443
-0.44960
cm 0.0112
-0.46914
cm 0.0078
C
' SO416 is the stream sulfate concentration, SO4_NRET is watershed sulfur retention, SOBC is the stream sum of base cation
concentration, ALKA11 is the stream acid neutralizing capacity, and PHEQ11 is the air-equilibrated stream pH.
, continued
8-187
-------
Table 8-70. (Continued)
Speanman Correlation Coefficients Significant at 0.05 / Significance Level / N = 31
Variable Unite SO416 SOMMRET SOBC ALKA11 PHEQ11
Geology
GEO_SEN
Geological weather. Index, mean
GEO_MAX
Geological weather, index, max.
GEO_GT4
Geological weather, index > 4 percent
Physiography
EL_AVG
Elevation, average m
EL.MAX
Elevation, maximum m
MAXREL
Maximum relief m
SLP
Slope, mean percent
ATNMEAN
In (a/tan 0), mean
GMP_FTN 0.51896 -0.43147
Footslope, toeslope, flood plain percent 0.0028 0.0154
LOW
Lowlands percent
-0.38149
0.0342
-0.40040
0.0256
•0.36411
0.0440
-0.40282
0.0247
0.40300
0.0246
0.50570
0.0037
-0.36072
0.0462
-0.40484
0.0239
-0.39194
0.0292
-0.43185
0.0153
0.47039
0.0076
-0.34798
0.0551
0.49028
0.0051
0.53204
0.0021
Hydrology
HYD.SLW
Hydrologic group C or D percent
DRN_SLW
Drainage class <= 3 percent
PERM
Permeability, mean cm/hr
PRM_SLW
Permeability class <= 3 percent
0.36819
0.0416
-0.40524
0.0237
0.46569
0.0083
0.37350
0.0385
-0.45887
0.0094
0.48854
0.0053
0.41309
0.0209
-0.36089
0.0461
0.56103
0.0010
continued
8-188
-------
Table 8-70. (Continued)
Spearman Correlation Coefficients Significant at 0.05 / Significance Level / N = 31
Variable
DEPTH
Bedrock depth, mean
BRDJ.T2
Bedrock class <= 2 (100 cm)
BRD.SHL
Bedrock <= 50 cm
IPOjSHL
Impermeable layer <= 50 cm
AREATER
Area, terrestrial
AREA_H2O
Area, water
DENSITY2
Drainage density, NSS
STRORDER
Stream order, maximum
Units SO416 S04 NRET SOBC ALKA11
m
0.37192
percent 0.0394
0.37383 -0.38794
percent 0.0383 0.0310
percent
ha
ha
0.35286
0.0515
PHEQ11
Land Use/Vegetation
FOREST
Forested land
CULTTV
Cultivated land
PASTURE
Pasture/grazed land
DISTURB
Disturbed land
WETLAND
Wetlands
VQT_CNF
Vegetation, coniferous
VGT.DCD
Vegetation, deciduous
VQT_DRY
Vegetation, dry open
VQTWET
Vegetation, wet open
•0.62067 -0.65401
percent 0.0002 0.0001
0.42443 0.49383
percent 0.0173 0.0048
0.54318 0.56287
percent 0.0016 0.0010
percent
ha
percent
percent
0.60522 0.61150
percent 0.0003 0.0003
percent
-0.60673
0.0003
0.53198
0.0021
0.49295
0.0048
0.55585
0.0012
8-189
-------
Table 8-71. Results of Stepwise Multiple Regressions for DDRP Lake and Stream Sulfate
Concentration (S0416) Versus Watershed Attributes
Soil Physical
Properties
Soil Chemical
Properties
Deposition/
Climate
Geology
Physiography
Hydrology
Vegetation
Variable15
SAND
CLAY
FRAG
THKA
SOILDEN
CACL
MG~CL
SBCCL
BS CL
CEC CL
AC 8ACL
PH~01M
AL"AO
AL~CD
AL~PYP
CTOT
S04 H2O
SO4~PO4
SO4~EMX
SO4~B2
SO4~XIN
CMtTD
NHJLTD
H LTD
SO4 LTD
PREC L
RNOrT
TMP AVG
COASTD
GEO MAX
EL AVG
MAXREL
SLP
ATNMEAN
ATKBMEAN
GMP FTN
PERM
DEPTH
BRD SHL
IPO SHL
AREATER
AREA~H2O
WAUT
DDENSITY
VGT CNF
VGT~ORY
VGfWET
Subreaion* Reaion
1AC 1B 1C 1D 1E NE SBRP
23 8
3 4
K-) 5(-)
7(-)
1
4(-) 4(-)
3 4
. 1
3
2
2(->
3
'I
1
2(-)
2(-) 4(-)
1 5
2W
2
6(-)
6
4(-)
9(-)
5(-)
4 5
3(-)
5
1 3 5
1H
0.45
0.76
0.83
0.78
0.64
0.64
0.73
* 1A Is the Adirondacks, 1B is the Poconos/Catskills, 1C is Central New England, 10 is Southern New England,
.and 1E is Maine.
" Variable labels and units are found in Table 8-69.
e Numbers indicate order of entry into stepwise model. (-) indicates a negative parameter estimate.
8-190
-------
Table 8-72. Results of Stepwise Multiple Regressions for DORP Watershed Sulfur Retention
(SO4NRET) Versus Watershed Attributes
Subreqion*
Region
Variable1'
1AC
1B
1C
1D
1E
NE
SBRP
Soil Physical
Properties
Soil Chemical
Properties
Deposition/
Climate
Geology
Physiography
Hydrology
Vegetation
SAND
GUY
FRAG
THKA
SOILDEN
CA CL
MCTCL
SBCCL
BS CL
CEC CL
AC SACL
PHT01M
AL~CD
AL~PYP
CTOT
S04 H2O
SO4~PO4
S04~EMX
SO4~B2
SO4"XIN
CMITD
NH4"LTD
H LTD
SD4 LTD
PREC L
RNOFT
TMP AVG
COASTD
GEO MAX
EL AVG
MAXREL
SLP
ATNMEAN .
ATKBMEAN
GMP FTN
PERM
DEPTH
BRD SHL
IPD "3HL
AREA TER
AREA*H20
WALA"
DDENSTfY
VGT CNF
VGTDRY
VGTWET
R8
5(-)
4
2(-)
4 2
3
2(-)
1 (•) 5(")
5(-) 2
S 4 1
3
4(-) 3
3(~) 2(-)
1
2(-)
3(-} 4W 1
5
2W
3(~)
4W 1
0.64 0.76 0.82 0.76 0.69 0.34 0.44
* 1A is the Adirondacks, 18 is the Poconos/Catskills. 1C is Central New England, 10 is Southern New England.
. and 1E is Maine.
" Variable labels and units are found in Table 8-69.
6 Numbers indicate order of entry into Stepwise model. (-) indicates a negative parameter estimate.
8-191
-------
8.10.4.1 Northeast
The coefficients of determination, or R2 values, range from 0.45 to 0.83 for sulfate and from 0.34
to 0.82 for sulfur retention in the NE. in general, watershed attributes that had higher bivariate
correlations with water chemistry were selected as explanatory variables in the stepwise regressions.
For the northeastern region as a whole, the strongest association is between lake sulfate concentration
and long-term total sulfate deposition. Precipitation amount and runoff are also highly associated, with
a negative sign, which probably indicates a dilution effect. The sulfate isotherm half-saturation
concentration is the most highly associated soil chemistry variable, consistent with the regression results
using only soil chemistry variables (Table 8-61). Watersheds having greater areas of poorly drained foot
and toe slopes and lowlands generally have lower lake sulfate concentrations; these may be areas of
sulfate reduction. Watersheds with shallow bedrock or shallow impermeable soil layers have higher lake
sulfate concentrations. Open dry vegetation is correlated with high lake sulfate; open wet vegetation is
correlated with lower lake sulfate. Sandy soils are associated with higher lake suffate. There are some
differences in the variables selected in the regressions for the subregions, but most are correlated with
those selected for the region as a whole.
These results are consistent with those discussed elsewhere in Section 8. Lake sulfate
concentration is largely dependent on atmospheric deposition of sulfate as modified by amount of runoff,
sulfate adsorption-desorption characteristics of the soil, and soil depth and texture. Sulfate reduction in
wetlands and/or flooded soils can also reduce sulfur concentrations and therefore affect budgets in some
northeastern watersheds. Sulfate retention or release resulting from wetting and drying of soils during
seasonal cycles or over longer periods of wet or drought years can substantially influence watershed
sulfur status based on measurements made at one point in time. The extent to which these processes
and thus sulfur budgets are in equilibrium with atmospheric deposition or are acting as long-term sinks
cannot be determined with certainty from these analyses; observed relationships suggest reduction may
provide long-term watershed sulfur sinks (See also Section 7).
8-192
-------
8.10.4.2 Southern Blue Ridge Province
The first variables selected by the stepwise regressions for the SBRP are identical to those selected
for the soil properties alone (Tables 8-61 and 8-62). These are exchangeable magnesium, water-
extractable sulfate, and adsorption capacity (negative) for stream sulfate, and exchangeable magnesium
(negative), base saturation, and adsorption capacity for sulfur retention. For stream sulfate, additional
watershed attributes selected are runoff and soil permeability (both negative), and slope, generally
consistent with relationships seen in the NE. No additional variables were selected for sulfur retention.
The bivariate correlations also show relationships between stream sulfate and precipitation (negative),
shallow soils, and foot and toe slope soils. The latter is the opposite of that seen in the NE; these areas
in the SBRP are much better drained than comparable areas of northeastern watersheds, and may be
retaining less sulfur through sulfate reduction. They also have high exchangeable magnesium and
calcium, and may be adsorbing less sulfate. In the SBRP, sulfate deposition data are not related to
stream sulfate data, as discussed in Section 8.2.
8.10.5 Ca plus Mq fSOBCL ANC. and pH
Results of stepwise multiple regressions for Ca plus Mg, ANC, and pH are given in Tables 8-73
through 8-75. This section summarizes the results and discusses potential cause-effect controls on
surface water chemistry. As discussed in Section 8.9.6, these water chemistry variables are highly
correlated with each other and often show similar relationships with explanatory variables in the multiple
regressions and in the bivariate correlations (Tables 8-69 and 8-70).
8.10.5.1 Northeast
The coefficients of determination, or R2 values, for multiple regressions on these water chemistry
variables range from 0.50 to 0.91 for the northeastern subregions. Percentage area on the watershed
having open dry vegetation is often selected as the first variable in the models. These are areas in
pasture, cultivation, urban, and other disturbed land. They generally coincide with high base saturation
and deep, relatively flat soils, all of which correlate well with these dependent variables. The open dry
areas often are disturbed, exposing fresh weathering surfaces, have been limed or fertilized, or are
8-193
-------
Table 8-73. Results of Stepwise Multiple Regressions for DORP Lake Calcium Pius Magnesium
Concentrations (CAMG16) and Stream Sum of Base Cations (SOBC) Versus Watershed Attributes
Soil Physical
Properties
Soil Chemical
Properties
Deposition/
Climate
Geology
Physiography
Hydrology
Vegetation
Subregion" Region
Variable" 1AC 1B 1C 1D 1E NE SBRP
SAND 4(-)
CLAY
FRAG
THKA 3(-) 5(-)
SOILDEN
CACL
MG*CL 2
SBC CL
BS CL
CEC CL
AC SACL
PKT01M 5 3
AL*AO 2 11{-)
AL'CD
AL'PYP 3(-)
CTOT
SO4 H2O 4
SO4~PO4
SO4~EMX 5(-> 3(-)
SO4~B2
S04~XIN 3 2(-) 4
CM ITD 4(-)
NH4~ LTD
H LTD
S04 LTD
PREC L 1{-) 2(-)
RNOF"T S(-)
TMP AVG 4(-) 1(-)
CQASTD 6
GEO MAX 1 12
EL AVG 2(-) 10(-)
MAXREL
SLP
ATNMEAN
ATKBMEAN
GMP FTN
PERM 3(-) 9(-) 5(-)
DEPTH 5 8
BRD SHL
IPD BHL 2 7
AREA TER
AREA~H2O 3(-) 4
WALA~
DDENSITY
VGT CNF 5
VGTT3RY 1 1 2 t 1
VGfWET 4
R2 0.88 0.91 0.83 0.81 0.83 0.68 0.85
' 1A Is the Adlrondacfcs, 1B is the Poconos/Catskills, 1C is Central New England, 10 is Southern New England,
and 1E t's Maine.
Variable labels and units are found in Table 8-69.
Numbers indicate order of entry into stepwise model. {-) indicates a negative parameter estimate.
8-194
-------
Table 8-74. Results of Stepwise Multiple Regressions for DDRP Lake and Stream ANC
(ALKA11, ALKANEW) Versus Watershed Attributes
Soil Physical
Properties
Soil Chemical
Properties
Deposition/
Climate
Geology
Physiography
Hydrology
Vegetation
Subreoion* Reaion
Variable" 1AC 18 1C 1D 1E NE SBRP
SAND 3(-) 4(-)
CLAY
FRAG 4
THKA S(-)
SOILDEN
CA CL 37
MG"CL
SBCCL
BS CL 2
CEC CL
AC 6ACL
PH~01M 43 4
ALTW3 6(-)
AL~CD 5{-)
AL~PYP
CTOT
S04 H20
SOTPO4 8(-)
SO4~EMX
SO4~B2
SO4~XIN 2 2(.)
CMITD
NH4" LTD
H LTD
S04 LTD 2
PREC L 4(-) 2(-) 2(->
RNOrT
TMP AVG 5(.) 1(-)
COA5TD
GEO MAX 1 5 3
EL MG 3(-)
MAXREL 5
SLP
ATNMEAN
ATKBMEAN
GMP FTN
PERM*
DEPTH
BRD SHL
IPD 5HL 2
AR6A TER
AREA~H2O 4(-)
WALA"
DDENSITY 5(-)
VGT CNF 3 1(-)
VGTDRY 1 1311
VGTWET 4
R2 0.85 0.84 0.76 0.87 0.78 0.61 0.82
* 1A is the Adirondacks, 18 is trie Poconos/Catskills, 1C is Central New England, 10 is Southern New England
. and 1E is Maine.
° Variable labels and units are found in Table 8.9.5-1.
e Numbers indicate order of entry into stepwise model. (-) indicates a negative parameter estimate.
8-195
-------
Table 8-75. Results of Stepwise Multiple Regressions for DDRP Lake and Stream Air Equilibrated
pH (PHEQ11) Versus Watershed Attributes
Soil Physical
Properties
Soil Chemical
Properties
Deposition/
Climate
Geology
Physiography
Hydrology
Vegetation
Variable"
SAND
CLAY
FRAG
THKA
SOILDEN
CA CL
MG~CL
SBCTCL
BS CL
CEC CL
AC BACL
PH~01M
AL~AO
AL~CD
AL'PYP
CTOT
804 H20
SOTPO4
S04~EMX
SO4~B2
S04~XIN
CMTTD
NH4~LTD
H LTD
S04 LTD v
PREC L
RNOF~T
TMP AVG
COABTD
GEO MAX
EL AVG
MAXREL
SLP
ATNMEAN
ATKBMEAN
GMP FTN
PERM
DEPTH
BRD SHL
IPD SHL
AREA TER
AREA~H2O
WAUT
DDENSITY
VGT CNF
VGT~ORY
VGTWET
R2
Subreaion* Reaion
1AC 1B 1C 1D 1E NE SBRP
B(-)
4
1 3412
2 3(-}
4
2(.)
'(•)
2(-) 2M
4/.\
2{*) 4(-)
4(.)
4(~) 1 (*)
1 5
5(-) 5(-)
6
3
5(-) 3
/(.)
2
3 5
K-)
3 3 1
0.65 0.87 0.77 0.56 0.85 0.50 0.65
8 1A is the Adirondacks, 18 is the Poconos/Catskills, 1C is Central New England, 10 is Southern New England,
and 1E is Maine.
0 Variable fabels and units are found in Table 8.9.5-1.
c Numbers indicate order of entry into stepwtse model. (•) indicates a negative parameter estimate.
8-196
-------
associated with land uses that contribute base cations to runoff. Precipitation amount is inversely
correlated with all three dependent variables, probably a dilution effect. Bedrock weatherability index is
related to these dependent variables as seen in several of the regressions and the bivariate correlations.
Soil pH, base status, extractable aluminum, and several sulfate isotherm parameters are also related,
along with soil texture. These have been discussed in more detail in Section 8.9.6.1. Watershed
attributes including soil depth, permeability, and area in low geomorphic position are commonly correlated
explanatory variables. Where water can infiltrate rapidly and follow deep flow paths to contact with high
base saturation soils or weatherable minerals, base cation supply is high. There does not seem to be
any single good index of these characteristics that is common to all subregions, but combinations of
several indices in the multiple regressions lead to reasonably good explanatory power for most
subregions. This suggests the importance of knowing hydrologic characteristics of a watershed to explain
water chemistry..
8.10.5.2 Southern Blue Ridge Province
In the SBRP as in the NE, the variable most strongly associated with these water chemistry
variables is area in open dry vegetation. As in the NE, the geologic weatherability index is related to
base cation supply. Runoff, Ca+Mg deposition (probably a surrogate for precipitation), permeability,
elevation, slope, and relief all have regression estimates or bivariate correlations with negative signs; the
more water that passes quickly through the watersheds into the streams, the lower the base cation
concentrations and pH. The R2 for these models ranges from 0.65 to 0.85.
8.10.6 Summary and Conclusions
8.10.6.1 Sulfate and Sulfur Retention
In the NE, sulfate deposition, precipitation amount, and watershed hydrologic characteristics have
the strongest associations with surface water sulfate concentration and watershed sulfur retention. In the
SBRP, soil chemical variables (sulfate adsorption capacity, exchangeable magnesium, base saturation, and
water-extractable sulfate) have the strongest associations with stream sulfate concentration and watershed
sulfur retention. For stream sulfate concentration, watershed hydrologic variables also entered the
8-197
-------
regressions in the SBRP. Since soils in the NE are near steady-state with respect to sulfur adsorption,
current sulfate deposition is a associated with lake sulfate as discussed in Section 8.2.3. Soils in the
SBRP appear to be actively adsorbing sulfate, and the stream sulfate is controlled by soil chemistry (see
Section 9.2).
8.10.6.2 Ca plus Mg (SOBC), ANC, and pH
For the base cation-related water chemistry variables in both the NE and the SBRP, the percent
of land with open, dry vegetation consistently is among the first variables selected in the regressions.
These are areas in pasture, cultivation, urban, and other disturbed-land uses. This variable is notably
absent In the regression for the Adirondack Subregion, which contains almost no land with open, dry
vegetation and also has the lowest mean lake ANC values. Conversely, the Poconos/Catskills Subregion,
with the highest proportion of open, dry vegetation, has the highest mean lake alkalinity. Area of open,
dry vegetation is the most strongly associated variable in the ANC models for the Poconos/Catskills and
Southern New England Subregions.
Areas of open, dry vegetation have usually been disturbed by the activities of man in some way.
The strong relationship between these areas and surface water base cations may result from the
disturbances (plowing, fertilization, liming, excavation leading to faster bedrock weathering, waste disposal)
or from characteristics that predispose the areas to disturbance (low slopes, fertile soils, etc.). Generally
these areas coincide with high base saturation, and deep relatively flat soils, all of which correlate well
with the dependent variables. The dependent variables are also correlated with bedrock weatherability
and surface water sulfate concentrations overall between the NE and SBRP. Such correlations within
each region, however, are weak or do not occur.
8.10.7 Summary Conclusions
• A significant proportion of the variability in surface water chemistry can be explained by
watershed and soil characteristics.
• Deposition alone does not explain the large variability seen in surface water chemistry.
Sulfate deposition is an important explanatory variable for surface water sulfate concentrations
8-198
-------
in the NE, but not in the SBRP. Additional information on watershed attributes is essential
for explaining index water chemistry.
* Variables found to be associated with surface water sulfate and watershed sulfur retention
include: sulfur deposition and soil solution sulfate concentration (in the NE); soii adsorption
capacity and base status (in the SBRP); watershed disturbances such as pits and quarries;
amounts of precipitation and runoff; extent of wetlands and flooded soils; and soil depth and
particle-size distributions.
• Variables found to be associated with surface water ANC, pH, and Ca plus Mg (sum of
base cations) include: bedrock type; watershed disturbances such as area of agriculture or
pits and quarries; levels of precipitation and runoff; soil base saturation and pH; soil sulfate
concentration; atmospheric deposition; and soii characteristics involving particle-size
distribution, permeability, and depth.
• Surface water chemistry may be significantly influenced by watershed disturbances or the
extent of sulfate-reducing and acidic organic soils. The Level II and III models do not deal
explicitly with these variables. One model assumption is that no land use change occurs
during the period being modelled; the available data and model structures do not permit
assessment of potential watershed changes that may occur as disturbed lands revert to
natural conditions as is happening today In many areas of the eastern United States. The
extent of sulfate-reducing wet soils is handled implicitly in model calibration as in-lake
reduction of sulfate, and the extent of acidic organic soils is handled by the aggregation of
soil chemistry through sampling classes.
• In general, the relationships found in the regressions are the postulated relationships that
are incorporated in the Level II and III models. Given the caveats discussed 1n this
document, the Level II and III models incorporate the variables that are most strongly
associated with surface water chemistry.
8-199
-------
SECTION 9
LEVEL II ANALYSES - SINGLE FACTOR RESPONSE TIME ESTIMATES
9.1 INTRODUCTION
Although a number of watershed processes are recognized as influencing surface water chemistry
(Sections 2 and 3), only a few are believed to represent the major controls on short- and long-term
changes in watershed response to acidic deposition. The MAS Panel on Processes of Lake Acidification
(NAS, 1984) focused on sulfate adsorption and base cation exchange by soils as critical time-varying
processes that might contribute to a delayed response to acidic deposition. The NAS Panel recognized
scientific uncertainties In the present and potential long-term role of these two processes. Mineral
weathering, cation uptake by vegetation, etc., are rate-limited processes, the magnitudes of which are not
likely to change substantially over the period of the DDRP projections (50 years). Sulfate adsorption and
cation exchange, on the other hand, are capacity-limited processes. As adsorption sites become
occupied or as exchangeable cations are leached from the soil, the buffering capacity of watershed soils
decreases, resulting in increased probability of acidification. The projected time frame of such changes
is believed to vary widely and Is thought to be a function of soil physical and chemical properties. In
watersheds with thin or very coarse-textured soils, buffering of acidic deposition by adsorption or
exchange would be very limited and some systems would respond almost Immediately. Alternatively,
watersheds with deep soils and high adsorption capacities and/or large exchangeable base cation pools
might experience significant changes in soil leachate chemistry only after decades to centuries of high
acidic deposition loadings.
This section presents results of Level II Analyses, which involve simulations of the temporal
-/
response of individual watershed processes considered in isolation. Sulfate adsorption and cation
exchange are examined as mechanisms contributing to delays in surface water acidification for
watersheds in the Northeast (NE) and Southern Blue Ridge Province (SBRP). The analyses are based
on models that consider only adsorption or exchange within the upper regolith (<. 1.5 m in the NE, <.
2 m in the SBRP). These analyses assess the influence of adsorption and exchange on present soil
9-1
-------
and/or surface water chemistry and project probable future changes in adsorption and exchange. Soil
chemistry data collected during the DORP Soil Surveys and models are used to project future (1) changes
in sulfate mobility controlled by sulfate adsorption and (2) changes in base cation leaching, soil pH, and
cation exchange pools controlled by base cation exchange. By considering base cation exchange but
not resupply (i.e., through mineral weathering), the models presumably overestimate the potential rate of
base cation leaching from the soil; this overestimate results in underestimates of response times for future
changes in soil and solution chemistry.
Because these analyses only consider temporal response of single processes to acidic deposition
in a portion of the watershed (i.e., the upper 1.5 • 2.0 m of watershed soils), model results should not
be interpreted as integrated projections of watershed response time. Rather, they represent a set of
bounding estimates of the relative importance, now and in the future, of the role of adsorption and
exchange within soils as delay mechanisms. The results of model simulations in some cases, however,
allow inferences about other processes not considered in the Level II Analyses (e.g., contributions of
mineral weathering to watersheds with ANC > 100 /ieq L"1). Section 7 provides a partial assessment
of the role of processes other than adsorption in mediating sulfate mobility in watersheds of the NE and
SBRP. Level HI modelling (Section 10) provides projections of changes in surface water chemistry based
on integration of adsorption and exchange with other processes.
9.2 EFFECTS OF SULFATE ADSORPTION ON WATERSHED SULFUR RESPONSE TIME
9.2.1 Introduction
As discussed in Sections 2 and 3, the DORP has focused on sulfate as the principal anion in
acidic deposition and as the major mobile anion affecting chronic surface water acidification at sites in
the eastern United States. The extent and duration of sulfate retention within watersheds varies widely
within and among regions, depending on deposition history and physical and chemical properties of soils.
Sulfate retention, therefore, has been identified as one of the most important variables influencing the
rate of watershed chemical response (i.e., changes in ANC) to acidic deposition (Johnson and Cole,
1980; Galloway et al., 1983a; MAS, 1984).
9-2
-------
At the start of the DDRP, soils in the glaciated northern areas of North America were generally
believed to have low sulfate adsorption capacity, resulting in negligible sulfate retention by soils and
watershed sulfate budgets at or near steady state. In contrast, watersheds in the southeastern United
States were believed to be characterized by high net sulfur retention, attributable to the moderate to high
sulfate adsorption capacities of deep, highly weathered soils (MAS, 1984). Site-specific and regional
analyses of watershed sulfur budgets (Rochelle et al, 1987; Rochelle and Church, 1987; discussed in
Section 7) have confirmed differences in regional sulfur budgets. These studies did not, however,
evaluate causal mechanisms, nor did they project a time frame for changes in the Southeast.
Studies of sulfur retention processes in watersheds, summarized by Church and Turner (1986)
and discussed in Section 3.3, suggest that adsorption is the most important net retention mechanism
in typical terrestrial systems in the NE and SBRP. Process studies have consistently identified
iron/aluminum hydrous oxide content and soil texture (clay content or surface area) as variables that are
positively correlated with adsorption and pH and organic content as variables that are negatively
correlated with adsorption. These findings, coupled with the observed differences in these soil variables
between the two regions, are consistent with (and have contributed significantly to the development of)
the paradigm that northeastern soils have low retention capacity and are near sulfur steady state, whereas
southeastern soils have high adsorption capacity and high watershed sulfur retention.
Previous regional soil comparisons (e.g., Johnson et al., 1980; Johnson and Todd, 1983)
documented regional differences in sulfate pools and adsorption and correlated them with differences in
soil pH and hydrous oxide and organic content of soils. These comparisons provided no direct basis,
however, for assessing sulfate dynamics in soils of a region and no means of forecasting response to
continued or altered loadings of sulfate. Within the DDRP, assessments of sulfur budgets (Section 7),
summary descriptions of soil chemistry data, and empirical linkages of soil chemical variables with surface
water chemistry (Section 8) provide Important incremental results and an Improved understanding of
processes controlling sulfate in these watersheds. These results are generally consistent with the
hypothesis that the mobility in watersheds of sulfate derived from acidic deposition is controlled by
9-3
-------
adsorption. The principal DDRP objectives, however, lie not just in identifying processes but in predicting
the dynamics of sulfate in study regions, specifically in projecting future changes in surface water sulfate
in response to continued sulfur deposition at current or altered deposition levels. Level II Analyses are
designed to project response of individual watershed processes; this section describes the procedures
for and results of projecting sulfur dynamics in soils of the NE and the SBRP.
9.2.2 Section Objectives
Analyses in this section are limited to consideration of changes In sulfate mobility in DDRP
watersheds and regions attributable to sulfate adsorption (and desorption) by soils. Controls on sulfate
by other processes (Section 3.3 and Section 7) of relatively minor importance in most DDRP watersheds
and in the regions as a whole. The goal of Level II Analyses of sulfate is to assess the importance of
sulfate adsorption In influencing delays in surface water acidification in the NE and SBRP. Specific
objectives of Level II Analyses are to:
• Characterize and compare sulfate poofs and sulfate adsorption capacity of soils in the NE
and SBRP.
• Estimate the response time of soils in DDRP watersheds to changes in sulfur deposition using
an adsorption-based model.
• Estimate time to steady state under current deposition loadings and project response time
to future Increases (SBRP) or decreases (NE) in deposition for systems not presently at
steady state, but for which sorption is regarded as an important control mechanism.
Extrapolate results to obtain regional projections.
• Summarize the contributions of sulfate adsorption to delays in surface water acidification
resulting from historic or future projected changes in deposition.
The results related to the fourth objective also provide data for evaluating and comparing the relative
importance of sorption and other processes considered by DDRP models (e.g., cation exchange). Such
comparisons, however, are not made in this section.
It is important to recognize that procedures and models used for this analysis treat sorption
processes in isolation. Processes affecting watershed chemistry other than those directly involving sulfate
9-4
-------
sorption are not considered, and except for sulfur deposition, watershed conditions (e.g., soil mass, soil
pH) and fluxes are assumed to be static over the duration of the projections, ft is equally important to
recognize that the projections and estimates of time to steady state made here apply only to sulfate.
Although change in sulfate mobility is one of the principal factors driving changes in base cations and
ANC, non-sulfur processes also play critical roles in such changes. Rates of change in ANC, and
particularly projected times to reach zero ANC (le., become acidic), are not necessarily coincident with
times to sulfur steady state. Systems can reach an acidic state prior to, concurrently with, or after sulfur
steady-state conditions are reached. The relationship between changes in sulfate and changes in ANC
is characterized as part of Level III Analyses and discussed in Section 10.
9.2.3 Approach
Level II sulfate analyses are based on model-based projections of future sulfate dynamics in soils
of watersheds in the DORP NE and SBRP Regions. Projections were made using soil chemistry data
generated by the DDRP Soil Surveys (Section 5.5). The principal soil variables used for these analyses
are sulfate adsorption isotherms generated for individual soils collected in the surveys and aggregated
to the watershed level. Projections were made using a modified version of the sulfate subroutine in the
Model of Acidification of Groundwater in Catchments (MAGIC) (Cosby et al., I985a, !986b).
9.2.3.1 Model Description
A critical early decision in this analysis was the selection of one or more models to describe
sulfate retention in watersheds. The DORP was conceived and developed as a relatively short-term
assessment project Consequently, project design dictated use of existing models rather than
development of new sulfur cycling models. This constraint restricted options for model selection; for
instance, no model available in 1985 effectively described sulfur cycling or net retention in soil organic
sulfur pools, and only very fragmentary data existed on transformation rates for organic pools.
Furthermore, many integrated watershed models were developed for systems with negligible sulfur
retention. For these models, terrestrial sulfur retention was set to zero (e.g., the Trickle Down Model,
Schnoor et al., 1984, I986b), or was described by empirical relationships that served principally to fit
9-5
-------
seasonal or hydrologically-driven variability in dissolved sulfate, without linkage to specific processes
(e.g., the Birkenes model, Christophersen and Wright, 1981). After consideration of available models that
have adsorption routines, the sulfate subroutine of the MAGIC was viewed as the most straightforward
and least data-intensive alternative and was selected for use.
The model uses a deterministic, mass-balance approach that considers only adsorption as a sulfur
retention process by soils (Cosby et al., 1985b,c; 1986b). Sulfate partitioning between dissolved and
sorbed phases is defined by an hyperbolic (Langmuir) isotherm. The original MAGIC subroutine has
been modified to accommodate multiple soil horizons (up to 10, although either 2 or 3 were used for this
study). Soil horizons are treated as a series of continuously stirred tank reactors (CSTRs); all inputs of
precipitation and sulfur (wet and dry) deposition are to the top mineral soil horizon (organic horizons are
not considered in the model, because sorption is negligible in the O horizon). Evapotranspiratlon
implicitly occurs in the top soil horizon. All flow is then routed sequentially through each soil horizon.
Data are input to the model using annual time steps. The projected surface water sulfate concentration
is defined by (set equal to) the equilibrated solution sulfate concentration in the lowest soil horizon.
Because sorption is essentially an instantaneous process, reaction kinetics are not considered and
equilibrium between solution and sorbed phases is assumed to occur in all cases.
For these analyses, model simulations were run starting 140 years prior to the base year (1984 for
NE lakes, 1985 for SBRP streams). Soil and streamwater surface water sulfate concentrations were
initialized at the start of simulations by assuming both to be at steady state with respect to deposition.
Simulations were run either 140 years (NE) or 300 years (SBRP) into the future, allowing projected sulfate
concentrations to reach steady state for all watersheds. Data sources for model simulations are described
below (Section 9.2.3.2).
9.2.3.2 Data Sources
Input requirements for the sulfur model include current sulfur inputs and outputs (precipitation,
runoff, total sulfur deposition, and sulfate concentration in runoff), scenarios of historic and future sulfur
9-6
-------
deposition, and soil variables to describe sulfate partitioning and adsorption capacity (adsorption
isotherms, son mass). Data sources are identified in Section 9.2.3.3; procedures for generation of
adsorption isotherms and for aggregation of soil chemistry data are described in Section 9.2.3.4.
Procedures used to estimate precipitation and sulfur deposition were described in Sections 5.6.3;
both typical year (TY, annual values) and long-term average (LTA) estimates of total sulfur deposition
were used for NE and SBRP watersheds. Runoff estimates, based on interpolation of 30-year average
USGS runoff maps, were generated as described in Section 5.7.1. Current lake sulfate concentrations
are from the EPA's Eastern Lake Survey (Unthurst et al., 1986a) and the Pilot Stream Survey (Messer
et al., 1986a) (Section 5.3).
Initial sulfate inputs (year 140) were set to 5 percent of current deposition; estimated sulfur
deposition between initial and base years (i.e., 1844 to 1984 in the NE and 1845 to 1985 in the SBRP)
was based on emission estimates of Gschwandtner et al. (1985). Estimates of historic deposition for the
NE and SBRP are based on emission estimates for Federal Regions I and II (CT, MA, ME, NH, NJ, NY,
Rl, VT) and Region IV (GA, NC, SC, TN), respectively; linear interpolation between the initial simulation
year and 1900 was used. The historic emission pattern was used as a scaling factor for each watershed,
a procedure that assumes that the relationship between regional emissions and site-specific deposition
over the last 140 years is constant.
Two scenarios of future sulfur deposition were used for each region as characterized in Section
5.6.1. The first scenario for each region was constant deposition through the entire simulation period.
For the NE, the alternative scenario is constant deposition for 10 years, followed by a linear decrease in
deposition for 15 years (by 2 percent per year), then constant deposition at 70 percent of current
deposition for the remainder of the simulation period. The alternative scenario for the SBRP also begins
with constant deposition for 10 years, followed by a linear increase in deposition between years 10 and
25, then constant deposition (at 120 percent of current levels) for the remainder of the simulation period.
9-7
-------
Mapping of soils and quantification of the areal extent of various soils on DDRP watersheds were
described in Section 5.4 Sampling and chemical/physical analyses of soils were described in Section
5.5. For each mineral soil horizon, sulfate adsorption data were used to compute adsorption isotherms
which were then aggregated with soil mass (computed from horizon thickness, bulk density, and coarse
fragment content) to obtain sample class and watershed values. Procedures for derivation of adsorption
isotherms and for aggregation of adsorption data are described in Section 9.2.3.4.
9.2.3.3 Model Assumptions and Limitations
Several critical assumptions are encompassed by the choice of model and by methods of data
collection. These in turn impose limitations on the scope of model projections. Key model assumptions
and their implications for data interpretation include:
• Sorption Is the only watershed process affecting sulfate mobility and watershed response
time. As noted previously, this decision was intentional and is believed to provide the most
effective means of assessing the significance of adsorption by soils as a process delaying
surface water acidification. To the extent that other terrestrial processes sequester or
generate sulfate on a net basis, model projections will under- or overestimate the time and/or
magnitude of the projected response. As noted earlier, the net role of other processes in
most DDRP watersheds is believed to be small. (The importance and influence of in-iake
processes on sulfur budgets and response time in northeastern lakes is addressed in Sections
7.2 and 10.)
• The analytical approach used to define sulfate partitioning in the soil (hyperbolic isotherms
defined by batch equilibrium methods using air-dried soils) adequately describes sulfate
partitioning by soils under field conditions. Recent findings (Hayden, 1987) support the use
of hyperbolic isotherms and batch equilibrium methods. A preliminary evaluation of effects
of soil drying suggested small, non-systematic effects on adsorption; however, subsequent
study (Hayden, 1987) suggests that the measured adsorption capacity of soils increases upon
drying. This issue is currently being thoroughly assessed by a separate EPA project.
* Soil and watershed conditions influencing adsorption (e.g., soil pH, Fe, Al, and organic
content) are static over the life of model projections. Potentially, pH is the most important
of these variables since adsorption is strongly pH dependent. If soil pH were to change
significantly, the projections of adsorption could be substantially altered. However, soil pH
is strongly buffered at low values in most of the NE and SBRP soils considered by DDRP,
and substantive changes in soil pH but are not expected.
• Hydrologlc routing is simple, representing the soil as a series of CSTRs; all flow is routed
sequentially through each horizon. The "perfect" hydrologic contact represented by a
simplified flowpath such as that used here does not realistically reflect how lateral flow,
macropore flow, etc., occur in the soil. However, data to objectively define alternate flowpaths
are lacking. The effect of flow bypassing upper or lower soil horizons under natural
conditions would result in projections of higher initial sulfate leaching (part of the input signal
would not be attenuated by sorption on the soil), but a more gradual (in terms of change
in concentration and time) subsequent sulfate response. The responses projected here
9-8
-------
represent an upper bound on initial response time assuming complete contact between the
soil and flow through the soil, and a lower bound on time to steady state.
• Because the model runs on an annual time step and uses identical precipitation and runoff
data from year to year, projections do not reflect the variability of natural systems. The lack
of "realistic' variability in the projections is recognized, but should have little effect on the
primary objective of projecting long-term changes attributable to chronic sulfur deposition.
If there are any long-term trends in precipitation or runoff, they would not, of course, be
represented by model projections.
9.2.3.4 Adsorption Data
Data describing suifate partitioning by soils, which are used to develop partitioning functions
(Isotherms) of suifate adsorption capacity of soils, were generated as part of the DDRP soil survey
(Section 5.5). Adsorption isotherms were developed for each soil, as were soil thickness, bulk density,
and coarse fragment adjustments. Adsorption isotherms were then aggregated from date for individual
soils to watershed values using a mass-weighting procedure. Determination of isotherm coefficients and
aggregation of data from individual soils to watershed values are described below.
In the design of the DDRP, emphasis was placed on projecting dynamics of suifate and other
ions at regional scales, rather than on a watershed-by-watershed basis. The design mandated that
procedures for sampling and aggregating soils data were targeted at describing soils for the region.
Using the sample classes described in Sections 5 and 8, soils for each sample class were sampled
approximately eight times across their area of occurrence, which in many cases included several states
and covered substantial sulfur deposition gradients. Aggregated sample class chemistry provides a
representative value for that sample class across the region, but probably does not optimally estimate
soil characteristics at the individual watershed level, and thus does not enable optimal projections for
Individual watersheds. As an example, suifate in northeastern lake systems is roughly at steady state
across the region; observed lake suifate concentrations are proportional to sulfur deposition and decrease
by over 50 percent from ELS Subregion 1B (Poconos/Catskills) to 1E (Maine) (Unthurst et al., 1986a).
Aggregated soil chemistry for sulfur variables in a sample class that extends from New York to Maine
are the same for all soils in the class, however, and thus presumably would underestimate concentrations
in New York while overestimating them for Maine.
9-9
-------
An alternative approach for sampling and data aggregation would have been to focus sampling to
enable characterizations of individual watersheds. This watershed focus would generate more intensive
sampling of points likely to be representative of each sampled watershed, but would have allowed
sampling of fewer catchments in the region, with the risk of describing less of the soil variability across
the region. A watershed focus also would have resulted in fewer sites for extrapolation of results to
obtain regional population projections. The watershed approach thus is regarded as less effective than
the regional sample class approach for describing most soils that occur on watersheds in each region
and for generating regional projections. While the sample class approach describes the central tendency
and most of the range of watersheds, it does not, however, provide precise watershed-level projections,
especially for extreme watershed values in the population. Soils data were mapped and sampled on
specific watersheds and aggregated to watershed values in order to allow explicit linkage between soil
characteristics and surface water chemistry. To deal with uncertainties in projections, uncertainty
estimates for major input variables for Level II models (soil mass and isotherm coefficients for Level II
sutfate analyses) were generated, and Monte Carlo analyses were used to describe uncertainty in model
projections for a subset of watersheds.
Adsorption isotherms were generated from data for soil-water slurries equilibrated with six different
amounts of sulfate (0, 2, 4, 8, 16, and 32 mg S L'1) described as SO4J), SO4_2, etc.. in Section
5.5.4.2.1. For each of the six samples, net sulfate adsorbed by the soil was computed from the change
in dissolved sulfate. For example, for the 8 mg L"1 sample:
SO4 8n - (SO4 ft - SO4 8,) * k (Equation 9-1)
— — i — i o
Where: SO4_8, - dissolved sulfate concentration prior to equilibration ( eq L*1)
S04_8f = final dissolved sulfate concentration after equilibration ( eq L"1)
L - volume of liquid (~ 0.050 L)
S = mass of volume of soil (~ 0.010 kg)
An extended Langmuir isotherm ("extended" by addition of a third variable to describe the non-zero
9-10
-------
Y-intercept) was then fit to the six data points for each soil (final dissolved sulfate and net adsorbed
sulfate) (Figure 9-1) (Hayden, 1987). The equation used to model sulfate partitioning has the form:
B * C + B, (Equation 9-2)
B2 + C
where: B1 = maximum sulfate adsorption (meq kg*1)
B2 = half saturation constant ( eq L'1)
"1
B3 = Y-lntercept (meq kg")
C « dissolved sulfate ( eq L"1)
Ec = net adsorbed sulfate at [SO42*] = C (meq kg*1)
The parameters Bv B2, and B3 were estimated using non-linear least squares, using the Fletcher-Powell
(1963) method to minimize the sum of squares function. The Retcher-Powell method uses a second
order algorithm that iteratively constructs an estimate of the inverse Hessian matrix. This matrix, in
conjunction with the residual sum of squares, provides an estimate of the covariance matrix for the
estimated parameters.
Several approaches were evaluated for aggregating data from individual soils, including weighted
averaging of isotherm coefficients or alternatively fitting a single isotherm to all data points for all soils
in an aggregation group (e.g., all individual soils in a pedon/master horizon or sample class/horizon).
Both approaches were rejected because they provided a poor description of the average partitioning
coefficient (isotherm slope) along the isotherm. As an alternative, after fitting Isotherms for individual
soils, values of net adsorbed sulfate corresponding to several concentrations of dissolved sulfate (0, 10,
25, 40, 75, 125, 200, 500, 1000, 2000 eq L*1) were computed for each soil. For each value of dissolved
sulfate, the mass-weighted average of the corresponding concentrations of adsorbed sulfate was
generated for all soils in an aggregation group (typically all soils with the same master horizon
designation in a sample class). Finally, a new isotherm was fitted to the set of weighted averages. This
isotherm was defined as the aggregate isotherm and was used to describe sulfate partitioning for that
group of soils.
9-11
-------
£
CO
~O
CD
€
8
So
I
EXTENDED LANGMUIR ISOTHERM
Where
E VC *B
0 = Ba + C 3
C = Dissolved Sulfate
EC- Net adsorbed sulfate at C
B, • Maximum Adsorption
B2 = Half-saturation constant
B3 = Y-intercept
ESSS (Equilibrium Soil Solution
Sulfate) = Dissolved Sulfate at
Figure 9-1. Schematic diagram of extended Langmuir isotherm fitted to data points from laboratory
soil analysis.
9-12
-------
This approach provides a very good estimator of the weighted average soil partitioning coefficient
(isotherm slope) over the range of dissolved sulfate of interest to this project (0 - 200 eq L'1), even for
groups of soils in which coefficients for individual soils are highly variable. Aggregation was conducted
in three steps, with any missing data assigned the aggregate average for other soils in its pedon/horizon:
(1) individual soil (sub)horizons to master horizon within a pedon (mass weighting); (2) pedon/horizon
to sample class/horizon (mass weighting); and (3) sample class/horizon to watershed/ horizon (mass
and area weighting of each sample class occurring on each watershed). For routine uses, data for soil
master horizons were used directly in the model and were not aggregated. For certain model
applications, data were aggregated to 1 or 2 horizons per pedon using a comparable mass-weighting
approach.
Because the aggregation approach was not conducive to direct computation of parameter
uncertainty, uncertainties for the original isotherm fits were retained; a Monte Carlo procedure was used
during each step of aggregation to generate estimates of uncertainty in aggregated coefficients at the
sample class and watershed level. The uncertainty in the sulfate isotherms was propagated through the
aggregation procedure using the Monte Carlo technique described in general in Section 6.3. Application
of the procedure to sulfate isotherm aggregation proceeded through steps similar to those used for
aggregation of other variables. The aggregation from individual subhorizons to sample class master
horizon was repeated 100 times, each time selecting a randomly perturbed set of coefficients for each
subhorizon isotherm. The perturbation of B1 was selected first from a normal distribution with a standard
deviation obtained from the residual sum of squares and the Inverse Hessian matrix from the nonlinear
least squares. The perturbed value of Bv along with the correlation of B1 and B2 from the inverse
Hessian matrix, were used to estimate the conditional expectation of B2 given Bv This conditional value
was then perturbed by a value drawn from a normal distribution with the conditional standard deviation
of B2 given Bv A similar procedure was used to perturb B3, except that the mean and variance were
adjusted for both B, and B2. The mean values, standard deviations, and correlation matrix of the
coefficients were summarized at the sample class level. These values were then passed to the watershed
level aggregation algorithm. The uncertainty calculation was conducted as above, except that the
9-13
-------
correlations were derived from the sample class Monte Carlo study rather than from an Inverse Hessian
matrix.
The rationale for the mass-weighting aggregation approach described above is consistent with the
common aggregation approach discussed in Johnson et al. (1988b). Several alternative approaches to
aggregation of soil chemical data were discussed in that document, including weighting schemes that
would represent watershed factors such as hydrologlc flowpaths, landscape position, etc. Ultimately,
alternative aggregations for capacity variables, including sulfate adsorption capacity were rejected. This
decision was based principally on the lack of objective criteria for setting weighting coefficients to
describe hydrologic routing or other watershed factors (including unsuccessful attempts to empirically
determine statistically significant coefficients). The mass weighting approach used here provides unbiased
estimates of the pools and/or capacities (e.g., sulfate adsorption capacity, exchangeable base cation
pod) for capacity variables in soils of the ODRP watersheds. Hydrologic routing, incomplete soil contact,
landscape position, etc., influence the degree of interaction between acidic deposition and the soil, and
might alter the rate at which soil pools or capacities are affected. In the absence of quantitative estimates
of routing coefficients, however, the unbiased pool estimates generated by the mass-weighting approach
provides the best description of soil pools and capacities for the Level II models used here.
9.2.3.5 Evaluation of Aggregated Data and Model Outputs
Several approaches were used to evaluate aggregated soil sulfate data and model outputs. An
initial assessment of isotherm data and aggregation procedures was made by comparing the equilibrium
soil solution sulfate of isotherms aggregated to the watershed level (B horizons) to measured surface
water sulfate concentrations in the NE and SBRP. If adsorption by the soil were the sole process
influencing sulfate mobility and if aggregation procedures were perfect, a 1:1 correlation between soil and
lake/stream sulfate concentrations would be obtained. Realistically (due to contributions of factors such
as hydrologic routing, heterogeneity of natural soils, uncertainties introduced by soil sampling and
analysis, and effects of regionally-focused aggregation), a high correlation between soil solution and
surface water sulfate was not expected. The purpose of this comparison was to evaluate whether the
9-14
-------
two sets of values were comparable and whether major biases existed that would invalidate the entire
approach.
Results of this comparison (Figure 9-2), show that for the most part agreement between computed
soil solution and surface water sulfate concentrations is good. Although the data have considerable
scatter, points for NE Subregions A, C, and E and for the SBRP generally plot near the 1:1 line. The
effects of aggregating data collected along a deposition gradient (noted in Section 9.2.3.4) are clearly
apparent for NE Subregions A, C, and E. Although the lake sulfate concentrations range from roughly
50 to 150 eq L*1, aggregate soil solution concentrations are clustered near 100 eq L"1. For watersheds
in NE Subregions B and D, computed soil solution sulfate concentrations are consistently biased high.
The difference between these watersheds and other systems in the NE and SBRP is believed to be
related to differences in soil type rather than geographic location. The difference might be an artifact of
soil handling (air-drying) procedures. This bias, although substantial, occurs in only a subset of the data
and, In any case, is not sufficiently large to invalidate the data or the aggregation approach. It is also
important to note that equilibrium soil solution sulfate is not used directly in the Level II models. Related
isotherm variables that affect mode) projections (reflected by isotherm slope) appear to be much less
sensitive to effects of air-drying.
In addition to the evaluation of aggregated data described above, several approaches were taken
to evaluate model outputs. Model projections were compared to observed surface water chemistry in
several ways. Model simulations start 140 years in the past and run through the present, allowing
projections for the base year to be compared to observed lake or stream chemistry, and means and
distributions of the two datasets to be compared for biases. Preliminary evaluation of model results for
soils on northeastern watersheds indicated very rapid time to sulfate steady state and showed that model
inputs (Isotherm coefficients) in many cases could be varied by almost an order of magnitude without
significantly changing the projected sulfate concentration for the base year. Evaluation of model
projections also was done for the SBRP. Using both mean values and sample distributions for the SBRP
9-15
-------
350
300
=L 250
o
§ 200
§
3
w 150
£
i 100
A
50-
D
O N.E. - Subregion A.C.E
n N.E. - Subregion B,D
A SBRP
50
100
150
200
250
300
350
Surface water sulfate (p.eq L *1)
Figure 9-2. Comparison of measured lake (NE) or stream (SBRP) sulfate concentration with
computed soil solution concentration.
9-16
-------
target population, modelled and measured sulfate concentrations were compared, as well as modelled
vs. observed percent sulfate retention. Projected rates of increase In dissolved sulfate for DDRP
watersheds also were compared to available data on measured rates of increase for sulfate in
southeastern watersheds.
9.2.3.6 Target Populations for Model Projections
For both the NE and SBRP, projected changes In sulfate are presented for lake (NE) and stream
(SBRP) populations at regional scales. Model runs were made using data for the DDRP watersheds in
the respective regions, then extrapolated to regional target population projections, using weights defined
by the National Surface Water Survey (LJnthurst et al., 1986a; Messer et al., I986a). In the NE, model
input data were prepared for all watersheds, and the model run for all watersheds In Priority Classes A
through G, i.e., lakes in classes H (seepage lakes) and I (significant internal sulfur sources) were deleted
from the analyses. (Priority classes are described in Section 10.4.) After initial assignment of priority
classes, the lake type of one northeastern lake (1D2-036) was changed from dosed to impoundment, and
four additional northeastern watersheds (1C2-068, 1E1-025, 1E1-040, and 1E3-040) were identified as
having probable significant internal sulfur sources. Data for lake 102-036 were then Included in the
analyses, and the four lakes with putative internal sulfur sources were deleted. The final dataset used
for generating watershed sulfur projections included 131 NE watersheds, representing a regional target
lake population of 3,314 lakes.
In the SBRP, ail of the 35 DDRP stream watersheds were included in the analysis, except a single
watershed -in Priority Gass E (2A08808), which has significant internal sulfur sources. Using weights
defined during the Pilot Stream Survey, results for the SBRP watersheds were extrapolated to describe
a regional target population of 1,492 stream reaches. '.
9-17
-------
9.2.4 Results
9.2.4.1 Comparison of Northeast and Southern Blue Ridge Province Isotherm Variables
Before presenting and discussing model projections generated as part of the Level II Analyses for
sulfate, a comparison of data used as model inputs is useful, including adsorption isotherm data for soils
of the two regions and secondary data derived from the isotherms. Table 9-1 summarizes isotherm
data by soil horizon and soil order for the NE and by soil horizon for the SBRP. In addition to isotherm
coefficients, data in the table include several derived variables that provide a more convenient basis for
comparing the potential for sulfate adsorption by soils in the two regions. Derived parameters were
computed using a dissolved sulfate concentration of 100 eq L'1 to facilitate comparison. The derived
parameters include isotherm slope (soil-water partitioning coefficient), adsorbed sulfate (change in
adsorbed sulfate per kg soil as sulfate concentration increases from 0 to 100 eq L*1), and adsorbed
sulfate for soil horizons, which couples adsorption with soil mass to describe potential sorption by the
pedon.
Examination of the isotherm data reveals differences in adsorption capacities of soils within the
NE and very pronounced differences between soils in the NE and SBRP. Within the Northeast, Entisols
have the lowest adsorption capacity, whereas potential sorption capacity of Inceptisols and Spodosds
is roughly equal. For all three northeastern soil orders, adsorption capacity and isotherm partitioning
coefficients are lowest in the poorly developed C horizon soils. Comparison of NE and SBRP data
consistently suggests higher adsorption by SBRP soils; maximum adsorption capacities are higher for
each SBRP horizon than for any of the northeastern soils, and the partitioning coefficient (slope) is two-
to tenfold higher for SBRP soils than for the same horizon in northeastern soils. These differences are
reflected in adsorbed sulfate pods; on a unit mass basis, sulfate pools at 100 eq L"1 are typically three-
to ten fold higher for SBRP soils than for those in the NE. When the greater mass of SBRP soils
(especially in the B horizon) is considered, the 100 eq L'1 adsorbed sulfate pool in SBRP soils is about
10 times as large as those for northeastern Inceptisols and Spodosols and 30 times that for northeastern
Entisols.
9-18
-------
Table 9-1. Comparison of Summary Data for Sulfate Adsorption Isotherm Data for Soils in the
Northeastern United States and Southern Blue Ridge Province.
Region/
Order
Northeast
Errtisols
tnceptisols
Spodosols
Soil
Horizon
A/E
B
C
A/E
B
C
A/E
B
C
Isotherm
B1 i
(meq kg"7)
2.37
1.05
0.76
3.15
3.68
1.63
2.72
5.13
1.19
Coef"
Halfsat'n
( eq I'1)
1641
997
994
1560
1017
1007
1117
893
970
Slope®
100 eq L"1
1.08
1.21
0.63
1.58
2.59
0.96
1.28
4.33
0.98
Adsorbed Sulfate
Pool @ 100 eq L'1"
(meq kg'1) (keq ha"1)
0.114
0.133
0.088
0.214
0.308
0.172
0.279
0.483
0.154
0.09
0.27
0.54
0.90
0.17
1.73
0.99
2.89
0.15
1.66
0.83
2.64
Southern Blue
Rldae Province
All
A/E
B
C
5.89
7.13
4.80
1199
322
361
3.39
12.18
6.45
0.541
2.657
1.837
0.98
20.90
5.86
27.74
a Coefficients for Langmuir isotherm of the form:
Adsorbed SO*2* = B1 ' C
where c
halfsat'n + C
dissolved sulfate concentration
Computed pools of adsorbed sulfate using the equation listed in footnote a.
9-19
-------
The observed differences in adsorption characteristics of northeastern and SBRP soils are generally
as expected. Retention capacity of soils in the SBRP, expressed as adsorbed suifate for the pedon, is
much higher than for soils in the NE. Two principal reasons for this difference are apparent. The first
is related to differences In suifate adsorption capacity of A and B soil horizons in the two regions.
Comparison of soil chemistry characteristics for the two regions suggests that differences are not, as has
been suggested (e.g., NAS, 1984), attributable solely to differences In soil age and degree of weathering.
Although upper horizons of northeastern soils have lower clay content than SBRP soils, the northeastern
soils do have substantial concentrations of extractable iron and aluminum. Extractable aluminum is often
higher in northeastern soils than in those of the SBRP. Northeastern soils, however, also have much
higher organic content than SBRP soils, and organic blocking is likely to reduce anion adsorption capacity
of northeastern soils substantially and to account for much of the regional difference in adsorption
capacity of upper soil horizons (Chao et al., 1964a; Johnson and Todd, 1983). The second factor
affecting total pedon adsorption capacity is explicitly tied to soil age and extent of weathering. Soils in
the NE have typically undergone significant weathering only to a depth of 30-50 cm; subsoils are
minimally weathered and have few days or hydrous oxides and thus little effective substrate for sorption.
In the SBRP, by contrast, most soils are extensively weathered to a depth of well over a meter, and
subsoils have abundant clays and hydrous oxides and very low organic content, resulting in high anion
adsorption capacity. SBRP soils thus not only have higher adsorption capacity per unit soil mass than
soils in the NE, but also have a much greater mass of those soils with high adsorption capacity. This
results in potential suifate retention capacities for SBRP soils that are 10- to 30-fold higher than for typical
northeastern soils and leads to differences in projected response times to sulfur deposition for the two
regions.
9.2.4.2 Model Results - Northeastern United States
Based on model projections using long-term average deposition data, sulfur response times for
soils in northeastern watersheds are very rapid in all cases. For typical systems in the NE, the projected
lag between changes in deposition and surface water response is on the order of a decade. For some
watersheds the delay is as short as five years, and the longest projected lags are less than 15 years.
9-20
-------
For all of the 131 northeastern watersheds modelled as part of the Level II Analyses for sulfate, response
times are sufficiently short that, during periods of higher deposition prior to 1975, sulfate concentrations
exceeded steady state with 1984 levels of deposition. Concentrations are projected to be declining
currently in response to reduced deposition over the past decade (Figure 9-3).
Based on the results shown in Rgure 9-3, It is apparent that the sulfate model used for this analysis
predicts very short lags In sulfate response time and thus significant deviation from sulfur steady state
for soils in northeastern watersheds only during periods when sulfur inputs are changing rapidly. When
deposition inputs are decreased, projected surface water sulfate concentrations are also projected to
decrease rapidly; during the period of re-equilibration to the lower deposition level, soils release (desorb)
sulfate and the watershed has negative sulfur retention (i.e., watershed output exceeds input; Figure 9-
4). Conversely, as Figure 9-4 also shows, during the lag phase when deposition is increased, soils adsorb
sulfate and there is positive retention by the watershed. As used in this section, steady state for sulfur
refers solely to sulfur input/output budget status; no inference is intended regarding stasis of the
biogeochemical sulfur cycles within the watersheds. Percent sulfur steady state is computed as
cutout
Percent Sulfur Steady State = ( — ) 100 (Equation 9-3)
Sinput
and is related to percent sulfur retention by
inout " outout
Percent Sulfur Retention = ( ) 100 " (Equation 9-4)
S!nput
= 100 - Percent Sulfur Steady State
9.2.4.2.1 Evaluation of base year data, calibration of model inputs -
Sulfur input-output budgets, calculated for DDRP lakes using ELS sulfate concentration data and
long-term average deposition data, were computed as described in Section 7.3. Percent sulfur retention
ranges widely among northeastern lakes, from -60 to +70 percent, with a mean of -2.5 percent (Table
9-21
-------
Northeast Lakes
180 ^
160-
140-
x
rf 120"
t_
^ 100-
03
§ 80'
OS
-------
SULFATE CONCENTRATION
c
o
o
O
-------
9-2). In contrast to the computed percent retention, the range of model projections for 1984 is much
narrower due to the short response times for northeastern watersheds. The short response times, coupled
with decreases in deposition since 1975, result in model forecasts for the 1984 base year (when
northeastern lakes were sampled for the ELS) of slight to moderate negative retention for all northeastern
lakes (Tables 9-2 and 9-3, Figure 9-5). For long-term deposition data, modelled retention in 1984 varied
from -19.3 to -1.3 percent, with a population median of -7.1 percent. Estimates using typical year
deposition data were almost identical, ranging from -18.9 to -0.1 percent retention, with a median of -
6.8 percent (Table 9-3).
Although computed and modelled percent sulfur retention differ considerably, the range and
distribution of measured and modeiied sulfate concentrations for 1984 are very similar, and are
comparable to steady-state sulfate concentrations (Figure 9-5). As indicated by percent retention data,
modelled concentrations slightly exceed steady-state concentrations for ail systems, whereas (on a
watershed-by-watershed basis) measured sulfate often deviates substantially and unsystematically from
steady-state concentrations. Because DDRP objectives are focused at the regional population level, the
close overlap of measured and steady-state sulfate concentrations is reassuring, in that it suggests the
sulfur data used for model projections provide a good representation of the regional population of lake
sulfate concentrations.
Results discussed in the preceding paragraphs are based on use of soils data without adjustment
or model calibration. Projections of sulfate concentration and percent sulfur retention are essentially
unbiased, but the range of percent sulfur retention projected by the models is much smaller than the
range of measured percent sulfur retention. Sensitivity analyses indicate that, because of the rapid
response times of northeastern systems, projected lake sulfate concentrations for 1984 remain near
steady state even if the principal model inputs (isotherm coefficients and/or soil mass) are adjusted by
a factor of two. Model projections of base year sulfate concentrations and percent sulfur retention
remained unbiased, but the distribution of percent retention again was small. Changes made in
deposition or rainfall/runoff ratios resulted in changes in the steady-state sulfur concentrations and in
9-24
-------
Table 9-2. Summary Statistics for Modelled Changes in Sulfate Concentration, Percent Sulfur
Retention, and Delta Sulfate for Northeast Watersheds Using Long-Term Average Deposition Data.
Sulfate Concentration
Scenario
Year Mean Std. Dev. MIn. P_25 Median P_75 Max.
Lake SO4 110.0 39.5 33.8 81.7 105.4 130.7 213.8
Constant Depn.
0
10
20
50
100
Steady State
Decreased Depn.
20
50
100
Steady State
120.3
111.6
110.7
110.5
110.5
110.5
103.8
77.5
77.4
77.4
46.7
41.0
39.9
39.7
39.7
39.7
38.4
28.0
27.8
27.8
54.7
51.1
50.8
50.8
50.8
50.8
47.8
35.6
35.6
35.6
83.7
77.9
77.5
77.5
77.5
77.5
72.6
54.3
54.2
54.2
114.3
106.2
106.0
106.0
106.0
106.0
99.1
74.2
74.2
74.2
142.7
126.0
126.0
125.5
125.5
125.5
118.0
87.9
87.8
87.8
249.3
218.8
211.7
209.6
209.6
209.6
204.1
148.6
146.7
146.7
Percent Sulfur Retention
Scenario
Year Mean Std. Dev. Min. P_25 Median P_75 Max.
LakeSO4 .2.5 24.9 -60.0 -20.9 -3.1 15.6 61.1
Constant Depn.
0
10
20
50
100
Decreased Depn.
20
50
100
-7.9
•0.7
•0.1
0.0
0.0
-17.0
•O.1
0.0
4.0
1.0
0.2
<0.1
0.0
1.7
0.3
<0.1
-19.3
-4.7
-1.1
>"0»1
>^,1
-21.9
-1.4
>4.1
•10.0
-0.8
-0.1
0.0
0.0
-20.9
-0.8
>-0.1
-7.1
•0.3
>-0.1
0.0
0.0
-19.4
>-0.1
0.0
-5.2
•0.1
>1.0
0.0
0.0
-15.2
>-0 1
0.0
-1.3
0.0
0.0
0.0
0.0
-13.1
0.0
0.0
Delta Sulfate (Change from Year 0 to n)
Scenario Mean Std. Dev. Min. P 25 Median P 75 Max.
Constant Depn.
0-10
0-20
0-50
0-100
-8.7
-9.6
-9.7
-9.7
6.5
7.8
8.2
8.2
-30.6
-37.6
-39.7
-39.7
-11.1
•12.0
-12.1
-12.1
•6.4
-6.8
•6.8
-6.8
-4.1
-4.2
-4.2
-4.2
•0.7
•0.7
-0.7
•0.7
Decreased Depn.
0-20 -16.5 8.7 -45.2 -19.0 -14.1 -10.6 -5.8
0-50 -42.8 19.0 -100.7 -52.5 -39.1 -29.4 -16.9
0-100 -42.9 19.3 -102.6 -52.5 -39.1 -29.5 -16.9
9-25
-------
Table 9-3. Summary Statistics for Modelled Changes in Sulfate Concentration, Percent Sulfur
Retention, and Delta Sulfate for Northeast Watersheds Using Typical Year Deposition Data.
Sulfate Concentration
Scenario
Year
Lake SCX
Constant Depn.
0
10
20
50
100
Steady State
Decreased Depn.
20
50
100
Steady State
Mean
110.0
127.8
118.7
117.7
117.6
117.6
117.6
110.6
82.7
2.6
82.3
Std. Dev.
39.5
54.7
48.3
47.2
46.9
46.9
46.9
45.0
32.8
32.6
32.9
Min.
33.8
52.4
50.2
50.1
50.1
50.1
50.1
47.3
36.2
36.2
35.1
Percent Sulfur
Scenario
Year
LakeSO4
Constant D6pn.
0
10
20
SO
100
Decreased Depn.
20
SO
100
Mean
4.2
-7.7
-0.7
>-0.1
-0.1
0.0
-17.4
-0.7
-0.6
Std. Dev.
19.9
3.8
0.9
0.2
<0.1
<0.1
1.6
1.1
1.1
Mln.
-53.6
-18.9
-4.4
-1.0
>-0.1
>-0.1
-21.7
-3.2
-3.2
P_25
81.7
79.6
75.7
75.7
75.7
75.7
75.7
71.1
54.2
54.2
53.0
Retention
P_25
-4.6
-9.5
-0.7
>-0.1
0.0
0.0
-18.5
-0.9
>-0.1
Median
105.4
118.2
111.6
111.2
111.2
111.2
111.2
103.7
77.8
77.8
77.8
Median
0.0
-6.8
•0.4
>-0.1
0.0
0.0
-17.4
-0.1
>-0.1
P_75
130.7
157.6
146.6
146.0
146.0
146.0
146.0
137.0
102.2
102.2
102.2
P_75
17.0
-5.1
-0.2
0.0
0.0
0.0
-16.1
>-0.1
0.0
Max.
213.8
281.3
247.6
240.3
238.8
238.8
238.8
231.3
168.7
167.2
167.2
Max.
68.6
-0.1
0.0
0.0
0.0
0.0
-13.3
0.0
0.0
Scenario
Mean
Delta Sulfate (Change from Year 0 to n)
Std. Dev. Min. P 25 Median
P 75
Decreased Depn.
0-20
0-50
0-100
-17.6
-45.1
-45.2
10.1
22.1
22.4
-50.1
-112.7
-114.5
•20.2
-55.2
-55.3
-14.5
-40.4
-40.4
-10.4
-27.7
-27.7
Max.
Constant Depn.
0-10
0-20
0-50
0-100
-9.1
-10.0
-10.2
-10.2
7.1
8.6
8.9
8.9
-33.7
-41.1
-43.0
-43.1
-12.0
-12.8
•12.9
-12.9
-6.6
-7.0
-7.0
-7.0
-4.1
-4.3
-4.3
-4.3
-0.1
•0.1
-0.1
-0.1
-3.3
-15.4
-15.4
9-26
-------
Northeast Lakes
Year 0 Sulfur
Deposition = Long Term Average, Constant
1.0
0.8
0.6
.
3 .
0.2
0.0
SO 100 150 200
Sulfate Concentration ftieq L*1)
250
0.8
LO.6-
a.
§
I
o
0.4
0.0
-80
-40 0 40
Percent Sulfur Retention
80
Figure 9-5. Comparison of measured, modelled and steady-state sulfate for Northeast lake systems
in 1984.
9-27
-------
modified projections of sulfate concentration, but the range of modelled percent sulfur retention was
again virtually unchanged. The net result of changing deposition/hydrologic fluxes was the introduction
of bias In projected sulfate concentrations, without an expansion in the range of projected percent sulfur
retention to match observed distributions. Systematic changes in soils or deposition data that increased
the ranges of sulfate response (percent retention In the base year) could not be identified without
Introducing bias in projected average sulfate concentrations or percent sulfur retention.
An alternative calibration approach of adjusting soli chemistry data for individual watersheds was
also considered in order to match model projections with measured sulfate concentrations and percent
retention. The success of this approach, attempted for a subset of northeastern watersheds, was
marginal. For watersheds with sulfur retention less than about -25 percent (output >. 125 percent of
Input), no combination of adsorption parameters could match observed retention, unless historic
deposition sequences were altered. For watersheds with positive computed retention, matching modelled
to measured sulfate concentrations required increases in adsorption capacity (Isotherm Emax and/or soil
mass) by a factor of 14 to 24. Concurrent sensitivity analyses for SBRP watersheds (Section 9.2.4.3.1)
indicated that no adjustment of isotherm parameters was necessary or appropriate. It was concluded
that model inputs should not be calibrated, based on these results, that is (1) the lack of bias in average
projections for the NE, (2) substantial adjustments to isotherm data required for matching mean values
and ranges of projected sulfate concentrations and percent retention with measured distributions, and
(3) absence of similar needs for such adjustments for the SBRP, suggesting that there were no systematic
biases). This conclusion, In turn, led to the conclusion that, to the extent that significant deviations from
steady state are currently observed for sulfur in the NE (especially positive retention), they should be
attributed to uncertainties in sulfur input/output budgets or to other retention processes such as in-lake
retention or sulfate reduction in wetlands.
9.2.4.2.2 Projections of future sulfate concentrations -
Projections of future sulfate concentrations and percent sulfur retention in soils of northeastern
watersheds, for periods ranging from 10 to 100 years, are presented in Figures 9-6, 9-7, and 9-8, and
9-28
-------
in Tables 9-2 and 9-3. As previously noted, the reliability of model projection decreases with longer
projection periods. Model projections for periods longer than 50 years are included principally to provide
bounds on potential change after the 50-year period that is the focus of DDRP. Results using both LTA
and TY deposition scenarios, and for both the constant and ramped future deposition sequences, are
Included. Projected sulfate concentrations and percent retention based on the LTA and TY deposition
datasets are very similar; In order to avoid redundancy, therefore, discussion is limited to results based
on the LTA deposition dataset, except to note differences between the two sets of projections.
As expected on the basis of discussion in the preceding section, projected changes for sulfate in
the NE are rapid and times to steady state are short. If current levels of deposition are maintained, the
only projected changes are small declines in sulfate concentrations as watersheds come to steady state.
Within 10 years, sulfate Is projected to decrease from a median of 107.9 to 100.7 percent of steady-
state concentration, and maximum concentration is projected to decrease from 119.3 to 104.7 percent
of steady-state concentration. The corresponding median and maximum declines in sulfate concentrations
are 8.7 jieq L"1 (from 114.3 to 106.2 //eq L"1) and 30.6 ^eq L'1 (249.3 to 218.8 jieq L*1), respectively.
Within 20 years, at constant deposition, sulfate in all northeastern systems is projected to be within 2 peq
L*1, or 1 percent, of steady state.
For the scenarios of a ramped 30 percent decrease in sulfur deposition, similarly short response
times are projected. Most watersheds are projected to have virtually reached steady state with current
deposition by year 10; they begin to respond almost immediately to the reductions in deposition. As
inputs decrease, watersheds begin to re-equilibrate by desorbing sulfate, and projected percent retention
becomes negative. In year 20, during the period of decreased deposition, projected watershed sulfur
outputs are roughly 15-20 percent above inputs. At year 25, when the decrease in inputs ends, however,
systems again quickly re-equilibrate, and at 50 years, projections of watershed sulfur retention are within
1 percent (and 2 jueq L*1) of the new, lower steady-state concentration for all northeastern watersheds.
At year 50, projected changes in sulfate concentration in the ramped deposition scenario are considerably
9-29
-------
Northeast Lakes
Percent Watershed Sulfur Retention
Deposition = Long Term Average, Constant
1.0
0.8.
0.6
0.4
0.2
Northeast Lakes
Percent Watershed Sulfur Retention
Deposition = Long Term Average
Ramped 30% Decrease
0.0
1.0,
0.8,
.0.6.
0.4
0.2.
0.0
-25 -20 -15 -10 -5
Percent Sulfur Retention
-25 -20 -15 -10 -S
Percent Sulfur Retention
Northeast Lakes
Changes In Lake Sulfate Concentration
Deposition « Long Term Average, Constant
Northeast Lakes
Changes In Lake Sulfate Concentration
Deposition s Long Term Average
Ramped 30% Decrease
1.0
0.8
1-0.6.
10,4.
02.
0.0
1-0 ,
0.8
I-0.6
I0.4
-12S
-100
-75 -50
A Sulfate frieqL-')
-25
0.0
-125
•100 -75 -50 -25
A Sulfate JieqL-1)
Figure 9-6. Projected changes in percent sulfur retention and sulfate concentration for soils In
northeastern lake systems at 10, 20, 50 and 100 years. Data are shown for long-term average
deposition for constant and decreased Inputs.
9-30
-------
250-
200—
li 160-
ST
8 10°-
SO-
SULFATE CONCENTRATION
10 20 SO 100
Simulation Yew
Constant Dtpoctfon
20 SO 100
Simulation Year
Decreased Deposition
80-
60-
-to-
PERCENT SULFUR RETENTION
1O 20
Simulation Y«w
ComMant OvpcwUo
GO 100
20 SO 100
Sfcnutadon Y«w
D*CTMM«| Deposition
-ao—
-eo—
CUMULATIVE CHANGE IN SULFATE
O-1O O-2O O-SO O>1OO
SknutaOon Year
O-2O O-6O
Simulation Y«
Figure 9-7. Box-and-whisker plots showing changes in sulfate concentration, percent sulfur
retention, and change in sulfate concentration for soils in northeastern lake watersheds, using long-
term average deposition data.
9-31
-------
300-
250-
200-
150-
100-
80-
SULFATE CONCENTRATION
'Measured
10
20 SO
100
Simulation Yew
Constant Depostipn
20
SO 100
Simulation Year
ecreasedDe
PERCENT SULFUR RETENTION
100-1
son
o-
'-50-
-100-
Measured
10 20 50
Simulation Year
Constant Oepostion
100
20 SO 100
Simulation Year
Decreased Deposition
CUMUUVTtVE CHANGE IN SULFATE
o-
-ao-
-100-
•120
0-10 0-20 0-50
Simulation Year
Constant Oopostion
0-100
0-20 0-50 0-100
Simulation Year
Decreased Deposition
Figure 9-8. Box-and-wtiisker plots showing changes in sulfate concentration, percent sulfur
retention, and change in suifate concentration for soils in northeastern lake watersheds, using TY
deposition data.
9-32
-------
larger than for the constant deposition scenario. The projected decrease in median suifate is 39.1
L"1, with a range of 16.9 to 100.7 /jeq IT1. Projected changes for model runs using the TY deposition
dataset occur over time frames comparable to those for LTA deposition and are slightly larger (median
and maximum of 40 and 114 /jeq L*1, respectively) due to the slightly higher sulfur inputs defined for
most watersheds by the TY dataset. These results indicate that if deposition were reduced, a rapid and
proportional decrease in suifate leaching In soils and reduced suifate flux to surface waters would occur
in the NE. Because projected changes in suifate concentration would result in equivalent increases in
ANC and/or decreases in base cation leaching from watersheds, decreased deposition would result in
substantial increases in ANC or deceleration of base cation removal.
Projected suifate concentrations and percent retention approach steady state asymptotically, and
thus the response times discussed here (although short) are overestimates. The projected annual
changes In suifate concentration and percent retention decrease exponentially as the systems come to
steady state, and rates of change become increasingly small for the last few years. Given the
uncertainties in hydrologic and suifate measurements and the annual variability in watershed suifate fluxes,
95 or 105 percent of steady state is regarded as Indistinguishable from steady state. Time to reach 95
or 105 percent of steady-state concentration is a useful means of describing and comparing watershed
response to altered sulfur deposition. For the current period, in which suifate concentrations are declining
In response to reduced deposition, 48 percent of the systems are projected to be within 5 percent of
steady state at the end of the base year, 75 percent within 2 years, and 100 percent within 9 years
(Figure 9-9). Following the decrease in deposition in the ramp scenario, the most rapidly responding
systems are projected to have suifate concentrations within 5 percent of steady state only 3 years after
the end of the reductions; projected median and maximum times are only 6 and 15 years, respectively,
after the end of the decrease in deposition.
9.2.4.2.3 Summary of results for the Northeast -
Model projections for the northeastern United States, using two deposition datasets and two
scenarios of future deposition, uniformly indicate rapid soil response to past and potential future changes
9-33
-------
Northeast Lakes
Time To Sulfur Steady-State
6 9
Year After 1984
12
1-0,
0.8,
o
I 0.6-
£
£ 0.4.
I
0.2.
0.0
3 6 9 12
Year After End of Ramped Decrease
B
15
Figure 9-9. Projected time to steady-state concentration for sulfate in northeastern lakes (A) at
current deposition and (B) after end of decreasing input in ramp scenario. Results for long-term
average deposition are shown.
9-34
-------
in sulfur deposition to watersheds. At present, watershed sulfur concentrations are projected to be slightly
higher than steady-state concentrations and are decreasing due to recent decreases In deposition.
About half of the watersheds in the NE are estimated to have soils within 5 percent of steady state in the
1984 base year, and all are projected to be within 5 percent of steady state within 9 years of the base
year. For a hypothesized future decrease in deposition, median time to reach steady state (within ± 5%)
was projected to be only 6 years following the end of the decrease in deposition, and maximum projected
response time was only 15 years. These projections lead to the following conclusions:
• To the extent that watershed sulfur budgets in the NE deviate significantly from steady state
(particularly if they are retaining sulfate), the deviations are probably not the result of sorption
reactions in soils, but should be attributed to uncertainties in sulfur input/output budgets, non-
sorption sulfur sources (e.g., sulfide mineral weathering), or alternative retention processes
(e.g., reduction in lakes or wetlands). It is emphasized that this Level II Analysis has
considered sorption by soils as the only process regulating sulfur mobility in watersheds.
Other processes are recognized as having the potential to influence sulfur budgets
significantly In at least a small proportion of watersheds, but their consideration is beyond
the scope of this analysis.
• Watersheds in the NE should be regarded as direct response systems in terms of sulfate
dynamics. Soils have low sorption capacities and therefore can buffer changes in sulfur
inputs for only a very few years. If deposition is reduced, watersheds are expected to
respond with a rapid and proportional decrease in sulfur output.
9.2.4.3 Model Results - Southern Blue Ridge Province
9.2.4.3.1 Evaluation of base year data, model calibration -
In contrast to sulfate chemistry and dynamics in northeastern soils of watersheds, stream systems
in the SBRP are characterized by a wide range of sulfate concentrations and wide variability in percent
sulfur retention. Figure 9-10 shows the deposition sequence used for SBRP watershed modelling, the
sulfate response of a typical SBRP watershed, and the range in projected sulfate responses for stream
systems In the region. The historic deposition sequence used for the SBRP differs considerably from that
used for the NE. Significant increases in sulfur input began relatively recently in the SBRP, and
deposition reached 50 percent of current levels only about 25 years ago. Unlike the historic deposition
scenario for the NE, historic sulfur inputs have never significantly exceeded current levels of deposition.
The lower cumulative deposition to SBRP watersheds and the high sulfate adsorption capacity of many
9-35
-------
SBRP Streams
120i
Deposition Input
Stream Output
Median Response
Range
1850 1900 1950 2000
Year
2050 2100
Figure 9-10. Historic deposition inputs and modelled output for soils in stream systems in the
Southern Blue Ridge Province. Note the much slower response compared to systems in the
Northeast, shown In Figure 9-3. Historic deposition input Is based on emission estimates of
Gschwandtner et al. (1985). Sulfur flux Is expressed on a relative scale, 1985 deposition flux =
100. The base year for projections (1985) is indicated by the arrow. Note that because annual
precipitation and runoff are constant throughout the simulation period, changes in flux correspond
to proportional changes in projected stream sulfate concentration.
9-36
-------
soils In the region are the most important factors affecting the current sulfur budget status of watersheds
in the region. Typical watersheds in the SBRP presently retain over 50 percent of sulfur inputs, but as
shown in Figure 9-10, sulfate concentrations in SBRP watersheds are now projected to be increasing at
a substantial rate (proportional to changes in sulfate flux from the watershed). Measured increases in
sulfate concentration, at rates comparable to those projected for SBRP watersheds in this analysis, have
been reported for several stream systems in the region and have been summarized by Church et al. (in
review). The range of watershed response rates is much broader than that for the NE: a few watersheds
are projected to be already close to steady state, while sulfate concentrations in others are just beginning
to increase and are not likely to reach steady state during the 140-year period.
An early issue in evaluating SBRP model projections was calibration of the model and data for the
region. Because both the response times projected for watersheds and the range of base year
projections are much wider in the SBRP than in the NE, the need for and effects of calibrating input
data are much more obvious for SBRP systems. As for the NE, the model runs were begun at -140
years, continued through the base year (1985, when SBRP streams were sampled in the Pilot Stream
Survey), and then for 300 years into the future. Starting with uncalibrated isotherm data, measured and
modelled sulfate concentrations and percent sulfur retention were compared to evaluate bias and their
distribution. The measured and modelled data were in dose agreement for both sulfate concentration
and percent sulfur retention (Table 9-4 and Figure 9-11). Model projections essentially are unbiased for
both parameters with the average modelled concentration differing by only 2 peg L*1 from the average
for measured data (39.0 vs 36.8 eq L'1) and average retention differing by only 3 percent. Ranges and
standard deviations of model projections also closely approximate those of measured data. Modelled
sulfate concentrations are slightly higher than measured concentrations over most of the observed range,
although modelled concentrations are slightly lower at the high and low ends of the distribution.
Corresponding relationships for percent sulfur retention indicate lower modelled retention over most of
the range. Overall, the two sets of data are very similar, and a small systematic adjustment could be
9-37
-------
Table 9-4. Comparison of Measured and Modelled Base Year (1985)
Sulfate Data for SBRP Watersheds, Using Long-Term Average
Deposition Data. Values Represent Population-Weighted Mean
± 1 Standard Deviation.
Parameter
Measured Value
Modelled Value
Sulfate concentration ( eq L'1)
Percent sulfur retention
36.8 ± 25.7
68.3 ± 16.0
39.0 ±21.0
64.8 + 17.5
9-38
-------
SBRP Streams
Year 0 Sulfur
Deposition = Long Term Average, Constant
1.0. Concentration
0.8 J
0.6.
0.4.
0.2.
0.0
SO 100 150
Sulfate fiaq L'1)
200
1.0. Percent Sulfur Retention
0.8.
0.6.
I '
1 0.4.
I
0.2.
0.0
25 50 75
Percent Sulfur Retention
100
Figure 9-11. Comparison of measured, modelled, and steady-state sulfate for stream systems in
the Southern Blue Ridge Province In 1985.
9-39
-------
made to one of the isotherm coefficients to completely eliminate bias. Using LTA deposition data,
differences between measured and modelled base year sulfate concentration and percent sulfur retention
are very small and are comparable to differences in projections of base year sulfate concentrations and
percent retention using different deposition datasets (LTA, TY). There is thus no compelling rationale
for adjusting either the data or the model. For all subsequent projections, therefore, isotherm data were
used without adjustment.
Comparison of modelled rates of increase in sulfate concentration (for base year 1985) in the DDRP
watersheds to measured rates of increase for watersheds in the region (Table 9-5) indicates generally
good agreement. The range of rates generated by the model for DDRP watersheds encompasses all of
the measured rates; observed rates (except those for watersheds 2 and 18 at Coweeta) are between the
25th and 75th percentiles of the 34 DDRP sample watersheds. The close agreement between observed
and modelled rates of increase provides additional support for the use of isotherm data without extensive
calibration and provides a useful check on the model projections generated from those data.
Concurrent with assessments of the need for data adjustments during model calibration, differences
in model projections resulting from use of different deposition datasets were evaluated. Comparison of
projections based on LTA and TY deposition data (Figure 9-12) reveals systematic but very small
differences. For year zero data (expressed as percent sulfur retention) retention is marginally higher for
the LTA data, whereas projected concentrations are 5 to 10 eq L"1 higher for the TY data; times to
steady state for the two sets of projections are very similar. Time to sulfur steady state is typically 3-4
years shorter for projections based on TY than for LTA deposition data. Given these small differences,
the balance of this discussion will focus only on the long-term results, except to note differences between
the two sets of projections.
9.2.4.3.2 Projections of future sulfate dynamics -
Projections of future sulfur dynamics for the SBRP differ in almost every respect from those for the
NE. Projected sulfate concentrations for the NE are slightly above steady state, and the projected
9-40
-------
Table 9-5. Comparison of Modelled Rates of Increase for [SO,2] in DORP Watersheds in the
SBRP with Measured Rates of Increase in Watersheds in the Blue Ridge and Adjoining
Appalachians.
Site
DDRP watersheds
Cataloochee Cr., NC
Coweeta, NC
WS 2
WS 18
WS 27
WS 36
Deep Run, VA
Madison Run, VA
Femow, WV
WS 4
Period of Record
model-based
estimates
1968-1981
1974-1983
H
H
H
1980-1986
1968,1982
1970-1985
[soA
( eq L'1)
15-119
26
13
13
29
24
100
70
85-90
Rate of
[S04 "] Increase
( eq L yr ) References8
median = 1 .21 this study
range 0.2-2.9
Qg = lift
1.0 a
0.7 b
0.6
0.8
0.8
1.7 c
1.3 d
1.0 e
a References: (a) Smith and Alexander, 1986; (b) Swank and Waide, 1988; J. Waide, personal communication; (c) P. Ryan,
Univ. of VA, personal communication; (d) USGS. 1969, 1970; Lynch and Dise, 1985; (e) D. Helvey, personal communication.
9-41
-------
SBRP Streams
(DDRP watersheds only)
Modelled year 0 percent sulfur retention
Long Term Average
TypSalYear
20 40 60 80
Percent Sulfur Retention
100
1.01
0.8
.g
o
§• 0.6
ft
.1
3
| 0.4
O
0.2
Modelled year 0 sulfate concentration
0.0
— Long Term Average
Typical Year
25 50 75 100
Modelled Sulfate Concentration
125
L'1}
150
Rgure 9-12. Comparison of forecasts based on two sulfur deposition datasets for soils in SBRP
watersheds. Modelled-sulfate concentration and percent sulfur retention for the 1985 base year
are shown for long-term average and TY data.
9-42
-------
response Is small decreases in sulfate concentration as systems move toward steady state over the next
decade. In contrast, most SBRP watersheds are presently far below steady-state concentrations; however,
moderate to large increases in sulfate concentrations are projected over time frames of several decades
to over a century. As previously noted, the reliability of model projections decreases duration of the
projections. The time interval of primary interest to DDRP is 0 to 50 years; projections for longer periods
(100, 140 years) are included principally to characterize the magnitude of potential change following the
50-year projection period. Results using LTA and TY deposition data (Figure 9-13, Tables 9-6 and 9-7),
indicate significant increases in sulfate concentration and corresponding decreases in percent sulfur
retention for most SBRP watersheds within 20 years. At current deposition, the projected increase in
median sulfate concentration at year 20 is 24 peq I'1, with an additional increase of 25 /jeq L'1 by year
50. The range of the increase is 5 to 48 pteq L'1 at 20 years and 15 to 93 peq L'1 at year 50. By year
100, when the average projected total increase for stream sulfate is 66 jueq L*1, most watershed sulfur
budgets are projected to be near steady state; increases after year 100 will be restricted to a small
subset of systems with very long projected response times.
Between years 0 and 20, percent sulfur retention decreases by about 20 percent for soils in most
watersheds, and only a few watersheds approach zero percent retention. Decreases during this period
appear to be controlled by deposition/sorption capacity relationships. After year 20, however, a
substantial number of watersheds are at or very close to steady state, and by year 50 over half of the
SBRP watersheds have less than 10 percent sulfur retention. By year 100, over 75 percent of the
watersheds are within 5 percent of steady state, and most have projected retention of 1 percent or less.
Only a few systems, with very long response times, remain below steady-state concentration by year 140.
Box and whisker diagrams (Figures 9-14 and 9-15) summarize changes in sulfate concentration,
percent sulfur retention, and delta sulfate between 0 and 140 years. These diagrams illustrate not only
the trends for these parameters, but also the relationships among them. In particular, sharp increases
in suifate concentration and in delta sulfate are shown at 20 and 50 years. The increases slow by year
100 as percent sulfur retention approaches zero, constraining further changes in sulfate concentration.
-------
SBRP Streams
Percent Watershed Sulfur Retention
Deposition = Long Term Average,
Constant
SBRP Streams
Percent Watershed Sulfur Retention
Deposition = Long Term Average,
Ramped 20% Increase
1.01
20
40 60
Pweent Sulfur Retention
100
0.0
20
40 60
Percent Sulfur Retention
100
SBRP Streams
Changes In Sulfate Concentration
Deposition = Long Term Average,
Constant
1.01
0.8
1.0.6
0.0
SBRP Streams
Changes in Suffate Concentration
Deposition = Long Term Average,
Ramped 20% Increase
1.01
YMrO-SO
YMfO-100
YMfO- 140
40
60
120
160
200
80 120
A Sulfate ftteq f)
Figure 9-13. Projected changes in percent sulfur retention and in sulfate concentration for stream
systems in the Southern Blue Ridge Province at 0, 20, 50, 100 and 140 years. Data for long-term
average deposition, at constant and Increased deposition, are shown.
9-44
-------
Table 9-6. Summary Statistics for Modelled Changes in Sulfate Concentration, Percent Sulfur
Retention, and Delta Suifate for Watersheds in the Southern Blue Ridge Province, Using Long-
Term Average Deposition Data.
Scenario
Year
Mean Std. Dev.
Sulfate Concentration
Min. P 25 Median
P 75
Max.
Stream SO*
36.8
25.7
14.7
19.8
23.6
40.8
119.2
Constant
0
20
50
100
140
Steady State
Increased
20
SO
100
140
Steady State
Scenario
Year
Stream SO4
39.0
62.7
88.2
104.8
108.9
110.5
63.0
100.9
126.3
131.4
132.6
Mean
68.3
1.0
31.0
33.7
28.6
25.8
24.7
131.3
40.1
33.8
30.4
29.7
Std. Dev.
16.0
12.0
17.2
31.0
65.7
69.5
69.5
17.2
33.1
82.0
83.4
83.4
Percent
Min.
23.7
21.7
36.7
65.4
86.2
86.7
94.9
36.6
72.8
103.5
106.7
113.8
Sulfur
P_25
65.1
35.3
62.3
89.6
103.1
103.6
103.6
62.4
101.4
124.2
124.3
124.3
Retention
Median
74.9
57.5
86.9
111.1
127.0
127.6
127.8
87.6
130.6
152.2
153.1
153.3
P_75
79.1
85.5
134.0
154.0
184.4
189.8
190.4
135.0
179.7
222.5
228.2
228.5
Max.
85.9
Constant
0
20
50
100
140
Increased
20
50
100
140
64.8
43.8
21.1
5.7
1.6
50.1
24.8
5.1
1.0
17.5
24.3
22.7
10.6
3.5
21.7
23.0
9.6
2.2
21.0
3.2
<0.1
<0.1
<0,1
13.0
1.2
<0.1
<0.1
54.3
26.2
4.3
0.2
<0,1
34.5
8.6
0.3
<0.1
69.1
42.8
9.6
0.6
0.1
49.3
13.2
0.6
0.1
78.9
65.0
32.5
3.9
0.8
69.0
36.9
3.4
0.6
83.8
81.9
67.4
30.8
10.0
84.0
71.0
28.0
6.3
Scenario
Mean
Delta Sulfate (Change from Year 0 to n)
Std. Dev. Min. P 25 Median P 75
Max.
Constant
0-20
0-50
0-100
0-140
23.7
49.2
65.8
69.9
11.5
20.4
23.6
24.3
5.2
14.6
14.7
14.7
15.0
39.0
48.2
48.3
23.6
43.8
67.0
71.3
29.4
60.7
73.5
74.4
48.4
93.2
149.1
154.5
Increased
0-20
0-50
0-100
0-140
24.0
61.9
87.5
92.4
11.7
25.3
27.9
8.1
5.2
21.1
28.6
28.6
15.1
48.1
66.4
66.5
24.0
58.4
85.2
94.8
30.4
78.5
101.8
101.9
49.5
113.9
187.2
192.9
9^5
-------
Table 9-7. Summary Statistics for Modelled Changes in Sulfate Concentration,
Percent Sulfur Retention, and Delta Sulfate for Watersheds in the Southern Blue Ridge Province,
Using Typical Year Deposition Data.
Scenario
Year
Stream SO*
Constant
0
20
50
too
140
Steady State
Increased
20
50
100
140
Steady State
Mean
36.8
43.5
71.1
99.8
116.8
120.5
121.7
71.6
115.0
141.2
145.3
146.0
Sulfate Concentration
Std. Dev. Min. P_25 Median
25.7
22.7
32.5
34.3
27.0
23.9
23.0
32.8
41.0
31.5
28.2
27.6
14.7
12.5
18.6
35.4
77.3
86.3
86.4
18.6
38.1
97.8
103.6
103.7
19.8
25.2
45.0
81.7
93.1
97.0
104.0
45.1
92.4
112.3
120.3
124.8
23.6
40.8
72.4
106.5
120.2
120.2
120.2
72.7
123.7
144.2
144.2
144.2
P_75
40.8
57.6
92.4
119.1
131.2
133.3
133.5
92.9
140.3
159.0
159.6
160.2
Max.
119.2
106.5
134.1
171.9
199.2
203.1
203.1
136.3
197.8
240.8
243.8
243.8
Percent Sulfur Retention
Scenario
Year
Stream SO*
Mean
70.8
Std. Dev.
16.7
Min.
17.3
P_25
66.0
Median
78.2
P_75
82.1
Max.
87.0
Constant
0
20
50
100
140
Increased
20
SO
100
140
64.4
41.9
18.8
4.5
1.1
48.4
22.1
3.8
0.6
17.9
24.8
22.4
8.9
2.4
22.2
22.7
7.5
1.3
19.6
2.6
0.3
<0.1
<0.1
12.3
1.0
<0.1
<0.1
54.6
21.6
^7
<0.1
<0.1
30.5
5.6
0.1
<0.1
68.4
38.9
7,0
0.3
0.1
45.8
10.9
0.3
0.1
79.6
64.2
31.1
3.1
0.4
68.4
33.4
2.4
0.3
88.0
82.2
66.0
25.7
6.8
84.2
69.5
21.7
3.7
Scenario Mean
Delta Sulfate (Change from Year 0 to n)
Std. Dev. Min. P 25 Median P
75
Max.
Constant Depn
0-20
0-50
0-100
0-140
27.7
56.4
73.4
77.1
12.3
22.4
25.4
26.3
6.1
16.8
16.9
16.9
19.3
41.3
61.4
61.7
30.9
56.5
66.9
80.8
34.8
74.1
88.0
88.4
49.9
122.5
164.3
168.5
Increased Depn.
0-20
0-50
0-100
0-140
28.2
71.6
97.7
101.8
12.6
27.1
29.0
29.5
6.1
25.7
34.1
34.1
19.9
60.2
81.7
81.8
31.1
69.8
93.2
102.1
35.4
89.8
113.4
113.9
50.5
148.3
205.9
209.5
9-46
-------
300-1
250-
200-
s
g-
A 150-
100-
50-
0
SULFATE CONCENTRATION
LJ
Measured
20 SO 100 140
Simulation Year
Constant Depostion
20 50 100 1<0
SmutattonYear
Incraaaed Deposition
100-]
8O-
i. 60-
40-
20-
PERCENT SULFUR RETENTION
'Measured
20 60 100 140
Simulation Year
Constant DeposSon
SO 100 140
Simulation Year
Increased Deposition
200-1
150—
c
I
a; 100-
SO-
CUMULATIVE CHANGE IN SULFATE
20 50 100 140
Sfmutetfon Year
Constant DepeeOon
20
50 100
SImutetton Year
Increased Deposition
140
Rgure 9-14. Box and whisker plots showing changes in sulfate concentration, percent sulfur
retention, and change in sulfate concentration for soils in watersheds of the Southern Blue Ridge
Province. Data are shown for long-term average deposition data.
9-47
-------
300-
250-
200-
Ij
A IM-
S'
100-
50-
"Measured
SULFATE CONCENTRATION
20 SO 100 140
Simulation Year
Constant Depogton
20 so
100
Simulation Ye v
Incrossfld Deposition
100-
80-
,60-
40-
20 -
L
5
j
. -
f
tHUtNl JSULr-UM Hfcl t
. -"
INI HUN
^
•
S •
Measured
300-
250-
20 50 100
Simulation Year
Constant Dopostfon
140 20 SO 100 140
Simulation Year
Increased Deposition
CUMULATIVE CHANGE IN SULFATE
f~ 200-
1
81S°
< 100-
80-
•
5 ^
LJJ J
r
— ~
1 F
J C
4
ri
J
»
Sbnutatton Year
SO 100
Simulation Year
rwnraaed Deposition
140
Figure 9-15. Box and whisker plots showing changes in sulfate concentration, percent sulfur
retention, and change in sulfate concentration for soils in watersheds of the Southern Blue Ridge
Province. Data are shown for TY deposition data.
9-48
-------
Using the ramped deposition sequence, no differences in status at year 0 are projected, and
differences in sulfate concentration between constant and ramped scenarios at year 20 are insignificant
(1 peq I'1 or less). Because increases in sulfur input are not matched by enhanced sulfur output at year
20, percent sulfur retention for year 20 Is higher for the increased deposition scenario than for constant
deposition. Major effects of the increased deposition are seen in year 50 projections. Projected sulfate
concentrations for year 50 are typically 12-15 /jeq L'1 higher for the scenario with increased deposition
than with constant deposition, whereas percent sulfur retention is only slightly higher for the projections
with increased deposition. By year 100, almost all of the increase in deposition can be observed as an
increase in projected sulfate concentration; percent sulfur retention is comparable to, and in most cases,
actually lower for the increased deposition scenario forecasts than for the constant deposition projections
(Tables 9-6 and 9-7).
Figure 9-16 illustrates projected time to sulfur steady state (+. 5 percent of steady state) for current
and increased deposition scenarios. At current deposition levels, soils in SBRP watersheds are projected
to reach steady state in as little as 16 years after the base year, with a roughly linear increase in the
proportion of systems at steady state until year 75, when about 75 percent will be at steady state.
Following year 75, the increase in the number of systems at steady state is slower, with about 95 percent
of watersheds reaching steady state by the final year of model forecasts, year 140. For the systems that
reach steady state in more than about 60 years, increased deposition negligibly changes times to steady
state. For those systems projected to reach steady state In less than 60 years, especially those that
respond most quickly, increased deposition delays time to steady state. Higher deposition, coupled with
modest delays in increased watershed sulfur output, maintain these systems below steady state for as
long as two decades. The results for these watersheds do not correspond to lower stream sulfate
concentrations. Higher input simply results in a higher input to output ratio; projected stream suffate
concentrations are in all cases the same or higher for the increased deposition scenario than for current
deposition.
9-49
-------
SBRP Streams
Time To Steady-State
1.(
0.8
Q
Q.
0.6
0.4
O
0.2
0.0
Deposition => LTA Constant
0 20 40 60 80 100 120 140
Years To 95% Steady-State
0.8
.2
,0.6
o_
.1
iO.4
0.2-
0.0
Deposition « LTA, Ramped 20% Increase
20 40 60 80 100
Years To 95% Steady-State
120 140
Figure 9-16. Projected time to 95 percent of steady-state sulfur concentration of Southern Blue
Ridge Province stream systems. Results for long-term average deposition, for constant and
Increased deposition scenarios, are shown.
9-50
-------
The magnitude and consequences of the projected changes in sulfate over the next 20 to 100 years
on overall stream chemistry In the SBRP are substantial. The projected changes represent 50 to 100
percent increases In sulfate concentration within 20 years and, on average, about a threefold increase
in sulfate concentration when systems reach steady state. Increases in suifate of this magnitude will
cause major changes in surface water base cations and/or ANC and will accelerate base cation leaching
from soils, as discussed in Sections 9.3 and 10.
9.2.4.4 Uncertainty Analyses and Alternative Aggregation Approaches
9.2.4.4.1 Uncertainty analyses -
As noted in Section 9.2.3.4. partitioning coefficients (isotherm slopes) for sulfate equilibrium between
soil and solution phases are nonlinear; because the coefficients of these nonlinear isotherms are highly
correlated, generation of weighted averages of isotherm coefficients is not an effective or appropriate
means of aggregating isotherm data for soils. The procedures for aggregation of isotherm data described
in Section 9.2.3.4 are not conducive to direct estimation of uncertainty for adsorption isotherm coefficients
or for derived variables such as isotherm slope. The development of uncertainty estimates for Level II
sulfate projections thus required an alternative approach to data aggregation and use in model runs.
Section 9.2.3.4 described a Monte Carlo procedure for generating uncertainty estimates for model
projections. The procedure, however, also involved derivation of new isotherm coefficients and model
projections for each DORP watershed that were developed independently of projections that used data
aggregated by the original mass-weighting approach. Because the two sets of independent coefficients
and the projections generated from them could significantly differ, the initial concern in uncertainty
analyses was to assess comparability of the two sets of mode! outputs. Direct comparison of coefficients
was not possible, since the uncertainty analysis generated a new aggregate isotherm for each of the 100
model runs. Moreover, such an analysis would have been inconclusive since two sets of very different
isotherm coefficients can describe virtually identical partitioning curves over the range of dissolved sulfate
concentrations of interest here (0 to 300 peq L*1). Instead, the mean and median values (which are
virtually identical) from the Monte Carlo simulations for each SBRP watershed were compared to
9-51
-------
projections generated using the aggregate isotherm from the original mass-weighting aggregation.
Comparisons were made for projected sulfate concentration and projections of time to sulfur steady
state for several reference years.
Results indicate very close agreement between the two sets of projections for the base year and
time to steady state (Figure 9-17). Results for sulfate concentrations at other time intervals also were
similar. For the comparisons of concentration at year 0 and for projected time to steady state, slopes
and intercepts of the two lines are virtually equal to 1.0 and 0, respectively, and the coefficients of
determination exceed 0.99 in both cases. These results are important for two reasons: (1) they
demonstrate no fundamental inconsistencies between the aggregation and uncertainty procedures used
to generate the two sets of projections and that the uncertainties developed using the Monte Carlo
approach can be used to characterize uncertainty for projections and summaries that use data
aggregated by the routine aggregation approach; and (2) they also suggest that the adsorption isotherms
and the projections generated from those isotherms are highly constrained, I.e., two different and
independent data aggregations generate virtually identical projections.
Mean values and confidence intervals for projected base year sulfate concentrations for stream
systems in the SBRP are shown in Figure 9-18. Uncertainties are generally modest in magnitude, and
upper and lower confidence intervals are almost symmetrical and are within 10 to 15 peq L*1 of the mean
sulfate concentration. Uncertainties increase very little with mean projected sulfate concentration. Only
uncertainties in sulfate adsorption capacity (Including those in both the original least squares fitting of
isotherms to raw data points and data aggregation) and soil mass were considered. Separate analyses
of the components of uncertainty for four SBRP watersheds Indicate that uncertainty in soil mass is the
primary contributor to the total variability in base year projections of sulfate concentration; upper and
lower confidence bounds are within 5 percent of the median sulfate concentration for the four sets of
Monte Carlo simulations in which soil mass was held constant.
9-52
-------
SBRP Streams
(DDRP watersheds only)
Deposition = Long Term Average, Constant
CO
o
100
D>
j
601
8 40-
20-
O
CO
Modelled Year 0 Sulfate Concentration
1:1
y = 0.103+ 1.0098X
r*n 0.9981
20 40 60 80 100
SO4 (|ieq L1 ) - Median of Monte Carlo Simulation
140
Q 12°
c
1 100
§> 80
i-
20
Time to Sulfur Steady-State
y = -0.258 +0.9918x
r2e 0.9923
20 40 60 80 100 120
Years - Median of Monte Carlo Simulation
140
Figure 9-17. Comparison of model simulation results for DDRP Southern Blue Ridge watersheds.
Data generated by the mass-weighting common aggregation approach and median projected values
from Monte Carlo uncertainty analyses are shown.
9-53
-------
SBRP Streams
Year 0 Sulfate
Deposition = Long Term Average, Constant
0.8-
.0
10.6
ol
05
I 0.4
O
0.2
0.0
Upper Bound
Projected Distribution
Lower Bound
20 40 60 80
Sulfate Concentration (M-eq L'1}
100
Rgure 9-18. Projected base year sulfate concentration with upper and lower bounds for 90 percent
confidence intervals for Southern Blue Ridge Province watersheds.
9-54
-------
Projections of mean time to sulfate steady state in SBRP watersheds (with 5 and 95 percent
confidence intervals) are shown in Figure 9-19. Similar to base year results, confidence intervals are
almost symmetrical; but in contrast, uncertainties in response time increase with mean projected
concentration. Relative uncertainties are smaller than those observed for the base year concentration,
averaging only about 20 percent of the mean time interval. As was the case for the base year evaluation,
uncertainty in soil mass is the largest contributor to uncertainty In the projections; confidence intervals
for projected time to steady state vary by less than 5 percent of the median value in Monte Carlo
analyses for which soil mass was held constant. This result provides additional confirmation that the
isotherm fits are highly constrained and also highlights the important influence of uncertainty in soil mass
on projections involving changes in capacity variables.
9.2.4.4.2 Alternative aggregation approaches -
Uncertainties in aggregated data associated with the method of aggregation also were considered.
Use of the sulfate subroutine in MAGIC requires aggregating data for the entire watershed into one
compartment per soil horizon per watershed. Aggregating data for a variety of soils with differing
chemistry, vegetation, hydraulic contact times, etc., inevitably introduces errors and uncertainty. Johnson
et al. (I988b) discussed the rationale for several aggregation procedures and described the mass
weighting approach based on soil sample classes that was used for routine data aggregation for this
analysis. They did not address the possibility of adjusting aggregated data to account for differences in
soil chemistry at intermediate spatial scales, nor did they address other approaches to describe
"watershed effects" on aggregated soil chemistry (Section 8.9). As one means of assessing possible
watershed effects, date for the pedons sampled on individual DDRP watersheds in the SBRP were
aggregated and were used with LTA deposition data to make projections. Measured chemistry and
projections using the standard aggregation approach described by Johnson et al. (I988b) differed
considerably (Figure 9-20). Projections using the pedons-on-the-watershed approach substantially
overpredicted year zero sulfate concentration (modelled mean S042" = 67.4 jueq L*1 vs. measured mean
O 4
SO4 ' = 36.0 fjeq L") and underpredicted percent sulfur retention (median retention of 38 percent vs.
9-55
-------
SBRP Streams
Time to Sulfur Steady State
Deposition = Long Term Average, Constant
0.8-
c
.2
8.0.6
2
Q_
-------
SBRP Streams
(DDRP watersheds only)
Deposition = Long Term Average, Constant
1.0
0.8
.0.6-
Jo 0.4-
3
I
o
0.2
0.0
Year 0 Sulfate Concentration
Measured
Modelled
Sample Class agg.
Pedon on Watershed agg.
20 40 60 80 tOO
Sulfate flieq L1)
120
140
160
1.0
0.8-
0.0
Years to Sulfur Steady-State
Sample Class agg.
Pedon on Watershed agg.
20 40 60 80 100
Years to 95% Steady-State
120
140
Figure &-20. Effects of data aggregation on simulated watershed sulfur response for soils in DDRP
watersheds of the Southern Blue Ridge Province. Results for the common (sample class)
aggregation procedure and for an alternative aggregation using pedons sampled on each watershed
are shown.
9-57
-------
75 percent for measured retention and 69 percent for standard aggregation projections). Projections
using the alternative approach also indicated that almost 20 percent of SBRP watersheds are already at
sulfate steady state, in contrast to measured watershed sulfur retention, which indicates that more than
20 percent are at steady state. Additionally, model simulations based on the standard aggregation
approach project all watersheds to have >20 percent retention In the base year and no watersheds to
reach steady state for 16 years.
The overpredictions using the alternate aggregation (pedon-based) result from the soil sampling
design in the SBRP. The design called for approximately equal numbers of samples for each soil
sampling class, even though the spatial areas covered by the classes are widely variable. As a result,
for this aggregation approach, in which data were arithmetically averaged, soils collected from sample
classes with below-average area! spatial coverage were assigned artificially high weights. Because sample
classes with small area! coverage (e.g., shallow or flooded soils) often have low sulfate adsorption
capacity, averaged adsorption capacities are biased low and corresponding projected response times are
short. This component of uncertainty is not introduced in projections using the common aggregation
approach, because data are area weighted; the alternative aggregation approach, moreover, raised
questions about sampling design and the magnitude of uncertainties in parameter estimates. For these
reasons, it was dropped from consideration for this analysis. The issue of watershed effects (watershed-
to-watershed differences in soil chemistry) is under active investigation, and changes in aggregation
procedures for future analyses remain a possibility.
A second question concerns the number of soil horizons used for aggregation. The mode!
formulation treats soils as a series of continuously stirred tank reactors (CSTRs) in which reactions
proceed to equilibrium. This treatment results in model projections that are sensitive to the number of
CSTRs. At one extreme, for a one-CSTR model, projected output concentrations respond immediately
to changes in inputs and responses are sustained over a long period of time. At the other extreme, a
model with an infinite number of small CSTRs (having a total soil mass equivalent to that for the one-
compartment model) would act much like a chromatography column: output remains constant until the
9-58
-------
breakthrough of the front through the soil column, at which time output concentration increases as a
square wave (ignoring dispersion) to steady state. The number of horizons, which can be varied for the
Level It Analysis, will affect the timing of projected changes as well as the concentration at any point in
time (base year in this case). Results of running the model with soil chemistry data aggregated to one,
two, and three horizons (A/E, B, and C) are displayed in Figure 9-21. Few differences between
projections for two- and three-horizon aggregations are evident, but one-horizon projections do differ.
The close agreement between two- and three-horizon projections was not unexpected; the A/E horizon
is thin and has relatively low adsorption capacity (i.e., the A/E horizon CSTR has a very short response
time), so combining It with B horizon data has little effect on projections. On the other hand, combining
ail data In one horizon results in a system that responds immediately to altered inputs (thus resulting in
projections of higher base year sulfate concentrations); it also results in a system that responds more
gradually than a multiple-horizon model, projecting longer times to steady state. Model data provide no
real basis for choosing among one-, two-, or three-horizon models for routine modelling efforts. Based
on differences in soil chemistry among A/E, B, and C horizons and on the good fit between measured
and modelled projections of base year sulfate concentration and percent sulfur retention using the three-
horizon model, the three-horizon model was chosen for routine analysis.
9.2.4.5 Summary of Results from the Southern Blue Ridge Province
The response of soils and surface waters in the SBRP to sulfate deposition represents a textbook
example" of delayed response watersheds. The response to major sulfur deposition increases that have
occurred over the last two to three decades has been high watershed sulfur retention, with only modest
Increases in stream sulfate concentrations for most SBRP watersheds. Measured data summarized by
Church et al. (in review) and model projections indicate, however, that the delay Is now ending and that
surface water sulfate concentrations are increasing at rates projected to accelerate over the next few
decades. Major increases in stream sulfate concentration are projected for SBRP streams in the next
20 years, with continued increases for at least 50 years for most watersheds. When SBRP watersheds
come to steady state for sulfate (at projected times ranging from 16 to > 140 years) sulfate concentrations
9-59
-------
SBRP Streams
(DDRP watersheds only)
Deposition = Long Term Average, Constant
1.0
0.8
.0.6
I 0.4
O
0.2
0.0
Modelled year 0 percent sulfur retention
3 Horizons A/E.B.C
2 Horizons A/BB.C
1 Horizon A^ftC
20 40 60 80 100
Percent Sulfur Retention
1.0-i Time to sulfur steady-state
0.8
0.6-
= 0.4
0.2
0.0
3 Horizons A/E.B.C
2 Horizons A/E/8.C
•"••••" 1
25 50 75 100 125 150
Years To 95% Steady-State
Figure 9-21. Evaluation of alternate soil aggregation procedures for soils in SBRP watersheds.
Modelled percent sulfur retention for the 1985 base year and time to 95 percent steady state for
soils aggregated to one, two, or three discrete horizons.
9-60
-------
will be, on average, about three times current concentrations. The projected changes in stream sulfate
will result in substantial changes in streamwater chemistry and could substantially accelerate base cation
leaching from soils.
The results of these analyses are generally consistent with those of other DORP analyses. Model
projections of base year sulfate in soils of northeastern and SBRP watersheds are consistent with, and
provide a mechanistic explanation for, analyses by Rochelle and Church (1987). Their analyses,
summarized In Section 7.3, show watersheds in the northeastern United States to be at or near sulfur
steady state, whereas SBRP watersheds have high net sulfur retention. The very short sulfate response
times projected for the NE are also consistent with results of regression analyses presented in Sections
7 and 8, which Indicate that deposition is the principal control on surface water sulfate in the NE and that
significant sulfur retention (where observed) is probably attributable to sulfate reduction in lakes and/or
wetlands rather than to sorption.
The short sulfur response times projected for northeastern soils in this analysis are comparable to
watershed response times projected by integrated models (Section 10). Projected response times for the
SBRP are roughly comparable to those generated for SBRP watersheds by MAGIC (Section 10.11),
although the projections of time to steady state generated in Level II Analyses for SBRP systems are
generally somewhat shorter than the MAGIC forecasts. Two factors are believed to contribute to the
differences in projections:
• Hydrologic routing in the two models is different. The Level II projections used a simplified
routing in which all water was routed through all soil horizons, while a substantial portion of
runoff in MAGIC simulations bypassed either the upper or lower soil compartment.
• Soil depth was treated differently by the two modelling efforts. Level II models considered
adsorption within the top 1.5 to 2 meters of the regolith while the Level III models assigned
the adsorption capacity (and other chemical properties) of the lower soil horizon to all
material between the B-C horizon boundary and the estimated depth to bedrock. The MAGIC
projections are therefore based on a larger mass of soil, having a larger integrated pedon
adsorption capacity, which ultimately results in a slower projected response to changes in
sulfur deposition.
9-61
-------
Despite the differences between the two modelling approaches (Level II limited to adsorption in the
developed soils as compared to Level III models which integrate hydrologic processes with
biogeochemical processes in the entire catchment), the magnitude of differences between the two sets
of projections was generally small. Results are viewed as mutually supportive, both of the two modelling
approaches and of the projections generated by them.
9.2.5 Summary Comments on Level II Sulfate Analyses
At the start of the DDRP. it was widely believed that soils in the northeastern United States had low
sulfate adsorption capacity, resulting in rapid suifate response to changes in sulfur deposition and further
resulting in watershed sulfur budgets near steady state. Conversely, observed sulfur retention in
southeastern watersheds was attributed to high sulfate adsorption capacity of soils in that region.
Measured sulfate data for the two regions and model forecasts summarized in Table 9-8 are consistent
with previous soil sulfate data and provide strong support for this paradigm of regional sulfur dynamics.
DDRP model projections also suggest fundamental differences in future sulfate dynamics of the two
regions. Northeastern watersheds are very close to steady state; assuming constant deposition at current
levels for the future, only small changes in suifate concentration are anticipated as systems reach
equilibrium with deposition inputs. If deposition were to change in the future, model projections suggest
very rapid response by watersheds in the region, with systems projected to reach steady state with the
altered deposition inputs in 5 to 15 years.
in the SBRP, sulfate adsorption by soils has delayed effects of acidic deposition, but model
projections indicate that soils and watersheds in the region are now moving into a more dynamic phase,
in which relative adsorption by soils will decline and stream sulfate concentrations will increase sharply
in the coming decades. Major changes in stream water sulfate are projected for the next 20-50 years.
If and when they occur, equivalent changes in surface water base cations or ANC are inevitable, and
enhanced teaching of soil base cations is likely to occur.
9-62
-------
Table 9-8. Summary Comparison of Watershed Sulfur Status and Model Forecasts
in the Northeastern United States and Southern Blue Ridge Province. Model Projections are Based
on Long-Term Average Deposition Data, Assuming Constant Future Deposition
NE Lakes
Median Range
SBRP Streams
Median Range
CURRENT STATUS
Steady State Sulfate
Lake/Stream Sulfate
Percent Sulfur Retention
MODEL FORECASTS
Sulfate Cone. 0*eq L"1)
Year 0
Year 20
Year 50
Year 100
Year 140
Percent Sulfur Retention
Year 0
Year 20
Year 50
Year 100
Year 140
106.0
105.4
-3.1
114.3
106.0
106.0
106.0
-7.9
-0.1
50.8 - 209.6
33.8 - 249.3
-60.0 - 61.1
54.7 - 249.3
50.8-211.7
50.8 - 209.6
50.8 - 209.6
-19.3
-1.1
-0.2
Delta Sulfate ( eq L
Year 0-20
Year 0-50
Year 0-100
Year 0-140
-i
-6.8
-6.8
-6.8
--1.3
- 0.1
- 0.1
- 0.1
-37.6 --0.7
-39.7 --0.7
-39.7 - -0.7
103.5
23.6
74.9
35.3
62.3
89.6
103.1
103.6
69.1
42.8
9.6
0.6
0.1
23.6
43.8
67.0
71.3
69.5 - 190.4
14.7 - 119.2
23.7- 85.9
12.0 - 85.5
17.2 - 134.0
31.0 - 153.9
65.7 - 184.4
69.5 - 189.8
21.0 - 83.8
3.2 - 81.9
<0.1 - 67.4
<0.1 - 30.8
<0.1 - 10.0
5.2 - 48.4
14.6 - 93.2
14.7 - 149.1
14.7 - 154.5
9-63
-------
In a qualitative sense, the differences the current status and projected future sulfate dynamics for
the two regions are unequivocal. Sulfur budget status and soil adsorption isotherm data document clear
differences in soil and surface water chemistry for the two regions, and projections of future response
times differ by roughly an order of magnitude. In making such comparisons, it is important to recognize
that the models embody a variety of assumptions and approximations and that the projections carry
significant uncertainty. Given the magnitude of the differences in projected responses for the two regions,
however, their responses to sulfur deposition undoubtedly are also very different. Sulfate retention
appears to have been a minor contributor to delays in surface water response to acidic deposition in the
NE, but has been and continues to be a critical process delaying effects of deposition in the SBRP. In
considering the projected responses for sulfur, especially in the SBRP, it is important to recognize that
projections presented here apply only to sulfate and are based on the assumption that adsorption and
desorptlon are the only processes influencing watershed sulfur retention. Finally, readers should be
cognizant of the complexity in the relationships among sulfate, base cations, and ANC, which are
influenced by several processes. The timing and magnitude of changes in ANC need not be directly
correlated with changes in sulfate. In particular, time to steady state for sulfate should not be equated
to time to zero (or to any other threshold value) ANC.
9.2.6 Conclusions
Watersheds in the northeastern United States can be characterized as direct response systems in
terms of sulfate dynamics mediated by sorption in soils. Northeastern watersheds are near sulfur steady
state and are projected to respond quickly to changes in sulfur deposition.
• In the base year (1984), median measured percent sulfur retention was -3.1 percent for LTA
deposition, 0 percent for TY deposition data.
• Modelled percent sulfur retention for the base year was slightly negative for both LTA and
TY data, -7.1 and -6.8 percent, respectively.
If deposition continues at current levels, all northeastern watersheds are projected to be within
steady state in less than 10 years. Median sulft
percent 2 /ieq L ) of steady state in 20 years.
*• W|^W*»IW> • Wl 1111 >MWW «* *f*A* • W« 1» *WVWIM) **II • IVI M «V*****»1
5 percent of steady state in less than 10 years. Median sulfur concentrations will decrease
to within one percent 2 weq L ) of steady state in
At current deposition, changes in median sulfate concentration as watersheds reach steady
state will be small (7 /*eq L ) with a maximum of 40 eq L (for LTA deposition); for TY
deposition, median and maximum projected changes in sulfate are 7 and 43 eq L'1,
9-64
-------
respectively. Changes will have little impact on overall water chemistry of most lakes in the
region.
* If deposition is decreased 30 percent, the magnitude of changes would be much greater, with
a decrease in median sulfate of 39 or 40 peq L (LTA and TY deposition, respectively).
Maximum projected decreases are 103 and 115 eq L"1 for the two deposition datasets.
• Projected response of northeastern lake sulfate concentrations to future changes in
deposition, such as the ramped deposition decrease, are rapid; model forecasts predict all
watersheds to be within 5 percent of steady-state sulfate concentrations within 15 years after
the end of the deposition decrease.
• Based on the rapid projected response of northeastern watersheds, sulfate adsorption has
played a minor role in the delay of surface water acidification in the NE.
• Large deviations from sulfur steady state observed for a few watersheds in the base year
cannot be attributed to sorption. Alternative explanations include uncertainties in sulfate
input/output budgets, internal watershed sulfur sources, and watershed retention processes,
principally sulfate reduction in wetlands and lakes.
Watersheds in the Southern Blue Ridge Province should be characterized as delayed response
systems. Sulfate adsorption by soils has minimized the effects of acidic deposition on surface water
chemistry in the region. Sulfate concentrations in SBRP watersheds are projected to increase significantly
in the next 20 to 50 years, however, as the adsorption capacity of soils is exhausted.
* Median measured sulfate retention in SBRP watersheds for the 1985 base year was 74.9
percent for LTA deposition, 78.2 percent for TY deposition data. The percent retention varies
from 24 to 86 percent for LTA deposition.
• Median modelled retention for the base year is also high: 69.1 and 68.4 percent for LTA and
TY deposition datasets, respectively. The range of modelled percent retention for LTA
deposition was 21 to 84 percent.
* Time to steady state at current deposition varies from 16 to more than 140 years; median
projected time is 61 years. At 20, 50, 100, and 140 years from the base year, projected
median percent sulfur retention is 43, 9.6, 0.6, and 0.1 percent. Maximum projected retention
for the same periods is 82, 67, 31, and 10 percent.
• As soils In SBRP watersheds reach steady state, average sulfate concentrations in watershed
runoff will increase roughly threefold. The median sulfate concentration is projected to
increase from 35 /*eq L1 to 62, 90. 103, and 104 eq L'1 at 20, 50, 100, and 140 years.
Maximum projected increases for sulfate for the same periods are 48, 93,149, and 155 eq
L-\
• For the 20 percent increase in deposition in the SBRP, times to steady state increase by up
to 20 years for the watersheds with short retention times, but are almost unchanged for most
watersheds. The increase in deposition has almost no effect on projected sulfate
concentrations at year 20, but results in significant increases in delta sulfate at later years
compared to the constant deposition scenario. Projected increases in median sulfate
concentration for the increased deposition scenario at years 50,100, and 140 are 58,85, and
95 peq L : maximum projected increases are 114, 187, and 193 eq L .
9-65
-------
* Model projections indicate that adsorption of sulfate by soils has played a major role in
delaying potential adverse effects of sulfur deposition on surface waters in the SBRP. Most
SBRP watersheds wili not reach sulfate steady state for several decades, but significant
increases in sulfate concentration are projected for the next 20 to 50 years.
• The large increases in sulfate concentration projected for the next 20 to 50 years will have
major implications for overall surface water chemistry and are likely to accelerate base cation
leaching from soils.
9.3 EFFECT OF CATION EXCHANGE AND WEATHERING ON SYSTEM RESPONSE
9.3.1 Introduction
During the development of the MAS panel report (MAS, 1984), much discussion was devoted to the
role of cation exchange and weathering in "protecting* watersheds from acidification. One group of panel
members argued that cation exchange in most watersheds has a huge capacity to buffer changes caused
by acidic deposition. Therefore, they argued, if cation exchange were an important process within a
specific watershed, then future changes as a result of acidic deposition were probably not a concern.
Another group of panel members argued that the buffering capacity of soils was finite, and that continued
exposure to current levels of deposition would have long-term, adverse effects on water quality in some
systems. The conclusion of the committee as a whole was that the rde of cation exchange in buffering
against the effects of acidic deposition is an area of considerable uncertainty, and that these processes
need to be considered when attempting to project future effects of acidic deposition on aquatic
ecosystems.
Toward this goal, the Level II base cation studies were designed to determine the role of base
cation exchange in controlling future changes in surface water chemical composition. The specific
objectives were to
• identify the role that base cation exchange has in determining current surface water
composition;
• determine the capacity of base cation exchange processes to buffer against future changes
in surface water composition as a result of acidic deposition; and
• make projections regarding the magnitude and extent of changes that could occur in
regionally representative soils and surface waters as a result of continued exposure to acidic
deposition.
9-66
-------
Background Information concerning weathering and base cation exchange processes was presented
in Section 3.4. Given the objectives stated above and relying on our current understanding of the
processes presented in Section 3.4, a number of hypotheses were developed that were used to guide
the investigations of the role of cation supply processes In regulating surface water chemistry In
representative watersheds in the DORP study regions.
9.3.1.1 Level II Hypotheses
Five hypotheses guided the Investigations In the Level II analyses:
(1) Cation exchange processes determine surface water composition.
(2) Soils delay surface water acidification.
(3) Increased deposition induces net cation leaching.
(4) Cation resuppty rate is slow.
(5) Soil chemistry is an Indicator of soil response to acidic deposition.
It was not possible in all cases to test the hypotheses with the survey data collected for the Project. For
example, testing of the fourth hypothesis requires time series data, the collection of which was beyond
the scope of the DDRP. As a result, this hypothesis was treated as a "system-level" assumption for the
analyses, the implications of which are discussed below.
9.3.1.1.1 Cation exchange processes determine surface water composition -
The first hypothesis, that cation exchange processes regulate observed surface water composition,
is designed to identify the primary process or processes that regulate surface water chemistry. In
systems that have attained steady state with respect to sulfate deposition (see Sections 7 and 9.2),
primary mineral weathering and biological uptake are probably the principal processes that modify the
composition of incident deposition. (Under steady-state conditions, the base cation exchange pool should
actively reflect the dynamic balance between these two important processes.) Regardless of their relative
importance, however, if soils are the media that regulate surface water ANC values, then this should be
reflected by the composition and chemical properties of the soil exchange complex.
9-67
-------
The hypothesis is tested by comparing surface water composition projected using soil cation
exchange models with observed values. A close correspondence between the observed and projected
values suggests that soil exchange processes have a major role in regulating surface water chemistry.
Major discrepancies between observed and projected values would provide Information regarding
alternative controlling processes. For example, if the outputs from the soil models suggest that soils are
strong hydrogen Ion buffers, i.e., if the aggregated model results fall into narrow ranges of pH and ANC,
this would suggest that other processes, such as primary mineral weathering, are serving as major
sources (or, for cation accretion into biomass, sinks) for base cations in the population of watersheds
being studied. Examining this hypothesis, therefore, provides bounds for arguments regarding which
processes are primarily involved in regulating observed surface water composition.
9.3.1.1.2 Soils delay surface water acidification -
The second major hypothesis is that soils will delay, but not prevent, the acidification of surface
waters. The concept behind this hypothesis Is that soils have a finite capacity to buffer against changes
in surface water chemistry caused by increased levels of acidic deposition. In essence, the chemical and
physical characteristics of a soil reflect a soil's response to some given set of environmental conditions.
Therefore, at a given level of deposition, vegetative uptake, mineral weathering, etc., the cation exchange
pool reflects a balance of the various sources and sinks for cations in that area. This balance is dynamic,
changing seasonally and with the shifting flow of cations among the various reservoirs.
When a perturbation such as acidic deposition is imposed on a system, the system (in this case
the soil) evolves toward a new state of balance. The rate at which changes take place depends both
on the sizes of the cation reservoirs in the system and on the flux of material between reservoirs. If the
transfer rates of material between reservoirs is slow, or if the mass of material in the affected reservoirs
is large relative to the transfer rates, then rates of evolution toward the new balance point tend to be
slow. Conversely, if reservoirs are small relative to the size of the flux between reservoirs, then
adjustments to a new system state can occur rapidly.
9-68
-------
We contend, with this hypothesis, that the pool of base cations on soil exchange complexes is
large relative to the rate of cation loss from the system by leaching. As a result, the rate of adjustment
of the exchange complex to the new deposition conditions should require years to decades before a new
steady-state, or dynamic balance, condition is attained.
To test the hypothesis, a model approach is used. Measured soil properties serve as inputs to the
various models. The Level II models, all of which have a mass balance component, track the loss of
base cations from soils at the specified levels of deposition. The models are primarily concerned with
exchange processes and do not explicitly include cation supply via weathering. Therefore, the computed
mass balances should correspond to the maximum leaching rates that could occur. The rate of change
of base cation status of the soils included in the study, then, should be related to the amount of time
over which the soils should delay acidification of surface waters.
9.3.1.1.3 Increased deposition induces net cation leaching -
The third major hypothesis is that increased levels of deposition, specifically Increased
concentrations of sulfate and nitrate In deposition, increase the rate of cation leaching from the soil
exchange complex by way of the mobile anion effect (Johnson et ai., 1980; Seip, 1980). Two factors are
considered when evaluating this hypothesis. First, the average base status of the soil exchange complex
represents a balance among the various supply and demand processes in the ecosystem. For example,
under steady-state conditions, weathering should supply sufficient base cations to meet the demands of
vegetative uptake while maintaining soil solution concentrations in equilibrium with the soil exchange
complex. Perturbations to the system, such as changed deposition, will after this steady-state condition.
Second, charge balance requirements need to be maintained between the soil exchange complex
and soil solutions. Maintaining charge balance, coupled with the increased anion loads provided by
acidic deposition, requires that total (acid plus base) cation concentrations in soil solutions increase. The
ratio of the base to acid cations will not change dramatically, however, at least during the initial stages
of leaching. The higher concentrations of base cations in soil solutions lead to a net depletion of base
9-69
-------
cations from the exchange complex. If this increased leaching is not matched by an increased level of
supply (e.g., from weathering), then the overall effect will be a net depletion of the base cations from the
exchange complex.
We have tested this hypothesis using a modelling approach. As for the second hypothesis, model
runs are conducted that enable the determination of whether the increased anion concentrations In
deposition will, indeed, result in an increased rate of leaching of the base cations. The mass balance
computations, in combination with the equilibrium mass action descriptions of the system, should permit
an unequivocal evaluation of this concept.
9.3.1.1.4 Cation resupply is slow -
The fourth hypothesis, that the rates of cation resupply to the soil exchange complex are slow
relative to the rates of base cation stripping, is not being tested directly in this study. Rather, the
hypothesis is being subsumed in the models as an assumption, or, more accurately, the assumption is
that exchange reactions provide sufficient buffering such that resupply rates are not an Issue for the time
scales of concern to the study.
One reason for using this approach is that, with current technology, no definitive method exists for
distinguishing the different sources of base cations to surface waters. Therefore, by assuming that all
base cations are derived from exchange sites, the modelling yields, effectively, "worst case* scenarios for
the depletion of the soil buffering capacity. If, under these circumstances, the results suggest an
extensive capacity of the soil to buffer against the effects of acidic deposition, then the resupply rate Is
not an issue of Importance in this study.
9.3.1.1.5 Soil chemistry as indicators of soil response to acidic deposition -
The final hypothesis is intended to provide the groundwork to use selected soil properties as
qualitative indices of soil "health" and the expected response to acidic deposition. Recently, several
attempts have been made to correlate soil properties with their anticipated response to acidic deposition
9-70
-------
(vanLoon, 1984; Stuanes, 1984; Lau and Mainwaring, 1985). Results from these studies suggest that soil
properties are useful indicators of how soils will respond to continued exposure to present or anticipated
levels of acidic deposition.
The hypothesis is being tested using two approaches. First, the results from the Level I statistical
analyses (see Section 8.8.4) have been presented. These results suggest significant relationships between
present day soil properties and observed surface water chemistry. These observations support the
contention of the relationship between basic soil properties and the response of the system to acidic
deposition. Second, as part of the Level II modelling activities, relationships will be examined between
current soil properties and the magnitude of projected changes in soil and surface water composition.
While these results will not integrate the roles of multiple processes, e.g.. weathering and ion exchange,
they should provide some additional evidence for examining the hypothesis.
9.3.1.2 Approach
As previously discussed (see Section 9.1), the approach used for Level II base cation analyses is
model-based. The primary processes believed to regulate exchange processes are known, and models
have been developed that describe these processes in internally consistent manners. As such, existing
model formulations are used extensively in conducting these studies.
Data used In running the models were collected specifically for this study. Section 5 provides
details of the type, quantity, and level of information gathered. In collating the data for use in the models,
certain decisions were made regarding how data from individual soils and watersheds would be
condensed, or aggregated, for use in the models. Because the primary goal of the DDRP is to make
regionally representative projections about future changes in surface water chemistry as a direct result
of acidic deposition, a decision was made to aggregate the date, first into groups of soil with similar
chemical and physical characteristics and, then, to the watershed level.
9-71
-------
Because this aggregation approach was used, projections regarding individual watersheds will not,
necessarily, be accurate reflections of the chemistry observed in that watershed. On a population basis,
however, the models should provide useful information about the anticipated behavior of the soils in the
DDRP study regions. Details regarding the soil aggregation procedures are outlined below (Section
9.3.1.2.2).
Finally, the rationale used to define the time scales over which simulations were executed is
presented. Model projections are inherently uncertain. As the durations of simulations increase, the
associated overall uncertainty increases. Therefore, there are practical limits to the usefulness of long
time frame projections. Section 9.3.1.2.3 provides a brief discussion about the trade-offs between
uncertainty and information gained.
9.3.1.2.1 Off-the-shelf models -
In designing the Level II base cation studies, one of the Issues considered was the selection of
models. As discussed in Section 2, a decision was made during the planning stages of the DDRP to use
only published, peer-reviewed, and publicly available models. A primary advantage of this decision was
that the data requirements for these models were known, so the field programs could be developed to
collect the appropriate data required. A second advantage was that minor modifications or improvements
could be incorporated Into the model codes In a timely manner. Because of concerns relative to field
design issues, and because the report from the MAS (NAS, 1984) indicated that models describing the
major soil processes controlling base cation dynamics were available, only published and publicly
available models were selected for application in the Level II base cation studies. The selected models
are described in Reuss (1983), Reuss and Johnson (1985), and Bloom and Grigal (1985) (see Section
9.3.2).
9.3.1.2.2 Aggregated soil chemistry data -
Having selected models for use in the Level II Analyses, the next major issue was preparation of
data for use in the models. Soil physical and chemical data were gathered on a representative sampling
9-72
-------
of soils in the NE and SBRP (Section 5.5.1). These data were obtained from individual pedons and soil
horizons. To transform these data into a form usable by the models, the data were aggregated to
produce information that was representative of whole watersheds.
Details of the aggregation procedures were presented by Johnson et al. (1988b). Briefly, the steps
taken to produce the aggregated data depend on the structure of the model to be applied. In general,
data are first averaged within the master horizons (i.e., 0, A/E, B or C horizons) of individual soil
sampling classes. Then, if required by the models (e.g., those that describe the soil as a single "box"),
results from the master horizons are averaged to yield values of parameters representative of the sampling
class as a whole. The procedures used to average soil chemical and physical properties at the horizon
and sampling class levels varied slightly in accordance with the model for which the data were being
developed. For models that use capacity variables as inputs, e.g., the Bloom-Grigal model, soil properties
were averaged using mass weighting procedures. For models using Intensity variables as inputs, an
intensity weighting scheme (Johnson et al., 1988b) was developed that preferentially weighted the lowest
subhorizon in generating values for master horizons, and then employed straight numerical averages to
produce sample dass/pedon data.
Finally, data from individual sampling classes were averaged, using areal weighting, to produce soils
data representative of the watershed as a whole. The weighting used in this last aggregation step was
strictly related to the relative occurrence of the sampling class on a particular watershed. The weighting
precludes bias based on the location of the soil on a watershed. For example, although it might be
argued that riparian soils have a greater influence on the composition of surface waters than do ridge-
crest soils, riparian zone soils and those soils immediately adjacent to the lakes are not preferentially
weighted relative to upland soils. The decision to use the uniform weighting approach was based
primarily on the difficulty of developing uniform, broadly based algorithms to apply preferential weighting
to specific soils based on geomorphic considerations.
9-73
-------
9.3.1.2.3 Scale of temporal forecasts -
Another decision to be made in implementing the Level II Analyses was the time scale over which
to run the model simulations. In the near term, dramatic, permanent changes to surface water
composition are not expected to occur on annual time scales. Acidic deposition is a phenomenon that
has probably affected eastern North America for at least several decades. Rapid responses to changing
deposition, if they were to occur, have probably already taken place.
For long time scale projections, the major factor determining the duration of simulations to be run
is the uncertainty associated with the major parts of the modelling efforts. As soil composition and
properties evolve with continued exposure to acidic deposition, the response of these soils is also
expected to change. We anticipate that, for longer time scales, projected changes will become more
dramatic. However, the larger changes are balanced by the Increases in the uncertainty of the analyses
for periods exceeding, e.g., 50 years.
Using these procedures as guidelines for bounding the time intervals to be modelled, simulations
for 20, 50, and 100 years were selected for the NE Region. For the S8RP, simulations for 20, 50, 100,
and 200 years were selected. The 20- and 50-year projections provide information about relatively near-
term changes that might be anticipated and are relevant time frames with regard to the implementation
of regulatory controls.
The 100- and 200-year projections are included as "worst-case* results. Such projections will allow
policymakers to understand the magnitude of changes that could occur. The 200-year simulations are
included for the SBRP primarily because major changes to sulfate mobility in soils in this region are
expected to occur during the next century (see Section 9.2). By extending the model simulations for an
additional 100 years, the full effect of changes in mobile anion concentrations will become evident
Inasmuch as the NE is, essentially, at steady state with regard to sulfur deposition (see Section 7), this
additional time is not required.
9-74
-------
9.3.2 Descriptions of Models
The model originally selected to conduct the base cation analyses was one developed by Reuss
(1983) and Reuss and Johnson (1985). This model uses a mass action approach to modelling soil
exchange processes. As such, the model requires a broad range of data as input, which was
incorporated into the design of the field program for the Project. An additional factor in the selection of
the Reuss model for use in the Level II Analyses was the fact that its data requirements are compatible
with those of some of the Level III models to be used in the study.
In addition to the Reuss model, one other model was incorporated into the Level II base cation
studies, a model developed by Bloom and Grigal (1985). This model describes soil exchange processes
based on observed relationships between the cation exchange pool and soil pH. This model, therefore,
not only expands the model base from which the Level II Analyses are conducted, but also provides
an alternative approach for describing soil exchange reactions.
9.3.2.1 Reuse Model
9.3.2.1.1 Model description -
The Reuss model was originally developed by Reuss (1983) and coworkers (Reuss and Johnson,
1985; Johnson and Reuss, 1985). The model Is an equilibrium-based, mass balance model in which the
solubility of a gibbsite-like phase is assumed to control the concentration of aluminum. Subsequently,
exchange reactions are used to partition the cations AI3+, Ca2*, Mg2*, Na+ and K* between the solid
and solution phases. Figure 9-22 presents schematically the processes considered in the model. The
mode) computes soil pH, soil solution ANC, and base cation and dissolved aluminum concentrations.
The model then "re-equilibrates" soil solutions with atmospheric carbon dioxide and computes surface
water composition.
Reuss's approach has several advantages for modelling exchange reactions in soil environments
over the use of simple exchange reactions. First, the charge balance requirement of the code makes the
model responsive to ionic strength. In forested soils, composition of the soil solutions have been shown
9-75
-------
Top
Horizon
n
Middle
Horizon(s)
Bottom
Horizon
Rgure 9-22. Schematic diagram of the principal process involved in the cycling of base cations
in surficial environments. Arrows indicate the major pathways through which ions are interchanged
among the reservoirs. No attempt is made to distinguish the relative fluxes among the different
reservoirs. The heavier lines, however, indicate those processes that serve as the focus of the
Level II modelling efforts presented here.
9-76
-------
to depend on ionic strength (Richter et al., 1988). Therefore, this aspect of the model permits a more
realistic simulation of natural exchange reactions than do the less involved computations. Second, the
model allows the user to specify the partial pressure of carbon dioxide (pCO2) in the soil gas. Although
the pCO2 in forested soils rarely exceeds about 1 percent (Fernandez and Kosian, 1987; Solomon and
Ceding, 1987), these levels can be high enough to significantly affect soil solution composition (Reuss
and Johnson, 1985,1986). Third, by relying on a glbbsite-like phase to regulate aluminum activities, one
degree of freedom in the solution composition is effectively constrained. Finally, the mass balance
constraints allows the user to track cation depletion from the exchange complex as a function of time,
hydrogen ion loading, and the imposed physicochemical environment
The Reuss model focuses .on soil exchange reactions. The model does not consider other cation
source/sink processes such as mineral weathering, nitrogen transformations, or afforestation, even though
these processes may have equal or greater influence in regulating surface water composition in certain
ecological settings. Models have been developed that include these processes, and thus yield an
integrated system response (Cosby et al., 1985a,c, 1986a,b; Gherini et al., 1985; Galloway et al., I983a;
Nikolaidis et al., 1988) to the imposed deposition. These integrated models, however, cannot be used
effectively to understand the contributions of individual processes, such as soil cation exchange, to the
buffering responses of watersheds.
9.3.2.1.2 Formulation -
The original versions of Reuss's code were written in BASIC. Table 9-9 lists the chemical species,
principal reactions, and related chemical definitions used in developing the computer code. The original
versions were written for a one-horizon setting in which the water flux, rather than time increment, was
used in scaling the step sizes for time-series simulations. Reuss's codes also ignored ammonium inputs
to soils, thus effectively using H+ as the surrogate for NH4+ deposition. In incorporating the Reuss
model into the DDRP, a number of modifications were made.
9-77
-------
Table 9-9. List of the Chemical Species and Reactions Considered Within the Reuss
Model Framework.
Aqueous Species: H^. OH, Na' K1. NH/, Cal\ Mg2', Cf, NO3, HCO3, SO/,
AI3\ AI(OH)2\ AI(OH)2*, and AI(OH)4"
Solid Phases:8 AJ(OH)3, Exch-AI, Exch-Ca, Exch-Mg, Exch-Na, and Exch-K
Mass Action Equations:
KW - {H*}{OH-}
KC - {H*}{HC03-}/(pC02)
K^gibb) - {AI3*}/{H*}3
K,* - {AI(OH)2*}{H*}/{AI3*}
Kg* = {AJ(OH)2+}{H+}2/{AI3+}
K^ = {AI(OH)4-}{H+}4/{AI3*}
= {Ca2+}3[X-AI]2/{AI3+}2[X-Ca]3
{Ca2+}[E-Na]2/{Na+}2[X-Ca|
{Ca2*}[E-K]2/{K+}2p(-Ca]
Charge Balance Equation:
H* + Na* + K* + NH.+ + AlfOH)2+
+ 2.0*(Ca2+ + Mg2* + AI(OH)2+) + 3.0* M3+ = „
OH' + HC03* + CT + NO3' + AI(OH)4' + 2.0 * SO4*
ANC or Alkalinity:
ANC - (OK) + (MOV) + (A1(QH)4- - (H*J
- (AI(OH)2*) - 2*(AI(OH)2*) - 3*(AI3
or
ANC » (Na*) + (K*) +• (NH4*) + 2*[(Ca2*) + (Mg2*)]
- (CT) - (N031) • f ='
Exch - A! = Exchangeable aluminum
Exch - Ca = Exchangeable calcium
Exch - Mg • Exchangeable magnesium
Exch - Na = Exchangeable sodium
Exch - K m Exchangeable potassium
9-78
-------
The model was adapted for this study by receding in FORTRAN, which enabled greater execution
speeds and, thus, an ability to handle more simulations. The formats of the input and output datasets
were revised to better accommodate the needs of this study. In recoding the model, a number of
operational changes were Implemented. For example, the data of May et al. (1979) instead of those
employed by Reuss and Johnson (1985) were selected to describe aluminum speciation. In addition, the
algorithms used to partition ions between solution and the solid-phase exchangers were modified to
provide more accurate mass action expressions for thin and low base saturation horizons. Given these
changes, rigorous one-to-one comparisons of results obtained from the FORTRAN and BASIC versions
of the model have not been possible, as the two models yield slightly different results. A more substantial
modification to die code entailed the use of the Vanselow exchange formulation. In the original versions,
the Reuss code employed the Gaines-Thomas formulation for cation exchange processes. Comparisons
of three exchange formulations (Hoidren et al., 1989), including the Gaines-Thomas, Vanselow, and Gapon
models, suggested the Vanselow model provided results more representative of field data than the other
two models. Differences among the three models, in general, were small, but significant.
The selectivity coefficients for the Vanselow formulation are based on a mass action expression of
the form:
{ Nn+ }m ' [ XM 1"
>" ' I XN ]m
(Equation 9-5)
where [XM] and [XNj are the mole fractions of the solid species indicated, m and n are the appropriate
stoichlometric coefficients; the species enclosed in the braces {i} indicate activity of the Ith aqueous
species. The specific mass action expressions for the exchange reactions considered are listed in Table
9-9.
Other modifications to the code were also incorporated. The model was expanded to allow
inclusion of up to four horizons. With this expansion, a provision was made to allow the user to route
soil water from any horizon directly to surface water. This "routed water" is assumed to not be
9-79
-------
equilibrated with soils deeper in the pedon. Rather, it is "mixed" on a volume-weighted basis with waters
derived from other horizons plus all water draining from the bottom of the pedon. New pH values and
aluminum concentrations are then computed for the "surface water" assuming equilibration with
atmospheric pCO2.
Three options were incorporated into our version of the model regarding the treatment of input
nitrogen chemistry. In the original versions of the model ammonium in deposition was ignored, and H*
effectively served as a surrogate for NH4*. This treatment was retained as one option in our code. The
second option (used in all model runs for this report) was based on the presumed reaction:
NO3" + NH/ = org-N (Equation 9-6)
in which the two nitrogen species are accreted into the organic nitrogen pool on an equivalent basis.
If nitrate concentrations exceed ammonium concentrations in deposition, then the excess nitrate is passed
through the soil as a mobile anion. Conversely, if ammonium concentrations exceed nitrate in deposition,
the excess ammonium is presumed to be replaced by H+. The third option is based on the reaction:
3 N03" 4- 5 NH4+ - 4 N2(g) + 9 H20 + 2 H* (Equation 9-7)
in which nitrate and ammonium combine to form nitrogen gas, water, and hydrogen ion. This process
originally was conceived to occur if the organic nitrogen pool attained steady state. As with the second
option, excess nitrate is passed on to the soil as a mobile anion, or the excess ammonium is presumed
to be replaced by H*.
These options are not sufficiently comprehensive for modelling of nitrogen species distributions or
concentrations in surface waters. As previously indicated, the purpose of the Reuss model is to examine
soil exchange phenomena with regard to the effects of acidic deposition, and not to provide detailed
information concerning the effects of nitrogen transformations in the soil environment The different
9-80
-------
routines, however, provide the user with some degree of flexibility in the treatment of nitrogen
transformations.
Finally, the time series computations were converted from deposition volume-controlled increments
to time related steps, largely as a matter of convenience for dealing with units and to facilitate use of the
model by others.
9.3.2.1.3 Assumptions -
As with any model, assumptions are necessary regarding certain processes, the soil environment,
and the characteristics of certain reactions. These assumptions and their justifications are outlined below,
along with an assessment of the effect they have on model predictions.
9.3.2.1.3.1 Gibbsite solubility controlling soil AI3+ concentrations -
One of the main features of Reuss's model is that the solubility of a gibbsite-like [AI(OH)3 ] phase
controls the concentrations of dissolved aluminum in soil solutions. Objections have been raised to this
assumption on several grounds. First, in many acidic forest soils, gibbsite is probably not present as a
separate or distinct phase, and thus cannot regulate concentration of an aqueous species. Second,
investigators have noted that aluminum activities, {Al3*}, in forest soil solutions do not behave according
to classic gibbsite solubility dynamics in response to changing H* activities (Johnson, 1986; Bloom et
al., 1979a,b). Theoretically, aluminum activities should decrease by three orders of magnitude for each
unit increase in pH, i.e.,
Iog10 {AI3+ } = C - 3 pH (Equation 9-8)
where C is an arbitrary constant related to the solubility product of gibbsite. In the studies cited,
however, aluminum activities appear to be independent of soil pH, or vary in ways different from the
above relationship. A third concern focuses on the variability In aluminum solubility behavior observed
in natural materials. In natural systems, gibbsite is expected to display a range of solubilities, based on
9-81
-------
the grain size, crystallinrty of the parent material, and conditions under which it was formed. This issue
is irrelevant in the context of the Reuss model, as apparent solubility products are computed on a
sample-by-sample basis using field data to constrain the aluminum behavior.
Despite these concerns, the solubility of a gibbsite-like phase is assumed in the present analyses
to control aluminum activities In soil solutions. Figure 9-23 shows pH vs. Iog10 {Al3*} (both measured
in 0.002 M CaCI2) for all samples collected during the ODRP NE Soil Survey. The solid line indicates the
theoretical solubility of gibbsrte (C - 8.774; May et al., 1979). In computing the aluminum activities, only
the hydroxide complexes of aluminum were included in the speclation model. Contributions from sulfate,
fluoride, or organic ton pairs or complexes were not incorporated into the speciation model because
data on the counter ion species were not available from the analytical solutions. The contributions of the
suifate, fluoride, and organic complexes to total dissolved aluminum concentrations increase with
increasing pH, so, effectively, aluminum activities should be increasingly overestimated at successively
higher pH values.
For soli samples with pH values greater than about 4.0, i.e., all B and C horizon samples, and
about half of the A/E horizon samples, gibbsite solubility appears to provide a reasonable model of
aluminum solubility. Regression of the data with pH values greater than 4.0 yields a slope of -1.4. For
soils with pH values between 4 and 5, predicted aluminum activities are generally within an order of
magnitude of measured values. Soils with higher pH values generally display high aluminum activities.
These results are attributed to the inability to incorporate organic complexes of aluminum into the
speciation model.
For soils with pH values less than 4.0, i.e., all O horizon soils and about half of the A/E horizons,
aluminum activities appear to be independent of soil pH. This observation is interpreted as an indication
that the mass of rapidly exchangeable aluminum available on soil exchange sites is limited. Most of the
soil buffering is expected to be derived from mineral horizons. Given the behavior of soils illustrated In
Figure 9-23, and considering the limitations of the aluminum speciation model used to estimate the
9-82
-------
-3
-4-
-5-
< -6-
o>
o
-7-
-8-
-9
Range of data regressed
4.0 < pH < 5.5
n = 826
m= 1.64
b « 2.42
r2 = 0.662
Q
D
.gibbsite
'solubility
4 5
Soil pH
Figure 9-23. Plot of the iog of the activity of Al3* vs. soil solution pH for individual soil samples
collected for DDRP. Both parameters were measured in a 1:2.5 soihsolution suspension of 0.002
M CaClf The cluster of points at pH < 3.9 are O and organic-rich A horizons and do not follow
a gibbsite-like mineral solubility behavior. The points on the right side of the graph suggest that
solutions are highly oversaturated with respect to gibbsfte. However, the speciatlon model used
to estimate activities included only hydroxy complexes. Chloride, sulfate, fluoride, and organic
complexes were not considered in these computations. As such, these points probably
significantly overestimate actual (Al3*) activities in these solutions.
9-83
-------
aluminum activities indicated, the evidence supports the model of a gibbsite-like solubility behavior to
describe aluminum availability in soils.
The use of the limited aqueous speciation model to compute aluminum activities potentially
introduces one other problem. If aluminum activities are significantly overestimated, then selectivity
coefficients computed using the artificially high values should similarly be too large. (The value is
calculated as: K^80 - {AI3+ }2 fX^ ]/{Ca2+ }3 [X^ ]. This effect, however, partially compensates
for an opposite effect, namely that induced by not considering rational activity coefficients for the solid-
phase exchangers (see next subsection).
9.3.2.1.3.2 Constancy in selectivity coefficients as functions of base saturation -
As detailed in Section 9.3.2.1.2, the Reuss model describes exchange reactions using Vanselow-
type mass action equations. The equations are developed and used based on data derived from the
soils. This approach is reasonable as long as the changes in the base saturation of the soils under study
are limited. Problems may be encountered, however, during time dependent simulations if significant
changes in base saturation are projected to occur.
Problems may arise because selectivity coefficients are not true thermodynamic constants. As
presently formulated, the selectivity coefficients do not incorporate rational activity coefficients for the
solid-phase exchangers. Therefore, the selectivity coefficients are only approximately constant, and then
only for narrow ranges of base saturation around those levels for which they were calculated. As base
saturation declines, the selectivity coefficients would be expected to vary accordingly.
As an example of this behavior. Figure 9-24 illustrates the relationship between log10(Kexac ) (the
selectivity coefficient for the Ca/AI exchange reaction) and base saturation for aggregated A/E horizon
samples used In the watershed runs. For base saturations between 30 and 40 percent, selectivity
coefficients average slightly more than 100. As base saturation decreases, selectivity coefficients increase
such that for samples with base saturations between 10 and 12 percent the constants have average
9-84
-------
values of about 1000. As indicated, this change Is a direct result of not having incorporated the rational
activity coefficients into the expression for the solid-phase exchangers.
Obviously, the apparent change fn the selectivity coefficients as a function of base saturation is of
concern in terms of the model results. To determine what the effects of varying selectivity coefficients
might be, a modelling experiment was undertaken in which the selectivity coefficients for the Ca/AI
exchange reaction were both increased and decreased by an order of magnitude for each of the master
horizons In the 145 watersheds. The model was run using these inputs, and the projected surface water
ANC was determined. Results for the present-day ANC values are summarized in Rgure 9-25.
Changing the selectivity coefficients by an order of magnitude introduces about a 10 eq L"1
change in the projected ANC for any particular system. This change is small, as the predicted total base
cation concentrations for most systems fall in the 100 to 200 eq L'1 range. Therefore, errors introduced
by having selectivity coefficients that are off by as much as an order of magnitude are approximately 5
to 10 percent. This change is small enough not to affect the long-term projection of depletion of
buffering capacity significantly in most systems.
9.3.2.1.3.3 Soil gas pCO2 -
The last major assumption made in running the Reuss model was the selection of an average
annual soil gas pCO2. For the model runs reported here, the partial pressure of CO2 for all horizons and
all soil classes was assumed to be 0.005 atm on an average annual basis. Soil pCO2 concentrations vary
with a number of factors, Including temperature, soil productivity, and moisture content. Unfortunately,
there are not enough available data for the range of soils included in the DORP to be able to model
values accurately. Data that are available (Solomon and Ceding, 1987; Fernandez and Kosian, 1987; Lam
et al., 1988) indicate soil gas CO2 can exceed 0.5 percent, and in some soils at some times of the year,
levels are less than 0.5 percent. A partial pressure of 0.005 atm was selected simply because it appears
to be a representative value for forested soils, based on available data.
9-85
-------
$
2-
10
20 30
% Base Saturation
40
50
Figure 9-24. Plot of the log of the selectivity coefficient for the calcium-aluminum exchange
reaction vs. the measured base saturation in A/E horizons in the NE. The increase In the
selectivity coefficients with decreasing base saturation is a direct result of not incorporating rational
activity coefficients Into the mass action expression used to estimate selectivity coefficients.
9-86
-------
50
0 50 100
Predicted ANC (p,eql_-1)
Figure 9-25. Histograms of the (unweighted for the population estimates) projected present-day
ANC values for lakes in the NE. The three curves were generated by varying the selectivity
coefficient for the calcium-aluminum exchange reaction by ± an order of magnitude from the value
estimated from the soils data. Varying the selectivity coefficient by a factor of 10 changes the
projected ANC values for any system by about 10 peq L'1 for present-day conditions. This is not
of sufficient magnitude to have a significant effect on the projected rates of depletion of buffering
capacity for the vast majority of these systems.
9-87
-------
Reuss model outputs are responsive to increases in soil gas CO2 concentrations. Figure 9-26
illustrates present-day surface water ANC values predicted using CO2 concentrations of 0.001,0.005, and
0.025 atm C02. As the partial pressure of C02 increases in the soil gas, the ANC of the associated
surface water also increases. It is possible, therefore, to adjust predicted present day surface water ANC
values up (or down) simply by adjusting the CO2. In making these adjustments, however, there Is a
trade-off. By increasing soil gas C02 concentrations to increase the predicted ANC values, the rate of
base cation leaching from the soil exchange complex is increased significantly. Therefore, the rate at
which soil buffering capacity becomes depleted is also increased. This is illustrated in Table 9-10, where
changes predicted for surface water ANC at 50 and 100 years are given for the three soil gas CO2
scenarios.
Finally, it should be clarified that no attempt was made, in selecting the 0.005 atm value for pC02,
to use this parameter to "calibrate" the Level II models. Admittedly, had the pCO2 been adjusted on a
watershed-by-watershed basis, a much better fit to observed soil pH values could have been obtained.
However, the purpose of this part of the modelling exercise was to determine the magnitude of possible
responses to acidic deposition. In the absence of more specific data on soil gas CO2 levels in individual
watersheds, the approach taken here yields the least controversial, and most widely applicable, results
possible.
9.3.2.1.4 Limitations -
The Reuss model focuses on soil exchange reactions. The model does not consider other
processes such as sulfate adsorption, mineral weathering, nitrogen transformations, or afforestation, even
though these processes may have equal or greater influence in regulating surface water composition in
certain ecological settings (Likens et al., 1977; Johnson et al., 1988b). The purpose of this part of the
study, however, was specifically to examine exchange processes and their contribution in regulating
surface water composition and buffering against changes caused by acidic deposition.
9-88
-------
70
Pco2 - 0.001 atnr
Pco2 = 0.005 atrr
Pco2 - 0.025 atm
0 50 100
Predicted ANC Qieq L -1)
Figure 9-26. Histograms of the (unweighted for the population estimates) projected, present-day
ANC values for lakes in the NE. The three curves were generated by varying the soil gas partial
pressure of carbon dioxide by plus or minus a factor of 5 from the assumed value of 0.005
atmospheres. Decreasing the partial pressure of COZ reduces the projected ANC by about 10
peq L"1 on the average, but does not dramatically affect the projected rates of depletion of buffering
capacity of the systems in the NE. Increasing the partial pressure dramatically increases both the
projected mean ANC values for the lakes and the rates of cation depletion from the soil exchange
complex.
9-89
-------
Table 9-10. Effect of pCO2 on Changes Projected to Occur in Surface
Water ANC Values at 50 and 100 Years Using the Reuss Model. Deposition
Used in the Model is LTA. Values Are Given as the Mean, Population-
Weighted ANC Values for the NE (see Section 9.3.3.1 for details)
Time Step ANC@ ANC@ ANC@
0.001 atm 0.005 atm 0.025 atm
Present day ANC -6.6 10.0 74.5
A-ANC @ 50 years -8.5 -13.7 -50.7
A-ANC @ 100 years -21.3 -32.1 -97.0
9-90
-------
Most of the other model limitations were alluded to in Section 9.3.2.1.3. The most significant
among these is that no provisions are made to consider organo-cations, and especially organo-aluminum
interactions. The data are not available to include these interactions in our modelling efforts.
Another limitation of the Reuss model is the Implicit assumption that the reactions considered can
be modelled on an equilibrium basis. Clearly, most ongoing chemical processes in watersheds are
subjected to rapidly and constantly changing chemical environments. Fluctuations in temperature, fluid
flow, and external cation demands occur on daily, weekly, and seasonal bases. Few if any processes
actually attain chemical equilibrium. Nonetheless, the Reuss model assumes exchange processes can
be modelled using an equilibrium approach. Given the relatively rapid nature of exchange reactions and
the annual time step used in most computations, this assumption is probably not unreasonable.
9.3.2.1.5 Model inputs -
9.3.2.1.5.1 Deposition and associated data -
The model requires deposition data including precipitation quantity (cm yr*1) and average annual
concentrations of the major tons In precipitation (SO42~, CT, NO3", NH4+, Ca2+, Mg2*, Na*. and K+).
The atmospheric flux of each ion was the combined wet plus dry average annual deposition. Evapo-
transpiration (% ET) data are required to adjust the concentrations of the non-reactive tracers (e.g. CQ
between deposition and runoff. This parameter also helps define the ionic strength of the soil solutions,
thereby influencing solution composition.
As described in Section 5.6, a number of deposition scenarios are used for model simulations. The
LTA deposition Is used as the baseline against which other results are compared. LTA data are the best
available estimates of total deposition occurring in each watershed. Typical year (TY) deposition data
have also been compiled for these watersheds and are used in model simulations. Two reduced dry
deposition scenarios have been examined as part of these efforts. The first scenario, long-term annual
average-reduced or LTA-rbc, assumes fluxes of base cations (Ca2+, Mg2*, Na*, and K+) in dry
deposition to be half the values used in LTA. This scenario was included because concerns had been
9-91
-------
raised over the large particle deposition rates incorporated into the LTA baseline estimates. The second
scenario, LTA-zbc, assumes zero flux for dry deposition base cations. The LTA-zbc deposition dataset
yields maximum H+ deposition estimates for each watershed and has been included in the Reuss model
analyses to ascertain the magnitude of error potentially caused by uncertainty in dry deposition data.
In addition to the above data, which assume constant depositional inputs to the watersheds over
the course of the simulations, ramped datasets have been'constructed. In these ramped datasets, total
(wet plus dry) sulfate and hydrogen ion depositional fluxes are varied during the course of each
simulation. Changes are assumed to occur over a 15-year period, from year 10 in the simulation to year
25. The change is linear during this period, and the value attained in year 25 is maintained to the end
of the simulation. In the NE Region, the ramp decreases sulfate depositional values to 70 percent of
current estimates; in the SBRP, the ramp increases sulfate fluxes to 120 percent of current estimates.
9.3.2.1.5.2 Soils data -
The model requires physical and chemical information about each of the horizons included in the
simulations. Required physical parameters are horizon thickness, bulk density, percent coarse fragments,
and a hydrologic runoff parameter. Required chemical parameters include cation exchange capacity
(CEC), base cation concentrations on the exchange complex, selectivity coefficients for the Ca/AI, Ca/Mg,
Ca/Na, and Ca/K exchange reactions, soil gas pCO2, the apparent solubility product for AI(OH)3(s), and
the stoichiometric coefficient for H* to be used in describing the dissolution of the aluminum solid phase.
Multiple-horizon versions of the model require the above information for each of the horizons to be
considered. Some minor adjustments were required to incorporate these parameters in the model.
Soil bulk densities in the DDRP database were entered on a coarse fragment-free basis. As a
result, two adjustments to the associated field data were necessary. First, the percent coarse fragments
parameter, was assigned a value of 0 in all cases. The contribution of coarse fragments is subtracted
from the bulk density and soil (horizon) thickness, since these fragments are essentially unreactive mass.
Second, to retain the proper reactive soil mass, the horizon thicknesses were adjusted to remove the
9-92
-------
contribution of the rock fragments. The fragments not only add mass to a horizon, but also contribute
to the overall thickness. Had this correction had not been made, the reactive masses of the individual
horizons would have been larger than those actually measured.
Another fixed parameter in the input datasets was the soil gas pC02 concentrations. As discussed
In Section 9.3.2.1.3, a uniform value of 0.005 atm (0.5 percent) was used for all model computations.
Finally, for this report, the stoichiometric coefficient for H* used to describe the dissolution of the
aluminum oxyhydroxkJe phase is assumed in all cases to be 3.00. Although the model can adjust this
parameter (e.g., In response to observed aluminum behavior in 0 and A/E horizons; see Section
9.3.2.1.3), we retained the gibbsite-like solubility behavior because of data limitations regarding aluminum
behavior in individual soil samples.
Other data used as model input were taken directly from the DORP soil chemistry database. For
most of the simulations discussed In the report, data were aggregated according to the procedures and
protocols presented in Section 9.3.1.2.2. That is, data from the six to ten pedons in each sampling class
were aggregated (Johnson et al., I988b) to a master horizon level (O, A/E, B and C horizons). Extensive
parameters, such as horizon thickness, were aggregated by simple arithmetic averaging. Intensive
parameters, such as soil pH or CEC, were aggregated using mass weighting procedures.
In addition to the sampling class-based design, a study was undertaken to evaluate an alternative
aggregation procedure. Results from some of the multtvariate analyses (see Section 8.3) suggested a
significant watershed-specific component to observed variances. Simulations were conducted using only
data collected on each individual watershed to model that watershed. Because soils were not sampled
on all watersheds, and because difficulties were encountered with some of the analytical data, complete
coverage of the DDRP watersheds was not possible using the alternative aggregation scheme. Data were
collected on 129 of the 145 systems included in the study, however. Results from this effort are used
to determine if changing the aggregation scheme would significantly affect conclusions.
9-93
-------
9.3.2.1.6 Model outputs -
For each simulation, the model generates two results files, one containing projections for surface
water composition and the other describing soil and soil solution composition for major chemical species.
Results are compiled for the first and final years of the computation and at user-specified intervals during
the simulation. For example, if the user were running a 13-year simulation and requested output at 5-
year intervals, the result files would contain data for years 1,5, 10, and 13.
Occasionally, results are not available for all soils or watersheds for the duration requested
(typically, 100-year simulations). This occurs when the model falls to converge with a particular set of
input parameters, which at most is about 7 percent of the simulations. Initially, the failure-to-converge
rate was considerably higher than 7 percent. However, by adjusting convergence criteria, the loss of
results minimized without sacrificing significant numerical accuracy.
Information on many variables Is retained in the two output files. For surface waters, data on pH,
ANC, SO42', N03", CT, Ca2*, Mg2+, Na*. K*. AI3+, sum-fAl)^, and ionic strength are captured. For
soils, Information on soil pH, base saturation, and exchangeable Ca, Mg, Na and K are retained for the
solids, and ANC, Ca2+, Mg2*, Na*. and K+ data are retained for the soil solutions. For this report,
analyses focus on a few select parameters, namely surface water pH and ANC and soil pH and base
saturations, because these parameters are believed to provide the most easily interpretable indicators of
system responses to continued exposure to acidic deposition.
9.3.2.2 Bloom-Grigal
In the DDRP, surface water Is the principal resource of interest. However, soils play a vital role in
maintaining the quality of surface waters because drainage waters entering lakes and streams pass
through soils. Soils can buffer drainage waters against changes In several ways, as discussed in other
parts of this report. If soils in the study regions were to change dramatically (e.g., become more acidic),
these changes would ultimately be reflected in the subtending surface waters and in the status and health
9-94
-------
of forest vegetation. Characterizing the status of the soils in the DORP regions and considering the effect
of chronic acidic deposition on them is, therefore, important.
Two very different simulation models have been included in the DDRP to assess the Impact of
acidic deposition on surface waters. The Reuss model was discussed in the previous section. This
section describes the Bloom-Grigal model.
9.3.2.2.1 Model description -
The Impact of acidic deposition on soils can be modelled following one of two approaches (Bache,
1983). The first approach (used in the Reuss model) is to view the interaction of precipitation with soils
as a perturbation of the equilibrium between Ions in the soil solution and ions on the soil ion exchange
complex. Following the perturbation, the system returns to equilibrium according to the theories of ion
exchange equilibria. The second approach Is to view this interaction as a simple mass action (non-
equilibrium) exchange reaction. Following this approach, the amount of acidity En deposition replaces an
equivalent amount of base cations in the soil. The Bloom-Grigal model is a form of this second
approach.
The Bloom-Grigal model estimates the loss of base cations on an annual basis using the following
equation:
S - I - A - C (Equation 9-9)
where S is the sum of base cations, I is the amount of effective acidity in deposition, A is the acid
leached from the soil, and C is the correct ion factor for the decrease in acidity due to protonation of
bicarbonate. The model Is presented graphically in Figure 9-27.
The Bloom-Grigal model is a simple semi-empirical computer simulation model created to project
the effects of acidic deposition on soils (Bloom and Grigal, 1985). The model tracks soil pH and base
saturation. Unlike the Reuss model, the Bloom-Grigal model is not formulated to project the chemistry
of subtending surface waters. The model does, however, follow the concentrations of aluminum in the
9-95
-------
soil solution during simulation runs which can serve as an indicator of possible changes in surface water
chemistry and forest health. The Bloom-Grigal model was initially formulated to assess the effects of
acidic deposition on forested soils in northeastern Minnesota. Because the model is based on widely
applicable principles, we believe that it can be meaningfully applied to project the effects of acidic
deposition on the soils In the DDRP study regions.
9.3.2.2.2 Model formulation -
The Bloom-Grigal model is formulated around the assumption that, in steady-state ecosystems,
acidic deposition depletes base cations on the soil ion exchange complex The model's simplicity lies
in the fact that soils are treated as a single homogeneous unit or compartment and all incoming
deposition reacts completely with the soil in the compartment. Soils, however, are much more complex.
The Bloom-Grigal model seems to be an appropriate tool for assessing the impact of acidic deposition
on forested soils.
The Bloom-Grigal model assumes that the acidity in deposition reacts completely with the soil. In
other words, the model makes no provision for deposition to be routed around the soil and directly into
the surface water or into the subsoil strata. The amount of exchangeable base cations removed from
the soil compartment is calculated as the difference between the input acidity and the output of H* and
AT3*, corrected for the protonation of bicarbonate. The amount of base cations lost is subtracted from
the pool of exchangeable base cations and a new base saturation is calculated. The Bloom-Grigal
model then calculates a new soil pH based on an equation that relates soil pH to base saturation. After
adjusting parameters, the model then simulates the next year of deposition (see Figure 9-27).
This model was created to assess the effect of acidic deposition on non-sulfate adsorbing soils.
Soils that adsorb sulfate have lower base cation removal rates than soils that do not. In this regard, the
Bloom-Grigal model is probably more appropriate for application to the soils in the NE than in the SBRP.
9-96
-------
Deposition
Prectp/runoff
rSoil chenpistry
r
y Modelling
dalaset
Record watershed ID, year,
soil pH, % base saturation
Increment year
End
Simulation
10 < Years 25
Calculate new deposition
Calculate input acidity
Calculate output acidity
Calculate profanation
of bicarbonate
Calculate loss of bases
Calculate new % base
saturation and soil pH
Figure 9-27. Flow diagram for the one-box BloonvGrigaf soil simulation model.
9-97
-------
Another feature of the Bloom-Grigal model is that it incorporates the input of nitrogen in deposition.
Because forested soils are generally deficient in available nitrogen, inorganic nitrogen in deposition is
removed by plants and organisms in the soil (Bloom and Grigal, 1985). When plants assimilate nitrogen
in the form of nitrate (NO3~), they release hydroxyis (OH*) to the soil, which is a non-acidifying reaction.
However, when plants assimilate nitrogen as ammonium (NH4*), they release protons (H+). Ammonium
uptake is an acidifying reaction. The biological oxidation of NH4* to NO3" produces one H* for every
molecule of NH4* oxidized. The Bloom-Grigal model incorporates these processes in calculating the net
or effective acidity of deposition.
The original versions of the Bloom-Grigal model were coded in FORTRAN and BASIC; the version
used in this analysis is coded in a high speed, compilable form of BASIC. In addition to optimizing the
code for speed, code has been added that allows the input data to be processed in a batch mode. The
output is now formatted to magnetic media to simplify the data reduction process. Additional lines of
code have been added to calculate the various deposition scenarios automatically as the model is
running. The fundamental equations in the original model have not been altered, however.
9.3.2.2.3 Assumptions -
A number of assumptions are made in modelling the effect of acidic deposition on soils with the
Bloom-Grigal model. Some are implicit to the model, others are made to meet the needs of our current
application. The assumptions used in implementation of the Bloom-Grigal model are itemized below
including additional explanatory discussion or comments.
9.3.2.2.3.1 Sulfate adsorption -
The Bloom-Grigal model assumes that sulfate is not adsorbed by the soil and is treated as a
completely mobile anlon. As mentioned previously, in soils that have net sulfate adsorption, this
assumption may lead to an overestimation of the amount of base cations actually leached from the soil.
Q.QA
57*370
-------
9.3.2.2.3.2 Input acidity -
The total effective acidity (H+tolal) in deposition is equal to:
H+totai = H* + NH4+ - N03* (Equation 9-10)
9.3.2.2.3.3 Extent of reaction -
The effective acidity in deposition reacts completely with the soil.
9.3.2.2.3.4 Depth Of soil -
The depth of reactive soil material equals the mean aggregated thickness of the soil sampling
classes represented by the types of soils on the specific watersheds.
In their original paper, Bloom and Grigal (1985) assumed that only the top 25 cm of soil are
affected by acidic deposition. We consider the effect, however, on the whole soil compartment. Our soil
chemistry input data are aggregated to represent the central tendency of the soil chemical characteristics
of the whole soil compartment. The effect of acidic inputs on data aggregated in this way, thus,
represents a mean effect. At the same time, this assumption allows for the water that flows in cracks
or root channels to lower soil horizons before reacting with the soil.
9.3.2.2.3.5 Volume of drainage water -
The volume of water moving though the watersheds in each year of simulation is equal to the long-
term annual average runoff.
9.3.2.2.3.6 Partial pressure of CO2 -
The partial pressure of ambient CO2 is approximately 0.0003 atm. Soil air is, however, enriched
with C02 due to biological respiration and is consequently elevated. In all of the Bloom-Grigal model
runs, the partial pressure of CO2 in the soil air is set at 0.005 atm, a value thought to be reasonable for
forested soils.
-------
9.3.2.2.3.7 Activity Of AI3+ -
To calculate the amount of input acidity that is converted to output acidity by aluminum buffering,
the activity of Al3* in soil solutions is calculated using the following equation:
2.60 - 1.66 * soil pH (Equation 9-11)
This equation Is the empirical part of the Bloom-Grfgal model. In developing their model, Bloom and
Grigal had a fundamental problem with using the solubility of AI(OH)3 to describe the variation in AI3+
with pH. They state that in very acidic soils, such as forested soils, AT3* is undersaturated with respect
to the precipitation of AI(OH)3. Therefore, AI(OH)3 solubility Is a poor model for the pH-AI3* relationship.
To establish a more realistic relationship between AI3+ and pH, they developed the above equation from
laboratory measurements of A13+ in artificially acidified soils. Although not appropriate for all soils, Bloom
and Grigal believe that model results from which their equation was generated were reasonable for
selected forested soils of northeastern Minnesota.
9.3.2.2.3.8 Relating soil solution pH to base saturation -
The pH of soil solutions is related to base saturation (BS) by the following equation:
pH = pKa + n * log [BS/(1 - BS)] (Equation 9-12)
where p^ is the apparent acidity constant for soil (i.e., aggregate watershed/soil compartment) and n
is an empirical constant. This equation is an extended form of the Henderson-Hasselbach equation.
The Bloom-Grigal model used here calculates pKa and n for each watershed using the input values
of soil pH and base saturation. These parameters describe the relationship between soil pH and base
saturation and are unique for each watershed.
9-100
-------
9.3.2.2.3.9 Base cation uptake -
The model assumes no net accretion of base cations in biomass. The uptake of base cations by
forest vegetation is an acidifying process by which H* is exchanged for an equivalent amount of base
cations to maintain charge neutrality. At the same time, through litterfall and decomposition, base cations
are released to soils. The Bloom-Grigal model only tracks the flux of base cations that are leached from
the soil. This no-accretion assumption implies that the uptake of base cations by vegetation is exactly
equal to the amount recycled to the soil.
9.3.2.2.3.10 Mineral weathering -
Mineral weathering is the ultimate source of base cations, and the Bloom-Grigal model has a
subroutine that calculates the contribution of base cations to the soil solution via mineral weathering. The
rate of mineral weathering for these simulations, however, is set to zero for two reasons. First, assuming
no base cation resupply a "worst-case" base cation loss scenario is evaluated, thereby bounding the
projections. Second, the relationships between weathering and soil solution pH are not sufficiently
established to provide accurate parameters for the weathering equations. One complication, in particular,
is that mineral weathering rates are a dynamic function of the chemical weathering environment.
9.3.2.2.3.11 Cation exchange capacity -
Cation exchange capacity (CEC) is constant throughout the period of simulation. Scientifically this
is not correct. Soil CEC is derived from two sources: (1) secondary clay minerals with permanent charge
due to isomorphous substitution of lower valent cations for cations in the clay crystal lattice, and (2)
variable charge sites on organic matter, para- and noncrystalline hydrous oxides, and edge sites on
permanently charged days. The variable charge CEC is a function of pH, i.e., the net soli CEC changes
as with changes in pH. As pH increases the variable charge CEC increases, and vice versa. Because
of scientific and data limitations, we have chosen to hold CEC constant.
9-101
-------
9.3.2.2.3.12 Time steps -
The time step for simulations is annual. For assessment purposes, yearly time steps are a useful
increment. From a modelling standpoint any shorter time step (e.g., daily) is data intensive and
computationally demanding. Shorter time steps may provide more accurate projections, however.
9.3.2.2.4 Limitations-
Soils are highly complex and no simulation models exist that accurately depict the flux of energy
and matter In soil systems. As with any attempt to project future events, the Bloom-Grigal soil simulation
model is not without limitations. Some of the limitations are due to the state of the science and others
are have been imposed by the DORP.
The scientific limitations center around the factors that control aluminum solubility and the
relationship of soil pH to base saturation. Bloom and Qrigal (1985) empirically developed equations to
describe this relationship for a selected set of northeastern Minnesota forested soils. As described In their
paper, the equations appear appropriate for forested soils in Minnesota. In the DDRP, the equations are
assumed to be widely applicable and they are not independently verified, it is doubtful, however, that
these equations are universally true due to vast differences in soils and vegetation.
Soils are dynamic systems. Soil properties fluctuate on a daily basis, and daily temperature and
moisture changes affect a broad range of soil processes. Broader seasonal changes also occur. The
use of annual time steps assumes that soils are static, possibly restricting the accuracy of the projections.
As mentioned above, however, shorter time steps are data and computationally restrictive.
Individual soil processes are Inextricably linked to a number of other processes and considering
a single process (e.g., base cation flux) in isolation may distort projections. In the DDRP Level II
Analyses, processes are Isolated in order to focus on the principal soil reactions associated with surface
water acidification. It Is recognized that some of the uncertainty in assessing effects Is due to this
approach of Isolating facets of the whole ecosystem.
9-102
-------
9.3.2.2.5 Model Inputs -
The Bloom-Grigal model was designed not to be data intensive. The data required to run the
Bloom-Grigal model fall into four categories: (1) deposition data, (2) precipitation data, (3) soil chemistry
data, and (4) fixed parameters. The deposition data are described in Sections 5.6 and 9.3.3. Table 9-
11 lists the specific data requirements.
The soil chemistry data used in these simulations has been aggregated to the single compartment,
watershed level. These procedures are described in detail in Johnson et al. (1988b). The capacity
variables, sum of base cations (SOBC) and CEC are capacity weighted. Soil pH is intensity weighted.
9.3.2.2.6 Model outputs -
The Bloom-Grigal model simulates soil processes relevant to the assessment of impacts of acidic
deposition on soils. During model simulation runs, soil pH, soil base cation status (le., base saturation),
and soil solution Al3* are tracked. Principal Interest for this analysis is soil pH and base saturation.
During 200-year simulations, soil pH and percent base saturation are recorded (see Figure 9-27)
at years 0, 20, 50,100, and 200. The results are converted to change in soil pH and change in percent
base saturation by subtracting the initial values from the projected values. Because the Initial values are
higher than the projected values, the reported results are all negative numbers, reflecting a decrease.
The projected changes in soil pH and percent base saturation are presented as cumulative
distribution functions (CDF) for graphical comparisons. The CDFs represent regionally weighted
projections for soils on the target population of watersheds. Summary statistics for the CDFs also are
presented for numerical comparisons.
9.3.3 Model Forecasts
Level II base cation analyses were conducted using Reuss's (Reuss, 1983; Reuss and Johnson,
1985) cation exchange model and Bloom and GrigaTs (1985) cation depletion model. Results from the
9-103
-------
Table 9-11. List of Input Data for the Bloom-Grigal Soil Acidification Model.
Input Variables Units
Annual average runoff cm
Annual H+, NH/, N03~, and SO48~ In wet deposition keq ha'1
Annual H+, NH/t N03', and SO42' in dry deposition keq ha*1
Soil pH . pH(H2O)*1
Sum of soil base cations (0.1 M NH4Q) keq ha*1
-i
Soil cation exchange capacity (O.t M NH4CI) keq ha'
Fixed Parameters Value
Length of simulation NE = 100 years
SBRP » 200 years
Partial pressure of CO2 0.005 atm
Activity coefficient of Al3* 0.82
Activity coefficient of Al(OH)2+ 0.92
Wet and dry SO*2" deposition are used to calculate alternative deposition scenarios.
9-104
-------
individual models are presented in this section along with a comparison of the projections made using
the two models.
9.3.3.1 Reuss Model
9.3.3.1.1 Data sources •
Summaries and examples of the various datasets used in running the Reuss model were presented
in Section 5. A brief summary of the data used for the simulations also is given below.
The data fall into two categories: deposition data and soils data. Four deposition datasets were
used in making population estimates of watershed responses. As described in Section 9.3.3.1.1.1, these
datasets were used in model simulation runs assuming constant levels of deposition for the future and
in conjunction with a ramping function that adjusted deposition downward by 30 percent in the NE and
upward by 20 percent in the SBRP (see Section 5.6). Similarly, soils data were aggregated using two
approaches. The sampling class-based aggregation described in Section 9.3.1.2.2 was used with each
of the deposition scenarios. In this approach, soils data were aggregated to master horizon/watershed
level. The second approach (watershed-based aggregation) was initially undertaken because some
preliminary Level I Analyses indicated a substantial "watershed effect." That is, some combination of local
variables indicated that a soil from a given watershed was more similar to other soils in the watershed
than it was to other soils in the region from the same sampling class. While this preliminary observation
was not substantiated by additional investigations (see Section 8.8.1), the watershed-based aggregation
procedure was further examined to determine whether substantial differences in the results would be
observed. Results from this examination are presented in Section 9.3.3.1.2.1.
Given the number of deposition scenarios and soils aggregation approaches available, 16 distinct
sets of results could be generated for the NE. Because the purpose of examining the scenarios and the
aggregation schemes was to determine the sensitivity of model results to different conditions, discussions
are limited to nine combinations of deposition scenarios and soils aggregation schemes. All of the
9-105
-------
constant and ramped deposition scenarios are run in conjunction with the master horizon/watershed soils
aggregation scheme. The two soils aggregations are run using the constant level, LTA deposition. Thus,
results obtained using LTA deposition and the master horizon/watershed soils aggregation scheme serve
as the baseline dataset against which other results are compared.
9.3.3.1.1.1 Deposition data -
Four deposition datasets were used. The dataset considered to be most representative of "actual"
deposition is the LTA dataset, derived from 5-year averages of species concentrations in deposition and
30-year averages of precipitation quantities (see Section 5.6).
Except for the TY dataset, which is based on data obtained from a year with mid-range depositional
values (see Section 5.6), other deposition datasets are variations of LTA. In constructing LTA, transport
and deposition of large particles (> 20 //m) were integral components of the dry deposition estimates.
The uncertainty in the long-range transport of these larger particles (concern that net H* fluxes to
watersheds might be underestimated) prompted construction of two additional deposition datasets. LTA-
rbc is essentially identical to LTA, except that the estimated dry deposition of base cations (Ca2*, Mg2+,
Na+, and K+) is reduced by 50 percent. LTA-zbc assumes zero net dry deposition of base cations.
Dataset LTA-zbc, as a result, yields the highest hydrogen ion fluxes to watersheds, and, in fact, probably
significantly overestimates net H+ fluxes. In this context, LTA-zbc can be viewed as a "worst-case"
deposition scenario.
9.3.3.1.1.2 Soils data -
Soils data were aggregated using one of two approaches. The primary aggregation scheme uses
the soil sampling class concept around which DDRP was designed (see Section 5.5). The other
aggregation was based on locale, and is described in more detail in this section.
The aggregation scheme routinely used in the DDRP is the master horizon/watershed aggregation.
Data representing each of the four master horizons (O, A/E, B, and C) are obtained. For each master
9-106
-------
horizon, data are first averaged to within sampling class using protocols described in Johnson et al.
(19885), which are then averaged using areal weighting to obtain estimates for a watershed. Hydrologic
routing of water is considered if at least one of the sampling classes has a lower-most horizon that is
anything other than a C horizon. For example, overland flow of water for the watershed is set equal to
the percentage of precipitation falling directly on rock outcrops and is routed directly to the surface water
without equilibration with any of the soil horizons. As another example, for watersheds having soils in
sampling class H01 (which has only an 0 horizon), that fraction of soil water equal to the areal
percentage cover of the watershed by H01 is routed to surface water after equilibrating with the 0
horizon. While this approach is oversimplified, watershed hydrologic characteristics are spatially
distributed, and adequate representation of the complexity in natural systems cannot be accomplished
in the current formulation. The hydrologic routing was established for these analyses in full cognizance
of its limitations. Bedrock outcrops tend to occur along ridgelines, so Incident precipitation will not run
off directly into the surface water.. Histic soils, on the other hand, tend to be concentrated in riparian
zones. Histic soils can have extremely low permeabilities, and unless they are dry, incident precipitation
will tend to run off from their surfaces. Nonetheless, the model equilibrates incident deposition with these
soils. The model also does not consider any aspect of lateral flow, and therefore downward percolation
is likely to be considerably, overestimated especially on steeper slopes. Considering the various trade-
offs, we feel that the hydrologic routing, as described, yields a reasonable approximation for modelling
these complex, spatially-related processes.
Second, a watershed-based aggregation of soils data was undertaken in order to obtain information
concerning the sensitivity of model results to the aggregation method. For this approach, only data from
those soils sampled on a particular watershed were used to describe the watershed. Therefore, if the
only two soils sampled on a watershed were a Histosol and a Spodosoi, the data from those two soils
were used to represent the watershed regardless of the actual areal coverage. The potential problem with
this aggregation is that, for watersheds on which sample classes are minor proportions of the total
watershed area, the soils sampled may not be representative of the actual local environment. As
described in Section 9.3.3.1, however, preliminary concerns had suggested that, even with this limitation,
9-107
-------
the aggregation might be more representative of the population of soils in each of the regions than is
the sampling class-based information (see Sections 5.2 and 8.8).
9.3.3.1.2 Projections of surface water ANC -
9.3.3.1.2.1 Results from the Northeast -
9.3.3.1.2.1.1 Prediction of current conditions -
The distribution of current surface water ANC values projected for the NE using the Reuss model
is Illustrated in Figure 9-28, along with upper and lower bounds for 90 percent confidence intervals
associated with the projection. The ANC values for each of these fakes, as measured by the Eastern
Lake Survey (Unthurst et al., 1986a), are listed in Table 5-3 for comparative purposes. An obvious
feature of these projections is the extremely tight clustering of the results in the range of -25 to +50
peq L"1. This clustering has been observed on virtually all model runs conducted to date, including those
runs using data aggregated at the watershed level and those conducted on individual sampling classes.
For the Individual sampling classes, the upper limit for ANC values exceeds 200 peq L~\ while for the
other 3? classes In the NE an upper limit of 80 jieq L"1 is observed.
These results are consistent with the hypothesis that soil exchange reactions can buffer soil and
surface water ANC values and that the buffering occurs in the low ANC range. Although surface waters
with higher ANC values occur In the NE, they are not typical of the region. Soil exchange reactions,
therefore, do not adequately explain the observed distribution of surface water ANC. Figure 9-29
illustrates the relationship between observed and projected ANC values. Clearly, the tight clustering of
the predicted values near zero indicate no significant correlation.
In order to explain the observed distribution of ANC values in the population of lakes sampled for
this study, it is necessary to invoke some mechanism other than base cation exchange to produce ANC
values greater than 100 fjeq L'1. Uptake of cations by aggrading vegetation is a possible mechanism,
9-108
-------
NE Lakes
Deposition = LTA
Year = 1
Model = Reuss
Upper Bound
Projected
Lower Bound
-25 0 25 50 75
ANC (jlieq L-1)
100
Figure 9-28. Cumulative distribution of projected, present-day ANC values for lakes In the study
population in the NE as projected using Reuss's cation exchange model. LTA deposition was used
in making these projections. The error bounds on the plot are the 90 percent confidence Intervals
and were obtained using a Monte Carlo approach, assuming that errors on individual input
parameters to the model are normally distributed, and that the only source of error is in those input
parameters.
9-109
-------
300
200-
CD
O
<
•
JO
to
O
100-
-100
n = 145
r2 = 0.03
-100
0 100 200 300
Measured ANC (\ieq L *1
Figure 9-29. Scatter plot of the projected, present-day ANC values for lakes In the NE, obtained
using the Reuss model vs. observed (ELS) values. The heavy diagonal line indicates the 1:1, or
perfect correspondence, line. As is apparent, the model projects that current ANC values should
cluster at values that are slightly in excess of 0 j/eq L*1. This is interpreted as indicating the
importance of mineral weathering In controlling observed surface water composition for the majority
of systems in the NE.
9-110
-------
but if cation uptake were a significant process In these watersheds, the observed ANC values would be
lower than those computed by the model. The other major process that could explain the distribution
is primary mineral weathering, which can significantly alter cation balance. Release of base cations and
ANC through reactions such as those listed in Section 3.4 can Increase surface water ANC to values well
above the 100 peq L"1 limit apparently imposed by soil exchange processes. Other processes that could
increase ANC to the levels observed in the lakes are presently unidentified.
For lakes exhibiting ANC values exceeding 100 peq L'1, mineral weathering apparently is the
dominant watershed process controlling ANC. For systems with ANC values less than 100 jieqL"1, either
mineral weathering or soil exchange processes could be regulating the observed levels. Given available
methods, however, determining which process accounts for the observed ANC values is not possible.
The implications of these findings are significant in terms of projected future changes in surface
water chemistry. If mineral weathering is, in fact, regulating ANC levels in those systems with ANC
greater than 100 /jeq L*1, then these systems probably will not experience significant future declines in
ANC at current levels of deposition. Inasmuch as present trends in the NE indicate stable or declining
hydrogen ion deposition, lakes with ANC values exceeding 100 peq L*1 are probably not at risk with
regard to future acidification.
Soil exchange processes might regulate ANC in systems exhibiting ANC levels less than 100 peq
L"1. If so some of these systems might currently be experiencing an increase in base cation leaching
rates in response to acid anion inputs from acidic deposition. In the future, these systems could
experience significant ANC decline. Unfortunately, given the current state of the science, distinguishing
between those systems in which ANC is controlled by mineral weathering and those in which ANC is
controlled by soil exchange processes is not currently possible.
To provide an upper bound on the number of systems that may experience additional declines in
ANC, summary Information from the Eastern Lake Survey can be examined (Unthurst et al., I986a). Data
9-111
-------
from this survey suggest that about 1,038 lakes (about 15 percent of the total ELS target population in
the NE) have ANC values in the range of 0 to 50 jjeq L"1. The largest population of lakes that might
be adversely affected by changes to the soil exchange buffering capacities is in the Adirondacks
(Subregion 1A), where 321 lakes (25 percent of the target population) have ANC between 0 and 50 peg
L1. The Poconos/Catskills Subregion (1B) has the fewest lakes in this ANC class: 116 lakes (7.8
percent). As noted, the proportion of these systems that may actually experience future declines in ANC
cannot be determined. Some proportion of the systems that currently have low ANC values, however,
will probably experience adverse changes.
An issue of concern regarding these conclusions is the sensitivity of the results to the input data
used in the simulations. To address this issue, several different versions of input data were used in
running the simulations: four deposition scenarios and two soil aggregation schemes (see Sections
9.3.2.1 and 9.3.3.1). Summary results from these model runs for projected present-day ANC values are
given in Table 9-12. For the four deposition scenarios, the differences among projected ANC are minimal,
with projected population-weighted, mean lake ANC values of 9 ±. 1 ^ieq L"1; medians, maxima, and
standard deviations are equally comparable. The largest differences are observed for the projected
minima. The LTA-zbc deposition scenario results in an ANC value that is 10 peq L*1 less than that
projected using the LTA and 15 peq L*1 less than that projected using the TY.
The greatest observed differences occur with the use of the different soil aggregation schemes.
For the data listed In Table 9-12, the columns under LTA and WBA were obtained using the same
deposition data, but different soils aggregation schemes. The data under the LTA column were obtained
using the master horizon/watershed aggregation scheme, whereas those under the WBA (or the
Watershed Based Aggregation approach) column were aggregated based on soils collected from
individual watersheds and used to describe only those watersheds. The WBA data indicate moderate
changes in the means and medians for the present-day ANC values. The extremes, however, represent
a much broader range of values than are actually represented by the field data. Figure 9-30 illustrates
the relationship between the observed and projected ANC values obtained using the WBA scheme.
9-112
-------
Table 9-12. Summary Statistics for the Population Estimates of
Current ANC Conditions for Lakes in the NE Region for Five Different
Deposition or Soils Aggregation Schemes (Refer to the text for explanation
of the different input scenarios.)
LTA TY LTA-rbc LTA-zbc WBA
Mean 10.0 8.3 9.4 8.8 35.5
Std Dev. 18.4 19.0 18.9 19.5 87.4
Median 8.3 7.4 7.8 7.4 18.9
P25 0.34 -1.8 -0.7 -0.2 0.3
P75 21.7 20.6 21.0 21.0 43.3
Max 70.8 67.1 70.7 70.6 863.7
MIn -52.0 -46.8, -56.7 -61.3 -121.1
9-113
-------
Fifteen of the 129 lakes in the sample have projected ANC values exceeding 100 peq L"1. Despite the
wider range of projected values, the WBA scheme does not improve the correlation between observed
and projected values. This finding is not surprising, since the soils sampled on any given watershed were
not selected to be representative of the soils on that watershed, but rather to be representative of a group
of soils In the region (see Section 5.2.4.1). Therefore, although the WBA scheme may more accurately
portray the variability of Individual soils in the regions, it does not demonstrably provide a more accurate
means for explaining observed surface water composition.
9.3.3.1.2.1.2 Projected future conditions -
In order to project the magnitude of changes in ANC that might occur in the NE, as well as the
time frame over which such changes might occur, the Reuss model was run using its mass balance
component. The mass balance component of the model tracks the loss (or gain) of base cations from
soil exchange sites through time. For these simulations, precipitation quantity (cm yr"1) as well as the
depositional fluxes were used to specify the total loadings of ions delivered to the soil. Annual time steps
were used In making these computations. For the NE, model simulations were run for a total of 100
years, with results of the computations being collected at 10-year intervals. Results are reported only
at the 20-, 50-, and 100-year time increments.
Projected, time-dependent changes in ANC values for the population-weighted results are illustrated
in Figures 9-31 and 9-32 and summary statistics are given in Table 9-13. The Reuss model considers
only the effects of the soil cation exchange process in making these projections. Mineral weathering
reactions would, in general, further delay the response of these systems to the effects of acidic
deposition. At 20 and 50 years, most systems in the NE are projected to experience minimal change
in ANC. Apparently, the soil buffering capacity in these systems is sufficient to moderate the effects of
acidic deposition over these time scales. Only a small percentage of the watersheds (about 10 percent)
is projected to experience losses of ANC that exceed about 25 j/eq L"1 within the 50-year time frame.
9-114
-------
300
200-
cr
CD
a
,0
CO
O
100-
-100
n = 129
r2 * 0.01
-100
0 100 200
Measured ANC (jieq L -1)
300
Rgure 9-30. Scatter plot of the present-day lake ANC values projected using the Reuss model
in conjunction with the Watershed-Based Aggregation (WBA) soils data vs. observed (ELS) ANC
values. The range of ANC values projected using this approach is much greater than obtained
using the sampling class/watershed-based approach. However, the correlation is not improved.
Three projected points with ANC values in excess of 300 fieq L*1 are not shown on this plot.
9-115
-------
1.0
O 0.8
o
Q.
O
0.6
CL
-------
1.0
O 0.8
O
CL
2 0.6
CL
CD
S5 0.4
O 0.2
0.0
-75
NE Lakes
Deposition = LTA
Year = 100
Model = Reuss
Upper Bound
Projected
Lower Bound
-50 -25
A ANC (jieq L-
Rgure 9-32. Cumulative distribution of the projected surface water ANC values projected for the
study population of lakes in 100 years in the NE. The model runs were conducted using LTA
(constant level) deposition.
9-117
-------
Table 9-13. Descriptive Statistics of the Population Estimates
for Changes in Lake Water ANC for Systems in the NE. Mean,
Median, Standard Deviations for the Population and the Maximum
Changes Projected Are Presented for Each of the Four Deposition
Scenarios at the Time Increments 20, 50, and 100 Years
LTA TY LTA-rbc LTA-zbc
ANC (0) (Mean) 10.0 8.3 9.4 8.8
ANC (20)
Mean -6.1 -6.4 -6.5 -6.9
Std 16.4 18.0 17.4 18.3
Median -2.0 -2.4 -2.3 -2.5
Max -101.9 -118.3 -107.1 -110.0
ANC (50)
Mean -13.7 -16.1 -15.5 -17.5
Std 23.6 26.4 26.4 30.0
Median -5.2 -6.0 -6.0 -6.4
Max -127.5 -138.8 -140.0 -160.0
ANC (100)
Mean -32.1 -43.1 -39.4 -44.7
Std 36.1 51.5 43.8 49.6
Median -13.9 -22.0 -16.4 -20.6
Max -185.4 -231.7 -207.5 -228.7
9-118
-------
The 100-year projections for changes in ANC (Figure 9-32) suggest a bimodal distribution in the
way watersheds respond to the effects of acidic deposition. About half of the watersheds in the region
are projected to experience minimal changes (<-l3 fteq L~1) over the 100-year time frame. The other
half is projected to experience a median change in ANC of about -50 peq L"1 and a maximum change
of almost -200 peq L*1. The magnitude of these changes is of concern, if mineral weathering reactions
do not control ANC. A closer examination of the results (Table 9-13) suggests that projected changes
in the ANC values through time are not linear, but rather accelerate to a point where the buffering
capacity of soils is depleted. Soils response to acidic deposition is analogous to a buffer effectively being
titrated by acidic deposition. As such, any given soil behaves In the same way a dissolved buffer in an
aqueous system behaves (Figure 9-33). Assuming that the system is not yet near to or beyond the
inflection point of the titration curve, the initial response of a soil to continued loadings of acidic
deposition will be a gradual, and almost linear, decline in projected ANC for some period of time. Once
the system reaches the inflection point, however, the rate of decline in ANC dramatically accelerates until
the buffering capacity of the system is depleted.
For the soils examined to date, these observations have two major implications. First, minimal
changes observed in lake water ANC values do not necessarily preclude the possibility that more dramatic
changes will occur in the future. If the buffering capacity of a soil is currently being depleted, the full
effect might not be Immediately apparent. Rates of change in system response can increase~with time,
unless the process is being moderated by mineral weathering. Second, for the soils included in DDRP,
dramatic changes in system response to acidic deposition are projected only for those systems with lower
ANC values. Most of the titration curves deviate from relatively flat slopes to steeper slopes as the
inflection points approach ANC in the range of -20 to +20 ^eq L*1. Therefore, the systems that are
most vulnerable to dramatic future changes in ANC are those that currently have an associated surface
water ANC of about 0 //eq L"1.
9-119
-------
20
cr-20
i
O
-40.
T3
-------
The alternative soils aggregation scheme, WBA, yields results that are qualitatively similar to those
obtained using the master horizon/watershed aggregation (Table 9-14). Quantitatively, the changes
projected using the WBA scheme are two to three times as large as those projected with the routine
aggregation method. Also, the WBA scheme projects substantial changes in a small number of
watersheds during the early phases of the simulations. These results substantially shift the mean values
of the changes to more negative numbers. Because there is a lower limit to values that ANC can attain
within the framework of this model, the magnitude of changes that can occur in the population means
is limited.
The last group of simulations addresses the effects that ramped deposition has on projected future
changes. As discussed In Section 9.3.3.1.1.1, the three LTA deposition datasets, as well as the TY data,
were modified using a ramp function that decreased sulfate and hydrogen ion depositions! fluxes by 30
percent in the NE between years 10 and 25 of the simulations. Differences between the projections made
using ramped and constant deposition are presented in Table 9-15. Not surprisingly, differences between
the two scenarios are minor at the 20-year point. By year 50, the median declines projected for ANC
using ramped deposition are only half as large as those projected using constant deposition. After 100
years, the differences In the medians are less. Ramped deposition results in changes in surface water
ANC that are two-thirds the magnitude of those for constant deposition. Differences in the means are
more uniform for both year 50 and year 100. At both years, ramped deposition results in changes that
are about 60 percent as large as those obtained using constant deposition.
Incorporation of mineral weathering effects into these results would suggest smaller differences
between the constant and ramped depositions than those reported here. A supply of cations from
weathering would tend to minimize the changes projected by both datasets, but such effects would be
larger for the constant deposition scenario than for the ramped deposition scenario.
9-121
-------
Table 9-14. Summary Statistics Comparing
the Projections Regarding Changes in Surface Water
ANC Values Obtained Using Different Soils
Aggregation Schemes .
LTAa WBA'
b
ANC (0) 10.0 35.5
A-ANC (20)
Mean -6.1 -25.1
Std -16.4 42.0
Median -2.0 -5.2
Max -101.9 -216.4
A-ANC (50)
Mean -13.7 -43.9
Std 23.6 55.1
Median -5.2 -14.3
Max -127.5 -241.4
A-ANC (100)
Mean -32.1 -66.9
Std 36.1 67.6
Median -13.9 -36.5
Max -185.4 -275.6
The UTA data have been obtained using a sampling-
class-based aggregation, in which soils from the whole
region are used to describe specific soils on the
watershed (see Section S.S.1).
The WBA is based on data obtained from
only those soils sampled on the watersheds being
described. The text contains details of the procedures
used.
9-122
-------
Table 9-15. Summary Statistics of the Differences Between the
Population Estimates for Future ANC Projections Made Using the
Constant Level and Ramped Deposition Scenarios8.
LTA TY LTA-rbc LTA-zbc
ANC (0)
Mean 0.0 0.0 0.0 0.0
A-ANC (20)
Mean 2.1 2.3 2.1 2.2
Std. Dev. 1.0 1.6 1.3 1.3
Median 1.7 1.9 1.6 1.7
Max 9.9 10.7 10.6 12.1
A-ANC (50)
Mean 5.7 7.0 5.9 6.3
Std. Dev. 4.8 6.6 5.5 6.4
Median 3.0 3.8 3.1 2.9
Max 25.8 27.5 29.2 33.2
A-ANC (100)
Mean 11.4 15.6 12.7 12.2
Std. Dev. 10.8 14.2 11.3 11.5
Median 4.7 11.3 5.4 4.5
Max 79.9 56.7 57.4 54.8
The values were computed as the difference between ramped and constant
deposition. The magnitude of the values can be compared to the descriptive
statistics presented in Table 9-13 to obtain estimates of the absolute values of the
changes incurred with the ramped datasets. Standard deviations are presented as
absolute values.
9-123
-------
9.3.3.1.2.2 Results from the Southern Blue Ridge Province -
9.3.3.1.2.2.1 Current conditions
The distribution of current surface water ANC values projected using the Reuss model for the SBRP
is illustrated in Figure 9-34. These values can be compared to the actual distribution of ANC measured
for these stream reaches during the Pilot Stream Survey (Messer et al., 1986a) (see Table 5-6). As with
the northeastern results, the extremely tight clustering of the results around an ANC value of zero is
notable. Mean and median values for each of the four deposition scenarios (Table 9-16) are between
2 and 4 eq L'1 , and the total range for the four scenarios Is about -15 to 23 eq L"1.
As for the northeastern data, these results are Interpreted as an indication that the soils of the
region are characterized by strong buffering. Additionally, the results suggest a dominant role for mineral
weathering in regulating the observed surface water composition, since neither sulfate adsorption nor
cation accretion into biomass can readily explain the differences between observed and projected ANC
values (Figure 9-35). Mineral weathering also might explain why the observed ANC values are
considerably higher than the model results.
9.3.3.1.2.2.2 Projected future conditions
As described for the NE, simulations of the time-dependent responses of the ANC in the study
population stream reaches in the SBRP were conducted. Annual time steps were employed for these
runs, and results were collected at 10-year intervals; data are summarized here for the 20-, 50-, 100-, and
200-year increments only.
Changes in the projected surface water ANC values are summarized in Table 9-17. During the first
50 years of these simulations, the Reuss model results suggest that changes, even the maximum
changes, are trivial relative to our ability to measure representative ANC values. As base cation supply
becomes depleted, these changes become much more dramatic, but this depletion is projected to occur
on a century-long time scale. Mean and median changes for this region are estimated to be -20 ± 5
9-124
-------
SBRP Stream Reaches
Deposition = LTA
Year = 1
Model = Reuss
1.0
O 0.8
o
CL
O
0.6
3= 0.4
.2?
=3
E
O O-2
0.0
-50
Upper Bound
Projected
Lower Bound
-25 0 25 50
ANC (jieq L-1)
75
100
Figure 9-34. Cumulative distribution of projected present-day ANC values for stream reaches in
the study population in the SBRP, as projections using Reuss's cation exchange model. Long-
term average (LTA) deposition was used in making these projections. The error bounds on the
plot are the 90 percent confidence intervals and were obtained using the parameter error estimates
developed for northeastern region soils. Then, as completed in the NE, a Monte Carlo approach
was used to obtain population estimates of the errors.
9-125
-------
Table 9-16. Summary Statistics for the Population Estimates of
Current ANC Conditions for Stream Reaches in the SBRP for Four Different
Deposition Scenarios (Refer to the text for explanation of the different
input scenarios.)
LTA TY LTA-rbc LTA-zbc
Mean 3.9 2.2 3.7 3.4
Std. Oev. 5.8 6.1 6.0 6.2
Median 2.9 2.3 2.9 2.9
P25 -0.55 -1.5 -0.55 -0.55
P75 7.0 4.7 6.7 6.5
Max 21.2 23.0 21.2 20.8
Mln -12.8 -14.1 -15.3 -17.7
9-126
-------
100
75.
7j 50.
CT
1
O 25.
4
O
-25.
-50
o
in
CM
8
CM
-1
8
8
CO
O
O
Measured ANC (jieq LT )
Figure 9-35. Scatter plot of the projected present-day ANC values for stream reaches in the SBRP,
obtained using the Reuss model, vs. observed (NSS) values. The heavy diagonal line indicates
the 1:1, or perfect correspondence, line. As is apparent, the model projects that current ANC
values should cluster at values that are slightly in excess of 0 peq L"1. This is interpreted as
Indicating the importance of mineral weathering in controlling observed surface water compositions
for the majority of systems in this region.
9-127
-------
Table 9-17. Descriptive Statistics of the Population Estimates for
Changes in Stream Reach ANC Values for Systems in the SBRP.
Mean, Median, and Standard Deviations for the Population and the
Maximum Changes Projected Are Presented for Each of the Four
Deposition Scenarios at the Time Increments 20, 50, 100, and
200 Years
LTA
TY
LTA-rbc
LTA-zbc
ANC(O) (Mean)
3.9
2.2
3.7
3.4
A-ANC (20)
Mean
Std. Dev.a
Median
Max.
A-ANC (50)
Mean
Std. Dev.
Median
Max.
A-ANC (100)
Mean
Std. Dev.
Median
Max
A-ANC (200)
Mean
Std. Dev.
Median
Max.
•1.2
0.4
-1.2
-1.8
-3.0
1.1
-2.9
-5.2
-14.6
6.0
-14.8
-27.4
-81.2
24.5
-77.8
-134.8
-1.2
0.7
-2.2
-2.2
-3.7
2.0
-4.0
-7.4
-23.0
16.2
-18.5
-58.8
-97.8
36.2
-103.3
-161.7
-1.2
0.5
-1.9
-1.9
-3.5
1.5
-3.7
-5.9
-20.3
12.8
-18.3
-48.0
-97.3
33.1
-103.9
-154.9
-1.2
0.4
-2.0
-2.0
-5.3
2.8
-4.8
-10.9
-33.7
23.6
-24.4
-80.4
-120.1
39.1
-122.7
-185.3
Standard deviations are reported as absolute values.
9-128
-------
eq L"1 on a 100-year time scale. Over two hundred years, these changes increase by a factor of 5 to
approximately -100 ± 20 eq L"1. These changes are projected to occur regardless of the selected
deposition scenario. These results are illustrated in Figures 9-36 and 9-37 for the LTA deposition.
Watersheds in the SBRP are projected to respond relatively uniformly to the different deposition
scenarios, unlike the NE, for which a range of responses to acidic loadings was displayed. This
observation can be explained by several factors. First, the watersheds in the SBRP were selected from
a geographically more limited area than those in the NE. Second, the number of stream reaches studied
In the SBRP is considerably smaller than the lake study population in the NE. This smaller subset of
systems will limit the observed variability simply because of the reduced sample size being examined.
In examining the changes projected for surface waters in the SBRP, it is important to remember
that the Reuss model Is a cation exchange model, and it does not consider the effects of increasing
anion mobility. At present, the soils in the SBRP are retaining significant percentages of sulfur being
deposited in the region (see Sections 7 and 9.2). As a result, rates of base cation leaching from the soil
exchange pool are probably less than those presented above because the total anion concentration in
soil solutions are lower than considered in the model. The rates of leaching will increase as the soils
approach zero net retention of sulfur and will approach the projected levels asymptotically. Therefore,
the magnitude of observed'changes should be some non-linear combination of the time frames involved
in base cation leaching and changes in sulfur retention.
Mineral weathering would even further delay any anticipated changes in observed surface water
ANC values. As weathering proceeds, additional cations are provided both to the exchange complex and
to surface waters. As in the NE, it is not possible with the data and models currently available to isolate
the separate effects of weathering and cation exchange. In a qualitative sense, however, we conclude
base cation-related changes in surface water ANC in the SBRP should occur only on century-long time
scales once the effects of weathering are incorporated into the projections.
9-129
-------
SBRP Stream Reaches
Deposition = LTA
Year = 50
Model = Reuss
to
O 0.8
O
d.
O
1_
Q_
CD
0.6
+3 0.4
JB
3
E
O 0.2
0.0
-75
Upper Bound
Projected
Lower Bound
-50 -25
A ANC (M-eq L-1)
Figure 9-36. Cumulative distribution of projected changes (at 50 years) in surface water ANC
obtained using the Reuss model for stream reaches in the SBRP. The deposition scenario used
In making these projections was LTA. Confidence intervals around the distribution are based on
uncertainty estimates of the individual parameters used in the model.
9-130
-------
SBRP Stream Reaches
Deposition = LTA
Year - 100
Model = Reuss
1.0
O 0.8
O
CL
2 0.6
CL
0)
*S 0.4
o
0.0
-75
Upper Bound
Projected
Lower Bound
-50 -25
A ANC (|ieq L-i)
Figure 9-37. Cumulative distribution of projected changes (at 100 years) in surface water ANC
obtained using the Reuss model for stream reaches in the SBRP. The deposition scenario used
in making these projections was LTA. Confidence intervals around the distribution are based on
uncertainty estimates of the individual parameters used in the model. The "choppiness" of the
curve Is due, in part, to the smaller number (n=20) of watersheds for which 100-year projections
were obtained.
9-131
-------
The last major issue concerns the effects of ramped deposition datasets on the response of
watersheds to acidic deposition. As discussed in Section 5.6, the ramping functions increased deposition
by 20 percent during the 10- to 25-year time interval of each simulation. This ramping function was
used in conjunction with each of the deposition datasets.
Differences In the projections of surface water ANC between the ramped and constant scenarios
are given in Table 9-18. Not surprisingly, projected differences are minor during the first 50 years of the
simulations, although the increased levels of deposition in the ramped dataset nearly double the median
projected changes at 50 years (from -2.9 eq L'1 to -5.0 eq L*1). At 100 years, the projections using
the ramped deposition are double those for constant levels of deposition. Median changes between the
deposition scenarios are not as large, but the ramped scenario projections result in changes that are
50 percent larger than those for the constant deposition. At 200 years, the medians of the population
projections for the ramped and constant deposition scenarios continue to diverge. However, the
differences in the population means have not changed substantially from those observed at 100 years,
suggesting that the limiting values proscribed by the composition of the deposition are being attained.
9.3.3.1.2.3 Comparison of results from the Northeast and Southern Blue Ridge Province -
Comparison of the effects projected by the Reuss model in the two regions indicates both
similarities and differences between the two regions. In both regions the soils behave initially as strong
buffers for surface water ANC. Also, in both regions the projected present-day ANC values are generally
substantially less than the actual observed values. These observations are interpreted to indicate the key
role that mineral weathering plays in regulating ANC in surface waters of the two regions.
The soils in the two regions, however, are projected to respond differently to continued exposure
to acidic deposition. At present levels of deposition, soils in the NE appear to be more susceptible to
significant changes in the future than are the soils in the SBRP. In a sense, this conclusion is counter-
9-132
-------
Table 9-18. Summary Statistics of the Differences Between the
Population Estimates for Future ANC Projections Made Using the Constant
Level and Ramped Deposition Scenarios for Stream Reaches in the SBRPa
LTA TY LTA-rbc LTA-zbc
ANC (0) Mean 3.9 2.2 3.7 3.4
A-ANC (20)
Mean
Std. Dev.a
Median
Maximum
A-ANC (50)
Mean
Std. Dev.
Median
Maximum
A-ANC (100)
Mean
Std. Dev.
Median
Maximum
A-ANC (200)
Mean
Std. Dev.
Median
Maximum
•0.9
0.1
-1.0
-1.1
-1.5
0.6
-2.1
-6.9
-8.7
7.1
-6.3
-41.1
-24.3
11.5
-29.1
-95.9
-1.0
0.0
-0.9
-1.1
-3.3
1.7
-3.2
-8.7
-20.6
17.3
-11.9
-94.5
-25.9
3.1
-19.6
-64.9
-1.0
0.0
-1.0
-0.9
-2.9
1.4
-2.2
-18.1
-15.6
8.9
-11.7
-35.9
-26.8
3.2
-20.7
-90.2
-1.0
0.0
-1.0
-0.9
-3.6
2.5
-2.6
-30.8
-25.8
12.0
-31.5
-53.3
-23.3
0.7
-18.3
-74.0
The values were computed as the difference between ramped and constant deposition. The magnitude
of the values can be compared to the descriptive statistics presented in Table 9-17 to obtain estimates
of the absolute values of the changes incurred with the ramped datasets. Standard deviations are
reported as absolute values.
9-133
-------
intuitive because the soils in the NE tend to exhibit higher levels of base saturation (see Section
9.3.3.1.3.1.3). The soils in the NE, however, are also younger than those in the SBRP, and as a result,
tend to have less clay-size materials. Because the bulk of the exchange capacity is associated with fine
particles (see Section 8.8.1) and because the soils in the NE tend to be shallower than those in the
SBRP, soils in the NE apparently have a lower overall capacity to supply base cations to surface waters
from exchange processes.
9.3.3.1.2.4 Summary -
Several conclusions can be drawn from the observations made using the Reuss model and the
projected behavior of watersheds in both the NE and SBRP.
• For lakes in the NE currently exhibiting ANC values in excess of 100 /*eq L*1, mineral
weathering Is probably the dominant watershed process controlling observed ANC values.
• At present levels of deposition, NE lakes with ANC values In excess of 100 peq L'1 will
probably not experience declining ANCs in the foreseeable future.
* For lakes in the NE currently exhibiting ANC values of less than 100 jieq L*1, soil exchange
processes may be regulating the observed ANCs, although In most systems, the observed
levels are probably controlled by a combination of cation exchange and mineral weathering.
• As an upper limit, over 1000 additional lakes in the NE region could become acidic (E.e.,
ANC <. 0 /*eq L ) within a 50- to 100-year time frame. This is four times the number of
lakes that are currently acidic. This number is considered to be extreme because the
contribution of weathering is not Included in these projections. However, some lakes are
expected to become acidic during the next several decades.
• In the SBRP, changes in observed ANC values due to changes In the base status of soils
during the next century should be minimal. Observed changes in this region will be driven
primarily by changes in anion mobility in these soils (see Sections 7.3.4 and 9.2.3.2.3).
9-134
-------
9.3.3.1.3 Results - predictions of soil pH and percent base saturation -
Another concern regarding the effects of acidic deposition is the changes in soil pH and base
saturation status. As discussed in Section 9.3.1.1, soils can be used as indicators of potential future
changes. As with the ANC results, these model results are presented on a regional basis.
9.3.3.1.3.1 Northeast-
Unlike the ANC projections, for which the correspondence between observed and predicted values
was only a secondary concern, the Reuss model should be able to predict observed soil pH values with
a reasonable degree of accuracy. (Present day base saturation Is an input to the model and, as such,
cannot be used in this type of an analysis.) Figure 9-38 Illustrates the correlation between the observed
and predicted soil pH values for all of the master horizon/watershed combinations considered in the NE.
Two features are immediately apparent from this plot
First, there is a high correlation between the observed and predicted values. In general, the model
tends to over-predict Individual observations. For measured pH values greater than about 4.0, the model
results exceed measured values by 0.20 ± 0.10 pH units. The divergence between the two increases
substantially at pH values below 4.0. Therefore, the Reuss model reasonably predicts the relative
differences in soil pH among soils (for pH values exceeding 4.0).
Second, the model predicts very few soil pH values of less than 4.0, and, in fact, the data appear
to reach a plateau at soil pH values of about 4.25 +. 0.25. Effectively, the lower limit to soil-water acidity
Is defined by hydrogen ion content of deposition after it has undergone evapotranspirath/e concentration.
This lower limit is about 3.8 in the NE region (precipitation with a pH of 4.2, concentrated by 40 percent
through evapotransplration). Within the Reuss formulation, no provisions are available to address acidity
generated by organic processes, and only limited acidity can be added to soil solutions by the exchange
of base cations in deposition for acid cations on soil exchange sites. For these reasons, the model has
difficulty predicting the extremely low pH values observed in most 0 horizons and in the organic-rich A
9-135
-------
6.0
5.3-
.O
Q4.7
4.0-
3.4
.ID
580
0.768
1.35
0.948
3.4
4.0
4.6
pH water
5.2
5.8
Figure 9-38. Comparison of measured vs. calculated soil pH values for the 580 aggregated master
horizons in the NE. The heavy diagonal line is the 1:1, perfect correspondence line. In general, the
model slightly over-projects soli pH values.
9-136
-------
horizons. For most other horizons, however, the relationship between observed and predicted soil pH
values are acceptable.
Projections regarding future changes in base saturation and pH of soils in the NE are listed in
Tables 9-19 and 9-20, respectively. The projections are illustrated in Figures 9-39 and 9-40 for 50 and
100 years, respectively. Mean and median changes in soil base saturation exhibit uniform rates of
depletion of about 0.75 percent ± 0.05 percent per year throughout the simulations regardless of the
deposition scenarios used. The rates of depletion are slightly higher for the reduced base cation loading
scenarios, as expected. Extreme values are only about three times the magnitude of the mean changes
observed for the population of systems being studied. Soil pH values show similar time-dependent
changes (Table 9-20). Soil pH values decline at a mean rate of about 0.04 pH units per year throughout
the simulation, with only minor, but consistent, differences projected among the different deposition
scenarios.
The data presented here are based on the results aggregated from mineral horizons only. An issue
of concern with these results, therefore, is the possible effect that organic horizons might have on the
magnitude or direction of changes projected by the model. To evaluate this issue, the model runs using
data aggregated both with and without the presence of organic layers would need to be conducted.
These model runs have not been performed. However, this issue, Is addressed In Section 9.3.3.2 for the
Bloom-Grigal model. The importance of organic horizons in regulating changes to soil chemistry are
presented there.
9.3.3.1.3.2 Soils as an indicator of possible future changes in ANC -
Soils may serve as indicators of future changes occurring because of acidic deposition. An
analysis of this hypothesis is useful for identifying those systems that are most susceptible to adverse
changes. This information also could be used in the design phases of a monitoring program.
9-137
-------
Table 9-19. Summary Statistics of the Projected Changes in Soil Base Saturations
in the NE Region, Obtained Using the Different Deposition Scenarios or Soil
Aggregation Schemes. The Time Increments Included in the Table Are 20, 50, and
100 Years
LTA TY LTA-rbc LTA-zbc WBA
% BS (Initial)
Mean 20.9 20.9 20.9 20.9 24.4
Std. Dev.a 10.5 10.5 10.5 10.5 20.3
Median 17.6 17.6 17.6 17.6 16.7
A-% BS (20 years)
Mean -1.4 -1.4 -1.4 -1.4 -1.7
Std. Dev. 0.9 0.9 0.9 0.9 1.6
Median -1.3 -1.3 -1.3 -1.3 -1.5
Max. -6.0 - -6.0 -6.0 -6.0 -5.0
A-% BS (50 years)
Mean -3.5 -3.7 -3.7 -3.8 -4.2
Std. Dev. 1.7 2.0 1.8 1.8 3.9
Median -3.4 -3.5 -3.5 -3.6 -4.3
Max. -11.0 -13.0 -11.0 -12.0 -20.0
A-% BS (100 years)
Mean -7.6 -7.9 -8.0 -8.4 -7.5
Std. Dev. 3.2 3.4 3.3 3.6 6.5
Median -7.5 -8.1 -7.9 -8.1 -6.4
Max -17.0 -21.0 -18.0 -20.0 -33.0
Standard deviations are reported as absolute values.
9-138
-------
Table 9-20. Summary Statistics of the Projected Changes in Soil pH in the NE
Region, Obtained Using the Different Deposition Scenarios or Soil Aggregation
Schemes. The Time Increments Included in the Table are 20, 50, and 100 Years
LTA
TY
LTA-rbc LTA-zbc WBA
Soil pH (initial)
Mean
Std. Dev.a
Median
A-Soil pH (20 years)
Mean
Std. Dev.
Median
Maximum
A-Soil pH (50 years)
Mean
Std. Dev.
Median
Maximum
A-Soil pH (100 years)
Mean
Std. Dev.
Median
Maximum
5.32
0.194
5.34
-0.075
0.140
•0.032
-0.68
-0.167
0.187
-0.086
-0.88
•0.355
0.278
-0.272
-1.10
5.30
0.206
5.33
-0.075
0.132
-0.040
-0.67
-0.181
0.198
-0.105
-0.90
•0.385
0.298
-0.326
-1.12
5.32
0.194
5.34
-0.078
0.143
-0.036
-0.71
-0.181
0.194
-0.108
-0.91
-0.389
0.295
-0.299
-1.12
5.32
0.194
5.34
-0.081
0.145
-0.037
-0.73
-0.192
0.203
-0.114
-0.94
-0.418
0.310
-0.344
-1.15
5.30
0.206
5.33
-0.046
-0.131
-0.011
-0.65
-0.116
0.177
-0.048
-0.84
-0.289
0.274
-0.210
-1.01
a Standard deviations are reported aa absolute values.
9-139
-------
NE Lakes
Deposition = LTA
Year = 50
Model = Reuss
C'.OH
o
Q_
O
a!
0.6 -
0.4
£
0.0
-30.00
95 % Conf. Limit
Predicted Distribution
5 % Conf. Limit
-20.00 -10.00
A Base Saturation (%)
o.oo
NE Lakes
Deposition = LTA
Year = 50
Model = Reuss
ci.o-
o
0.8 -
Q.
O
O
o.o
95 % Conf. Limit
Predicted Distribution
5 % Conf. Limit
B
-0.75
-0.50
A pH
-0.25
0.00
Figure 9-39. Cumulative distribution of projected (a) base saturations and (b) soil pH values for
soils in NE. Projections were made using the Reuss model in conjunction with the LTA (constant
level) deposition. The results are presented for 50 years.
9-140
-------
NE Lakes
Deposition = LTA
Year = 100
Model = Reuss
Q.
O
-0.6H
E
3
O
0.0
-30.00
95 % Conf. Limit
Predicled Distribution
5 % Conf. Limit
-20.00 -10.00
A Base Saturation (%)
0.00
NE Lakes
Deposition = LTA
Year = 100
Model = Reuss
a.
o
£
-------
To conduct this analysis, aggregated, watershed-level estimates of mineral horizon base saturations were
obtained for the 145 watersheds in the NE. These data were plotted against (1) the observed lake water
ANC values for each of the lakes and (2) the projected changes in ANC at 20, 50 and 100 years. Figure
9-41 shows the relationship observed between aggregated soil base saturations and surface water ANC.
These data support a significant relationship between these variables (see Section 8.8.1). Although there
is considerable scatter in the results, lakes with lower ANC values tend to have soils with lower
aggregated base saturations.
The relationship between current base saturation and projected changes in ANC is more
pronounced, as illustrated in Figure 9-42. In this analysis, the projected magnitude of change in ANC
at 20-, 50-, and 100-year intervals is related to the current, aggregated watershed base saturation. At each
of these time steps, watersheds with aggregated soil base saturations in excess of 20% exhibit little or
no significant decline in projected ANC over the course of the simulations. As the base saturations
decrease below 20 percent, however, there is a marked increase in the magnitude of the response of
individual systems to the effects of acidic deposition. These results suggest that systems with aggregate
base saturation of less than 20 percent should be most susceptible to the effects of acidic deposition,
at least in terms of projected changes in surface water ANC.
An alternative approach is to examine changes in soil base saturations as a function of the current
state of the systems. Figure 9-43A shows the relationship between current, aggregated, mineral soil base
saturations and projected changes in base saturation at 50 years. A significant relationship does not exist
between the magnitude of the projected changes and the current base saturation of the systems being
studied. This result is interesting, especially in light of the rather strong relationship observed between
base saturation and the projected change in surface water ANC. The observation suggests that the
largest changes in soil base saturation (in the absence of weathering) occur independently of the present
base status of soils. The magnitude of the changes may be mediated by physical factors, such as the
thickness or bulk density of the soils. There are some chemical limitations on these changes as well.
9-142
-------
o
tn
400;
300-
"200-
100-
o-
0 a
a
o a
D
a a
a
a
a
a o °
a
a G "
n
a
aa
aa a °
a a
a a a D
a
a o
a a Q a
^ •>*- ° ••".-•_- *
a a D o
n a % a „ n"
jpjO.0 3.° a of D
*r° Q1 a a^a a a "
tP a
n D °
0 aa B
a
a
a
a
a
a
a
Aggregated
Mineral Horizon
Only
n = 145
m= 6.10
b = 29.3
r2 = 0.310
20 40
%BS
60
Figure 9-41. Plot of the measured (ELS) ANC values for lakes in the NE vs. the estimated,
watershed-level base saturations for mineral horizons in those watersheds.
9-143
-------
T -5
O
-1O
to
•o
-1S
. * *--"
*".* - - r
•* * " '
Aggregated
Mineral Horizon
Only
20 Years
20
4O
60
%BS
-2O
-4O
-80
Aggregated
Mineral Horizon
Only
60 Years
B
20
%BS
6O
—. -SO-
•5 -1OO
-g -ISO-
-200
Aggregated
Mineral Horizon
Onty
1OO Years
20
%BS
60
Rgure 9-42. Plot of the changes in surface water ANC values at (a) 20, (b) 50, and (c) 100 years
as projected by the Reuss model vs. the estimated, present-day, watershed-level base saturations
for mineral horizons in those watersheds. The deposition used in computing these differences is
the LTA deposition.
9-144
-------
U"
-1
-2
-3
-4-
8 *
**•
-8
-9
-10-
-11-
-12-
£
a ° a
'*' ••"•"* V „ „ '
;-*•*>;"• • • •
a • n_ * ' a a
<-BS • "
-------
Current base saturation is related to the relative magnitude of changes expected to occur in the
base status of these soils (Figure 9-43B). Although the data are scattered somewhat (probably due to
differences in soil physical parameters and to variations in soil exchange properties) the lower the Initial
base saturation, the greater the projected relative depletion of base cations from the soil exchange
complex. This result is consistent with the observations concerning surface water ANC changes and
demonstrates that the soils are behaving in an internally consistent manner.
As noted throughout this section, the Level II models are, by and large, single-process models,
used in this context to determine the contribution of individual processes to the integrated responses of
watersheds as complete systems. The suggestion that systems with base saturations in excess of 20
percent are at minimal risk to future change needs to be considered in the context of the complete
system. Therefore, watersheds with higher aggregate base saturation could experience significant
acidification if other processes, such as hydrologic routing of water within the soils and ground water,
restrict the degree of interaction between soils and soil-water. Similarly, the present base saturation status
of soils probably plays only a limited role in regulating episodic acidification (as opposed to chronic
acidification, the principal issue of concern in this report).
Conversely, soils with base saturations of less than 20 percent might not experience significant
chronic depletions in ANC if related processes, such as mineral weathering, were able to sustain current
base saturations. The above analysis, however, suggests that these systems are more susceptible to
adverse changes. Programs designed to monitor future changes should consider using soil base
saturation status as one criterion for site selection.
9.3.3.1.3.3 Southern Blue Ridge Province -
Summaries of the results for the SBRP are given in Tables 9-21 and 9-22 and in Figures 9-44 and
9-45. Soils in the SBRP currently have base saturations that are half as large as those in the NE (see
Section 5.5.1.3). For this reason, it is reasonable to expect both larger and more rapid responses to the
9-146
-------
Table 9-21. Summary Statistics of the Projected Changes in Soil
Base Saturations in the SBRP, Obtained Using the Different Deposition
Scenarios. The Time Increments Included In the Table Are 20, 50, 100, and
200 Years
LTA
TY
LTA-rbc LTA-zbc
% BS (initial)
"Mean 10.5 10.5 10.5 10.5
Std. Dev.a 5.7 5.7 5.7 5.7
Median 9.3 9.3 9.3 9.3
A-% BS (20 years)
Mean -0.49 -0.55 -0.59 -0.70
Std. Dev. 0.27 0.29 0.30 0.32
Median -0.44 -0.51 -0.51 -0.59
Max -1.09 -1.18 -1.18 -1.26
A-% BS (50 years)
Mean -1.89 -2.41 -2.37 -2.94
Std. Dev. 0.37 0.58 0.37 0.49
Median -1.90 -2.52 -2.42 -2.96
Max -2.80 -4.26 -3.28 -3.82
A-% BS (100 years)
Mean -5.16 -6.04 -6.00 -7.16
Std. Dev. 0.76 1.14 0.70 0.99
Median -5.06 -5.64 -5.84 -7.15
Max -7.24 -9.22 -7.93 -8.90
A-% BS (200 years)
Mean -8.83 -9.03 -9.36 -9.44
Std. Dev. 0.99 1.38 1.45 1.32
Median -8.78 -9.10 -9.44 -9.68
Max -12.41 -12.41 -13.03 -12.41
Standard deviations are reported as absolute values.
9-147
-------
Table 9»22. Summary Statistics of the Projected Changes In Soil pH
in the SBRP, Obtained Using the Different Deposition Scenarios. The
Time Increments Included in the Table Are 20, 50, 100, and 200 Years
LTA
Soil pH (initial)
Mean
Std. Dev.a
Median
A-Soil pH (20 years)
Mean
Std. Dev.
Median
Maximum
A-Soil pH (50 years)
Mean
Std. Dev.
Median
Maximum
A-Soil pH (100 years)
Mean
Std. Dev.
Median
Maximum
A-Soil pM (200 years)
Mean
Std. Dev.
Median
Maximum
5.15
0.10
5.13
-0.03
0.01
-0.03
-0.06
-0.10
0.03
-0.10
-0.19
-0.34
0.09
-0.35
-0.49
-0.66
0.15
-0.67
-0.82
TY LTA-rbc
5.12
0.10
5.12
-0.03
0.01
-0.03
-0.06
-0.13
0.05
-0.15
•0.21
•0.40
0.14
-0.47
-0.64
-0.65
0.16
-0.64
•0.81
5.15
0.10
5.12
-0.03
0.01
-0.04
-0.06
-0.13
0.05
-0.14
-0.21
-0.41
0.12
-0.45
-0.58
-0.69
0.15
-0.71
-0.84
LTA-zbc
5.15
0.10
5.13
-0.04
0.01
-0.04
-0.06
-0.19
0.06
-0.20
-0.27
-0.52
0.14
-0.57
-0.68
-0.74
0.17
-0.81
-0.86
Standard deviations are reported as absolute values.
9-148
-------
SBRP Stream Reaches
Deposition = LTA
Year = 50
Model = Reuss
o 0-8 -
0_
O
1-0.6 H
Q_
£
95 7. Conf. Limit
Predicted Distribution
5 % Conf. Lfmit
-30.00
-20.00 -10.00
A Base Saturation (%)
o.oo
SBRP Stream Reaches
Deposition = LTA
Year = 50
Model = Reuss
o
r
o
a.
o
0.8 -
D
30.2-)
E
95 7. Conf. Limit
Predicted Distribution
5 % Conf. Limit
B
-0.75
-0.50
A pH
-0.25
0.00
Rgure 9-44. Cumulative frequencies of changes in (a) soil base saturation and (b) soil pH for the
population of soils in the SBRP. The projections are for year 50 and have been computed using
LTA deposition data.
9-149
-------
SBRP Stream Reaches
Deposition = LTA
Year = 100
Model = Reuss
C 1-0 i
o
?OJI
CL
o
£0.61
-------
effects of acidic deposition in the SBRP compared to the NE. Examination of the model results, however,
suggests that the soils in the SBRP respond more slowly to acidic deposition than do the soils in the NE.
At 50 years, the average base saturations in SBRP soils have declined by between 20 percent and
30 percent, depending on the deposition scenario considered. These declines are equivalent to absolute
changes in base saturation of 2 to 3 percent. By 100 years, the average base saturation for the soils
in this region have declined by 60 percent +. 10 percent, and by 200 years, by approximately 90 percent.
Again, these projections are made with the assumption that weathering is not supplying base cation to
the soils of the region. Clearly, primary mineral weathering supplies base cations to these soils and,
hence, soil acidification will be slower than the rates projected here.
Changes in soil pH projected using the Reuss model are parallel to those projected for the soil
base cations. Changes are minimal at 20 years, with an absolute magnitude of the projected changes of
0.04 pH units, regardless of the deposition scenario used. By 50 years, the changes are significant.
Within this time frame, soil pH values have declined by an average of about 0.13 pH units, depending
on the deposition scenario. The rate of decline in soil pH increases between 50 and 100 years. At this
point, the soil is projected to be losing much of the buffering capacity, with a resultant drop in soil pH.
By 200 years, when much of the soil buffering capacity has been depleted, the average soil pH has
declined to values near the minimum that can be reached in the context of the Reuss model framework.
9.3.3.1.3.4 Comparison of results from the NE and SBRP -
Results from the Reuss modelling effort have led to many observations concerning the soil behavior
in the two regions and how that behavior affects the ANC of waters passing through those soils (see
Table 9-23). First, the absolute rate of cation depletion is slower in the SBRP than it is in the NE. Within
the first 50 years, mean base saturations have declined by about 3.5 percent in the NE, while they have
declined by only slightly less than 2 percent in the SBRP. However, in terms of the percentage of
available cations, cation depletion is severe in the SBRP. After 50 years, between 20 and 30 percent of
the cations on soil exchange sites have been lost through leaching, whereas in the NE only about 15-
9-151
-------
Table 9-23. Comparison of the Changes in Soil Base Saturation and Soil pH that
Are Projected to Occur in the NE and SBRP. The Projections Have Been
Obtained Using Reuss's Cation Exchange Model and Are Presented for Two
Deposition Scenarios, the LTA and TY Depositions
LTA TY
a Standard deviations are reported as absolute values.
NfSBRP NE SBRP
A_ANC (year 50)
Mean
Std. Dev.a
-13.7
23.6
-2.96
1.05
-16.1
26.4
-3.7
2.0
AJfcBS (Year 50)
Mean
Std. Dev.
-3.5
1.7
-1.9
0.4
-3.7
2.0
-2.4
0.5
A^Soil pH (Year 50)
Mean
Std. Dev.
-0.17
0.19
-0.10
0.03
-0.18
0.20
-0.13
0.05
A_ANC (Year 100)
A_
A_
Mean
Std. Dev.
% BS (Year 100)
" Mean
Std. Dev.
_Soil pH (Year 100)
" Mean
Std. Dev.
^32.1
36.1
-7.6
3.2
-0.36
0.28
-14,6
6.04
-5.2
0.8
-0.34
0.09
-43.1
51.5
-7.9
3.4
-0.39
0.30
-23.0
16.2
-6.0
1.1
-0.40
0.14
9-152
-------
20 percent of the available cations are lost during the same time period. These trends continue at 100
years. In the NE, base saturations have declined by about 7.5 percent, or about one third of the total
supply of available cations on soil exchange sites. In the SBRP, base saturations have declined by only
slightly more than 5 percent. However, this decline constitutes more than half of the available buffering
capacity.
For soil pH, parallel trends to those describe above are observed. In the NE, soil pH values decline
by about 0.2 pH units during the first 50 years of the simulations, while In the SBRP, the average change
Is on the order of 0.1 pH units. However, because soil pH values In the NE are initially higher than those
In the SBRP by an average of about 0.15 pH units, the observed differences result primarily in a lessening
of the disparity between the two regions in terms of their characteristic pH values. By 100 years, changes
In soil pH in the SBRP have started to accelerate such that the absolute magnitude of the differences
observed between the two regions are, again, equal to about 0.15 pH units. We interpret this observation
as an indication that the loss of buffering capacity occurs later in the SBRP relative to the NE This
difference Is attributable to differences in soil physical properties, such as soil thickness and bulk density,
rather than to differences in soil chemical characteristics. The absolute magnitude of the changes
projected for the two regions is equal to about 0.35-0.4 pH units (depending on the deposition scenario
considered).
Results from the Reuss model suggest that, in the absence of mineral weathering, both regions will
sustain substantial losses of base cations from their soils. In translating these changes into the effects
on surface water chemistry, the model results suggest that the largest effects (on the time scale of 100
years) will be observed in the NE. Larger changes are projected to occur much earlier in the NE. For
example, after 50 years, the mean change in projected surface water ANC in the SBRP is less than -3
/jeq L"1, whereas it is more than -13 /ueq L*1 in the NE. At 100 years, the rate of ANC decline has
increased in the SBRP. At this point, the projected change for the SBRP is about -15 peq L'1 (using the
LTA deposition; this change is about -23 peq L"1 for the TY deposition). However, this change is still only
half the magnitude of that projected to occur in the NE, regardless of the deposition scenario. Therefore,
9-153
-------
larger relative changes in the base cation pool are projected for the soils in the SBRP, the larger
projected effects of those changes appear in the surface waters of the NE region.
9.3.3.1.3.5 Summary —
A number of observations and conclusions can be drawn from results obtained using the Reuss
model to evaluate changes in soil pH and base saturations in the NE and SBRP regions.
• In the absence of mineral weathering, significant depletions of base cations are projected for
the soils of both the NE and SBRP regions.
• The absolute magnitude of base cation depletion is greater in the NE than it is in the SBRP.
The relative projected changes, however, are greater in the SBRP.
• Current base saturation of soils in the regions can be used as indicators of potential future
change in surface water ANC. Soils with base saturations currently in excess of about
20percent appear to undergo minimal changes on the time scale of the next 100 years. For
soils with base saturations less than 20 percent, however, projected changes in surface water
ANC appear to increase with decreasing aggregate base saturation. This effect is more
pronounced in the NE region than it is in the SBRP.
• Current base saturation can be used as an indicator of the anticipated relative changes that
might occur in the soil base status over the next tOO years. The percentage decline in base
saturation increases with decreasing base saturation, although other factors, such as soil
thickness or bulk density, probably influence the relationship as well.
9.3.3.2 Bloom-Grigal Model
9.3.3.2.1 Data sources -
in the DDRP, the basic unit of investigation is the watershed. Instead of characterizing the effects
of acidic deposition on individual soils, the research focus is the integrated effect of the soils on a
particular watershed. Consequently, all of the Bloom-Grigal modelling input data are at the watershed
level. Because the DDRP sample of watersheds serve as the basic link to the target population of
watersheds, watershed level results can be extrapolated to the target population of watersheds.
9-154
-------
The data required to run the Bloom-Grigal model include total annual wet and dry deposition, total
annual runoff, and selected soil chemistry data. All of these data were collected as a part of the DDRP
and are discussed in detail in Section 5.
9.3.3.2.1.1 Deposition data -
The deposition data are from four sources: (1) Typical Year (TY), (2) Long-Term Annual Average
(LTA), (3) LTA Reduced Base Cation (LTA-rbc)-LTA with a 50 percent reduction in dry base cations, and
(4) LTA Zero Base Cation (LTA-zbc)--LTA with a 100 percent reduction in dry base cations. Both of these
reductions in dry base cations are offset by concomitant increases in dry H*. The details on the
acquisition/generation of the DDRP deposition data sets are given in Section 5.6.
A summary of the regionally weighted median deposition inputs in the four deposition data sets
(LTA, LTA-rbc, ITA-zbc, and TY) used in the Bloom-Grigal modelling is presented in Table 9-24 by region.
In the NE there appears to be little difference between LTA and TY. A priori, we expect to see only
minor differences in the forecasts made with these two deposition data sets. The S8RP TY median value
of H* is 22 percent greater than the LTA value. The NH4+ is, however, lower and NO3" is greater.
Consequently, the total effective acidity {H*tota| = H* + NH4+ - NO3") is only slightly larger.
The largest differences in H+to{aJ are between the LTA and the reduced (LTA-rbc) and zero (LTA-
zbc) deposition data sets. In the NE the difference between the median H+tota| in the LTA and median
value of H+tota| in the LTA-zbc is 0.19 keq ha"1, while in the SBRP this difference is 0.24 keq ha"1. Such
differences should result in differences in projections, especially for the higher levels of H*tota].
9.3.3.2.1.1.1 Deposition scenarios -
The Level II base cation models are run with three deposition scenarios. The scenario common
to both the NE and SBRP is the constant deposition scenario. In this scenario the annual load of
deposition is held constant for the duration of the simulation.
9-155
-------
Table 9-24. Regionally Weighted Median Values of Initial Annual Deposition Inputs to the Bloom-
Grigal Model for the Northeastern Region and the Southern Blue Ridge Province8.
H+ NH4+ NO3~ Total Acid lnputb
NE
LTA
LTA-rbc
LTA-zbc
TY
SBRPLTA
LTA - rbc
LTA - zbc
TY
0.71
0.79
0.91
0.78
0.67
0.82
0.97
0.82
0.15
0.15
0.15
0.14
0.22
0.22
0.22
0.16
0.44
0.44
0.44
0.45
0.42
0.42
0.42
0.46
0.43
0.49
0.62
0.44
0.47
0.61
0.77
0.51
a Values are in keq ha'1 yr*1
b Total Acid Input - [H* + NH4+ - NOT!
9-156
-------
9.3.3.2.1.1.2 Northeast -
In addition to the constant deposition scenario in the NE, a ramp down scenario is used to simulate
a 30 percent decrease in wet and dry SO42" deposition. Deposition is held constant for the first ten
years of the simulation. Beginning with the eleventh year, deposition is decreased by 2 percent per year
for fifteen years for a total decrease of 30 percent. This new level is then held constant for the duration
of the simulation.
9.3.3.2.1.1.3 Southern Blue Ridge Province -
In addition to the constant deposition scenario in the SBRP, a ramp up scenario is used to simulate
a 20 percent increase in wet and dry SO42" deposition. Deposition is held constant for the first ten years.
Beginning with the eleventh year, deposition is increased by (20/15) percent per year for fifteen years
for total increase in deposition of 20 percent. This new level is then held constant for the duration of the
simulation.
9.3.3.2.1.2 Soils data -
The Bloom-Grigal model uses one value for the following soil chemistry variables to depict the soil
chemistry of a particular watershed: soil pH, cation exchange capacity (CEC), and the sum of
exchangeable base cations (SOEBC). To obtain results that represent the central tendency of the DDRP
regions, a large number of observations for these van'ables were aggregated to obtain values for each
of the DDRP watersheds. Combining or aggregating these data can be accomplished in several ways.
It is not correct to use a simple average for all variables; rather, capacity and intensity variables should
be weighted differently. Of the variables used in the Bloom-Grlgal model simulations, soil pH was
aggregated using an intensity variable aggregation method, whereas CEC and SOEBC were aggregated
using a capacity variable aggregation method. The details of these methods are provided in Johnson
et al. (I988b).
9-157
-------
To evaluate the role of soil organic horizons (Oa, Oe, and Oi) in the chemistry of soils, the soils
data for the Bloom-Grigal data were aggregated two ways: (1) including organic horizons and
(2) excluding organic horizons.
A summary of the regionally weighted median values of the Bloom-Grigal soil chemistry input data
(aggregated with and without organic horizons) is presented in Table 9-25. In the NE, Inclusion of the
organic horizons decreases the median pH by 0.30 and base saturation by slightly more than 1 percent.
In the SBRP the changes are even more negligible. Although the pH and SOEBC values are similar
between the regions, CEC in the SBRP is more than twice that in the NE. Simply stated, the soils in
the SBRP have greater exchangeable acidity than those in the NE with similar SOEBC.
The regional initial soil pH and percent base saturation with and without organic horizons are
presented in Figure 9-46 as cumulative distribution functions (CDFs). This manner of presentation allows
interregional and intraregional differences to be easily observed. The soil pH in the SBRP is less affected
by the exclusion of the organic horizons than in the NE.
9.3.3.2.2 Model projections -
In all, Bloom-Grigal model simulations representing more than 300,000 years were needed to obtain
the results for the four deposition data sets and different deposition scenarios. A subset of these is
presented below by region, and a regional comparison follows in Section 9.3.3.2.3.
9.3.3.2.2.1 Northeast region -
The results of the Bloom-Grigal simulations in the NE with LTA, LTA-rbc, and LTA-zbc are presented
in Figures 9-47 and 9-48 for the change in soil pH and percent base saturation, respectively. Statistical
summaries of the CDFs are presented in Tables 9-26 and 9-27.
The projected changes in soil pH and percent base saturations using the constant LTA deposition
scenario are quite small (Figure 9-47). The median change after 100 years is only -0.04. Of the systems
9-158
-------
Table 9*25. Regionally Weighted Median Values of Annual Initial Soil Chemical Values Input Into
the BloonvGrigal Model for the Northeastern Region and the Southern Blue Ridge Province3.
NE
SBRP
PH
SOEBC
CEC
BS
With
W/0
With
W/O
PH
4.62
4.92
4.85
5.01
SOEBC CEC
40.04 183.8
34.11 177.4
40.42 433.3
40.62 436.4
BS
2198
20.60
9.22
9.20
=• intensity weighted soil pH
« mass weighted sum of exchangeable base cations
= mass weighted cation exchange capacity
base saturation [{SOEBC/CEC)*100]
1 All values in keq ha*1 except BS which is percent.
9-159
-------
0<>-«1
o
Q_
OQ.t
o
0.0
Soil pH
Organic Horizons Included
NE
S8RP
4.5 S.O S.S
Soil pH
6.0
Percent Base Saturation
Organic Horizons Included
i.o
§o.e-
o
a.
oo.s-
a)
2 0.4
o
0.0
NE
SBRP
0 10 20 3O 40 SO CO
Base Saturation (%)
1.0
r
o
a
00.6
0.0
4.0
Soil pH
Organic Horizons Excluded
NE
SBRP
4.5 3.0 S.5
Soil pH
6.0
1.0 -i
O
a.
o
0.0
Percent Base Saturation
Organic Horizons Excluded
NE
SBRP
10 20 30 40 SO 60
Base Saturation (%}
Rgure 9-46. Cumulative distributions of aggregate initial soil pH and percent base saturation in the
NE and SBRP, with arid without organic horizons.
9-160
-------
NE Lake Watersheds
Deposition = LTA Constant
Organic Horizons = Included
Dry Base Cations = 100 %
Jo..
r
o
Q.
O0.»
O
3
E.
0*
Yr. 20
Yr. SO
™_ Yr. 100
-O.SO -0.23
A Soil pH
ooo
HE. Lake Watersheds
Deposition = LTA Ramp 30% Decrease
Organic Horizons = included
Dry Base Cations = 100 %
r
o
a.
O0.6
o
yr. JO
Yr. so
Yr. 100
-0.23
A Soil pH
0.00
NE Lake Watersheds
Deposition = LTA Constant
Organic Horizons = Included
Dry Base Cations = SO %
I"
o
a.
go.«
O.
— Yr. 10
Yr. SO
—- Yr. 100
NE Lake Watersheds
Deposition = LTA Ramp 30% Decrease
Organic Horizons = Included
Dry Base Cations = 50 ?J
•c
o
a.
oo.«
_— Tr. 20
»». SO
Tr. 100
A Soil pH
-OJO -O.Z5
A Soil pH
0.00
NC Lake Watersheds
Deposition = LTA Constant
Organic Horizons = Included
Dry Base Cations = 0 %
I"
o
a.
— Yf. 20
• - Yf. »
— Yr. 100
-0.23
A Soil pH
NE Lake Watersheds
Deposition = LTA Ramp 30% Decrease
Organic Horizons = Included
Dry Base Cations = 0 %
—— Yr. JO
Yr. 90
Yr, 100
O
a.
po.«
0.0
-0.30 -0.23
A Soil pH
Figure 9-47. Regional CDFs of the projected change in the pH of soils on NE lake watersheds
under constant and ramp down (30 percent 4) deposition scenarios after 20, SO, and 100 years of
LTA, LTA-rbc, and LTA-zbc deposition. Organic horizons are included.
9-161
-------
NE Lake Watersheds
Deposition = CTA Constant
Organic Horizons = Included
Dry Base Cations = tOO %
i-
0
-10.00 -1CJOO
A Base Saturation (%)
NE Lake Watersheds
Deposition = LTA Ramp 30% Decrease
Organic Horizons = Included
Dry Bose Cottons = 100 %
- - - vJ;IS
tr. 1•) deposition scenarios after 20, 50,
and 100 years of LTA, LTA-rbc, and LTA-zbc deposition. Organic horizons are Included.
9-162
-------
Table 9-26. Bloom-Grigal Model Regional Projections for the Change in Soil
pH in the Northeastern United States. Projections Made Using LTA, LTA-rbc, and LTA-zbc
Deposition with Constant and 30% Ramped Down Deposition Scenarios at Three
Levels of Base Cations in Dry Deposition. Results Reported for 20-, 50-, and
100-Year Projections. Organic Soil Horizons Included
YEAR
YEAR
YEAR
YEAR
Deposition - Constant ** Dry Base Cations = 100% ** LTA
MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.02
-0.04
•0.06
0.02
0.05
0.07
-0.10
-0.19
-0.35
-0.03
•0.06
-0.10
-0.01
-0.03
-0.04
0.00
0.00
0.00
0.00
0.00
0.00
Deposition = Constant ** Dry Base Cations » 50% ** LTA - rbc
MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.03
-0.06
-0.09
0.03
0.05
0.07
-0.13
-0.22
-0.38
-0.04
-0.09
-0.13
-0.02
0.05
-0.08
-0.01
-0.02
-0.04
0.00
0.00
0.00
Deposition = Constant ** Dry Base Cations « 0% ** LTA - zbc
MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
SO
100
-0.05
-0.10
-0.14
0.04
0.06
0.08
-0.17
-0.27
-0.44
-0.08
-0.14
-0.18
-0.05
-0.09
-0.14
-0.03
-0.06
-0.10
0.00
0.00
0.00
Deposition = 30% Decrease
MEAN STD DEV MIN
Dry Base Cations = 100% ** LTA
P 25 MEDIAN P 75 MAX
20
50
100
-0.01
-0.02
-0.02
0.02
0.02
0.03
-0.07
-0.11
-0.17
-0.02
-0.03
-0.03
-0.01
0.01
-0.01
0.00
0.00
0.00
0.00
0.00
0.00
continued
9-163
-------
Table 9-26. (Continued)
Deposition = 30% Decrease ** Dry Base Cations = 50% ** LTA - rbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.02
•0.03
-0.04
0.02
0.03
0.04
-0-11
-0.14
•0.22
-0.03
-0.04
-0.05
-0.02
-0.02
-0.03
-0.01
-0.01
•0.01
0.00
0.00
0.00
Deposition = 30% Decrease ** Dry Base Cations » 0% ** LTA - zbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.05
-0.06
-0.07
0.03
0.04
0.05
-0.15
-0.17
-0.27
-0.07
-0.07
-0.10
-0.04
-0.05
-0.07
-0.02
-0.03
-0.04
0.00
0.00
0.00
9-164
-------
Table 9-27. Bloom-Grigal Model Regional Projections of the Change in Percent
Base Saturation in the Northeastern United States. Projections Made Using LTA,
LTA-rbc, and LTA-zbc Average Deposition with Constant and 30% Ramped Down Deposition
Scenarios at Three Levels of Base Cations in Dry Deposition. Results Reported for 20-, 50-, and
100-Year Projections. Organic Soil Horizons are Included.
Deposition = Constant ** Dry Base Cations = 100% ** LTA
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.97
•2.00
-3.15
1.11
2.34
3.79
-5.00
-11.26
-18.70
-1.58
-3.49
-5.83
-0.57
-1.05
-1.46
0.00
0.00
0.00
0.00
0.00
0.00
Deposition = Constant ** Dry Base Cations = 50% ** LTA - rbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-1.43
-2.90
-4.45
1.20
2.47
3.95
-5.56
-12.11
-19.88
-2.09
-4.14
-6.96
-1.25
-2.63
-3.68
-0.49
-0.96
-1.26
0.00
0.00
0.00
Deposition = Constant ** Dry Base Cations = 0% ** LTA - zbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-2.32
-4.52
-6.58
1.41
2.80
4.33
-6.63
-13.72
-21.94
-3.05
-5.74
-8.82
-2.24
-4.29
-6.16
-1.50
-2.70
-3.36
0.00
0.00
0.00
Deposition = 30% Decrease ** Dry Base Cations = 100% ** LTA
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.74
-0.90
-1:14
0.90
1.20
1.69
-4.38
-6.86
-10.03
-1.32
-1.43
-1.57
-0.38
-0.38
-0.38
0.00
0.00
0.00
0.00
0.00
0.00
continued
9-165
-------
Table 9-27. (Continued)
Deposition = 30% Decrease ** Dry Base Cations = 50% ** LTA - rbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-1.18
-1.49
-1.91
1.07
1.54
2.26
-4.94
-8.36
-12.66
-1.69
-2.25
-2.55
-1.00
-1.03
-1.03
-0.31
-0.31
-0.31
0.00
0.00
0.00
Deposition = 30% Decrease ** Dry Base Cations = 0% ** LTA - zbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-2.02
-2.68
-3.49
1.29
1.97
2.95
•6.02
-10.01
-15.27
-2.71
-3.57
-4.83
-1.86
-2.50
-2.85
-1.19
-1.21
-1.21
0.00
0.00
0.00
9-166
-------
In the target population, less that 25 percent of the watersheds have a projected decrease in soil pH
greater than -0.10. The largest decrease is projected to be -0.35. Most of these changes are probably
within the uncertainty of the model and are not significant. With the LTA-rbc and LTA-zbc deposition
larger decreases in soil pH in a greater proportion of the systems is evident. Yet, even at the highest
level of acidic input (LTA-zbc), the median change in soil pH is only -0.14. Larger decreases are evident
in a few systems (<10 percent).
Ramping the deposition down by 30 percent reduces the projected declines significantly. With LTA-
zbc the median decline in soil pH is -0.07, one half that projected for the constant LTA scenario. The
results for the projected change in base saturation are similar to those for pH. However, when the initial
median base saturation is only 17 percent, a decrease of 6 percent (LTA-zbc) after 100 years (to 11
percent) is projected. The 30 percent decrease In deposition results in smaller changes.
Excluding the organic horizons results in an amplified decrease in soil pH and base saturation
(Figures 9-49 and 9-50, Table 9-28 and 9-29). Without the contribution of the organic horizons, the
median change in soil pH and percent base saturation after only 20 years is nearly equal to or greater
than the 100-year projections for soils with organic horizons. This result Is misleading, however. The
initial median pH of the soils without the organic horizons is 4.92, and after 100 years of LTA deposition
the median change is -0.21. For the soils with the organic horizons the initial median pH is 4.62, and
after 100 years the median change is only -0.04. Thus, although pH of the soils without the organic
horizons had greater projected changes, their pH values were still projected to be higher at the end of
the 100-year simulation.
As for pH, the decrease in percent base saturation for the soils with the organic horizons is greater
than for the soils without the organic horizons. However, because percent base saturation is initially
lower for the soils without the organic horizons, the projected percent base saturation is much lower
than for the soils with the organic horizons.
9-167
-------
N£ Loke Watersheds
Deposition = LTA Constant
Organic Horizons = Excluded
Dry Base Cations = 100 %
°
I
s
I,
O
r£ so
—w, Yr. l«
A Soil pH
ftie
NC Lak« Watersheds
Opposition = LTA Ramp 30% Decrease
Organic Horizons ~ Excluded
Dry Base Cafions = tOO %
\
0.
J*
o
Tr. 20
»r. 50
— Tr. 100
-(US
Soil pH
NE Lak« Watersheds
Deposition = LTA Constant
Organic Horizons = Excluded
Dry Base Cations = 50 %
Jo.
o
gjp -O.H
A Soil pH
NE Lake Watersheds
Deposition = LTA Ramp 30% Decrease
Organic Horizons = Excluded
Dry Base Cations = 50 %
~— it. 10
Yr. So
r
o.
A Soil pH
I
I-
3
I..
NE Lake Watersheds
Deposition = LTA Constant
Organic Horizon* = Excluded
Dry Base Cations = 0 3E
NE take Watersheds
Deposition = LTA Ramp 30% Decrease
Organic Horizons = Excluded
Dry Base Cations = 0 %
I-
A Soil pH
A Soil pH
Rgure 9-49. Regional CDFs of the projected change in the pH of soils on NE lake watersheds
under constant and ramp down (30% 4) deposition scenarios after 20, SO, and 100 years of LTA,
LTA-fbc, and LTA-zbc deposition. Organic horizons are excluded.
9-168
-------
NE Lake Watersheds
Deposition = LTA Constant
Organic Horizons = Excluded
Dry Case Cations s 100 %
r
g.
go.i
a.
NE Lake Watersheds
Deposition = LTA Ramp 30% Decrease
Organic Horizons = Excluded
Dry Base Cations ~ 100 %
go,
a.
5.
§«
o
— Tr. M
- - Yr. 30
~ Yr. 100
Base Saturation
-lOjOO -10JJO
Base Saturation (%)
NE Lake Watersheds
Deposition = LTA Constant
Organic Horizons - Excluded
Dry Base Cations = 50 X
NE Lake Watersheds
Deposition = LTA Ramp 307 Decrease
Organic Horizons = Excluded
Dry Base Cations = 50 7.
.2
I
O
o*
o
a.
r
13
1-
-10.08 ejo
Base Saturation (%)
^— w. M
- - Tr. SO
— tr. 100
-MM -IOJO
& Base Saturation
0.00
NC Lake Watersheds
Deposition = LTA Constant
Organic Horizons = Excluded
Dry Base Cations = OX
^—»*. M
NE Lake Watersheds
Deposition = LTA Ramp 30% Decrease
Organic Horizons = Excluded
Dry Base Cations = 0 %
4"
o
a.
o
a.
s«
L
o
-MM
_, _
A Base Saturation (X)
—80iW —ItkOft
Base Saturation (X)
Figure 9-50. Regional CDFs of the projected change in the percent base saturation of soils on NE
lake watersheds under constant and ramp down (30% 4.) deposition scenarios after 20, SO, and
100 years of LTA, LTA-rbc, and LTA-zbc deposition. Organic horizons are excluded.
9-169
-------
Table 9-28. Bloom-Grigal Model Regional Projections of the Change in Soil pH
in the Northeastern United States. Projections Made Using LTA, LTA-rbc, and LTA-zbc Deposition
with Constant and 30% Ramped Down Deposition Scenarios at Three Levels of Base
Cations In Dry Deposition. Results Reported for 20-, 50-, and 100-Year Forecasts.
Organic Soil Horizons Are Excluded
Deposition = Constant ** Dry Base Cations = 100% ** LTA
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.08
-0.16
•0.22
0.07
0.12
0.15
-0.33
-0.45
-0.55
-0.13
-0.27
-0.34
-0.05
-0.12
-0.21
-0.02
-0.06
-0.10
0.00
0.00
0.00
Deposition = Constant ** Dry Base Cations = 50% ** LTA - rbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.10
-0.21
-0.28
0.07
0.13
0.14
-0.38
-0.47
-0.60
-0.16
-0.33
-0.40
-0.09
-0.18
-0.27
-0.04
-0.08
-0.15
0.00
0.00
0.00
Deposition = Constant ** Dry Base Cations = 0% ** LTA - zbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.14
-0.27
-0.35
0.09
0.15
0.16
-0.45
-0.53
-0.69
-0.22
-0.41
-0.48
-0.14
-0.31
-0.39
-0.05
-0.11
-0.19
0.00
0.00
0.00
Deposition = 30% Decrease ** Dry Base Cations = 100% ** LTA
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.06
-0.07
-0.08
0.05
0.06
0.07
-0.28
-0-28
-0.29
-0.10
-0.11
-0.12
-0.04
-0.06
-0.08
-0.02
-0.03
•O.03
0.00
0.00
0.00
continued
9-170
-------
Table 9-28. (Continued)
Deposition = 30% Decrease ** Dry Base Cations - 50% ** LTA - rbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.09
-0.12
-0.15
0.06
0.08
0.09
-0.37
-0.37
-0.37
-0.13
-0.19
-0.22
-0.08
-0.10
-0.13
-0.03
-0.06
-0.08
0.00
0.00
0.00
Deposition = 30% Decrease ** Dry Base Cations » 0% ** LTA - zbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-0.13
-0.19
-0.24
0.08
0.10
0.11
-0.44
-0.44
-0.48
-0.19
-0.28
-0.33
-0.13
-0.20
-0.25
-0.05
-0.09
-0.13
0.00
0.00
0.00
9-171
-------
Table 9-29. Bloom-Grigal Model Regional Projections for the Change in Percent
Base Saturation in the Northeastern United States. Projections Made Using LTA, LTA-rbc, and LTA-
zbc Average Deposition with Constant and 30% Ramped Down Deposition Scenarios at Three
Levels of Base Cations in Dry Deposition. Results Reported for 20-, 50-, and 100-Year Projections.
Organic Soil Horizons Are Excluded
Deposition = Constant ** Dry Base Cations = 100% ** LTA
YEAR
20
50
100
MEAN
-2.40
-4.79
•6.93
STD_DEV
1.58
3.11
4.54
MIN
-6.74
-14.34
-23.19
Deposition = Constant ** Dry
YEAR
20
50
100
MEAN
-3.12
-6.05
-8.56
STD_DEV
1.57
3.02
4.39
MIN
-7.64
-16.04
-24.40
Deposition = Constant ** Dry
YEAR
20
50
100
MEAN
-4.14
-7.66
STD_DEV
1.89
3.51
-10.36
MIN
-8.79
-18.03
4.97
Deposition = 30% Decrease **
YEAR
20
50
100
MEAN
-1.96
-2.45
-3.07
STDJ3EV
1.31
1.83
2.65
MIN
-5.98
-9.04
-14.13
P_25 MEDIAN
-3.56
•6.51
-8.95
Base Cations
P_25
-4.21
-7.87
-9.99
Base Cations
Pj25
-5.55
-9.72
-26.41
-2.16
-4.38
-7.08
= 50% **
MEDIAN
-2.81
-5.59
-8.65
= 0% **
MEDIAN
-4.04
-7.51
-12.25
P_75
-1.41
-2.72
-3.87
LTA - rbc
P_75
-2.04
-4.47
-6.06
LTA - zbc
P_75
-2.49
-5.28
-9.86
MAX
0.00
0.00
0.00
MAX
0.00
0.00
0.00
MAX
0.00
0.00
-7.030.00
Dry Base Cations = 100% ** LTA
P_25
-2.66
-3.20
-4.37
MEDIAN
-1.87
-2.19
-2.46
P_75
-1.06
-1.07
, -1.08
MAX
0.00
0.00
0.00
continued
9-172
-------
Table 9-29. (Continued)
Deposition = 30% Decrease ** Dry Base Cations = 50% ** LTA - rbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-2.79
-3.87
-5.06
1.43
2.16
3.12
-7.00
-11.30
•16.92
-3.83
-5.03
-6.35
-2.50
-3.48
-4.68
-1.85
-2.64
-2.97
0.00
0.00
0.00
Deposition = 30% Decrease ** Dry Base Cations = 0% ** LTA - zbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
-3.81
-5.69
-7.52
1.75
2.68
3.78
-8.13
-13.55
-19.69
-5.07
-7.34
-9.06
-3.68
-5.34
-7.21
-2.32
-3.90
-5.01
0.00
0.00
0.00
9-173
-------
There are two principal explanations for the above results. First, soils without organic horizons have
higher initial pH values. At higher pH values less Al is available to buffer the losses of base cations.
Recalling Equation 9-9 (S = I - A - C), the tendency of a system to lose bases (S) increases if the inputs
of acidity (I) are held constant and the buffering of Al (A) and protonation of bicarbonate (C) are
decreased. Such is the case at higher pH values. Second, the large decreases in soil pH result from
low base saturation, as reflected by the equation that relates soli pH to base saturation (see Equation
9-12). For low base saturation (<20 percent), the slope of the pH versus percent base saturation line
increases dramatically and small changes in base saturation result in large changes in pH. Because the
systems without organic horizons have higher pH values, their base cation losses are greater than for
other soils with lower pH values (e.g., the soils with the organic horizons) assuming all other soils
characteristics are the same. The loss rate of base cations decreases, however, as the soil pH
decreases. Turchenek et al. (1987) and Turchenek et al. (1988), also using the Bloom-Grigal model,
demonstrated similar results.
The median change in base saturation after 50 years of constant LTA deposition on soils without
organic horizons is -4.38, and the pH change is -0.12. After an additional 50 years, the percent base
saturation decreases by an additional -2.44 and the pH by -0.09.
Organic horizons apparently influence the soil chemistry in at least two ways. First, because
organic horizons have abundant base cations, they increase the size of the exchangeable base cation
pool (see Table 9-25). Because of the concomitant addition of CEC, however, the relative magnitude
of the median percent base saturation remains the same. Second, because organic horizons are
inherently acidic, the lower soil pH values decrease the rate of base cation removal from the soil cation
exchange complex. At lower soil pH values, potentially toxic acid cations, such as Al3* and Mn2*
become more prevalent and may be transported in drainage waters to surface water or ground water.
The Bloom-Grigal modelling results using the TY deposition in the NE are similar to those using
the LTA deposition. For this reason they are not presented here.
9-174
-------
9.3.4.2.2 Southern Blue Ridge Province •
Although the median aggregated soil pH values are higher in the SBRP target population of
watersheds than In the NE, SBRP soils have dramatically lower percent base saturation. Because of
these chemical properties, and for the larger reasons described in the preceding section, the soils in
the SBRP are projected to experience decreases in pH and percent base saturation than soils in the NE.
The changes projected for the soils without the organic horizons differ only slightly from those for
the soils with the organic horizons. Unlike the NE, omitting the organic horizons does not appreciably
affect the initial aggregate soil pH and percent base saturation. As for the NE, the forecasts using the
TY deposition data are only slightly higher than those using the LTA. (These data are not presented.)
The CDFs for the projected changes in soil pH and percent base saturation using the LTA, LTA-
rbc, and LTA-zbc deposition data sets are presented in Rgures 9-51 and 9-52. The summary statistics
for these CDFs are presented in Tables 9-30 and 9-31. These results are for the soils with the organic
horizons included.
After 50 years under the constant deposition scenario, the median predicted change in soil pH is
-0.16. After 100 years it is -0.24. From year 100 to year 200 the change is only -0.07. The change in
percent base saturation after 100 years is -3.22, and after 200 years of the change is only -3.39. These
results imply that between year 100 and 200 the buffering mechanism these soils shifts with the latter
mechanism buffering soil pH to more acidic levels.
Projected changes with the increased acid loadings of the LTA-rbc and LTA-zbc are much more
rapid. After 50 years under constant LTA-rbc deposition, the projected change in soil pH equals that
under the LTA deposition after 100 years. With the LTA-zbc, an equivalent projected change occurs in
less than 50 years. The 20 percent ramped increase in deposition further increases the rates of projected
change: increased acid inputs increase the initial rate of change, i.e., the decrease in base saturation
and soil pH. The convergence of the CDFs for 50, 100, and 200 years demonstrates these results.
9-175
-------
i«
T
O
SBRP Stream Watersheds
Deposition = LTA Constant
Organic Horizons = Included
Dry Base Cations = 100 %
-oio -oj
A Soil pH
oA»
SBRP Stream Watersheds
Deposition = LTA Ramp 20% Increase
Organic Horizons = Included
Dry Base Cations = 100 %
o
I*
0.
— Tr. 30
--- Tr. SO
---- Tr. tOO
Tr. 200
UO -0.11
A Soil pH
t
o
Q.
SBRP Stream Watersheds
Deposition = LTA Constant
Organic Horizons = Included
Dry Base Cations = 50 %
-'
SBRP Stream Watersheds
Deposition = LTA Ramp 20% Increase
Organic Horizons = Included
Dry Base Cations = SO %
It. M
c«j§-
r
o
a.
n
I-
"5
o
r-''
f ' , -
A Soil pH
-ajo -045
A Soil pH
OLOO
SBRP Stream Watersheds
Deposition = LTA Constant
Organic Horizons = Included
Dry Base Cations = 0 %
I"
I
o
SBRP Stream Watersheds
Deposition = LTA Ramp 20% Increase
Organic Horizons = Included
Dry Base Cations = 0 %
2"1
n.
l«
S
A Soil pH
A Soil pH
Figure 9-51. Regional CDFs of the projected change in the pH of soils on SBRP stream watersheds
under constant and ramp up (20% t) deposition scenarios after 20, 50, 100, and 200 years of LTA,
LTA-rbc, and LTA-zbc deposition. Organic horizons are included.
9-176
-------
SBRP Stream Watersheds
Deposition — LTA Constant
Organic Horizons = Included
Dry Base Cations s 100 %
ija -
• Proportion
2 8
|*4
1-
YrT SO /•
•--- Tr. I0» »'
Tr. 2M f 1
|
il
•i '
SBRP Stream Watersheds
Deposition = LTA Ramp 20% Increase
Organic Horizons — Included
Dry Base Cations = 100 %
1.0-
I"-
a.
go...
|L
f"
i-
on.
Tr. 20 -
Tr. SO ft
Tr. 100 (,
Tr. 200 J ,
!'
I f
I
> 1
f f
II
/;
--•**" ' »' /
*•
_T
A Base Saturation (%)
-MM -VIM -1040 9.00
a Base Saturation (%)
SBRP Stream Watersheds
Deposition = LTA Constant
Organic Horizons * Included
Dry Base Cations = 50 %
o
a.
I'
Tr.
Tr.
*
-nee -nun -IOJM
A Base Saturation (X)
SBRP Stream Watersheds
Deposition = LTA Ramp 20% Increase
Organic Horizons = Included
Dry Base Cations = SO %
|«
o
a.
|"
I-
Tr. M
Tr. SO
-« Tr. IOO
Tr. MO
—20.00 — lOjOO OjOO
A Base Saturation (%)
SBRP Stream Watersheds
Deposition — LTA Constant
Organic Horizons = Included
Dry Base Cations = 0 %
Tr. M
Tr. M
J"
I
S"
a.
.
— — Tr. ISO
Tr. 20*
»
i
>;
-MuOO -«0j«0 t&
Base Saturation (X)
SBRP Stream Watersheds
Deposition = LTA Ramp 20% Increase
Organic Horizons = Included
Dry Base Cations = 0 %
|JO-
c
r
Ji
a.
1
O
"SIS -r
Tr. 100 j
Tr. MO {,
!•
If
1.
I1
M
i*
fl
i
ft
i.
I1
j
/ i
-ZUQ -IOJW .0.00
Base Saturation (%}
Figure 9-52. Regional CDFs of the projected change in the percent base saturation of soils on
SBRP stream watersheds under constant and ramp up (20% t) deposition scenarios after 20, 50,
100, and 200 years of LTA, LTA-rbc, and LTA-zbc deposition. Organic horizons are included.
9-177
-------
Table 9-30. Bloom-Grigal Model Regional Projections for the Change in Soil
pH in the Southern Blue Ridge Province. Projections Made Using LTA, LTA-rbc, and
LTA-zbc Deposition with Constant and 20% Ramped Up Deposition Scenarios at Three
Levels of Base Cations in Dry Deposition. Results Reported for 20-, 50-, 100-, and
200-Year Projections. Organic Soil Horizons Are Included
YEAR
YEAR
YEAR
Deposition = Constant ** Dry Base Cations = 100% ** LTA
MEAN STD DEV MIN P 25 MEDIAN P 75
Deposition = Constant ** Dry Base Cations = 50% ** LTA - roc
MEAN STD DEV MIN P 25 MEDIAN P 75
Deposition = Constant ** Dry Base Cations = 0% ** LTA - zbc
MEAN STD DEV MIN P 25 MEDIAN P 75
MAX
20
50
100
200
-0.07
-0.16
-0.24
•0.28
0.03
0.06
0.09
0.10
-0.14
-0.32
-0.55
-0.62
-0.10
^).21
-0.29
-0.32
-0.07
-0.16
-0.24
-0.27
-0.05
-0.11
-0.17
-0.19
-0.02
-0.05
-0.08
-0.08
MAX
20
50
100
200
-0.11
-0.25
-0.34
-0.36
0.04
0.08
0.10
0.09
-0.21
-0.48
-0.61
-0.66
-0.15
-0.32
-0.40
-0.42
-0.11
-0.24
-0.31
-0.35
-0.09
-0.20
-0.27
-0.29
-0.03'
-0.06
-0.12
-0.19
MAX
20
50
100
200
YEAR
20
50
100
200
•0.16
-0.33
-0.41
-0.43
Deposition -
MEAN
-0.08
-0.24
-0.35
-0.38
0.05
0.10
0.10
0.09
* 20%
STD
0.03
0.08
0.11
0.10
-0.29
-0.61
•0.66
-0.68
Increase ** Dry
_DEV MIN
-0.15
-0.46
-0.68
-0.71
-0.22
-0.42
-0.47
-0.50
Base Cations
P_25
-0.11
-0.30
-0.42
-0.44
-0.14
-0.32
-0.41
-0.43
= 100% **
MEDIAN
-0.08
-0.23
-0.33
-0.36
-0.13
-0.29
•0.36
-0.37
LTA
P_75
-0.06
-0.18
•0.28
-0.30
-0.03
-0.07
•0.14
-0.26
MAX
-0.02
-0.06
-0.13
-0.21
continued
9-178
-------
Table 9-30. (Continued)
Deposition = 20% Increase ** Dry Base Cations = 50% ** LTA - roc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
200
-0.13
-0.33
-0.43
-0.45
0.04
0.10
0.10
0.09
-0.23
•0.60
-0.72
-0.73
-0.16
-0.42
•0.49
-0.51
•0.12
-0.32
•0.41
•0.43
-0.10
-0.28
-0.37
-0.37
-0.03
-0.07
-0.15
-0.27
Deposition = 20% Increase ** Dry Base Cations = 0% ** LTA - zbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
200
-0.17
-0.41
-0.49
-0.50
O.Q6
0.12
0.11
0.09
-0.31
-0.71
-0.75
-0.75
-0.23
-0.51
•0.55
-0.57
•0.16
-0.41
•0.49
-0.49
-0.14
-0.37
-0.43
-0.43
-0.03
-0.08
-0.17
-0.33
9-179
-------
Table 9-31. Bloom-Grigal Model Regional Projections for the Change in Percent Soil Base
Saturation in the Southern Blue Ridge Province. Projections Made Using LTA, LTA-rbc, and LTA-
zbc Deposition with Constant and 20% Ramped Up Deposition Scenarios at Three Levels of Base
Cations in Dry Deposition. Results Reported for 20-, 50-, 100-, and 200-Year Projections. Organic
Soil Horizons Are Included
YEAR
Deposition » Constant ** Dry Base Cations = 100% ** LTA
MEAN STD DEV MIN P 25 MEDIAN P 75
MAX
20
50
100
200
YEAR
20
50
100
200
YEAR
20
50
100
200
YEAR
20
50
100
200
-1.20
-2.44
-3.50
-4.14
Deposition
MEAN
-1.75
-3.43
-4.53
-5.09
Deposition
MEAN
-2.32
-4.31
-5.28
-5.77
Deposition
MEAN
-1.32
-3.35
-*.67
-5.28
0.35
0.79
1.33
2.20
= Constant **
STDJ5EV
0.35
0.78
1.41
2.55
= Constant **
STD_DEV
0.40
0.84
1.61
2.89
-2.21
-4.67
-7.27
-11.16
Dry Base
MIN
-2.82
-5.55
-7.71
-13.94
; Dry Base
MIN
-3.41
-6.34
-9.19
-16.35
-1.30
-2.98
-3.96
-4.10
Cations
P_25
-1.99
-3.73
-4.59
-4.62
Cations
P_25
-2.59
-4.67
-5.16
-5.17
-1.23
-2.43
-3.22
-3.39
= 50% ** LTA -
MEDIAN
-1.72
^3.36
-4.11
-4.23
= 0% ** LTA -
MEDIAN
-2.33
-4.25
-4.70
-4.75
- 20% Increase ** Dry Base Cations = 100% **
STD_DEV
0.36
0.84
1.48
2.74
MIN
-2.39
-5.65
-8.19
-15.15
P_2S
-1.45
-3.75
-4.94
-4.97
MEDIAN
-1.34
-3.20
-4.29
-4.39
-0.88
-1.83
-2.69
-2.77
• rbc
P_75
-1.39
-2.71
-3.52
-3.62
zbc
P_75
-1.99
-3.65
-4.27
-4.31
LTA
P_75
-0.98
-2.69
-3.53
-3.71
-0.46
-0.68
-0.74
-0.74
MAX
-1.15
-1.54
-1.57
-1.57
MAX
-1.67
-2.02
-2.03
-2.03
MAX
-0.57
-1.5
-1.67
-1.67
continued
9-180
-------
Table 9-31. (Continued)
Deposition = 20% Increase ** Dry Base Cations = 50% ** LTA-rbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
200
-1.89
-4.27
-5.40
•5.93
0.36
0.84
1.65
3.08
-2.99
•6.45
-9.24
-17.57
-2.13
-4.64
-5.29
-5.30
-1.87
-4.16
-4.78
-4.91
-1.52
-3.59
-4.32
-4.37
-1.28
-2.08
-2.10
-2.10
Deposition = 20% Increase ** Dry Base Cations = 0% ** LTA - zbc
YEAR MEAN STD DEV MIN P 25 MEDIAN P 75 MAX
20
50
100
200
-2.46
-5.00
-5.91
-6.41
0.42
0.92
1.88
3.43
-3.57
-7.01
-10.86
-19.96
-2.74
-5.21
-5.78
-5.78
-2.48
•4.97
-5.28
-5.31
-2.12
-4.34
-4.84
-4.85
•1.80
-2.41
-2.42
-2.42
9-181
-------
These results are explained by the initial conditions: the greater the pH and the lower the base
saturation, the faster the base cation depletion rate (see Section 9.3.3.2.2.2).
This convergence of the CDFs may represent the limit of change, at least for the next two centuries.
In order to consider the limit of change, assume that the median value of change represents the central
tendency for change. The limit for change in soil pH, therefore, is approximately -0.40 and for base
saturation is about -4.75. The results of subtracting these values from the current median values suggest
that the new median pH value will be about 4.5 and for percent base saturation about 4.5. Both of these
values are quite low considering that they represent aggregate values-i.e., the weighted average of all
soil horizons.
9.3.3.2.3 Regional comparisons
Soils in the NE currently are somewhat lower in pH than soils in the SBRP. Soils in the SBRP,
however, have much lower percent base saturation. These two differences lead to very different
projections of the change in soil pH and percent base saturation with the Bloom-Grigal model. The
median estimates of total effective acidity (H*tota, = H+ + NH4+ - N03") inputs in LTA deposition
datasets for the NE and SBRP are similar (see Table 9-24). The output from the simulations using these
two datasets, therefore, can be compared (Table 9-32).
The regional response of soils to acidic deposition (changes in soil pH and percent base saturation)
differ. Because the soils in the SBRP are older and more extensively weathered, their initial percent base
saturation is markedly lower than that of the younger, less weathered soils in the NE. Aggregate soil
pH values for the SBRP are, at the same time, slightly higher, which may be due to lower organic matter
content. These two conditions, moderate to high pH (high for forested soils) and low percent base
saturation result in rapid and severe projected decreases in soil pH and percent base saturation in soils
that are already low in base cations. Only minor changes in soil pH and percent base saturation are
projected for the NE.
9-182
-------
Table 9-32. Summary of the Bloom-Grigal Projected Changes in Soil pH and
Percent Base Saturation in the NE and SBRP Under Constant LTA Deposition.
Change in Parameter
After Selected Years
Region Parameter Initial
Value 20 50 100
NE pH 4.62 -0.01 -0.03 -0.04
SBRP pH 4.85 -0.07 -0.16 -0.24
NE % B.S. 21.98 -0.57 -1.05 -1.46
SBRP % B.S. 9.22 -1.23 -2.43 -3.22
9-183
-------
A series of buffer ranges proposed by Ulrich (1983), assist, in part, with the interpretation of these
results. He suggested that soil-water pH is indicative of the mineral phases that buffer the soil. He
proposed five distinct buffer ranges:
(1) Calcium carbonate (pH > 6.2)
(2) Silicate (pH 6.2 - 5.0)
(3) Cation exchange (pH 5.0 - 4.2)
(4) Aluminum (pH 4.2 - 2.8)
(5) Iron (pH 3.8 - 2.4)
In the NE, the soils are generally in the Al buffer range (as defined by Ulrich), which is consistent with
the model predictions. In the SBRP, soils are principally in Ulrich's cation exchange buffer range. When
the pool of exchangeable base cations is depleted, the cation exchange buffer is exhausted, and the
buffering of the system becomes controlled by Al. In Section 9.3.3.2.2.2, the convergence of the CDFs
were suggested as bounding the change in soil pH and percent base saturation, and pH 4.5 was
proposed as the limit. The apparent buffering of the NE against changes in soil pH (despite significant
acid inputs) suggests that soil pH values near 4.5 are likely to be in the Al buffer range rather than the
cation buffer range. The Al buffer range should be extended from pH 4.2 - 2.8 (as suggested by Ulrich)
to a range of pH 4.5 - 2.8.
9.3.3.2.4 Summary and conclusions
Based on model projections, the soils in the NE appear to be buffered against changes in soil pH
and percent base saturation by an Al buffering mechanism. Soils in the SBRP may experience significant
decreases in soil pH and percent base saturation because of their current status and the level of acid
inputs. While currently buffered against changes in pH via cation exchange buffering, the effectiveness
of this buffer will be exceeded with the current levels of acid input. pH of these soils is projected to
decrease until changes in soil pH become controlled by the Al buffering system.
9-184
-------
Such changes are likely to affect surface waters. While AJ buffering prevents the occurrence of
even lower soil pH values, AI3+ and other acid cations (e.g., Mn2"1" and Fe3"1") will become the dominant
cations in the soil. These elements are toxic to plants and soil microbes and are also potentially toxic
in the aquatic environment.
These projections may represent the worst case estimates of the effects of acidic deposition on
soils of the NE and SBRP. Several key points should, however, be reiterated: (1) these projections
were made in the absence of mineral weathering and biomass accretion; (2) sulfate was treated as a
completely mobile anion; (3) the projected changes are sensitive to the relationships between soil pH and
percent base saturation, and these relationships were empirically derived for a selected subset of soils
outside the DDRP regions; and (4) many types of soils were aggregated to derive a single value for initial
soil pH, cation exchange capacity, and the sum of base cations.
The major conclusions of this analysis using the Bloom-Grigal model are:
« Organic horizons contribute sufficient base cations to increase the size of the base cation
pool, which slows the rate of acidification.
* In the NE, organic horizons contribute acidity and base cations, which results in lower cation
leaching rates.
* Soils in the NE are buffered against changes in soil pH and percent base saturation via an
Ai buffering mechanism.
• Soils in the SBRP may experience significant decreases in soil pH and percent base
saturation. The median soil pH could decrease as much as 0.5 pH units, and the median
percent base saturation may decrease from its current level of 9.2 percent to 4.5 percent.
The are thought to be worst case estimates because sulfate is considered to be a mobile
anion in this analysis. The extent of change in the SBRP soils will be limited by the AI buffer
range.
• The soil pH buffer ranges by Ulrich (1983) provide a good basis for interpreting the model-
based projections.
9.3.4 Comparison of the Bloom-Griaal and Reuss Model Projections
Results from two soil cation exchange models have been presented in detail. The behavior
modelled by the two formulations is remarkably different in some respects and more comparable in
9-185
-------
others. A summary of the median, mean, and maximum changes in percent base saturation and soil
pH in the NE is presented in Table 9-33 for 50- and 100-year projections. Specific comparisons between
the models can be made at two levels. First, model results can be compared for the entire population
of lakes in the NE or stream reaches in the SBRP. Because the primary purpose of the DORP Is to
obtain such regional estimates, this comparison is of particular importance. On a more detailed level,
model results can be compared for individual lakes or stream reaches. While a high degree of
correspondence between the model outputs for individual systems should be expected, the comparison
at this level may help to increase our understanding of the behavior of the individual models.
With respect to the estimates for the changes expected at the population level, several observations
are of interest here. The dynamics of the two models are quite different. Initially, the Bloom-Griga! model
projects substantially larger changes in percent base saturations than does the Reuss model (Table 9-
33 and Figure 9-53). Both the mean and median values for changes in percent base saturation projected
using the Bloom-Grigal model are larger at 20 and 50 years than those projected using the Reuss
formulation. At 100 years, however, the relative magnitude of the changes projected by the two models
is reversed. At 100 years, both the mean and median changes projected by the Reuss model are larger
than those projected by the Bloom-Grigal model. Overall, the CDFs for the projected changes in soil
base saturation for systems in the NE using the two models are reasonably similar.
In contrast to model behavior for percent base saturation, the models project quite different
distributions for the response of soil pH to acidic deposition. Results from the Reuss model suggest that
the rate of change in soil pH increases over the course of the 100-year simulation. The Bloom-Grigal
model results, on the other hand, suggest that any changes in soil pH over this period are generally
linear. Another major difference between the two models is that, with the Reuss model, a small number
of systems experience extreme changes in soil pH during the simulation period. The Bloom-Grigal model
results, in contrast, suggest that extremes should not be observed. The effect of the longer tail
9-186
-------
Table 9-33. Comparison of the Results from the Reuss and Bloom-Grigal Models
with Regard to the Magnitude of Changes in Soil pH and Base Saturation Projected
in Soils of the NE. Results Are Shown for 50 and 100 Years
Median
Mean
s.d.a Maximum
A % BS (20 years)
" Reuss -1.3 -1.4 0.9 -6.0
Bloom-GrigaJ -2.2 -2.4 1.6 -6.7
A % BS (50 years)
~ Reuss -3.4 -3.5 1.7 -11.0
Bloom-Grigal -4.4 -4.8 3.1 -14.3
A % BS (100 years)
" Reuss -7.5 -7.6 3.2 -17.0
Bloom-GrigaJ -7.1 -6.9 4.5 -23.2
A soil pH (20 years)
~ Reuss -0.03 -0.08 0.14 -0.68
Bloom-Grigal -0.05 -0.08 0.07 0.33
A Son pH (50 years)
~ Reuss -0.09 -0.17 0.19 -0.88
Bloom-Grigal -0.12 -0.16 0.12 -0.45
A Soil pH (100 years)
" Reuss -0.27 -0.36 0.28 -1.10
Bloom-Grigal -0.21 0.22 0.15 -0.55
Standard deviations are reported as absolute values.
9-187
-------
NE Lakes
Deposition = LTA
Year = 50
Model - Reuss
o"
a
o
•0.6 -
o
30.2-1
£
o
0.0
-50.00
95 r. Conf. Umil
Predicted Oiilrlbulion
5 7. Can), limit
-10.00 -10,00
A Base Saturation (%)
o.oo
NE Lokes
Deposition = LTA
Year = 100
Model = Reuss
c i
o
o
a.
o
0.8
o
3 0.2-|
E
\ 0.0 •
-Jooo
95 X Conf. limit
Pradlctod Olilrlbullon
5 X Conl. Limit
B
-20.00 -10.00
Base Saturation (%)
0.00
NC Lake Watersheds
Deposition = LTA Constant
Organic Horizons = Excluded
Dry Base Cottons = 100 %
OH.O
r
o
Q.
O0.6
£0.4
O
-30.90
It. 20
- - Tr. JO
. Tr. 100
-ZO.OO -10.00 0.00
Base Saturation (%)
Rgure 9-53. Cumulative distributions of changes in soil base saturation for the population of
watersheds In the NE: (A) illustrates changes projected by the Reuss model at SO years; (B)
Indicates those changes projected after 100 years, again using the Reuss model; and (C)
shows the results at 20, 50, and 100 years, as projected using the Bloom-Grigal formulation.
9-188
-------
on the Reuss model population distributions is an increase in the mean projected change in soil pH over
a 100-year period. Although the population medians, as projected by both models, are more similar than
the medians, they still differ significantly as illustrated for the NE (Figure 9-54).
Comparison of the results for individual systems in the NE supports and reinforces the information
obtained from the population-level evaluations. Figure 9-55 shows a scatter plot of the changes in
percent base saturation projected for individual systems by the two models at 50- and 100-year intervals.
Surprisingly, no correlation between the two model outputs is evident. Clearly, the two approaches
used to model cation loss from the soil exchange complex differ. Nevertheless, when integrated over
the population of systems in the NE, the differences between the models become sufficiently small to
yield similar population estimates.
A greater degree of correlation appears to exist between the models for soil pH projections (Figure
9-56). At 50 years, the Reuss model appears to project smaller changes in soli pH for most of the
systems. However, the Reuss model projects some extreme changes for small number of systems,
relative to the magnitude of changes projected by the Bloom-Grigal model. After an additional 50 years,
the Reuss model projections have increased in magnitude relative to those made with the Bloom-Grigal
model. The number of systems projected to have extreme changes in soil pH also increases.
The patterns for individual systems and for populations observed in the NE also are observed in
the SBRP. Because the simulations were extended to 200 years in the SBRP, however, some of the
differences are more pronounced. Table 9-34 summarizes the results for changes in soil pH and percent
base saturation obtained with the two models for this region. In the SBRP, the Bloom-Grigal model
initially projects larger changes in both soil pH and percent base saturation than does the Reuss model.
As the simulations progress, however, the changes projected by the Reuss model increase more rapidly,
so that by 200 years, substantially larger changes for both percent base saturation and soil pH are
projected. Figures 9-57 and 9-58 illustrate the changes projected for the population of soils in the SBRP
at 50 and 100 years.
9-189
-------
NE Lakes
Deposition = LTA
Year = 50
Model ~ Reuss
95 Z Con(. Limit
Predicted Distribution
5 % Conl. Limit
-0.75
-0.50
A pH
-0.2S
NE Lakes
Deposition = LTA
Year = 100
Mode! - Reuss
ct.o-|
Q.
O
-0.6
o
30.2-
E
3
0.0 •
-0.7S
95 X Conl. Limit
Predicted Distribution
S X Con(. Limit
B
-0.50
A pH
-0.25
0.00
NE Lakes Watersheds
Deposition = LTA Constant
Organic Horizons = Excluded
Dry Base Cations = 100 %
1.0
Jo..
o
a.
00.6
out
Yr. »
Tr. SO
Tr. tOO
-0.50 -OJ5
A Soil pH
Rgure 9-54. Cumulative distributions of changes In soil pH for the population of watersheds
In the NE: (A) illustrates changes projected by the Reuss model at 50 years; (B) indicates those
changes projected after 100 years, again using the Reuss model; and (C) shows the results at
20, 50, and 100 years, as projected using the Bloom-Grigal formulation.
9-190
-------
_ -5
at
0X0
o £
£•0
03
-10
-15
..-« *
dd%BS
Reuss Model
-6
cd
ga -1*1
1*
-18
m
-24
B
-17 -15 -13 -11
-7 -5
-1
del%BS
Reuss Model
Rgure 9-55. Scatter diagrams of the projected changes in base saturation for individual
systems (not population weighted) In the NE obtained from the Reuss and BIoonvGrfgal
models: (A) compares results from the two models after a 50-year simulation and (B) illustrates
the relationship observed after 100 years.
9-191
-------
0.0
-0.1
I'
-0.3
eg
o g
•«.4<
-0.6
•°'4
del soil pH
Reuss model
O.o
O.o
-0.2
-0.4
-0.6
0 1
«• •
lOOYoan .
n
m
. 116
- 0222
b - 4.157
1* - 0.191
B
-1.0
-0.8
-0.6
-0.4
•0.2
0.0
del soil pH
Reuss model
Figure 9-56. Scatter diagrams of the projected changes in soil pH for individual systems (not
population weighted) in the NE obtained from the Reuss and Bloom-Grigal models:
(A)compares results from the two models after a 50-year simulation and (B) illustrates the
relationship observed after too years.
9-192
-------
Table 9*34. Comparison of the Results from the Reuss and Bloom-Grigal Models
with Regard to the Magnitude of Changes in Soil pH and Base Saturation Projected
in Soils of the SBRP. Results Are Shown for 50 and 100 Years
Median Mean s.d. Maximum
A % BS (50 years)
~ Reuss -1.9 -1.9 0.4 -2.8
Bloom-Grigal -3.1 -3.1 0.8 -5.1
A % BS (100 years)
~ Reuss -5.1 -5.2 0.8 -7.2
Bloom-Grigal -3.9 -4.4 1.3 -7.9
A % BS (200 years)
~ Reuss -8.8 -8.8 1.0 -12.4
Bloom-Grigal -4.4 -5.1 2.2 -11.7
A Soil pH (50 years)
~ Reuss -0.10 -0.10 0.03 -0.19
Bloom-Grigal 0.23 -0.23 0.07 -0.40
A Soil pH (100 years)
" Reuss -0.35 -0.34 0.09 -0.49
Bloom-Grigal -0.35 -0.35 0.10 -0.67
A Soil pH (200 years)
~ Reuss -0.67 -0.66 0.15 -0.82
Bloom-Grigal -0.38 -0.39 0.10 -0.75
9-193
-------
SBRP Stream Reaches
Deposition = LTA
Year = 50
Model = Reuss
ci.o-
gO.8 -
Q.
O
£06-
J0.4-
"o
30.2-
E
95 X Cool. UmH \"J.
Pradictad Oiilribulion :/ .'
S S Cenl. limit }.•
;
;'
7
L
Ou.u
-SO 00
-20.00 -10.00
A Base Saturation (%)
SBRP Stream Reaches
Deposition = LTA
Year = 100
Model = Reuss
c'-o
Q.
0°f,
SS X Canf. LlmH
Pr«tle<«d Distribution
$ X Conf. Umll
MOO
SBRP Stream Watersheds
Deposition = LTA Constant
Organic Horizons = Excluded
Dry Base Cations m 100 %
|I
O
o.
o.
Cumulatlv
Z t
-M
- - Tr! M (
:~RJS I;
|,
li
j;
jj
/; '
,s^\ '•
IM -acLeo -iieo «jio
& Base Saturation (%)
B
-20.00 -10.00
A Base Saturation (%)
o.oo
Figure 9-57. Cumulative distributions of changes in soil base saturation for the population of
watersheds in the SBRP: (A) illustrates changes projected by the Reuss model at SO years; (B)
indicates those changes projected after 100 years, again using the Reuss model; and (C) shows
the results at 20, 50, and 100 years, as projected using the Bloom-Grigai formulation.
9-194
-------
SBRP Stream Reaches
Deposition = LTA
Year = 50
Model = Reuss
O°-B-I
Q.
O
-0.6H
30.2-|
E
75
O
,0.0
95 7. Cent. Urn!!
Predicted Distribution
5 Z Cent. Limit
-0.75
-0.50
pH
-0.25
SBRP Stream Reaches
Deposition = LTA
Year = 100
Model = Reuss
ci.o
o
0-8
E
3«
95 7. Conf. limit
Predicted Distribution
5 X Conf. Limit
B
-0.75
-0.50
A pH
-0.25
0.00
SBRP Stream Watersheds
Deposition = LTA Constant
Organic Horizons = Excluded
Dry Base Cations = 100 %
go..
r
o
a.
oo.«
o
0.0
A Soil pK
(LOO
Figure 9-58. Cumulative distributions of changes in soil pH for the population of watersheds In
the SBRP: (A) illustrates changes projected by the Reuss model at 50 years; (B) Indicates those
changes projected after 100 years, again using the Reuss model; and
-------
In summary, soil cation exchange models were used to explore possible changes in soil chemical
properties potentially occurring as a result of acidic deposition. Overall, the two models yield similar
results with regard to projected changes for the NE and the SBRP. The major differences between the
models appear to be that the Bloom-Grigal model projects more rapid initial changes to the soil chemical
environment, whereas results obtained using the Reuss model indicate that changes should occur more
rapidly as the soil exchange pool becomes depleted. Information needed for more critical evaluation
of the two models currently is not available.
9.3.5 Summary and Conclusions
Results from two soil cation exchange models have been presented. These models focus on the
role of cation exchange processes in regulating pH and percent base saturation in soils. The models do
not consider processes such as primary mineral weathering, uptake of cations by vegetation, suifate
dynamics, or detailed hydrologic flow regimes; nor do they address the deep regolith (e.g., soil depths
> ~2 m). Consequently, these model results are not directly comparable to the integrated watershed
process models presented in Section 10. The models do, however, provide considerable information
concerning how base cation pools may respond to continued acidic deposition.
The two models provide slightly different types of information about the soils and their associated
surface waters. The Reuss model projects changes in both surface water chemistry and soil chemistry.
In contrast, the Bloom-Grigal model provides information about the magnitude of projected changes only
in soil chemical properties. The models employ markedly different algorithms in making these projections.
The Reuss formulation uses a mass action approach. This approach allows each of the soil reactions
to proceed independently, while simultaneously allowing individual soil properties to vary in an internally
consistent manner. The Bloom-Grigal model relies on empirically-derived relationships to define time-
varying behavior of individual soil parameters. Each approach has certain advantages, making it
important to determine how the two models differ in projected changes to the population of systems in
the NE and SBRP regions.
9-196
-------
While these models do not explicitly consider many processes, it is possible to understand
qualitatively how non-modelled processes would affect the projections presented here. For example,
cation accretion in biomass is a net base cation sink, and thus has an acidifying effect on the soils.
Conversely, mineral weathering is a net source for base cations. Incorporation of a weathering term in
these models would delay the projected response times of individual systems. Unfortunately, regionally
based estimates of the magnitude of these processes is unavailable. Despite these limitations, model
results do provide information to possible watershed responses. For systems with long projected
response times, future changes in the quality of surface waters likely will not be large. However, for
those systems with short projected response times, additional information about the magnitude of other
potential sources or sinks for base cations is essential for describing the responses of these systems
accurately.
Detailed results from the models have been presented in Section 9.3.3. Major findings, first for
surface waters and then for soils, are summarized below for both the NE and the SBRP.
For lakes in the NE currently exhibiting ANC values in excess of 100 peq L"1, mineral
weathering is probably the dominant watershed process controlling observed ANC values.
At present levels of deposition, lakes in the NE with ANC values in excess of 100 peq L~1
will probably not experience declining ANCs in the foreseeable future.
For stream reaches in the SBRP projected to exhibit ANC values in excess of 50 ^eq L"1
(after having attained a state of net zero sulfate retention), mineral weathering will probably
be the dominant watershed process controlling ANC values for systems with chemistry
currently dominated by sulfur dynamics.
Stream reaches in the SBRP with projected ANC values in excess of 50 peq L'1 (after having
attained a state of net zero sulfate retention) will probably not become acidic (ANC <. 0 peq
L'1) at current or slightly elevated levels of deposition. The capacity of weathering processes
to mitigate the effects of acidic deposition could be overwhelmed in those systems with
marginal (ANC <_ 100 peq L ) contributions from weathering, substantial increases in the
levels of acidic deposition were to occur.
For lakes in the NE exhibiting ANC values of less than 100 /ieq L"1, soil exchange processes
may be regulating the observed ANCs, although in most systems, the ANC is probably
controlled by a combination of cation exchange and mineral weathering.
9-197
-------
• As an upper limit, approximately 15 percent or over 1000 lakes (four times the number of
currently acidic lakes) in the NE with current positive ANC values could become acidic (i.e.,
ANC <_ 0 /teq L ) within 50 to 100 years. The projection is extreme, because the
contribution of weathering is not considered. However, some fraction of this number of lakes
will probably become acidic during the next several decades.
• In the SBRP, changes in observed ANC values that occur because of changes in the base
status of soils during the next century should be minimal.
• For systems in the NE and SBRP that have ANC values in the range of 0 to 50 peq L*1,
rates of system response are projected to Increase with continued exposure to acidic
deposition. The increased rates coincide with the depletion of soil buffering capacity.
• In the absence of mineral weathering, significant depletion of base cations is projected to
occur in the soils of both the NE and SBRP regions.
• The absolute magnitude of base cation depletion is greater in the NE than It is in the SBRP.
The relative projected changes are, however, greater in the SBRP.
• Current percent base saturation of soils in the regions can be used as indicators of potential
future change in surface water ANC. Soils with base saturation currently in excess of about
20 percent appear to undergo minimal changes on the time scale of the next 100 years. For
soils with base saturation less than 20 percent, however, projected changes in surface water
ANC appear to increase with decreasing aggregate percent base saturation, an effect that
is more pronounced in the NE than in the SBRP.
• Current percent base saturation can be used as an Indicator of the anticipated relative
changes that might occur in the soil base status over the next 100 years. The relative
percentage decline in percent base saturation [(current - projected)/current] x 100 increases
with decreasing percent base saturation, although other factors, such as soil thickness or bulk
density, probably also influence the relationship.
9-198
------- |