EPA/620/R-94/015
March 1994
ENVIRONMENTAL MONITORING AND ASSESSMENT PROGRAM
ARID ECOSYSTEMS 1992 PILOT REPORT
by
W. G. Kepner, R. O. Kuehl, R. P. Breckenridge, J. M. Lancaster, S. G. Leonard,
D. A. Mouat, T. G. Reinsch, K. B. Jones, A. C. Neale, T. B. Minor, and N. Tallent-Halsell
Environmental Monitoring Systems Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Las Vegas, Nevada 89193-3478
Printed on Recycled Paper
-------
ENVIRONMENTAL MONITORING AND ASSESSMENT PROGRAM
Arid Ecosystems Resource Group
William G. Kepner, Technical Director
Arid Ecosystems 1992 Pilot Report
NOTICE
The U.S. Environmental Protection Agency (EPA), through its Office of Research and
Development (ORD), funded and collaborated in the research described here. It has been peer
reviewed by the Agency and approved as an EPA publication. Mention of trade names or
commercial products does not constitute endorsement or recommendation for use.
Proper citation of this document is:
Kepner, W. G., R. 0. Kuehl, R. P. Breckenridge, J. M. Lancaster, S. G. Leonard,
D. A. Mouat, T. G. Reinsch, K. B. Jones, A. C. Neale, T. B. Minor, and N. Tallent-Halsell.
1994. Environmental Monitoring and Assessment Program: Arid Ecosystems 1992 Pilot
Report. EPA/620/R-94/015. U.S. Environmental Protection Agency, Washington, D.C.
-------
ACKNOWLEDGMENTS
The preparation of this plan has been a combined effort requiring the contributions of a
number of scientists from various universities, research institutes, public interest groups, and
federal agencies. Contributors are listed below. This manuscript has benefitted from the review
and comments of Michael M. Borman, U.S. Department of the Interior, National Biological Survey;
Charles F. Hutchinson, Office of Arid Lands Studies, University of Arizona; Richard Inouye, Idaho
State University; Wesley M. Jarrell, Oregon Graduate Institute of Science and Technology; and
Walter G. Whitford, U.S. Environmental Protection Agency, Environmental Monitoring Systems
Laboratory, Las Vegas; Brian A. Schumacher, U.S. Environmental Protection Agency, Environmental
Monitoring Systems Laboratory, Las Vegas, and Gregory E. Huber, U.S. Soil Conservation Service.
Appreciation goes to John R. Baker, Deirdre O'Leary, Rod L. Slagle, and Donna W. Sutton,
Lockheed Environmental Systems & Technologies Co. for their contributions and technical editing;
and Jan Aoyama and Pat Craig (Lockheed Environmental Systems & Technologies Co.) for
document processing and production.
CONTRIBUTORS
Section 1: William G. Kepner1 and Robert P. Breckenridge3.
Section 2: William G. Kepner1 and Robert P. Breckenridge3.
Section 3: Robert P. Breckenridge3 Judith M. Lancaster4, Stephen G. Leonard5, David A. Mouat5,
and Thomas G. Reinsch6.
Section 4: Robert O. Kuehl2.
Section 5: Stephen G. Leonard5, Judith M. Lancaster4, David A. Mouat5, and Thomas G. Reinsch6.
Section 6: Timothy B. Minor4.
Section 7: Anne C. Neale1 and Nita Tallent-Halsell5.
Section 8: William B. Kepner1, Robert P. Breckenridge3, and K. Bruce Jones1.
1 U.S. Environmental Monitoring Systems Laboratory, Las Vegas, Nevada.
2 College of Agriculture, University of Arizona, Tucson, Arizona.
3 EG&G, Idaho National Engineering Laboratory, Idaho Falls, Idaho.
4 Desert Research Institute, University of Nevada, Reno, Nevada.
U.S. Bureau of Land Management, National Soil and Range Team, Reno, Nevada.
6 U.S. Soil Conservation Service, National Soil Survey Center, Lincoln, Nebraska.
HI
-------
ABSTRACT
The U.S. Environmental Protection Agency and its collaborators have initiated a long-term,
policy-relevant research project, the Environmental Monitoring and Assessment Program, focused
on evaluating ecological conditions on regional and national scales. In 1992 the Arid Ecosystems
Resource Group (one of a number of EMAP ecosystem monitoring and research groups) conducted
a pilot study in the southeastern Utah portion of the Colorado Plateau. This report describes this
first field activity for arid ecosystems, an element of the Environmental Monitoring and Assessment
Program. The 1992 pilot study was developed to evaluate sampling plot design and the sensitivity
of selected indicators. The study focused on four objectives related to plot design, indicator
development, sampling frame material, quality assurance, information management, and logistics.
The primary categories of indicators selected for evaluation in the 1992 pilot study were vegetation
composition, structure, and abundance; spectral reflectance; soil properties; and soil erosion. Data
were collected on 29 sites within two major resource classesdesertscrub and conifer woodland.
This report describes the indicator measurement methods and the study results for each of the four
objectives. Each of these sections includes recommendations based on the 1992 study. The final
section summarizes the major conclusions and recommendations drawn from the 1992 pilot study,
draws implications from the study results, and discusses planned future studies.
IV
-------
TABLE OF CONTENTS
Notice ii
Acknowledgments Hi
Abstract jv
Figures vii
Tables jx
Abbreviations and Acronyms x
Section Page
1 INTRODUCTION 1
1.1 1992 Pilot Study 1
1.2 Report Overview 3
2 COLORADO PLATEAU PILOT STUDY 6
2.1 Study Objectives 6
2.2 1992 Pilot Study Area , 6
2.3 Site Selection . . 8
2.4 Sample Plot Design . . 8
3 INDICATOR MEASUREMENT METHODS 11
3.1 Vegetation Measurements 12
3.2 Remote and Ground-Based Spectral Measurements . 14
3.3 Soil Measurements 17
3.3.1 Soil Properties 18
3.3.2 Erosion Index Indicator Measurements 19
4 ASSESSMENT OF SAMPLING VARIANCE . 21
4.1 Analysis and Results .24
4.1.1 Vegetation Measurements 25
4.1.2 Ground Spectral Measurements 26
4.1.3 Soils Measurements 32
4.1.4 Variance Component Estimates 32
4.2 Recommendations 36
5 INDICATOR SENSITIVITY 39
5.1 Vegetation Indicator 40
5.1.1 Analysis and Results 40
5.1.2 Recommendations 45
-------
TABLE OF CONTENTS (CONTINUED)
Section
Page
5.2 Spectral Properties Indicator 45
5.2.1 Results 47
5.2.2 Recommendations 57
5.3 Soils Indicator 57
5.3.1 ResultsSoil Profile Description 57
5.3.2 Results-Soil Quality 59
5.3.3 ResultsErosion Index Measurements 70
5.3.4 Recommendations 72
6 FRAME MATERIALS '.73
6.1 Results 73
6.2 Recommendations 76
7 QUALITY ASSURANCE, INFORMATION MANAGEMENT, AND LOGISTICS 77
7.1 Quality Assurance 77
7.1.1 QA Program Overview 77
7.1.2 Results 79
7.1.3 Recommendations 90
7.2 Information Management 91
7.2.1 Results .92
7.2.2 Recommendations 93
7.3 Logistics 93
7.3.1 Results 93
7.3.2 Recommendations 94
8 CONCLUSIONS, RECOMMENDATIONS, AND ADDITIONAL STUDIES . 96
REFERENCES 99
APPENDIX
A Soil Classification 104
B Soil Description 113
VI
-------
FIGURES
Figure
1-1
1-2
2-1
2-2
3-1
4-1
4-2
4-3
4-4
4-5
4-6
4-6
5-1
5-2
5-3
5-4
5-5
5-6
5-7
5-8
5-9
5-10
.5-11
5-12
5-13
Page
Biogeographic provinces of North America used in the EMAP-Arid Program 2
The elements of designing and implementing a monitoring program 4
Location of the 1992 pilot study area within an outline of the Colorado Plateau. . 7
Diagram of indicator sampling areas at EMAP-Arid 1992 Pilot Study sites 9
Typical reflectance curves for vegetation, soil, and water with Landsat TM
wavebands (numbered). Band 6 (10.5 to 12.5 fim not shown). (After Lillesand
and Kiefer, 1987.) 15
Theoretical effects of extraneous variability on indicator response illustrating
how high extraneous variability can inhibit the utility of a condition indicator
that has some response to yearly transient climatic fluctuations 23
Analysis of variance residual plots for shrub cover and total vascular plant
cover using quadrat measures or transect means 27
Box plot of tree cover analysis of variance residuals by site 28
Histogram of NDVI measurements from circular plots and transects 30
Analysis of variance residual plots for NDVI measurements on circular plots,
transects, and for means of circular plots and transects 31
Analysis of variance residual plots for soils measurements (Page 1 of 2) 33
Analysis of variance residual plots for soils measurements (Page 2 of 2) 34
Total number of plant species versus total vascular plant cover at a site 41
Exotic plant species cover versus total vascular plant cover 42
Number of poisonous plants versus total number of plants at a site 43
Palatable plant species cover versus total vascular plant cover , 44
Total vascular plant cover versus the ratio of percent sand and percent clay 46
Correlation between NDVI derived from Landsat TM data and that from PSII data. 52
Relationship between NDVI PSII and total vascular plant cover 53
Relationship between NDVI TM and total vascular plant cover 54
NDVI TM versus the sum of interspace gravel, cobble, stone, and bare
soil cover 55
NDVI TM versus the sum of interspace moss, lichen, and cyanobacteria cover. . . 56
Clay-difference between SIR estimate and amount measured at site 63
SIR organic carbon versus measured OC at site 64
No significant difference between SIR and measured values for cation exchange
capacity 65
VII
-------
FIGURES
Figure Page
5-14 Difference between SIR and measured values for electrical conductivity at
1992 pilot study sites .66
5-15 Difference between SIR and measured values for sodium adsorption ratio at
1992 pilot study sites 67
5-16 Difference between SIR and bulk measured values for bulk density at
1992 pilot study sites 69
VIII
-------
TABLES
Tables
3-1
3-2
3-3
3-4
3-5
4-1
4-2
4-3
4-4
5-1
5-2
5-3
5-4
5-5
5-6
5-7
5-8
6-1
7-1
7-2
7-3
7-4
7-5
7-6
7-7
7-8
Page
Modified Daubenmire Cpver Classes 13
Information Recorded within Quadrats and Circular Subplots . . . 13
Principal Applications of Landsat TM Spectral Bands 16
Methods for Soil Sample Analyses in the EMAP-Arid Program 19
Measured and Estimated Data Needed to Calculate K and C Factors 20
Sites Selected and Sampled in the 1992 Colorado Plateau Pilot Study 22
Variables, Measuring Units, and Measures Used for Variance Component
Estimates , 35
Analysis of Variance for Measures on Subplots at Each of
Several Sites 35
Subplot Variance Component Estimates for the Vegetation, Spectral,
and Soils Measures 37
The Number of Spectral Measurements Made at Each Site 48
Dominant Vegetation and Soil Conditions at Sites From Which
Spectral Measurements Were Obtained During EMAP-Arid 1992 Pilot 49
Total Vascular Plant Cover at Each Site 50
NDVI Values Derived From Landsat TM Data for
EMAP-Arid 1992 Pilot Study 50
NDVI Values Derived From Personal Spectrometer II Data
Collected During EMAP-Arid 1992 Pilot Study 51
Soil Surfaces Analyses gO
Quartiles For Soil Surface Analyses3 61
SIR3 and RUSLE3 Factors and Annual Average Erosion Rates 71
Classification Error Matrix of the Preliminary GAP Data . 74
Measurement Quality Objectives for Vegetation 79
Measurement Quality Objectives for Soil Analyses 80
Cover Class Comparability 81
EMAP-Arid Crew Comparability for Tree Measurements 83
Between-Crew Comparisons for Field Measurements 86
Summary Statistics for Reference Samples 88
Summarized Results of Laboratory Replicate Analyses 89
Summary of Results of Field Duplicate Analyses . 89
IX
-------
ABBREVIATIONS AND ACRONYMS
AVHRR
CEC
CV
DQO
EC
EMAP
EMAP-Arid
EPA
GAP
GIS
GPS
IM
INEL
MLRA
MQO
NDVI
OC
PSII
QA
RUSLE
SAR
SCS
SIR
TM
USDA
USFWS
USLE
WEPP
WRD
advanced very high-resolution radiometer
cation exchange capacity
coefficient of variation
data quality objective
electrical conductivity
Environmental Monitoring and Assessment Program
Environmental Monitoring and Assessment Program Arid Ecosystems
resource monitoring and research group
U.S. Environmental Protection Agency
U.S. Fish & Wildlife Service Gap Analysis Pyrogram
geographic information system
global positioning system
information management
Idaho National Engineering Laboratory
major land resource areas
measurement-level quality objective
normalized difference vegetative index
organic carbon
personal spectrometer II
quality assurance
revised universal soil loss equation
sodium adsorption ratio
U.S. Soil Conservation Service
soil interpretation record
Landsat thematic mapper
U.S. Department of Agriculture
U.S. Fish and Wildlife Service
universal soil loss equation
water erosion prediction project
water retention difference
-------
SECTION 1
INTRODUCTION
The primary goal of the Environmental Monitoring and Assessment Program (EMAP) is to
monitor status and trends in the Nation's ecological resources (Thornton et al., 1993). It is the
intent and purpose of the EMAP Arid Ecosystems Resource Group (EMAP-Arid) to measure and
report on the extent, condition, and trends of several resource classes in the biogeographical
provinces of nearctic and neotropical North America within the United States (Brown et al., 1979;
Figure 1-1). Two resource classes, desertscrub and desert conifer woodlands, are representative of
resources in the EMAP-Arid program. These resources were studied in a 1992 pilot study
conducted in the southeastern Utah portion of the Colorado Plateau. This pilot study was the first
EMAP field study in arid ecosystems.
Regionally important issues in the Colorado Plateau and in other arid ecosystems (Kepner et
al., 1991) have been identified as biodiversity, livestock grazing, desertification, water resource
management, air quality, and global climatic change. Three societal values related to these issues
are currently identified as significant to arid ecosystems and have served to focus the conceptual
development of the monitoring and research strategy for EMAP-Arid, especially relative to framing
assessment questions and the selection and use of indicators. These values are:
Biological integrityspecies composition and structure (abundance and spatial arrangement)
and their associated functions (ecological processes) at various levels of biological
organization (i.e., genetic, species, population, community, ecosystem, and landscape).
Aestheticsbroadly defined as attributes that affect human perception and appreciation of
the environment.
Productivitythe quantity and quality of products or ecological services provided by arid
resources and their capacity for long-term maintenance.
Several assessment questions related to these values were developed by arid ecosystem
researchers. These questions guided the development of the 1992 pilot study and provided overall
direction. One such question was, "What proportion of desertscrub and conifer woodlands in the
Colorado Plateau with unacceptable species (e.g., presence of exotic or unpalatable species) and/or
have soils with unacceptable soil erosion or salinity values?" Questions of this type will be
answered under full regional implementation.
1.1 1992 PILOT STUDY
The National Research Council (NRC, 1990) has defined an approach for designing and
implementing program-level monitoring programs (Figure 1-2) that EMAP-Arid is using as
-------
S3 Great Basin
O Mohavian
H Sonoran
Californian D Chihuahuan
Arctic D Plains
Colorado Plateau
Figure 1-1. Biogeographic provinces of North America used in the EMAP-Arid Program.
2
-------
a template. Pilot or exploratory studies, such as the EMAP-Arid 1992 pilot study, are an important
beginning step in the interactive process of defining and redefining expectations, goals, and
strategy for implementing a monitoring program (Figure 1-2). The EMAP-Arid 1992 study is at this
critical step in the NRC approach and was undertaken to assess a common sample plot design and
to evaluate indicator measurements recommended by the scientific community for arid ecosystems.
The EMAP-Forest plot design was selected for evaluation to determine if a common
sampling design could be used across EMAP terrestrial resource groups. The EMAP-Forest
researchers had previously evaluated the variance component of selected soil variables (e.g.,
particle size and carbon content) (Riitters et al., 1991) for plot design and these variables were
similar to several of the soil properties under consideration by the arid group. However, there was
no information about the EMAP-Forest sampling design related to the standard rangeland soil and
vegetative sampling techniques and the spectral analysis approaches recommended by the scientific
communities (Breckenridge et al., 1993; Mouat et al., 1992) as indicator measurements for EMAP-
Arid.
The EMAP-Arid researchers selected a limited number of indicator measurements for
testing. Indicator selection was based on review of the literature (Fuls, 1992; Bryant et al., 1990)
and discussion with researchers who were conducting studies on arid ecosystems at Long-Term
Ecological Research sites (Whitford, 1986) and in other areas with well-established data bases on
ecological processes (e.g., experimental stations and U.S. Department of Energy sites). This
indicator selection process, described in Breckenridge et al. (1993), led the EMAP-Arid resource
group to focus on measurements or attributes of vegetation, spectral properties, soil properties, and
soil erosion as measures of stress (e.g., from grazing) and as potential indicators of productivity and
biological integrity (Schlesinger et al., 1990).
The 1992 pilot study was developed to evaluate sampling plot design and the sensitivity of
selected EMAP-Arid indicators using the EMAP design for site selection. The design uses a
randomly placed systematic grid to identify sampling site locations (White et al., 1992; Overton et
al., 1990). Comprehensive assessment of the ecological condition of the Colorado Plateau, the
ultimate goal for EMAP resource groups, was beyond the scope of this initial study. The pilot study
was also not intended to derive conclusions about those values, issues and questions previously
identified as significant in arid ecosystems.
1.2 REPORT OVERVIEW
The conceptual approach, objectives, rationale, and processes that led to the choice of
study sites and selection of indicator measures to be tested are described in the EMAP-Arid
Colorado Plateau Pilot Study-1992 Implementation Plan (Franson, 1993). The four objectives
established in the implementation plan for the 1992 pilot study form the organizational basis of this
report. These objectives are summarized in Section 2 which also includes a description of the pilot
study area and the sampling plot design. Section 3 summarizes the indicator measurements.
Sections 4 through 7 present results and recommendations relative to the four study objectives
summarized in Section 2. The knowledge gained during the 1992 pilot has helped significantly
-------
Slept
Define
Expectations and Goals
Step 3
Conduct Exploratory
Studies if Needed
Step 2
Define
Study Strategy
Refine
Questions
Step 4
Develop
Sampling Design
Refine
Objectives
Can
Changes Be
Detected
Rethink
Monitoring
Approach
Steps
Implement Study
Make Decisions
Step 6
Produce Information
Step 7
Disseminate
Information
Is Information
Adequate?
NXCIS90)
Figure 1-2. The elements of designing and implementing a monitoring program (NRC, 1990).
4
-------
to define questions relative to these objectives and to establish new pilot study objectives. Section
8 summarizes the conclusions and recommendation from this study and describes ongoing and
proposed studies.
-------
SECTION 2
COLORADO PLATEAU PILOT STUDY
The Colorado Plateau Pilot Study was designed to consider issues critical to the success
and implementation of the EMAP-Arid program. These issues included plot design, indicator
development, sampling frame material, quality assurance, information management, and logistics.
This section describes the objectives of the pilot study, provides an overview of the study area, and
describes the site selection process and sample plot design.
2.1 STUDY OBJECTIVES
The assessments of sampling variance and indicator sensitivity were the primary objectives.
Evaluation of the sampling frame and extent and of implementation activities related to logistics,
quality assurance, and information management were also considered important issues and were
established as additional objectives. The specific objectives were:
Assessment of sampling variance. Evaluate the EMAP-Forests Resource Group
sampling plot design relative to the selected EMAP-Arid indicators.
Indicator sensitivity. Evaluate the sensitivity of selected indicator measures to
independent evaluations of site condition as designated by various land management
agencies.
Sampling frame and extent. Evaluate the utility of using classified TM imagery and
other data acquired from the U.S. Fish and Wildlife Service Gap Analysis Program
(GAP) to select frame materials for the pilot study and future studies and to provide
data for extent estimation of arid ecosystems.
Quality assurance, information management, and logistics. Evaluate the quality
assurance, information management, data analysis, logistical, and reporting
requirements and constraints based on the pilot study area.
2.2 1992 PILOT STUDY AREA
The Colorado Plateau is an arid and semi-arid tableland in the southwestern UnitedrStates
(Franson, 1993). The entire 130,000-square mile region of the Colorado Plateau is much more
extensive than needed to fulfill the requirements of the intended indicator evaluation pilot;
therefore, only a portion of the Plateau was chosen for data collection (Figure 2-1). This
-------
Grand
j Junction
b
Pilot
Study
Area
50 100 Miles
' 0 50 100 150 Kilometers
Figure 2-1. Location of the 1992 pilot study area within an outline of the Colorado Plateau.
. 7
-------
region is predominantly managed as federal lands, i.e., Bureau of Land Management, National Park
Service, and Forest Service, and part of Navajo Nation, state, and private lands. The diversity of
ownership and jurisdiction promoted interagency participation in the pilot implementation. The area
is bisected by the Colorado River and includes many canyon lands that allowed for evaluation of
logistical requirements in some of the most difficult terrain that EMAP-Arid activities will likely face.
The area includes two resource classes which are prevalent within the Colorado Plateau (i.e.,
desertscrub and conifer woodlands) that were chosen for indicator evaluation.
2.3 SITE SELECTION
Sampling points for the 1992 pilot were shifted approximately 13.5 km to the northeast of
the original EMAP grid point. This strategy allowed the EMAP grid to be maintained, but did not
compromise the original sampling points that will be used for full-scale implementation. This shift
was performed using geographic information system (GIS) software to densify the original EMAP
digital grid and select the new grid. The new grid was then overlaid onto a Landsat-derived
vegetation plot developed by the GAP to identify which plant communities were associated with
the new, shifted sample points.
The resultant shift of the original points yielded 63 points within the study area for the
1992 pilot study. The design was structured such that if the sample site fell within either
desertscrub or conifer woodland, sampling would proceed. If the sample site fell into some other
subpopulation (e.g., a grassland), it would not be sampled. It was expected that sampling would
occur on approximately 30 sites, with about 15 in each subpopulation. In fact, the overall process
resulted in 21 desertscrub, 12 conifer woodland, and 6 mixed vegetation (e.g., grassland,
agriculture) or nonsites (e.g., water bodies) being selected for sampling in 1992. Of these 39 sites,
10 were identified as not fitting the resource class requirements, leaving a total of 29 sites included
fn the final study.
2.4 SAMPLE PLOT DESIGN
The sample plot design selected for evaluation during the 1992 pilot study resembled that in
use by the EMAP-Forest resource group (Kucera and Martin, 1991). This design (Figure 2-2) was
modified for the methods used for arid ecosystems (Franson, 1993). The sample plot consisted of
specific plot designs for each indicator, overlaid on one another, resulting in a hexagon-shaped plot
(approximately 1 hectare in area). A central circular subplot was centered on each designated
sampling point. Six satellite subplots were located with their centers 40 m from the center point
and oriented at 0, 60, 120, 180, 240, and 300 degrees relative to compass north. Each of these
circular subplots was 7 m in radius, with an area of 154 m2. Radial transects, AR,,BR, and CR
were extended from the center point to the centers of subplots A1, B1, and C1, respectively.
Exterior transects, AE, BE, and CE were extended between centers of subplots A1 and A2, B1 and
B2, and C1 and C2, respectively. Soil sampling locations were placed at AP, BP, and CP, each 20
m from the associated subplot center point and oriented to enhance association between soils and
the two respective vegetation transects.
Shrubs and trees greater than 1.5 m in height within subplots MD, A1, B1, and C1 were
identified and measured as a part of the vegetation composition, structure, and abundance
indicator. Shrubs less than 1.5 m in height were measured in 1- by 2-m quadrats, aligned with
8
-------
Vegetation/Spectral Quadrats
(10 per transect)
1 x 2m
MD = center circular subplot
A, B, C = appendages
1 = circular subplot off radial
transect
2 = circular subplot off external
transect
R = radial transect
E = external transect
P = soil pit (clockwise from radial
transect)
Figure 2-2. Diagram of indicator sampling areas at EMAP-Arid 1992 Pilot Study sites.
9
-------
their long axis parallel to each of the six transects. The quadrats along each transect were
separated by 1-m intervals, with 10 quadrats sampled along each of 6 transects, for a total of 60
quadrats sampled on the plot. Within each of the 1- by 2-m quadrats, a 50- by 50-cm subquadrat
was measured for vegetation composition, structure, abundance of forb and grass ground cover,
and surface soil attributes.
At one of the three soil sampling areas (AP, BP, and CP), a soil pit was dug to evaluate
characteristics of the soil profile and collect samples to be sent to the laboratory for analysis of
physical and chemical soil properties. At the two remaining areas, soil was extracted and described
to 50 crn and augered and described below this depth to 1.5 m or bedrock. Soil samples were
collected from the top two horizons for laboratory analysis. If one or both of these areas belonged
to a different soil series than the first, then a soil pit was also excavated at that site. The above
approach was used at 22 of the 26 study plots (permission was not granted for excavations at the
remaining three sites). At the other four plots, all three soil sampling areas were excavated and
described to a depth that included the top two horizons.
Information on spectral properties was collected at 13 sites. Because only one instrument
was available for surface measurements, it was not possible to visit all the sites. Measurements
were made within each of the seven circular subplots on a grid, with vertical and horizontal spacing
between sampling points of 3 m, centered on the subplot center. In addition, along six vegetation
transects, spectral measurements were evenly spaced in the 1 - by 2-m area in every other quadrat,
beginning with the first quadrat, for a total of '6 quadrats on a transect.
The entire sample plot represents a "conceptual hectare." If one imagines that each circular
subplot represents an area surrounding it that has a radius of 20 m (one half the distance between
the center points), then the entire plot represents either a circle of radius 60 m and area of 11,310
m2 or a hexagon with 60 m from the center to each vertex, with an area of 9,350 m2. This is an
important conceptualization, especially for the spectral properties indicator, as Landsat Thematic
Mapper pixels are 30 m on a side, so that a 3 by 3 cluster of pixels represents 8,100 m2 and a 4
by 4 pixel cluster covers 14,400 m2 (Franson, 1993). Thus, the conceptual hectare of the sample
plot can be linked with remotely sensed spectral data. The plot structure represents a nested
design with one plot at each site, several subplots for each indicator, and potentially several
samples within each subplot. This structure allowed for estimation of the variance for sampling
design components for each indicator (Franson, 1993).
10
-------
SECTION 3
INDICATOR MEASUREMENT METHODS
Vegetation composition and structure have been evaluated for decades in arid ecosystems
and are well established as important indicators of ecosystem condition. The difficult decision for
the 1992 pilot was to determine what type of measurement technique best fit within the EMAP
approach. Numerous vegetative sampling techniques were evaluated including plot sampling, belt
transects, and line intercept techniques (Brower et al., 1989). Most of the vegetative sampling
techniques have been developed for measuring forage supplies for big game and other herbivores.
The literature was reviewed and the EMAP-Arid researchers decided to use a modified Daubenmire
(1968) approach because it provides the ability not only to measure vegetation attributes about
species richness and diversity, but also to keep open options for relating this information to wildlife
habitat in future pilots (Anderson and Ohmart, 1986).
A remote sensing approach to collect information about a site offers a number of
advantages for indicator development such as producing spatially explicit estimates of ecological
condition over entire regions in a cost-effective manner. A number of researchers have developed
strong relationships between measurements and indices derived from remote sensing and
ecosystem variables. The Normalized Difference Vegetation Index (NDVI) is such an index and
researchers have shown very high relationships between NDVI developed from satellite and ground
measurements and leaf area index. The leaf area index correlates strongly with a number of other
extremely important ecosystem variables such as primary productivity and biomass (Hobbs and
Mooney, 1990; Running, 199O). The NDVI was selected as a candidate indicator for the 1992
pilot study because it has both a demonstrated relationship to vegetation parameters and a lack of
sensitivity to atmospheric conditions (Anderson et al., 1993; Holben et al., 1-980; Holben and
Fraser, 1984; Holben, 1986; Hobbs and Mooney, 1990; Peters et al., 1993); in addition, it has
been used to monitor phenologies! (vegetation) variables on regional, continental, and global scales
(Tucker et al. .1986; Townshend and Justice, 1986; and Justice et al., 1985). The Landsat TM
satellite data were used to determine NDVI because the waveband location for deriving information
concerning vegetation parameters is superior to multispectral scanner (MSS) data and the pixel size
of 30 by 30 m correlates more easily with field-based measurements than does the 1.1- by 1.1-km
pixels of the Advanced Very High Resolution Radiometer.
Soil properties were selected because they were determined to be critical in evaluating
ecosystem health and interpreting vegetative information. Articles by Schlesinger et al. (1990),
Fuls (1992), Holmgren (1988), Lugo and McCormick (1981), Grossmat and Pringle (1987), and
West (199O) provided the rationale for looking at soil parameters (physical, chemical, and biological
[i.e., crusts]) and focusing on their implications to management options, plant growth and the
water balance. Soil erosion was also included in the 1992 pilot because most of the data required
for estimating errosion were collected in the soil profile. These data could then be used as inputs
to the Revised Universal Soil Loss Equation (RUSLE) erosion models for evaluating the relationship
11
-------
between soil erosion and site condition. Several researchers have identified a positive correlation
between increased runoff and erosion with decrease in the serai stage of arid ecosystems
(Schlesinger et al., 1990; Renard and Simanton, 1991) and evaluated the sensitivity of the models
{Renard and Ferreira, 1993).
The following sections provide a summary of the field methods used in the 1992 pilot
study. Sections 3.1 and 3.2 discuss vegetation and spectral methods respectively. Section 3.3
discusses methods for soil properties and soil erosion. For those cases where published references
are available for the method, discussion is limited and the reader is referred to the primary
reference.
3.1 VEGETATION MEASUREMENTS
The composition, structure, and abundance of vegetation have been recognized as
important measures of the condition of arid ecosystems. These measurements are considered
diagnostic in determining changes in biological condition at the organism, population, community,
and ecosystem levels. They particularly reflect the relationships of cover and height to water
availability and production in the arid West (Neman! and Running, 1989; Tausch and Tueller, 1990;
Tausch and Nowak, 1991). In addition, ground-based cover measurements can be used to assess
the accuracy of remotely sensed spectral properties. During the 1992 survey, related site and soil
characteristics were observed and recorded simultaneously with vegetation sampling. Specific
measurements included:
Vegetation Cover-Percent vegetation cover by species was determined using the
Daubenmire cover class method. The method was modified as described by Baily and
Poulton (1968) by adding a seventh cover class (<1 percent) to estimate trace
occurrences of ground cover (Table 3-1).
Species FrequencyQuadrat sampling methodology provided for the determination of
frequency by plant species.
Ground CoverGround cover as total vascular plant cover, litter, rock, bare soil, and
cryptogams was measured and provided important information for soils and erosion
analyses. Ground cover for all but total vascular plant cover was determined both for
the total quadrat and for the canopy interspace area.
Species Composition-Through both quadrat sampling and site survey, the species
composition of each site was determined and recorded.
Essential Complementary DataInformation collected included the description of
topography and landforms surrounding the sample location, the slope and aspect of the
site, and land use in the area.
Vegetation sampling followed a procedure adapted from Rangeland Monitoring Trend
Studies (USDI, 1985). Trees greater than 1.5 m in height and widely spaced and large shrubs
greater than 1.5 m in height were measured in four fixed area subplots. Subplots were 14 m in
diameter with one centered on the plot center and the centers of the remaining three placed 40 m
12
------- An error occurred while trying to OCR this image.
-------
Voucher specimens for any unknown species were collected and preserved in a plant press
for later identification. Such species were assigned a sample number for use on data forms until
species identification could be confirmed.
To fully characterize the vegetation composition of the site, a general search of the area
was made to identify any species present on the site but not encountered in the vegetation
transects or circular subplots. These species were only listed and no other data was recorded for
them.
3.2 REMOTE AND GROUND-BASED SPECTRAL MEASUREMENTS
In the 1992 pilot study, spectral data from the Landsat Thematic Mapper (TM) were
compared with concomitant ground-based spectral measurements of vegetation and soils obtained
through the use of a portable handheld spectroradiometer. A TM digital image file from an
overpass on August 20, 1992, was acquired for the 1992 pilot study from path 036, row 034.
The file is geocoded and covers all but the most southerly two sites of the pilot study sampling area
in southeastern Utah.
The Landsat TM sensor collects data, which can be converted to indicate radiance from the
Earth's surface, every 16 days at the same time of day, about 10:00 a.m. in the case of
southeastern Utah. Data are collected in pixels, which measure 30 by 30 m on the ground, as a
radiance value for each of seven broad wavebands. Bands 1, 2, and 3 correspond to blue, green,
and red visible light. Band 4 is a near infrared band. Bands 5 and 7 are in the middle of the
infrared spectrum, and Band 6 is in the thermal part of the spectrum (Figure 3-1, Table 3-3).
During the developmental phase of the TM sensor, a deliberate decision was made to position
wavebands in such a way as to derive the maximum information possible concerning vegetation
parameters (Lillesand and Kiefer, 1987). As a result of this decision, five of the seven wavebands
are located in wavelengths where significant vegetation information may be obtained {Table 3-3).
All natural objects reflect solar radiation to varying degrees at different wavelengths,
resulting in a spectral curve that may contain distinctive features and may be used to identify
certain physical properties of the object (Lillesand and Kiefer, 1987). Healthy green vegetation
produces a spectral curve that is characterized by a reflectance peak at about 550 nm, a narrow
absorption trough at 680 nm, a sharp increase in reflectance between 680 and 760 nm (the "red
edge"), a plateau between about 760 and 1,100 nm, and then two broader absorption features at
1,400 and 1,900 nm, with interspersed peaks of diminishing reflectance (Figure 3-1).
Red soil, a characteristic of the sites investigated during the 1992 Pilot Study, also has a
distinctive spectral curve, gradually increasing in reflectance from 400 to about 750 nm and then
leveling into a plateau. The chemical composition of soil is varied and will usually be mirrored by
the peaks and troughs in reflectance at specific wavelengths that are characteristic of the different
chemicals involved.
14
-------
Dry Bare Soil (Gray-Brown)
Vegetation (Green)
Water (Clear)
0
s\s /< i r s\ \ s\/
0.4 0.6 0.8 1.0
1234
1.2 U4 16 1.'8
Wavelength (pm)
5
TM Bands
Figure 3-1. Typical reflectance curves for vegetation, soil, and water with Landsat TM wavebands
(numbered). Band 6 (10.5 to 12.5//m not shown). (After Lillesand and Kiefer, 1987).
15
-------
TABLE 3-3. PRINCIPAL APPLICATIONS OF LANDSAT TM SPECTRAL BANDS
(from Lillesand and Kiefer, 1987)
Band
Wavelength
Nominal spectral
location
Principal applications
0.45-0.52
Blue
0.52-0.60
0.63-0.69
0.76-0.90
1.55-1.75
10.4-12.4
2.08-2.35
Green
Red
Near-infrared
Mid-infrared
Thermal infrared
Mid-infrared
Designed for water body penetration,
making it useful for coastal water
mapping. Also useful for soil/vegetation
discrimination, forest type mapping, and
cultural feature identification.
Designed to measure green reflectance
peak of vegetation for vegetation
discrimination and vigor assessment. Also
useful for cultural feature identification.
Designed to sense in a chlorophyll
absorption region aiding in plant species
differentiation. Also useful for cultural
feature identification.
Useful for determining vegetation types,
vigor, and biomass content, for
delineating water bodies, and for soil
moisture discrimination.
Indicative of vegetation moisture content
and soil moisture. Also useful for
differentiation of snow from clouds.
Useful in vegetation stress analysis, soil
moisture discrimination, and thermal
mapping applications.
Useful for discrimination of mineral and
rock types. Also sensitive to vegetation
moisture content.
8 Bands 6 and 7 are out of wavelength sequence because band 7 was added to'the TM late in
the original system design process.
Thematic mapper wavebands vary in width between 60 and 270 nanometers, and radiance
values are averaged within each band to give a single value. Several factors affect the quality of
data collected by the sensor, some of which, such as spacecraft motion, are corrected on board the
satellite. In addition, atmospheric conditions greatly affect data quality, especially for shorter
wavebands, and it is necessary to radiometrically calibrate radiance values before any image
processing or interpretation takes place (Sabins, 1987). The sensor collects radiance values at 256
levels of brightness, which can, in turn, be used to discriminate the objects being sensed.
16
-------
Chavez (1989) has developed two methods to perform radiometric calibration of Landsat
data, one using ground-based measurements of calibration targets and the other using a relative
power law model to predict the atmospheric haze values for the multispectral bands, based on a
starting band haze value that is selected for each image on an individual basis. The Landsat images
used for the 1992 Pilot Study were radiometrically calibrated using the second method developed
by Chavez (1989). Bands 1 through 4 were calibrated individually, based on their maximum
reflectance for a specific pixel and with a different stretching parameter applied to each, in order to
optimize the dynamic range of radiance. These converted numbers (for example, in band 1 an input
radiance of 100 was reset to 79 and an input of 230 was reset to 251) were then multiplied by a
constant to convert them to reflectance. For Band 1 this constant was 10, for Bands 2 and 3 it
was 7, and for Band 4 it was 5. The conversion to reflectance values was carried out in order to
enable Landsat TM data to be directly comparable to ground-based spectrometer measurements,
which were obtained in a reflectance rather than radiance mode.
The ground-based spectral measurements were made between late June and late August
with a portable Personal Spectrometer II (PSII). The instrument incorporates a PC-compatible
computer with an LCD screen, to allow for real-time data display, and a battery that lasts 2 hours.
Calibration procedures followed those given in the Personal Spectrometer Procedures Reference
Manual (Analytical Spectral Devices, 1993). Measurements are made using an optical fiber 2 m in
length, terminating in a hand-operated gun with an adjustable field of view that permits
measurements of targets from 1 mm to many meters in size depending on the instrument height
above object level. For the pilot study, the instrument was held 1 m above the surface to be
measured resulting in the acquisition of reflectance values from a circular area 28 cm in diameter.
In the case of pinyon and juniper trees or large shrubs where it was impossible for the gun to be
held 1 m above the target, a surrogate branch or shrub of the same species was chosen in the
same phonological state as the original target. Reflectance data from the PSII were collected in
4,095 brightness levels in 512 wavebands, each about 1.4 nm in width, between 350 and 1,050
nm.
The NDVI was calculated for both Landsat TM and PSII reflectance data collected for the
1992 pilot study and is discussed in the results in sections 4 and 5. The NDVI makes use of the
contrast between the high reflectance of vegetation in the near infrared (760 to 900 nm) and the
low reflectance in /the visible red wavelengths (630 to 690 nm). It is normally derived from either
Landsat TM or Advanced Very High Resolution Radiometer (AVHRR) satellite data. For Landsat
TM, this index uses data collected in Bands 4 and 3; thus,
NDVI = b4-b3/b4 + b3
3.3 SOIL MEASUREMENTS
Soil quality directly influences the amount, timing, and distribution (lateral and vertical
movement) of soil moisture available for plant growth. Soil infiltration properties and surface
characteristics also directly affect erosion processes, including overland flows (runoff) and transport
of suspended and dissolved solids. Disturbances and stresses to surface and subsurface soil can
influence flow velocity, routing, soil detachment, and deposition. The result is accelerated soil
17
-------
erosion that further affects moisture infiltration rates and patterns. Ultimately, physical changes to
vegetation communities may result. An altered soil moisture regime, in conjunction with changes in
other soil properties through erosion, can result in degradation to soil productivity, landscape
features, and vegetation composition and abundance.
3.3.1 Soil Properties
Soil quality was not measured directly, but it is associated with a number of soil properties
measured during the pilot study. The three soil property measurements or attributes were: (1) soil
profile (description and analysis)-the characterization of a vertical section of the soil through all its
horizons and extending into the parent material; (2) the top two mineral soil horizons (description
and analysis); (3) and soil surface attributes-description of attributes of the topmost soil surface
including vascular vegetation, rock fragments, cryptogams, bare soil, litter, surface type, and
surface roughness. The three soil properties that were measured in the pilot study control both soil
moisture and susceptibility to erosion processes. The following soil indicator measurements were
collected during the 1992 pilot study:
Full soil profile description and analysis from one hole dug to 1.5 m or bedrock.
Auger hole description of soil profile and soil-surface description and analyses of surface
soil at remaining two locations.
Description of selected surface soil attributes along six 40-m transects (these data were
collected by the vegetation group).
Soil scientists used materials extracted from soil pits and auger holes at each site to
describe soil characteristics, draw a soil map, collect samples, and compare the described soils to
the soil survey map unit and components previously developed for the site. The soil series or
family at each site shown in Appendix A was determined from the soil descriptions. The soil series
phase was used as a key to access the soil interpretation record (SIR) of the USDA Soil
Conservation Service. The SIR contains estimates of soil properties and interpretive information. A
soil series may have multiple SIRs, and thus the soil series must be matched to the appropriate
critical phase criteria on the SIR. Because the appropriate SIR number was not recorded at the time
of sampling, a match was made later according to the scientist's best judgment. In the future, the
critical phase criteria should be recorded at the time of sampling along with the appropriate SIR and
phase identification. Soil maps were used to determine the percentage of soil at each site. When
different soil series were noted for the site, data were weighted by multiplying values by the
percent soil of the series at the site.
Soil samples were collected from at least one soil pit. When possible, the soil pit was
located at the edge of a canopy boundary. This location was chosen to provide consistent
sampling protocol and to try to obtain a mixed sample of interspace and undercanopy soil. Table 3-
4 summarizes the soil sample analyses methods used in the laboratory.
18
-------
TABLE 3-4. METHODS FOR SOIL SAMPLE ANALYSES IN THE EMAP-ARID PROGRAM
Parameter
Selected method3
Sample preparation
Carbonate claysb
Cation exchange capacity
Electrical conductivity
Organic carbon
Sodium adsorption ratio
Bulk density
Particle size analysis
Water retention difference
pH soil-water suspension
1B1
3A1d
5A8
81
6A2
5E
4A1
3A1
4C1
a United States Department of Agriculture-Soil Conservation Service. 1992 Soil Survey
Laboratory Methods Manual. Soil Survey Investigations Report No. 42. Version 2.0.
U.S. Government Printing Office, Washington, D.C.
b Carbonate clay analysis only performed when indicated by effervescence in 1 N HCI drop test.
3.3.2 Erosion Index Indicator Measurements
Soil erosion is almost universally recognized as a serious threat to the productivity and
sustainability of ecosystems as evidenced by the fact that most governments in the world give
active support to programs of soil conservation. Soil erosion from wind and water can be measured
directly or calculated using a variety of models. Because accurate field measurements often require
extensive instrumentation and sampling of multiple plots (Larson et al., 1983; Breckenridge et al.,
1991), EMAP-Arid researchers decided to use modeling as an alternative approach to determine site
erosion. The Revised Universal Soil Loss Equation (RUSLE) (USDA-ARS, 1991) was selected to
evaluate erosion during the pilot study. This model was evaluated for the ability to construct an
erosion index for a nonagricultural site. The EMAP-Arid researchers used the RUSLE formula to
determine soil erosion rates. The RUSLE formula (USDA-ARS, 1991) is:
19
-------
A = RxKxLSxCxP
where:
A = average annual soil erosion (t/ac/yr)
R = rainfall and runoff factor
K = soil credibility factor
LS = slope-length and slope-steepness factor
C = cover and management factor
P = support practice factor.
The R factor was determined using nearby climatic stations. The P factor used for all sites
was 1. The LS factor was measured at each site. The K and C factors were calculated from
measured and estimated data (Table 3-5). The K factor can also be obtained from the USDA-Soil
Conservation Service soil interpretation record (SIR). The range of SIR K factors values are 0.15 to
0.45.
TABLE 3-5. MEASURED AND ESTIMATED DATA NEEDED TO CALCULATE K AND C FACTORS
Measured
Estimated
Slope length
% slope
% very fine sand
%silt
% clay
% organic matter (1.72 x organic carbon)
Surface structure
Rock fragments within the soil
% ground cover (with lichen and moss)
% canopy
Above ground biomass
Permeability
Field roughness
Root mass in top 4 in.
20
-------
SECTION 4
ASSESSMENT OF SAMPLING VARIANCE
The primary resource classes that must be sampled by the EMAP-Arid group can be broadly
identified as extensive or as resources that do not have distinct boundaries. Extensive resources
such as desert woodlands are generally not easily identified as populations. The geographic
distribution of extensive resources can be fragmented and discontinuous over the regional
landscape. The geographic continuity of extensive arid resources is frequently interrupted by urban
areas, areas of agricultural development or other natural systems such as forests; thus, these
resources do not always have well-defined boundaries.
Two extensive resources, conifer woodlands and desertscrub, were studied during 1992 to
assess indicator measurement variance. Table 4-1 summarizes those sites selected and sampled.
A variance component for a resource condition indicator, estimated from the sample data, describes
the variability in the resource condition over its geographic extent. Various other statistical
descriptors such as the cumulative distribution function and quartile frequencies can be used to
depict that variability for assessment and interpretive purposes. The spatial variability of the
measured indicator among the sampled sites expresses the actual variability in the resource
condition if no other extraneous variation interfered with the sampling process.
The process of collecting samples can produce extraneous variability in the indicator
measurements in addition to the variability associated with the resource condition. The EMAP
survey design protocols include annual visits to sampling sites throughout the region and will
require multiple sampling crews to procure the measurements within adequate time frames. The
utility of indicators of resource condition to some extent depends upon the degree to which these
extraneous sources of variation inhibit the ability of the indicator measurement to describe resource
characteristics. Knowledge of these variance values is necessary not only to construct confidence
intervals for the measured indicators but also to evaluate the viability of the measurements as
indicators. The magnitude and influence of each of these components of variability must be
evaluated by the EMAP-Arid program as it progresses through the indicator development process.
Variance components that continue to require a high level of investigation include those associated
with the year, crew, measurement, and plot design.
Year VarianceThe year component of variance arises from yearly fluctuations of the
indicator measurements about some central value that may be expected regardless of the presence
or absence of an overall trend in the condition of a resource from year to year. Evaluation of this
natural variation in the .condition of a resource as a result of climatic and other associated natural
changes will be important to the detection of trends of a more fundamental nature to the systems.
The graphs in Figure 4-1 illustrate how high extraneous variability can inhibit the utility of a
condition indicator that has some response to yearly transient climatic fluctuations. The dashed
21
-------
TABLE 4-1. SITES SELECTED AND SAMPLED IN THE 1992 COLORADO PLATEAU PILOT STUDY
Site
Resource Type
Indicator Measurements
Sampled*
20752
20753
20755
20758
20907
20908
20911
20913
21059
21060
21061
21062
21064
21065
21210
21211
21212
21213
21214
21215
21216
21217
21361
21362
21363
21364
21365
21366
21367
COF101
Conifer woodlands
Desertscrub
Conifer woodlands
Desertscrub
Conifer woodlands
Conifer woodlands
Conifer woodlands
Desertscrub
Desertscrub
Desertscrub
Desertscrub
Desertscrub
Conifer woodlands
Desertscrub
Desertscrub
Conifer woodlands
Desertscrub
Desertscrub
Conifer woodlands
Conifer woodlands
Conifer woodlands
Desertscrub
Desertscrub
Desertscrub
Conifer woodlands
Desertscrub
Conifer woodlands
Desertscrub
Desertscrub
V SL
V SL SP
V SL SP
V SL SP
V SL
V SL SP
V SL SP
V SL SP
V
V SL SP
V SL SP
V SL
V SL
V
V SL SP
V SL
V SL
V SL
V SL
V SL
V SP
V SL SP
V SL
V SL
V SL
V SL SP
V SL
V SL
V
V SL SP
* V = Vegetation
SL = Soils
SP = Spectral
line shows a fundamental trend in the resource condition via the indicator. The solid line shows the
trend in the indicator measurement from field observations. The cyclical nature of the measured
indicator around the trend in panel B as a result of yearly fluctuations greatly exceeds that shown in
panel A. Although the measured indicator in panel A fluctuates around the fundamental trend, a
general trend is still evident from the observations. In panel B, the response of the measured
indicator to the climatic cycles is sufficient to mask evidence of a fundamental change. Each of the
other sources of extraneous variability in the following discussion can have equally important
effects on the measured indicator.
22
-------
100 -
80 -
60 -
40
20
CONDITION
TREND
-OBSERVED
INDICATOR
TREND
100
80
60 |
40
20 +
B
CONDITION
TREND
-OBSERVED
INDICATOR
TREND
01234
YEAR
567
3 4
YEAR
H
7
Figure 4-1. Theoretical effects of extraneous variability on indicator response illustrating how high
extraneous variability can inhibit the utility of a condition indicator that has some
response to yearly transient climatic fluctuations.
23
-------
The annual variation in indicator value may differ among subregions in the monitored region.
This variation is a response of the resource to local climatic and other effects that differ among
subregions of the resource and is indicative of subregion variation in addition to that associated
with annual fluctuations over the entire region. These differences in the annual variations among
the subregions are represented as a location-by-year interaction component of variance in the
monitoring program that must be evaluated for its effect on the detection of overall trends in the
resource condition.
The activities associated with acquisition of data for indicator measurements produce
variability beyond that associated with the natural transient variability in the condition of a resource
over years or subregions. These sources of variation also must be evaluated for the magnitude of
their influence on the statistical confidence of indicators. The following is a partial list of those
components of the sampling system that can be targeted to reduce overall variance.
Crew VarianceThe magnitude of the monitoring program will require multiple field crews
with each crew collecting data at different sites than those visited by the other field crews using
the same sampling protocols. The magnitude of crew-to-crew variability can be controlled through
training/experience, and crew evaluation during the sampling period in addition to using as few
crews as possible to complete the field work within the required time period.
Measurement VarianceMeasurement variances are produced during measurement of the
physical sampling units. Quality assurance and quality control protocols address measurement
variance. These protocols track and sequence activities from sample correction to the validated
and verified data base and incorporate procedures to minimize this component of variation.
Plot Design VarianceThe measurements of indicators at each site require a response design
at each sampling site. The response plot design consists o"f sampling units of various sizes and
orientations from which to collect the observations required at a monitoring site to measure the
response of a resource to the environment. The sampling units for extensive resources may consist
of subplots or transects, such as those used in,the 1992 EMAP-Arid pilot study, within which
subsamples or quadrats are used for measurement purposes. The variability among these physical
measuring units affects the precision of the indicator measures at the site. Thus, a characterization
of the variance properties of these sampling units is essential to the development of a response
design at monitoring sites that provides efficient statistical estimates of indicator measurements.
4.1 ANALYSIS AND RESULTS
The evaluation of indicator measurement variances associated with the sampling units that
potentially could be used in a common plot design for monitoring EMAP-Arid extensive resources
was a primary objective of the 1992 pilot study. Year and crew components of variance were not
considered for investigation in the 1992 pilot study and will be determined from larger and long-
term studies in the future. Measurement variances are discussed in Section 7.
Measurements considered most influential for the spectral, vegetation, and soils indicator
categories were selected as candidates for evaluation of their variance properties. A single spectral
measure, the Normalized Difference Vegetation Index (NDVI), was selected for the analysis. The
24
-------
variables selected for vegetation measurements were total vascular plant cover, shrub cover, and
tree cover. Soil variables analyzed were the clay, silt, sand, and very fine sand percentages;
organic matter; the soil credibility factor (K); and the length-slope (LS) steepness factor.
The variables for this analysis also were selected to represent measurements acquired from
different components of the plot design. The basic sampling units in the plot design at the sites
were the 50- by 50-cm quadrats, 1- by 2-m quadrats, circular subplots, and soil subplots. The
selected variables are each representative of uniquely different types of statistical variables which
affect how the indicator variables are used in the analyses. Each category of indicator
measurements will be discussed separately.
4.1.1 Vegetation Measurements
The three vegetation cover measures (total vascular plant, shrub, and tree cover) were
measured on different measurement units in the plot. Total vascular plant cover was measured on
50- by 50-cm quadrats nested along the 10 quadrats. Shrub cover was measured on ten 1- by 2-m
quadrats nested within each transect. Tree cover was measured in the fixed area circular subplots
(see Figure 2-2).
Total vascular plant cover and shrub cover were scored in one of seven Daubenmire cover
classes (see Table 3-1). Thus, the discrete random variables have seven possible values, expressed
as the Daubenmire cover range midpoint. The linear statistical model to describe the variation for
these two variables is a nested effects model:
-j(i)
(1)
where / = 1,2,..., s sites;/ = 1,2,..., ttransects per site; and k 1, 2, ... , q quadrats per
transect. Y/jk is the measurement (shrub cover or total cover) on the quadrat, // is the average
cover for all sites in the study, s,- is the difference of a particular site from the overall average cover,
fay is the difference of a transect on a site from the average for a site, and qk(f-, is the difference of
a quadrat on a particular transect from the average for a transect.
The sites, transects within sites, and quadrats located on the transects are considered
random samples of their respective representative populations for the purpose of this analysis. The
objective of the analysis was to estimate the components of variance for the respective random
effects in the model s/f tj(it, and qk(i:t.
Estimates of the components of variance are useful for planning future sampling strategies
at the monitoring sites. The estimates are only useful if they exhibit some consistency^
magnitude across the sampled population. Thus, an assumption of homogeneous variance of
quadrats on transects and of transects within sites is necessary for useful component estimates.
Both shrub cover and total cover are discrete variables measured in the seven Daubenmire
categories on the quadrats. Discrete variables tend to have strong relationships between means
and variances and can be expected to exhibit heterogeneous variance.
25
-------
The residuals from an analysis of variance, where the residual equals the observed value
minus the fitted model value, reveal the characteristics of the observations on the basic sampling
unit. In the case of shrubs and total cover, the basic sampling unit is the quadrat. The plots of
residuals versus fitted values for shrub and total cover on quadrats shown in Figure 4-2 reveal the
patterned residuals for each of the measurements. Clearly the patterned residuals preclude the
assumptions necessary for valid variance component estimates for the quadrats. Regardless of the
magnitude of cover, represented by fitted values, only several categories of values appear for the
residuals. The total cover residuals most clearly reveal the pattern of seven categories of cover. A
patterned residual plot occurs regardless of whether the measurements were made on 1 - by 2-m
quadrats for shrubs or 50- by 50-cm subquadrats for total cover.
Ten quadrats were measured on each of the transects. The variability patterns for transects
within sites can be evaluated from the transect means by using the average of the ten quadrats as
a transect observation. An analysis of variance was conducted on the transect means with the
nested model
(2)
where Ytik is the transect mean, // and sf have the same interpretation as the model in equation (1),
and ftp) is the difference of the transect mean from the average for a site. The residual plots from
the analysis of transect means for shrub cover and total cover shown In Figure 4-2 reveal a
considerable contrast to those for the quadrat measurement. The patterned residual plot has been
dispersed through the smoothing provided by the means. In addition no distinct variance pattern is
apparent from the plot. A test of transect variance homogeneity among the sites with the Levene
(median) test {Brown and Forsyth, 1974) did not reject the hypothesis of homogeneous variance; P
= 0.12 for total cover and P = 0.07 for shrub cover.
These results indicate that a useful measure of shrub cover or total vascular plant cover for,
development of sampling plans is the average of groups of quadrats rather than the individual
quadrat.
Tree cover was measured on four 7-m circular subplots at each site. Tree cover was
determined from height and crown "diameter measurements; thus, cover is a continuous measure on
the plot. A box plot of the residuals by site from an analysis of variance of tree cover is shown in
Figure 4-3. Although some sites exhibited what appeared to be considerably large or small
dispersion among the four subplots, a Levene (median) test for homogeneity of subplot variances
among the sites did not reject the hypothesis of homogeneous variances (P = 0.49). Suitable
estimates for subplot variances can, therefore, be obtained to plan sampling designs for tree cover.
4.1.2 Ground Spectral Measurements
Spectral measurements with the portable spectrometer were obtained in two different plot
designs at each of the sites. One design used the quadrat sample design for vegetation cover. The
second design used the circular subplots used for tree cover. The indicator measurement derived
from the spectral reflectance measures was the NDVI. The spectral data are of questionable quality
26
-------
Analysis of Variance Residuals
for Shrub and Total Vascular Plant Cover
Quadrat Shrub Cover
Residuals
100
50
«
0 -
-50 -
*
20 40 60
Fitted
30
20
10
0
-10
Transect Shrub Cover
Residuals
10 20 30
Fitted
Quadrat Total Cover
Residuals
50
0 20 40 60 80
Fitted
Transect Total Cover
Residuals
30
20
10
0
-10
-20 -
t* '. ' ;
* uf * > 1
j-£:^: :: j: ... *
t:*[-° '* ;" .:
* *
10 20 30 40 50 60
Fitted
Figure 4-2. Analysis of variance residual plots for shrub cover and total vascular plant
cover using quadrat measures or transect means.
27
-------
Tree Cover Residuals vs Site Number
Residuals
60 -
40 -
20 -
0 -
-20 ~
-40 -
§
«o m
«
t
\0
Site Number
Figure 4-3. Box plot of tree cover analysis of variance residuals by site. Brackets indicate range,
shaded areas are 25th and 75th percentile values, and the central white area is the
median.
28
-------
because the field calibrations of the spectrometer to a white reference standard were performed
incorrectly. This problem and its effect on data quality are discussed in more detail in Section 7.1.
The NDVI variable is a bounded continuous variable with 0 < NDVI <1. Histograms of
all observations collected on circular plots and the linear transects in the study are shown in
Figure 4-4. The NDVI observations are highly skewed with most observations near the lower
boundary value for NDVI.
Three NDVI measurements were obtained from each of the quadrats located on the
vegetation transects. The linear statistical model for these NDVI measurements is
(3)
where Yjjk, is a single NDVI measurement and all other components have the same description as
those for the model in equation (1) with the additional component e,(ijk) for the difference of a single
NDVI measurement within a quadrat from the quadrat value.
Sixteen NDVI measurements were made in a 4 by 4 array with a 3-m spacing within each
of the circular plots. The linear, statistical model is
Yijk
(4)
where Y/jk is the NDVI measurement, fj is the overall average for all sites, s,-is the site average
deviation from the overall mean, pi(i) is the difference of the subplot mean from the mean of all
subplots on the site and ekfljj is the difference of an individual measurement from the subplot mean.
Residual plots from an analysis of variance for NDVI measured on transects or circular plots
are shown in Figure 4-5. The variance patterns for the individual measurements within the circular
plots exhibit a typical pattern for bounded variables when many of the observations are near the
lower bound. The observations from quadrats on the vegetation transects exhibit an even more
extreme residual variance pattern than those from the circular plots. The more extreme pattern is
the result of having only 3 NDVI measurements within each basic measuring unit (the quadrat) as
opposed to 16 NDVI measurements within the circular plots.
If the average of all NDVI measurements on the circular plots or the transects is used for
the analysis (see Equation 2), a much different variance pattern is apparent in Figure 4-5. The "
means possess a distribution more amenable to analysis for estimation of variance components
similar to that for vegetation cover. Thus, planning for sampling designs must be based on the
means of groups of NDVI measures in either the transect or rectangular sampling arrays rather than
the smaller sampling units.
29
-------
Circular Plots
Transect Plots
o
O -i
*n
.
o
o -
CO
o
o -
o -
1
III....
~
o
0 -
-
o
o -
co
0
o -
o -
1
ll.
0.0 0.2 0.4 0.6 0.8
0.0 0.2 0.4 0.6
NDVI
NDVI
Figure 4-4. Histogram of NDVI measurements from circular plots and transects.
30
-------
NDVI Analysis of Variance Residuals
from Circular Plots and Transects
Circular Plot NDVI
Residuals
Circular Plot Means
Residuals
0.6
0.4
0.2
0.0
-0.2
. . t* 0.10
-*:" . " : ?"t
"" *": i« *' s o-os
-. " .. Kr
A!^J '1&* ' ' °'°
'I^KKllj »l>flu *
^^^^Bl fl!i f * *i $ -O.os
ti?lt5 j *!'*
'«i -0.10
. » t
: . .
^. «. . ,
1 fc »
** X *
*
0.0 0.05 0.15 0.05 0.10 0.15
Fitted Fitted
Transect NDVI Transect Means
Residuals Residuals
0.4
0.2
0.0
-0.2
0.4
0.15
- .t , 0.10
(
!»t** «* * . 0.05
v i* . : J 1 1
* . * * o.o
*'<
* . -0.05 :
t -0.10 -
I
1 ': .«
i li * '
* * »i *
*
-
0.0 0.2 0.4 0.6
Fitted
0.04 0.08 0.12 0.16
Fitted
Figure 4-5. Analysis of variance residual plots for NDVI measurements on circular plots, transects,
and for means of circular plots and transects.
31
-------
The Levene (median) test for homogeneity of circular plot or transect plot variances among
sites did not reject the null hypothesis for homogeneous variance; P = 0.46 for circular plots and P
=* 0.21 for transects.
4.1.3 Soils Measurements ,
The soils data were collected in three subplots (pits) at each site. The measurements
considered for variance analyses were the percentages of clay, silt, sand, very fine sand and
organic matter as well as the soil erodibility factor, K, and the length-slope and steepness factor,
LS. The statistical model for the soil measurements is
Y±J=
(5)
where V«is the soil measurement, jj is the overall mean, s,-is the difference of the site mean from
the overall mean, and p,^ is the difference of the subplot value from the site mean. The component
of variance of interest for a soils measurement is the variance among subplots within sites. The
analyses of variance residuals did not exhibit any unusual properties and the analysis of variance
estimates of the variance components were computed for each of the soils measurements.
Residual plots for the seven soil measurements are shown in Figure 4-6. The Levene
(median) tests for homogeneity of subplot variances were nonsignificant for percent clay (P =
0.47), percent sand (P = 0.16), percent very fine sand (P = 0.30), organic matter (P = 0.41),
erodibility (P = 0.13), and length-slope (P P = 0.16), with percent silt somewhat significant (P =
0.03).
4.1.4 Variance Component Estimates
The vegetation, soils, and spectral measures used for the variance component estimates are
shown in Table 4-2. In most cases the variables are averages of measurements on the subplots.
The subplots for the vegetation measures were the transects for all cover measures. Subplots for
the spectral measure, NDVI, were the transects or circular subplots. Soils measures were used
from the three soil pit subplots at each site.
The linear statistical model for the analysis of variance is
Y1:j=
(6)
where / = 1, 2, ... , s sites;/ = 1,2,..., /?,- subplots at site /; // is the overall mean; s/ is the
difference of the site mean from the overall mean; and p-(il, is the difference of the subplot from the
average of subplots at a site. It is assumed the subplot effect, p-(il, is a random effect with mean 0
and variance on .
32
-------
Clay
Silt
10
5
0
-5
Residuals
\
5 10 15 20 25
Fitted
Residuals
10
0
-10
-20
0 10 20 30 40 50 60
Fitted
Sand
Residuals
20 40 60 80
Fitted
Very Fine Sand
Residuals
20
10 ' ./ > :
** t f» *
0 * * * g^f * fm
* * *
10
t . .
15
10
5
r\
"
.5
-10
*
* * ^ ^ #;
* *
. . ,
10 20 30 40 50
Fitted
Figure 4-6. Analysis of variance residual plots for soils measurements (Page 1 of 2).
33
-------
Soil Erodibility-k
Length-Slope
Residuals
0.3
0.2
0.1
0.0
-0.1
-0.2
-0.3
Residuals
0.1 0.2 0.3 0.4 0.5
Fitted
1.5
1.0
0.5
0.0
0.5
1.0
»*
v* !
4r^ 8
'* *. !
. '
0.5 1.0 1.5 2.0
'' - .
Fitted
Organic Matter
6
4
2
0
-2
-4
Residuals
-«>:
0
24
Fitted
Figure 4-6. Analysis of variance residual plots for soils measurements (Page 2 of 2).
34
-------
TABLE 4-2. VARIABLES, MEASURING UNITS, AND MEASURES USED
FOR VARIANCE COMPONENT ESTIMATES
Variable
Measuring unit
Variance estimate
Total vascular plant cover
Shrub cover
Tree cover
NDVI
NDVI
All soils measures*
50- by 50-cm quadrat
1 - by 2-m quadrat
7-m circular plot
3 samples on 50- by 50-cm
quadrat
16 samples on circular plot
1 to 3 soil pits
Transect average
Transect average
7-m circular plot value
Transect average
Plot average
Soil pit average
* Clay, silt, sand, very fine sand, organic matter, K (soil credibility), and LS (length-slope and
steepness)
The variance components were estimated from the analysis of variance for each variable.
The analysis of variance in Table 4-3 shows the source of variation degrees of freedom and
expected mean squares. Variance components can be estimated by equating observed mean
squares to expected mean squares. The variance component of interest for the current study is
that for subplots within sites. Thus, the component estimate is the observed mean square for
subplots within sites (MSW) for each of the measures in Table 4-2 or azp = MSW.
The subplot variance component estimates for the vegetation, spectral, and soils measures
are shown in Table 4-4. The standard deviation for subplots, the overall study mean, and the
percent coefficient of variation for each measure is included in Table 4-4. Also included in
Table 4-4 are the half lengths of the 95 percent confidence interval estimates for a site mean for
each of the measurements. The half length of the interval was computed as:
where t is the Student t for a two-sided 95 percent confidence interval, s2 is the estimate of the
subplot variance, and n is the number of subplots at a site for the 1992 pilot study. The degrees of
freedom for the Student t are those shown for the subplot variance estimate in Table 4-3.
The measured variable exhibited a considerable range in percent coefficients of variation.
They ranged from 10.9 percent for percent sand to 88.1 percent for soil organic matter. However,
a majority (seven) of the variables had coefficients of variation between 19.9 percent and 46.7
percent. The half lengths of the 95 percent confidence interval estimates in the last column of
Table 4-4 showed differences analogous to those of the coefficients of variation among the
variables as expected. As a percentage of the overall mean, the half lengths ranged from 13
percent of the mean for percent sand to 102 percent of the mean for soil organic matter. Again a
35
-------
TABLE 4-3. ANALYSIS OF VARIANCE FOR MEASURES ON SUBPLOTS AT EACH
OF SEVERAL SITES8
Source
Degrees of
freedom
Mean
square
Expected
mean square
Among sites
S- 1
MSA
Subplots within site
M
r/-1)
MSW
= MSW
M
3 S
MSA
MSW
N
number of sites in pilot study.
mean square among sites.
estimate of population.
number of repetitions within site,-.
observed mean square for subplots within sites.
total number of samples collected for all sites.
majority of the variables, seven, had half lengths between 23 percent and 38 percent of the mean.
They were total cover, shrub cover, NDVI circular plots, percent clay, percent silt, percent very fine
sand, and soil credibility.
These results indicate considerable differences in precision among the measured variables
with the 1992 plot design. In some cases an increase of 50 percent more observations would be
required to obtain confidence interval lengths of a more modest nature of 15 percent of the mean,
whereas other variables would require at least a doubling of the number of observations to achieve
a decent order of precision to estimate the site mean.
4.2 RECOMMENDATIONS
The variance patterns exhibited by the discrete vegetation measures and the bounded NDVI
measure indicate that viable variance component estimates for response design planning can only
be obtained from the averages of clusters of measurements such as the transects or circular plots.
Valid variance component estimates were obtained for those variables, such as tree cover and soils
measures, whose basic measurement unit was the subplot. They in turn can be used for usual
sample size considerations.
36
-------
TABLE 4-4. SUBPLOT VARIANCE COMPONENT ESTIMATES FOR THE VEGETATION,
SPECTRAL, AND SOILS MEASURES
Variable
Total cover
Shrub cover
Tree cover
NDVI transect
NDVI circular
Clay
Silt
Sand
Very fine sand
Organic matter
Ka
LSa
Subplots
6
6
4
6
7
3
3
3
3
3
3
3
Degrees
of
freedom
145
129
57
52
54
48
48
48
48
48
48
48
Subplot
variance
87.78
67.93
467.62
0.002398
0.001954
7.04
44.45
60.93
30.93
3.69
0.007495
0.13724
Standard
deviation
9.37
8.24
21.62
0.049
0.044
2.65
6.67
7.81
5.56
1.92
0.087
0.37
Overall
mean
27.78
1 7.63
34.36
0.086
0.095
8.31
20.33
71.36
28.01
2.18
0.335
0.634
Coefficient
of
variation
33.7
46.7
62.9
57.2
46.3
31.9
32.8
10.9
19.9
88.1
26.
58.4
Interval
half
length
7.57
6.66
21.62
0.04
0.033
3.08
7.74
9.06
6.45
2.22
O.1
0.43
a K = soil erodibility
LS = length-slope and steepness
The measurements considered for this variance study showed considerable discrepancy in
precision among the variables with the current plot design. Considerable increases in the numbers
of observations would be required to have more precise estimates of site means for many of the
variables. Clearly, the adaptation of the EMAP-Forest plot design to site sampling for arid resources
needs to be reconsidered in light of these results.
The patterns of spatial variation for sampling units are not well known for arid communities
and may differ considerably from other types of natural communities. In addition, a response plot
design that integrates measurements for three indicator categories-vegetation, spectral and soils
needs to be determined. This need necessitates a study to determine efficient clusters of units to
obtain measurements for the discrete vegetation and bounded NDVI variables that can be integrated
with those measurements which already provide viable variance component estimates.
Such a study would require the ability to evaluate the variabilities associated with different
sizes and shapes of the measurement unit clusters, more so than was possible with the current
response plot design. Some insight into the behavior of variances was obtained with the current
design, especially with measurements on NDVI. Clearly clusters of three observations from
quadrats on the linear transects were inferior to clusters of all 18 observations on the linear
transects. Also, some differences of variability were evidenced in the clusters of 18 NDVI
37
-------
measurements on the linear transects from those with clusters of 16 NDVI measurements in a
rectangular array from the circular plots.
It is recommended that a study be conducted to determine an optimal integrated response
design for EMAP-Arid monitoring. Such a study should be conducted in the manner of a uniformity
sampling study that allows a wide range of arrangements of the basic measurement Units from
linear transects of varying lengths to varying shapes and sizes of rectangular arrays of the units.
The relationships of the arrangements to their respective variances can be used to craft efficient
sampling designs at a site. Also, this type of study would result in data to estimate the level of
spatial correlation that can be expected from the measurements. Knowledge of the spatial
correlation would indicate the need for any spatial separation among the measurement units to
increase the amount of independent information acquired from the units.
38
-------
SECTION 5
INDICATOR SENSITIVITY
One of the primary aims of the indicator evaluation process is to evaluate the degree to
which individual indicators represent a range of ecological condition (Hunsaker et al., 1990; Frost et
al., 1992). This is often referred to as evaluation of indicator sensitivity. Two general types of
indicator sensitivity are commonly evaluated: the grouping or clustering of indicator values across
an environmental gradient (Ludwig and Reynolds, 1988) and the degree to which an indicator varies
within a known range of conditions (NRC, 1994).
The first type of sensitivity analysis normally involves recognition of patterns or clusters
(e.g., pattern recognition or detection) of values of indicators across an environmental gradient.
The study is designed to determine if indicator values will separate or cluster into one or more
groups and whether the groups correspond to the environmental gradient. This design allows an
evaluation of indicator sensitivity to a range of environmental conditions, even if standards (e.g.,
desired conditions) for evaluating condition are not known.
The second type of sensitivity analysis generally involves selecting sample sites based on a
range of "known" or "desired" conditions and evaluating the degree to which indicators vary across
those conditions. This type of sensitivity analysis requires an a priori agreement on what
constitutes condition (e.g., nominal, marginal, subnominal) and knowledge of the geographic range
of the condition (so that representative sites can be selected).
Initially, the EMAP-Arid researchers had intended to evaluate indicator sensitivity relative to
known or desired conditions as determined by existing information available from federal land
management agencies. The EMAP-Arid team decided to conduct this initial pilot study in the
Colorado Plateau due to the wealth of information available from this area (Kepner, 1991).
biscussions were held with a number of management agencies and these discussions led to the
understanding that EMAP-Arid could obtain congruous determinations of site condition for the
Colorado Plateau area. However, the EMAP team discovered significant differences in agency
descriptions of the condition of a site. This difference was substantial enough in several cases that
no consistent rating of a site could be established. Recently, similar concerns have also been
reported by the NRC in their review of rangeland health (NRC, 1994). As a result of these factors,
the 1992 pilot study was not able to address the objective to evaluate indicator sensitivity against
sites of "known" condition. Results presented in this section are only indicative of patterns in the
Colorado Plateau and the range in condition, delineated in these patterns is not known. However, it
is reasonable to assume the sites were different and represented at least a partial range in
condition.
39
-------
5.1 VEGETATION INDICATOR
As noted above, different agencies value different components of arid ecosystems and
condition ratings are not consistent among agencies. However, given the assumption that different
condition classes would be encountered in data collections at 29 different sites, some evaluation of
indicator sensitivity is possible from this pilot study. It is expected that the parameters measured in
the vegetation component of this study can serve as indicators of condition whether alone or
combined into ratios or indices.
5.1.1 Analysis and Results
This analysis is limited to the preparation of plots of site data to illustrate the applicability of
some vegetation parameters and their comparative ratios as indicators. The plots are of three basic
types, i.e., plots of total vascular cover on a site with other vegetation parameters or ratios, plots
of total vascular cover with soil parameters, and plots of total number of species on a site with
other vegetation parameters.
For instance. Figure 5-1 contrasts the number of species encountered at a site with total
vascular plant cover on a site. Of the sites examined in the 1992 pilot (Table 4-1} two sites were
notable deviations from the general pattern (outliers) suggesting that these sites may have been
subjected to stress. These sites are 21210 and 21062. Each had a large amount of cover (greater
than 60 percent foliar cover) and a considerable exotic species component in terms of cover.
These two sites plus site 20758 also appear as outliers in the plot in Figure 5-2. This plot
contrasts total vascular plant cover and exotic plant cover. The desertscrub site, 21210, is
dominated by Halogeton qlomeratus, an introduced annual forb that is poisonous to sheep. Both
desertscrub sites, 20758 and 21062, have large amounts of Bromus tectorum, an introduced
annual grass which can cause soremouth in livestock and wildlife species due to stiff awns in its
seedhead.
Figure 5-3 contrasts total vascular plant cover with a ratio of poisonous and physically
injurious plants to total plants on a site in terms of relative frequency. The ratio of poisonous and
physically injurious plants to total number of plants is generally expected to remain low in a
relatively undisturbed site. This plot discriminates the same three sites as do plots in figures 5-1
and 5-2, thereby suggesting these parameters may be good discriminators of subnominal versus
nominal sites in a regional index. When palatable (for cattle and sheep only) species cover is
considered (Figure 5-4), the relationship still held up as well because some exotic species are highly
palatable at some time of the year. These three outlier sites have several things in common. Each
has a relatively high amount of vascular plant cover. At site 20758 vascular plant cover is the
lowest of the three, which is why this site is not an outlier in Figure 5-1. Each has a high ratio of
exotic species which can be harmful or highly palatable to sheep or cattle at some point during the
plant growth cycle.
40
-------
40
CO 30
CO
CD
"O
&
CO
Q_
"b
CD
£ 20 -
10 -
o =Conifer Woodland
=Desertscrub
* =Desertscrub Outlier
o
o
o
o
00
o
I ' ' ' ' ' ' ' ' ' I '
0 10
20
30
40
50
60
n-i j
70
Total Vascular Plant Cover (%)
Figure 5-1. Total number of plant species versus total vascular plant cover.
41
-------
40
30
I
o
_cg
QL
20 -
O = Conifer Woodland
=Desertscrub
* =Desertscrub Outlier
10 -
o .
o
: i i i I i i I I I I i i i i 1 I I I t I i i I I I i [ i i i I I i I i I I
0 10 20 30 40 50 60 . 70
Total Vascular Plant Cover (%)
Figure 5-2. Exotic plant species cover versus total vascular plant cover.
42
-------
M
en
130
120
110
100
90
80
70 -\
I 60 H
CO
CO
O
O
CO
50 -
30-
20 -
10 -
0 -
o = Conifer Woodland
atOesertscrub
* =Desertscrub Outlier
o
o
o
o
o
o
o
o o
o
o
o
1 1 1 1 1 r
~\ 1 1 1 1 1 r
100
200
300
Total Species Occurence on Site
Figure 5-3. Number of poisonous plants versus total number of plants at a site.
43
-------
60
50-
^T 40 -
1 :
o
CB
CL ;
eo 30 -
1 :
C/D
i
1 :
-
;
-
10
-
0-
O =rConifef Woodland
atQesertscrub
* ssDesertscrub Outlier
.*
*
o
v *
° * *
o
o
o °
0
° *
o
o
o
0 10 20 30 40 50 60
Total Vascular Plant Cover (%)
Figure 5-4. Palatable plant species cover versus total vascular plant cover.
70
44
-------
Comparisons of total vascular plant cover with soil organic carbon showed a general pattern
of increasing total vascular cover with increasing levels of organic carbon for desertscrub sites.
This pattern was not evident for conifer woodland sites. Electrical conductivity versus total
vascular plant cover discriminated site 21210 which is the desertscrub site dominated by Haloqeton
glomeratus. This forb is tolerant of saline and alkaline conditions so the parameter is somewhat
descriptive in that sense.
Figure 5-5 is a comparison of the sand/clay ratio with total vascular plant cover on each
site. This plot discriminates several groups. Lower values of desertscrub and pinyon-juniper on the
x axis indicate decreasing sand percentages at the sites. Sites that have a ratio greater than 20
(i.e., high sand relative to clay content) are interesting. Sites with more sand relative to clay had
about the same total vascular cover, but sites with higher clay content fell into three groups. The
sites with the highest vascular plant cover were sites with lower sand-to-clay ratios and were
dominated by exotic annuals. The sites with moderate vascular plant cover, 37 to 45 percent,
were mixed as to species composition. Site 20758 was dominated by Bromus tectorum; the other
two sites had virtually no exotic species.
Each of these ratios indicates differences between some sites in this study. The ratios also
indicate that disturbance has occurred on some sites. Deviation from expected values in any one
analysis is not necessarily an indication of unacceptable conditions; however, deviations in multiple
parameter ratios are probably indicative of criteria that will distinguish nominal conditions from
subnominal ones. It should be noted that soil erosion (Section 5.3) was not excessive on any of
the sites using the current erosion indicator.
5.1.2 Recommendations
The analyses of the data from the 1992 pilot study demonstrate^ that some of the kinds of
data collected are useful as indicators. It is recommended that additional sensitivity studies be
conducted. With a larger data base and calibration data sets from sites of known condition, the
kinds of analyses presented in this report will be expanded to include multivariate techniques and
robust analysis based on non-normal distributions.
Existing soils, plant, and ecological site ancillary data should be compiled as part of this
larger data base. This data base would include data such as palatable species for livestock and
wildlife, poisonous plants, rare plants, soil maps, soil chemistry, and other pertinent data that are
available. These types of data will be important in indicator development and interpretation of
results.
5.2 SPECTRAL PROPERTIES INDICATOR
Investigations for the spectral properties indicator were conducted by different methods
aimed to test and evaluate the use of spectral measurements to derive information about the
spectral properties of vegetation and soils in arid ecosystems. The purpose of this study was to
compare Landsat TM spectral data with concomitant spectral measurements obtained through the
use of a portable handheld spectroradiometer (PSII). Thus, spectral properties of objects .in arid and
45
-------
70 -
60 -
50^
£ :
1 i
0 40 :
* '
DL :
0 30 -
CO
£ ;
H
js
20 -
10
0
O = Conifer Woodland
=Desertscrub
0
o
-------
semiarid environments could be compared to measured properties of those same objects to
determine if remote sensing can be used to derive significant and meaningful information about
environmental factors. Data obtained through remote means have the potential to allow
evaluations to be conducted on scales that would be cost effective in the vast areas of the western
United States.
5.2.1 Results
Spectral measurements were made at 13 sites; one of these sites was measured twice as a
crew comparability exercise and only the second set of measurements was included in the data set
for analysis. Overcast conditions resulted in low numbers of spectra being obtained from two sites,
and no measurements were made at three other sites (Table 5-1) due to problems with the
spectrometer necessitating the acquisition of a replacement instrument. During collection of
ground-based spectral measurements the initialization procedure for the Personal Spectrometer was
done in such a way that the white reference standard was saturated. To compensate for this error,
the data set was recalculated using values for reflectance in red and infrared wavelengths from a
soil spectrum from each transect or circular subplot. As a result of this error, the PSII data are
biased low and have more variables than normal. This data quality issue is discussed in more detail
in Section 7.
The Landsat scene did not include sites 20755 and COF101; therefore, the NDVI values
obtained from the groundbased and TM data were compared for a total of 11 sites. Seven of the
11 sites used to compare satellite and ground-based NDVI values were desertscrub with a
composition of varying species and the remaining four were conifer woodland with different shrub
components. Overall site productivity was determined by correlating vegetation indicator species,
physical site characteristics (e.g., elevation and slope), and soils data with potential vegetation
species and potential production in normal years as documented on Soil Interpretation Records in
the Soil Conservation Service National Soil Database. Sites were grouped by vegetation
productivity classes of > 1,000 kg/ha, 500 to 1,000 kg/ha, and <500 kg/ha and are listed in Table
5-2 with dominant plant"species, productivity class, and soil characteristics. Total vascular plant
cover was measured in alternate 1-by 2-m quadrats along each of the six transects per site and an
average site value obtained (Table 5-3).
The NDVI was ^calculated for the 1992 pilot study using both Landsat TM and PSII
reflectance data. In the case of the Landsat TM data, 16 pixels in a 4 by 4 matrix covered 1
hectare of the sample site. The NDVI was calculated for each' pixel and averaged to give a site
value (Table 5-4). For PSII measurements, seven circular subplot values of NDVI were calculated,
each an average of 16 spectral measurements. Six transect NDVI values, each an average of 18
spectral measurements, were also determined from PSII measurements. These 13 NDVI values
were averaged to provide a site NDVI (Table 5-5).
One of the primary intentions of this study on spectral properties was to examine the
correlations between satellite and ground-based spectral measurements using NDVI, as well as
47
-------
TABLE 5-1. THE NUMBER OF SPECTRAL MEASUREMENTS MADE AT EACH SITE
Site number
Number of spectra
acquired
20753
20755
20758
20907
20908
21911
20913
21060
21061
21210
21213
21216
21217
21361
21362
21364
21366
COF101
220
18
220
0*
220
120
220
172
184
220
0*
46
220
0*
0**
124
0**
220
* No data due to instrument failure
* * No data due to overcast conditions
between NDVI values and vegetation and soil variables. It is important to establish the correlation
between satellite and ground-based measurements to determine whether ground-based
measurements can be used in future studies to calibrate satellite data. The expected results would
be high correlation between ground-based and satellite derived NDVI and correlation of both TM
and PSII NDVI, to varying degrees, with ground-based measurements. Further, it was expected
that the conifer woodland sites would have higher NDVI values than desertscrub sites, but that
some overlap in values would exist as a result of different productivity levels. It was originally
planned to compare NDVI with site productivity classes. However, the low sample number in each
productivity class precluded performing this analysis.
The NDVI from both TM and PSII were plotted against each other, and, as expected, show
a high correlation (r2 = 0.71} (Figure 5-6). Although the PSII values are biased low as previously
discussed, these results are encouraging and show that ground-based spectral measurements may
be used to calibrate satellite data. Figure 5-7 shows the relationship between the PSII derived NDVI
and total vascular plant cover. Figures 5-8 through 5-10 show the relationship between the TM
derived NDVI and total vascular plant cover; gravel, cobbles, stones, and bare soil cover; and moss,
lichen, and cyanobacteria cover, respectively. Only one graph showing the relationship between
PSII-denved NDVI and selected parameters is shown as the relationship between PSII NDVI and
selected variables is similar to that of TM NDVI and those same selected variables. Low
productivity desertscrub sites have the lowest NDVI values for both TM and PSII, as expected.
48
-------
TABLE 5-2. DOMINANT VEGETATION AND SOIL CONDITIONS AT SITES FROM WHICH
SPECTRAL MEASUREMENTS WERE OBTAINED DURING EMAP-ARID 1992 PILOT
Site 20753, Dolores Point: medium productivity desertscrub dominated by Artemesia spp.r Stipa
columbiana, Agropyron smith!/, Vicia americana, Gutierrez/a sarothrae, Chrysothamnus spp., Pinus
edulis, and Juniperus osteosperma. The soil surface is dark and light red/brown in color, and was
moist at the time ground-based spectral measurements were made.
Site 20755, Island Canyon: low productivity conifer woodland dominated by Pinus edulis,
Juniperus osteosperma, Artemesia ludoviciana, and Opuntia spp. The soil surface is light red with
white sandstone cobbles and was dry when spectral measurements were made.
Site 20758, Tin Cup Mesa: medium productivity desertscrub dominated by Gutierrezia sarothrae.
Hi/aria, Salix spp., and Bromus tectorum. The soil surface is light red/brown in color, and was dry
when measurements were made.
Site 20908, La Sal Junction: low productivity conifer woodland dominated by Pinus edulis,
Juniperus osteosperma, and Artemesia spp. The soil surface is dark red/brown in color and was
very wet when spectral readings were taken.
Site 20911, Mustang Flat: medium productivity conifer woodland dominated by Pinus edulis,
Juniperus osteosperma, Artemesia spp., Gutierrezia sarothrae, and Vulpia octoflora. The soil
surface is medium to dark red/brown in color and was dry when measurements were made.
Site 20913, Gray Spot: medium productivity desertscrub dominated by Ephedra spp., Oryzopsis
hymenoides, Salix spp., Chrysothamnus spp., and Gutierrezia sarothrae. The soil surface is light to
medium red/brown in color and was dry when spectral readings were made.
Site 21060, Kane Springs Mesa: low productivity desertscrub dominated by Coleogyne
ramosissima, Gutierrezia sarothrae, Ephedra spp., and Juniperus osteosperma. The soil surface is
red and was dry at the time when spectral measurements were made.
Site 21061, Six Shooter Peak: medium productivity desertscrub dominated by Hi/aria, Gutierrez/a
sarothrae, Oryzopsis hymenoides, Sphaeralcea coccinea,*- and Atrip/ex confertifolia. The soil surface
was dry when measurements were made.
Site 21210, Valley City: medium productivity desertscrub dominated by Sarcobatus vermiculatus,
Halogeton glomeratus, Hi/aria, and Brassica spp. The soil surface is red/brown and light gray/brown
and was dry when readings were taken.
Site 21216, Cedar Mesa: conifer woodland dominated by Pinus edulis, Juniperus osteosperma,
Chrysothamnus spp., and Ephedra spp. The soil surface is light, medium, and dark red/brown and
was moist when the spectral readings were made.
Site 21217, Halchita: low productivity desertscrub dominated by Gutierrezia sarothrae. Hi/aria,
Ephedra spp., Coleogyne ramosissima, and Compositae. The soil surface is light red/brown in color
and was dry when measurements were made.
Site 21364, Cataract Canyon: low productivity conifer woodland dominated by Pinus edulis,
Juniperus osteosperma, Coleogyne ramosissima, Gutierrezia sarothrae, and Bromus tectorum. The
soil surface is light to medium red/brown in color and was dry when spectral readings were taken.
Site COF101, Dove Creek: low productivity conifer woodland dominated by Pinus edulis, Juniperus
osteosperma. Hi/aria, and Chrysothamnus spp. The soil surface is yellow/brown, gray/brown, and
red/brown and was dry when spectral measurements were made.
49
-------
TABLE 5-3. TOTAL VASCULAR PLANT COVER AT EACH SITE
Site
number
20753
20758
20908
20911
20913
21060
21061
21210
21216
21217
21364
COF101
Vascujar
plant
cover
43.11
37.54
18.13
15.96
34.75
15.77
26.18
62.00
18.25
14.98
21.48
35.16
Sample
standard
deviation
5.98
5.04
8.69
6.79
6.72
6.63
7.39
10.54
6.85
5.18
7.06
15.25
TABLE 5-4. NDVI VALUES DERIVED FROM LANDSAT TM DATA FOR
EMAP-ARID 1992 PILOT STUDY
Site
number
20753
20758
20908
20911
20913
21060
21061
21210
21216
21217
21364
Average
site
NDVI
n = 16
0.367
0.243
0.287
0.105
0.219
0.173
0.210
0.199
0.285
0.199
0.194
Site
sample
standard
deviation
0.044
0.014
0.032
0.016
0.006
0.012
0.010
0.014
0.009
0.007
0.008
50
-------
TABLE 5-5. NDVI VALUES DERIVED FROM PERSONAL SPECTROMETER (I DATA
COLLECTED DURING EMAP-ARID 1992 PILOT STUDY
Site
number
20753
20755
20758
20908
20911
20913
21060
21061
21210
21216
21217
21364
COF101
* Only three averaged
* * M r\ r\ at a r\ 1 1 a +r\ » * it*.
-------
.38
__
§
O = Conifer Woodland
^Desertscrub
TM NDVI =» 1.086(PS II NDVQ + .145
Unexplained low TM value
O Not used in regression
I I i i i I i i i I i i I >< I i i i I
.12 .14
, 10 -
PS II NDVI
Figure 5-6. Correlation between NDVI derived from Landsat TM data and that from PSII data.
.16
52
-------
CO
Q_
.38
.36
.34
.32
.30
.28 :
.26 :
.24
.22
.20
. 18
. 16
. 14 -
. 12 :
.10 -
.08 :
.06
.04 H
.02
O = Conifer Woodland
=Desertscrub
o
o
o
o
10 20 30 40 50 60 70
Total Vascular Plant Cover (%)
'"f
Figure 5-7. Relationship between NDVI PSII and total vascular plant cover.
53
-------
.38
.36
.34
.32
.30
.26
.24
.22
.20
. 16
. 14
12'
. 10
.08
.04
02
O ssConifer Woodland
Desertscrub
~
10
' 1 p
20
' 1 '
30
' ! '
40
50
60
11
70
Total Vascular Plant Cover (%)
%
Rgure 5-8. Relationship between NDVI TM and total vascular plant cover".
54
-------
.38
.36
.34
.32
30
28
26
.22-
20
. 18 :
.16-
. 14 -
. 12 -
. 10 :
.08 :
04
O
o
O
02
= Conifer Woodfand
=Desertscrub
o
I I I I I I I I I I I I I I I I I I I | I I I I I I ! I I I I I I I I I I I I I I I I I I I I I I I I I I I
I I I I I I I : I I I
0 10 20 30 40 50 60 70 80
Sum of Gravel, Cobble, Stone, and Bare Soil Cover (%)
Figure 5-9. NDVI-TM versus the sum of interspace gravel, cobble, stone, and bare soil
cover.
55
-------
-g
1
1
j
.37-
.36 ;
.35-;
.34 '-.
.33J
.32-J
.31-;
.30-;
.29-;
.28-;
.27-J
.26-;
.25-;
.24 \
.23-;
.22-1
.21 '-.
.20 \
.19-;
.17:
. 16:
. 14
.13
.12
.11
. 10
.09
.08 -
.07 -
'.06-
.05 -
.04 -
03 -
02 -
O = Conifer Woodland
=Desertscrub
0 Q
9
'" ':, ' " -
* :
* -^ o .
o
i
!
1
i
!
t I 1 1 t 1 1 1 1 1 [ 1 1 1 1 1 1 1 1 1 j 1 1 1 1 1 1 1 1 1 | 1 I i i i i i i | ''"'' ! ' | I 1
n 10 20 30 40 50 60 70 8C
Sum of Moss, Lichen, and Cyanobacteria Cover (%)
Figure 5-10. NDVI TM versus the sum of interspace moss, lichen, and cyanobacteria cover.
56
-------
5.2.2 Recommendations
During planning for the 1992 pilot study field work, the EMAP-Arid researchers decided that
the acquisition of spectral measurements should be confined to the transects along which
vegetation assessments were made and to a series of measurements in a square grid in the circular
subplots where tree data were collected. In addition, with only one spectral technician, the arid
team felt that this individual should not acquire detailed descriptions of surface cover
commensurate with the spectral measurements or specific measurements of land cover
components. As a result of these decisions no reference spectra of soils, rocks, cyanobacteria, or
representative plant species were made, and only minimal notes were taken concerning the
composition of each individual spectrum. This procedure limited the correlation that could be made
between spectral measurements and other variables collected in the field and therefore diminished
the ecological significance that could be drawn from ground-based spectral data.
The strong correlation between the TM and PSII NDVI-derived for the 1992 study indicates
that the overall concept of remote sensing as an indicator of ecosystem condition is valid, but that
more research is needed into indicator selection and integration. In addition, results of the study
suggest that NDVI alone may not be the most sensitive indicator for arid and semiarid regions due
to sparse vegetation cover. Other vegetation indices have been derived from spectral data (Huete
and Jackson, 1987, Perry and Lautenschlager, 1984), such as the soil adjusted vegetation index
(SAVI). The SAVI, in addition to total surface reflectance (albedo) and other indices, will be
evaluated as part of the future studies. Additional on-site information should be collected for
reference spectra of soils, rocks, cyanobacteria, and representative plant species.
5.3 SOILS INDICATOR
The sensitivity of soil as an indicator of ecological condition was evaluated as part of the
1992 pilot study. A preliminary attempt was made to compare soil profile characteristics and soil
physical and chemical properties collected at the study sites to the USDA Soil Conservation Service
(SCS) Soil Interpretation Racords (SIR). Since the SIR was determined in the field at the time of
initial mapping of the soil by the SCS, it provides an historical basis and independent assessment
comparison with present day measured features through time. As a result of this investigation,
many issues concerning the use of ,soils as a means to interpret vegetation and spectral indicators
and as an indicator of condition were uncovered. The following subsections describe the results of
this investigation and give recommendations for further study.
5.3.1 ResultsSoil Profile Description
Eighty-one soil pits at 26 sites were described and characterized as a part of the 1992
EMAP-Arid Pilot Study. These soils were categorized into four major land resource areas (MLRAs).
The MLRAs are geographically associated land resource units used in agricultural planning
(USDA-SCS, 1981). Most sites were in MLRA 35, Colorado and Green River Plateaus, but a few
were in MLRA 34, Central Desertic Basins, Mountains, and Plateaus, MLRA 39, Arizona and New
Mexico Mountains, and MLRA 48A, Southern Rocky Mountains. Forty-six pits were located on
mesa and bench summits, 27 on sideslopes of alluvial fans, hillsides, and fluvial terraces, and 8 pits
were located on local interfluves. Most of the soils were classified in the Aridisol soil order, and a
few were classified in the Entisol, Alfisol, and Mollisol soil orders. All of the sites had a mesic
57
-------
temperature regime except site 20752 which had a frigid temperature regime. Many of the sites
had an aridic moisture regime, seven of the 26 sites had an ustic moisture regime, and a few were
transitional between aridic and ustic. The elevation of the sites ranged from 1,300 to 2,300 m.
The average site slope was about 10 percent. The steepest sites were the woodland sites with
slopes of 3 to 40 percent. Most of the desertscrub sites had slopes less than 7 percent.
At least one soil pit and three auger holes were excavated at each site to describe soil
characteristics, draw a soil map, collect samples, and compare the described soils to the soil
survey map unit and components. The main goal of this effort was to obtain baseline soil
information and determine how well site-specific data compare to values in the soil interpretation
record (SIR) of the USDA-Soil Conservation Service. The soil descriptions used to determine the
soil series or family phase at each site are shown in Appendix B. The soil series phase was used as
a key to access the SIR. The SIR contains expert estimates of soil properties and interpretive
information about each soil series phase and their associated soil horizons and can be used to
evaluate field measured properties. The soil map was cut with scissors along soil boundaries if
more than one soil series phase was found at a site and each portion of the total map weighed to
determine the percentage of soil series at each site. The estimated SIR soil properties were
weighted by multiplying by the estimated soil series percentage for each site. The weighted data
were summed to obtain the average property value for the site.
The field soil classification was verified by collecting field samples, performing physical,
chemical and mineralogical analyses, and evaluating the resulting laboratory data. Thirty-two of the
approximately 18,000 United States soil series were recognized in the field. If the same soil series
was identified in the field at a given site at the second or third soil sampling area, only the first two
soil horizons were sampled resulting in incomplete verification of classification of 25 percent of the
pedons due to insufficient horizon data (Appendix B). However, unverified soil identifications were
presumed to be correctly identified in the field and were presented in the data set as the field
identified soil series. Nine series were described and sampled multiple times. These soils occurred
at multiple sites and represent soils of large areal extent. For example, the Rizno soil was sampled
13 times. The Rizno series, which is shallow and has two horizons or less, is a commonly
occurring soil in arid environments.
To evaluate the possibility of using published soil survey information to determine soil
parameters and to develop baseline data, the soils described at the site by the field crews were
compared to published soil surveys, SIRs, and series descriptions. Three sites were in areas where
soil surveys were not published. Of the 23 remaining sites, 16 included at least one soil series or
component of the published map unit. (A map unit is a conceptual group of delineations identified
by the same name in a soil survey that represents similar landscape areas comprised of one or
several soil series, plus inclusions.) Two sites were in miscellaneous rock land units with
insufficient soil information. Four sites had soils which were different from those indicated in the
published soil survey. More than half of the sites were in soil complexes where different soils are
mapped together in a map unit. Therefore, since the soils mapped in soil surveys are not mapped in
sufficient detail to be site specific, a trained soil scientist is needed in the field to correctly identify
the soils at each EMAP-Arid sampling site. The value of correct identification of soil is that
information about the soil which cannot be obtained by sampling may be located in the literature.
58
-------
5.3.2 Results-Soil Quality
Soil quality is best defined as the capacity of a soil to promote the growth of plants; protect
watersheds by regulating the infiltration and partitioning of precipitation; and prevent water and air
pollution by buffering potential pollutants such as agricultural chemicals, organic wastes, and
industrial chemicals. The quality of a soil is determined by a combination of physical, chemical, and
biological properties such as texture, water-holding capacity, porosity, organic matter content, and
depth (NRC, 1993).
Soil properties near or at the soil surface were evaluated during the 1992 pilot study as
candidate indicators. Soil samples were collected from three subplots located to intersect the drip
line of the canopy cover. The drip line was chosen as a sampling area to provide consistent
sampling protocol and try to obtain a mixed sample of interspace and undercanopy soil. The top
two mineral soil horizons were sampled and analyzed for particle-size distribution, organic carbon,
cation exchange capacity, soluble salts, bulk density, and water retention {Table 5-6). For each of
these parameters, a site average was computed by averaging the two mineral horizons from each
pit and multiplying by the percent soil series at each site. Table 5-7 shows the maximum,
minimum, median, and 75th and 25th quartiles for all horizons.
In order to provide a preliminary assessment of the soil quality at each site, the measured
properties were compared to pre-existing soil property values for the soil series encountered. The
published soil property values were obtained from the soil interpretation record (SIR) of the
USDA-Soil Conservation Service. For each soil taxon that is recognized by the SCS through
correlation in a soil survey area, important physical and chemical properties and indices of each
major soil horizon or combination of similar horizons have been estimated (USDA-SCS, 1983). The
SIR contains expert estimates of soil properties and interpretive information about each soil series
and the associated soil horizons. A soil series may have multiple SIRs; therefore, the soil series at
each site must be matched to the appropriate SIR. Unfortunately during this pilot, the appropriate
SIR number was not recorded at the time of sampling. Therefore, a match was made
during sample analyses according to the best judgment of the soil survey staff at the National Soil
Survey Center Laboratory in Lincoln, Nebraska. During future EMAP-Arid field surveys, the SIR will
be recorded at the time of sampling.
Estimated soil properties in the SIRs were compared to the measured soil parameter.
Averages for the estimated soil properties were computed for each site by matching the soil series
phase to its appropriate SIR, extracting the parameter of interest, and adjusting for the percentage
of that series phase present at the site. The SIR properties are reported as a range of values. The
midpoint of that range for a given property was determined and used in the comparisons.
The difference between the average SIR and measured property values was evaluated using
the t-test (Snedecor and Cochran, 1967) to evaluate the potential use of SIR values to indicate
status and trends in soil quality. The results are discussed for particle size distribution, organic
carbon, cation exchange capacity, soluble salts, bulk density, and water retention difference in the
following paragraphs. Analytical methods are given in Table 3-3.
59
-------
TABLE 5-6. SOIL SURFACES ANALYSES8
Site
20752
20755
20907
20908
20911
21064
21211
21214
21215
21363
21365
COF101
20753
20758
20913
21060
21061
21062
21210
21212
21213
21217
21361
21362
21364
21366
Form
C
C
C
C
C
C
C
C
C
C
C
C
D
D
D
D
D
D
D
D
D
D
D
D
D
D
Bulk
Density
g/cnn3
1.14
1.30
nd
1.35
1.38
1.33
1.40
1.18
1.16
1.36
1.37
1.22
1.32
1.40
1.52
1.44
1.48
1.36
1.36
1.43
nd
1.33
1.44
1.36
1.47
1.34
CEC/
Clay
0.62
0.29
0.04
0.51
0.63
0.57
0.51
0.66
0.52
0.62
0.23
0.38
0.60
0.73
1.00
0.26
0.65
0.42
0.39
0.35
0.61
0.67
0.48
0.36
0.74
0.69
CEC
cmol/kg
18.83
7.75
3.53
11.47
13.13
4.16
5.55
10.08
19.46
2.95
8.29
15.82
13.65
11.10
3.94
5.46
4.74
7.33
13.08
4.80
2.22
4.23
3.21
5.16
2.51
4.26
Clay
%
13.42
7.24
4.15
12.24
16.45
3.88
6.41
9.52
18.90
2.85
20.16
15.37
15.96
12.81
3.45
11.69
6.97
8.60
24.68
14.32
3.32
6.52
8.02
13.27
2.65
5.86
EC
dS/m
0.92
1.20
0.98
0.81
1.52
1.14
2.77
0.90
1.12
0.90
0.82
1.69
1.25
0.69
0.59
0.98
0.98
1.44
4.32
1.49
0.00
1.04
0.81
0.52
0.00 .
1.27
SAR
0.12
0.14
0.13
0.11
0.11
0.12
1.04
0.07
0.05
0.18
0.26
0.21
0.07
0.22
0.15
0.07
0.17
2.11
25.42
5.91
0.00
0.06
1.06
0.17
0.00
1.35
OC
kg/m2-dm
1.88
2.44
1.49
1.67
1.35
0.80
1.32
1.69
2.75
0.45
1.29
2.65
1.80
0.72
0.33
0.79
0.46
1.55
1.97
0.26
0.41
0.34
0.14
0.17
0.53
0.20
WRD
cm/dm
0.81
0.51
nd
1.04
1.53
1.10
0.70
1.07
1.19
0.96
0.75
1.17
1.93
1.20
2.51
0.81
0.89
1.93
1.14
1.00
nd
1.32
0.60
0.74
0.86
0.37
Form: Vegetative forms are C-conifer woodland and D-desertscrub; CEC = cation exchange
capacity; CEC/clay = corrected for CEC of organic carbon; EC = electrical conductivity; SAR =
sodium adsorption ratio; OC = organic carbon; WRD = water retention difference; nd = not
determined.
60
-------
TABLE 5-7. QUARTILES FOR SOIL SURFACE ANALYSES0
Quantity
Maximum
75 quartile
Median
25 quartile
Minimum
g/cm3
1.71
1.46
1.37
1.29
0.72
CEC/
Clay
1.86
0.67
0.52
0.33
0.01
CEC
cmol/kg
40.1
10.0
5.7
3.9
0.7
Clay
%
36.9
14.1
8.9
4.4
0.4
EC
dS/m
19.80
1.24
0.97
0.71
0.45
SAR
132.1
0.3
0.1
0.1
0.0
OC
kg/m2-dm
6.09
1.47
0.59
0.21
0.00
WRD
cm/dm
3.79
1.33
1.02
0.62
0.19
aCEC = cation exchange capacity; EC = electrical conductivity; SAR = sodium adsorption ratio;
OC = organic carbon; WRD = water retention difference.
5.3.2.1 Particle Size Distribution (Soil Texture)--
Particle size distribution is the fraction of sand, silt, and clay in the soil. The clay fraction
is defined as the particle-size class less than 0.002 mm. Physical and chemical activities of a soil
are related to the kind and amount of clay (USDA-SCS, 1983). If clay is lost preferentially (due to
wind and water erosion) from a soil leaving behind higher contents of sand and silt, then the
nutrient and water-holding capacity of a soil would decrease. Soil texture influences plant growth,
via water holding capacity and nutrient supply capacity, and is often used as a clue to how soils
have formed. Further, the content of the particle size separates is often used as model parameters
and is an essential part of site characterization. The proportions of particle sizes indicate sediment
history and differential erosion rates.
In the 1992 pilot study there was a significant difference (alpha = 0.01) between the SIR
and measured clay content values. The SIR overestimated the amount of clay measured at nearly
every site regardless of whether the vegetation class was woodland or desertscrub (Figure 5-11).
In four cases, where clay content differences were greater than 10 percent, a degradation of the
system is indicated above and beyond the error commonly associated with estimating particle
distribution or texture by feel. In contrast, sites 21215 and 21210 appear to be accumulating clay
and may be located in a depositional topographic setting. One possible explanation is that
preferential removal of clay by wind and water erosion has occurred at the sites since the original
properties were estimated for the SIR. Another possibility may be that the soil series phase was
mismatched with the SIR or that inaccurate estimates are reported in the SIR. Because of these
discrepancies, it appears that, for future EMAP-Arid field surveys, soil samples should be collected
so that measurements of site-specific particle size distribution can be made.
5.3.2.2 Organic Carbon-
Organic carbon (OC) in soil samples is measured to estimate the organic matter composed
of plant and animal matter in various stages of decomposition in the soil. Organic matter is
estimated by multiplying the OC by the constant 1.72. The OC generally is the most chemically
reactive fraction of the soil and is capable of holding moisture within the soil. High OC indicates a
large nutrient pool, sustainable fertility, and C-sequestration. The OC generally correlates with
61
-------
Clay
Difference of SIR and Measured
co
0)
CO
jo
O
20.0
15.0
10.0
5.0
0.0
-5.0
-10.0
t
I
h
woodland-
desert scrub
20752 20907 20911 21211 21215 21365 20753 20913 21061 21210 21213 21361 21364
20755 20908 21064 21214 21363 COF10 20758 21,060 21062 21212 21217 21362 21366
SIR - Soil Interpretation Record 5 it8
Figure 5-11. Clay-difference between SIR estimate and amount measured at site.
62
-------
vegetation types and amounts and integrates long-term trends in vegetation. Data collected for
each horizon were converted to a volume basis for the first 10 cm (1 dm) or to the depth of the
soil, if shallower, using the following formula:
OC (kg/m2-dm) = OC x bulk density x % < 2 mm x 0.1 x horizon thickness.
The range of values was from 0 to 6.09 kg/m2-dm. Conifer woodlands were higher than
desertscrub due to higher biomass at the woodland sites. When compared to the measured OC,
the SIR OC contents significantly underestimated the OC content at the sites indicating the possible
accumulation of organic matter since the SIR was originally determined (Figure 5-12). This
difference is noticeable for the woodland sites because of the accumulation of pine needles even
though the organic horizons were excluded from the calculation. In contrast, the OC content is
higher on desertscrub sites where vegetative cover is higher than average. The measured ,OC value
as compared to the SIR indicates a slight degradation (i.e., loss of OC) of the desertscrub sites,
except at sampling locations 21062 and 21210 (perhaps due to several drought years in the
Colorado Plateau). The SIR estimated average value for OC is not a good indicator of trend or
status. Thus, for future EMAP-Arid surveys, the soil surface horizon should be sampled and OC
measured to determine status and trend.
5.3.2.3 Cation Exchange Capacity (CEC)--
The CEC can be defined as the sum of the exchangeable cations that a soil or soil
constituent can absorb at a specific pH. Decreases in CEC indicate a loss of nutrient storage.
Additionally, the CEC-to-clay ratio can be used to estimate clay mineralogy and clay activity.
Changes in the CEC-to-clay ratio indicate weathering and soil development. There was no
significant difference between the SIR and measured values for all sites regardless of vegetative
cover type (Figure 5-13). However, for most desertscrub sites, the measured CEC was lower than
reported in the SIR. In two cases, where conifer woodlands were identified, the SIR markedly
underestimated the CEC. At both of these sites, the SIR also underestimated the organic carbon
content and clay content (figures 5-11 and 5-12) indicating the strong direct relationship between
these three parameters. Thus, for first visits to a site, the CEC should be measured as baseline
data. Any changes in CEC would be detected by changes in clay and organic carbon contents
because they can be correlated with CEC identified on subsequent visits.
5.3.2.4 Soluble Salts-
Electrical conductivity (EC) and the sodium adsorption ratio (SAR) are indicators of salinity
(soluble salts) or sodium (Na) accumulation. The lower limits of EC and SAR that affect salt
sensitive plants are 0.4 Siemens per meter and 13, respectively. Soils with a high SAR and low EC
have poor soil structure that adversely affects plant growth. Only the Ravola soil series at site
21210 had high EC and SAR levels. The high EC values were reported on the SIR but not the high
SAR values (Figure 5-14 and Figure 5-15). In general, the EC was underestimated in woodland
sites and overestimated for desertscrub sites indicating that either (1) salts are accumulating in the
woodlands and being lost in the desertscrub environments or that (2) the original estimates were
inaccurate. When comparing the EC and SAR measured values to the SIR values, the degradation
63
-------
Organic Carbon
Difference of SIR and Measured
1.0
0.5
W
o> o.o
CO, -0.5
E
£ -1.0
8-1.5
-2.0
I
woodland
1
1
I
desert scrub
20752 20907 20911 21211 21215 21365 20753 20913 21061 21210 21213 21361 21364
20755 20908 21064 21214 21363 COF10 20758 21060 21062 21212 21217 21362 21368
SIR - Soil Interpretation Record Olte
Figure 5-12. SIR organic carbon versus measured OC at site.
64
-------
Cation Exchange Capacity
Difference of SIR and Measured
6.0
4.0
^ 2.0
^
-------
1.5
1.0
f °-5
3
U>
So.o
£-0.5
CO
-1.5
-2.0
Electrical Conductivity
Difference of SIR and Measured
i
nt
IO1
woodland
desert scrub
20752 20907 20911 21211 21215 21365 20753 20913 21061 21210 21213 21361 21364
20755 20908 21064 21214 21363 COF10 20758 21060 21062 21212 21217 21362 21366
SIR - Soil Interpretation Record
Site
Figure 5-14. Difference between SIR and measured values for electrical conductivity at 1992 pilot
study sites.
66
-------
Sodium Adsorption Ratio
Difference of SIR and Measured
10.0
5.0
"8 o.o
CO
05
-5.0
a:
-20.0
-25.0
Ajll
woodland <^ ^> d<
1
ert scrub
20752 20907 20911 21211 21215 21365 20753 20913 21061 21210 21213 21361 21364
20755 20908 21064 21214 21363 COF10 20758 21060 21062 21212 21217 21362 21366
SIR-Soil Interpretation Record
Figure 5-15. Difference between SIR and measured values for sodium adsorption ratio at
1992 pilot study sites.
67
-------
of site 21210 is clearly visible as salts are accumulating at this site. Additionally, increased salt
content was identified at site 21211, but not of the sodium-based salts, indicating a potential for
degradation (assuming continued salt increase and accurate SIR estimates) at that site. In the
future, field methods should be used to measure EC at each site since increasing EC contents
indicate a threat to plant communities.
5.3.2.5 Bulk Density-
Bulk density is the mass of oven-dry soil per unit bulk volume. Volume is determined
based on a sample equilibrated at a moisture tension of -33 kPa. Bulk density is used to convert
weight-based data to volume-based data. Bulk density is related to soil texture. Bulk density
values for mineral soils typically range from 1.3 g/cm3 for clayey soils to 1.7 g/cm3 for sandy soils.
Bulk density values higher than those in the typical range indicate soil compaction which alters
water transmission rates and impedes root penetration. For example, 90 percent of roots are
impeded if the sand content is greater than 75 percent and bulk density is greater than 1.8 g/cm3
or if the clay content is greater than 35 percent and bulk density is greater than 1.50 g/cm3 (Jones,
1983). No horizons had a bulk density high enough to impede root growth.
In the pilot study, bulk density was not measured at all sites for all samples due to failure
to obtain a natural clod or loss of sample integrity during shipping or analysis. Where bulk density
was missing (two sites as indicated by nd in Table 5-6), values were estimated using DRAINMOD
algorithms (Baumer, 1989) for comparison purposes with the SIR values.
A comparison of the bulk density between SIR and measured data indicates that 77
percent of the sites were within 0.1 g/cm3 (Figure 5-16). The SIR values overestimate the
measured bulk density for conifer woodland sites while they underestimate the bulk density for
desertscrub sites. One possible explanation for this finding is that the increased organic carbon
contents (relative to the SIR estimates) at the woodland vegetation sites is leading to an
improvement in the soil at these sites (a "loosening" of the soil). In contrast, the relative loss of
OC at the desertscrub sites has led to an overall increase in the bulk density in this ecosystem.
More research is needed to determine significant threshold differences between SIR and measured
values. However, the characterization data are adequate to determine baseline status. To
determine trends and status, bulk density measurements need to be made at or near the soil
surface.
5.3.2.6 Water Retention Difference-
The portion of water in soil that can be absorbed by plant roots is defined as the available
water. This portion is estimated by determining the difference of water contents at 33 kPa and
1,500 kPa (WRD). The WRD affects plant growth and species composition and is affected by
anthropogenically and naturally induced stresses such as erosion and compaction. A reduction in
WRD would indicate that the ability of the soil to provide water to plants is being adversely
impacted. The actual amount of water stored is a function of soil storage capacity and climate.
The water retention difference computations were converted to a volume basis to a depth of 10 cm
(1 dm) or to the depth of the soil, if shallower, using the following formula (where Db is bulk
density):
68
-------
Bulk Density
Difference of SIR and Measured
0.4
0.2
§
CO
CO
0)
0.0
CO
.a
Q
-0.2
-0.4
LI
i
woodland
desert scrub
I
20752 20907 20911 21211 21215 21365 20753 20913 21061 21210 21213 21361 21364
20755 20908 21064 21214 21363 COF10 20758 21060 21062 21212 21217 21362 21366
SIR - Soil Interpretation Record
Site
Figure 5-16. Difference between SIR and bulk measured values for bulk density at
1992 Pilot Study sites.
69
-------
WRD - (33 kPa - 1,500 kPa) x Db x horizon thickness x %<2 mm
The range of WRD capacity for the top 10 cm at 1992 pilot study sites was from 0.19
cm/dm to 3.79 cm/dm. There is no significant difference between the SIR WRD and measured
WRD values. However, at site 20913 a marked increase in the difference between the SIR and
measured WRD value was noted indicating an improved water holding capacity at this site. Use of
the SIR to evaluate the status of the WRD of soil quality is an acceptable approach. While no
significant difference was determined, the range of SIR WRD values may be too wide to detect any
trends. In order to determine if the WRD is increasing or decreasing, repeated measurements
should be made. The WRD should be correlated with the vegetation data to evaluate the plant
community with respect to the potential water resource.
5.3.2.7 Summary of Soil Quality Results-
In general, use of the SIR to evaluate status of individual properties of soil quality is an
acceptable approach to obtain a reference value of the soil characteristics at EMAP-Arid sites.
More research is needed to combine individual properties into a single indicator of soil quality.
However, to establish trends in soil quality changes, repeated sampling of the same sites at
specified intervals of time will be required.
5.3.3 ResultsErosion Index Measurements
The erosion rates calculated for the 1992 pilot study sites are given in Table 5-8. The
cyanobacteria (cyano) was not considered in the computation of C, the cover and management
factor. The average C factor calculated by including half of the cyanobacteria amount is reduced
by about 40 percent. All the calculated erosion rates are substantially less than the soil T values in
the SIR indicating that acceptable erosion levels were occurring at the site and that little
degradation due to erosion was occurring. The T factor is an estimate and is traditionally used to
estimate acceptable soil loss for cropland. It may be inappropriate to use the value as a benchmark
in rangeland. Ten of the 26 sites were judged to be moderately eroded by soil scientists using
evidence of erosion such as rills, gullies, and desert pavement. The soil loss estimated by RUSLE
failed to identify these sites. The C factor was lower than expected and was the most important
(actor in yielding low erosion rates. Further investigation is warranted to assess the validity of the
C factor model. The SCS uses the RUSLE as the best current estimate of erosion rates; however it
is a lumped-model designed to meet long-term soil conservation planning needs on cropland and
may not be sensitive to single condition measurements as proposed by EMAP. For example, the C
factor varies temporally and calculated erosion rates will be higher when canopy cover is less.
Further testing of the C factor is recommended. The use of process-oriented models, such as the
Water Erosion Prediction Project (WEPP) (Lane and Nearing, 1989), should be tested to determine if
thoy are more suitable to determine trends and status.
70
-------
TABLE 5-8. SIR0 AND RUSLEa FACTORS AND ANNUAL AVERAGE EROSION RATES
Site
20752
20755
20907
20908
20911
21064
21211
21214
21215
21363
21364
21365
21366
COF101
20753
20758
20913
21060 .
21061
21062
21210
21212
21213
21217
21361
21362
Formb
C
C
C
C
C
C
C
C
C
C
C
C
C
C
D
D
D
D
D
D
D
D
D
D
D
D
Kc
(SIR)
0.24
O.25
0.17
0.29
0.31
0.23
0.27
0.29
0.21
0.16
0.24
0.17
0.14
0.24
0.31
0.18
0.29
0.25
0.19
0.33
0.45
0.15
0.34
0.29
0.29
0.25
Tc
(SIR)
2.50
1.00
1.00
1.50
5.00
1.50
1.49
2.15
2.78
1.10
1.36
1.04
1.00
1.25
5.00
3.32
5.00
1.00
4.80
5.00
4.57
3.09
3.56
1.00
1.00
1.27
Ad
(KSIR)
(t/a/yr)
0.02
0.01
0.02
0.03
0.04
0.05
0.02
0.01
O.O1
0.03
0.04
0.01
0.03
0.02
0.02
0.01
0.14
0.03
0.04
0.01
0.03
O.01
0.11
0.01
0.04
0.07
Kc
Eros
0.28
0.13
0.08
0.30
0.55
0.19
0.20
0.31
0.36
0.24
0.19
0.36
0.69
0.29
0.51
0.50
0.26
0.11
0.31
0.50
0.42
O.24
0.28
0.55
0.37
0.47
LSd
2.33
0.57
1.35
0.67
0.30
0.61
0.43
0.63
0.64
1.23
0.37
0.76
0.71
0.73
0.34
0.31
0.58
0.33
0.35
0.22
0.12
O.42
1.30
0.50
0.28
1.10
Cd
w/o
cyano
0.00
0.01
0.01
0.02
0.07
0.05
0.03
0.01
0.01
0.02
0.04
0.01
0.04
0.01
0.02
0.02
0.09
0.04
0.06
0.01
0.06
O.O1
0.04
0.01
0.05
0,02
cd =
w/0.5
cyano
0.00
0.01
0.01
0.01
0.01
0.01
0.01
0.00
0.00
0.01
0.01
0.01
0.03
0.01
0.00
0.01
0.02
0.02
0.01
0.01
0.06
O.OO
0.02
0.01
0.02
0.01
Rd
10.00
10.00
13.40
10.00
16.60
11.20-
16.50
10.00
9.20
13.20
10.00
10.00
10.80
10.00
10.00
10.00
10.00
10.00
9.60
10.00
10.00
8.5O
10.00
10.00
10.00
9.60
pd
1.00
1.00
1.34
1.00
1.66
1.12
1.65
1.00
0.92
1.32
1.00
1.00
1.08
1.00
1.00
1.00
1.00
1.00
0.96
1.00
1.00
O.85
1.00
1.00
1.00
0.96
Ad
(t/a/yr)
0.03
0.01
0.01
0.03
0.07
0.04
0.01
0.01
0.02
0.05
0.03
0.03
0.15
0.02
0.03
0.04
0.13
0.01
0.07
0.01
0.03
O.O1
0.07
0.02
0.05
0.13
Soil
erosion
class6
1
1.5
2
1
2
1
1
1
1.5
2
2
2
1:7
1
1.5
1
2
2
2
1
1
1
2
1.3
2
2
a SIR is the USDA Soil Conservation Service Soil Interpretation Records. RULSE is the Revised Universal Soil Loss
Equation.
b C = conifer woodland; D = desertscrub.
K = soil erodibility factors; T = soil loss tolerance; K(SIR) and T(SIR) are from the published SIR.
d A = average annual soil erosion; A(KSIR) is the soil loss calculated with IC(SIR) and LS, C, R, and P;
cyano = cyanobacteria; t/a/yr = tons per acre per year; LS = slope-length and slope-steepness factor;
C = cover and management factor; w/o = without; w = with; R = rainfall and runoff factor; P = support
practice factor.
e Soil Erosion Classes: 1 = slight, 2 = moderate, 3 = severe (estimated by soil scientists).
71
-------
5.3.4 Recommendations
Excavating as many as three pits per site to depths greater than 1 meter proved time
consuming. Fortunately, the soil identification and sampling at depth will not need to be repeated
if the same site is remeasured. It is anticipated that only surface horizons will need to be sampled
to show changes in the arid ecosystem at a given site. We recommend identifying the soils at
each site by auger observations, description,, and construction of a soil map of the site. The
dominant soil should be excavated, sampled, and analyzed to provide baseline data. This process
allows verification of the soil series classification, establishment of baseline condition, and
development of soil-vegetation relationships. Furthermore, the data provide inputs for models such
as the Water Erosion Prediction Project. The auxiliary areas should be excavated only by auger to
confirm and verify the soil series classification at the site. The SIR was a good estimator of several
surface soil properties (e.g., CEC, available water, bulk density, and salinity) and thus can be used
as a benchmark to assess condition and evaluate trends or as a source of input data for models. To
improve these interpretations, the critical phase criteria should be recorded on the field data sheet
and the appropriate SIR and phase identified for a more accurate assessment of the measured
properties as compared to the estimated properties. The surface soil property data indicated that
there were no significant differences between SIR and measured properties for cation exchange
capacity, water retention difference, bulk density, or salinity. Significant differences were,
however, measured between organic carbon and clay content indicating that these parameters in
soil samples need to be evaluated and analyzed to provide useful data.
The use of published soil maps to identify the kind and percentage of soils at a given site is
an essential tool to provide overall assessments of the sampled sites. A knowledge of soil and
landscape relationships is necessary and thus requires the use of SCS (or other professional) soil
scientists. Future surveys need to determine the number of soil samples required to make reliable
estimates; develop a method of evaluating surface soil compaction; and make a stronger effort to
relate soil properties and vegetation communities.
Other recommendations include:
Infiltration measurements are important indicator measurements for water balance and
should be incorporated into the 1993 pilot study (Kepner et a!., 1993).
Sites should be located on single landscape positions to reduce the soil variability. This
would eliminate the estimation of average property values for a site and enhance the
interpretation of vegetation and soil relationships.
72
-------
SECTION 6
FRAME MATERIALS
The Biotic Communities Map of North America, developed by Reichenbacher and Brown
(1992), was evaluated to determine how well this data base identified resource classes at the 39
selected sites (29 of these sites were sampled; see Section 2). Similarly, a preliminary evaluation
of the U.S. Fish and Wildlife Service (FWS) Gap Analysis Program (GAP) information was
conducted to determine how well the satellite-derived data base identified plant communities found
at the pilot study sample points.
The 1:8,000,000 scale Biotic Communities Map data base was digitized and imported into
GIS software for correlation with the EMAP-Arid sample grid points. Although the community-level
descriptions provided some utility in characterizing resource classes, the relatively coarse scale and
inaccuracy (as much as six degrees off longitudinally) of the map prevented any accurate resource
class identification in the pilot study area.
The GAP data set, developed by the Utah State University Fish and Wildlife Cooperative
Research Unit, consists of a map of classified plant communities derived from georeferenced,
Landsat TM imagery combined in a mosaic of the entire state of Utah. The construction of the
GAP data was a first attempt by the FWS to classify the vegetation of Utah using Landsat TM data.
The accuracy of the data set was largely unknown. An accuracy assessment of rangeland
formations (resource classes) had not been conducted at that time.
6.1 RESULTS
To provide the preliminary evaluation of the accuracy of trie GAP data, vegetation cover
assessments were made at each of the 39 EMAP-Arid sites. It is well documented in the literature
(Hay, 1979; Congalton, 1991) that at least 50 samples for each landcover category in a
classification are required to perform a valid accuracy assessment. A limited number of vegetation
assessments were taken at the 39 EMAP-Arid sites in hopes of providing some early evaluation
results to the GAP program. The 28 landcover categories used in the GAP data set were broken
down into six classes-desertscrub, grassland, pinyon/juniper woodland, water, agricultural crops,
and slickrock/barren. Based on the assessments performed at each sample site, the error matrix
shown in Table 6-1 was constructed. The error matrixes an array which expresses the number of
sample units assigned to a particular category, in this case the landcover classes derived from the
Landsat-based GAP data, relative to the actual category found on the ground (Congalton, 1991).
The columns in Table 6-1 represent the ground-based data while the rows indicate the classification
generated from the GAP data. The accuracy of the data set is computed by dividing the sum of the
major diagonal by the total number of sample units. Accuracies for individual categories are
computed by dividing the total number of correct sample units by the total number of sample units
in either the corresponding row or corresponding column. An evaluation of the "producer's
73
-------
Q
Q.
<
O
UI
cc
a.
UJ
z
u.
O
X
E
cc
o
cc
cc
UI
2
o
o
uZ
(O
(0
<
u
(O
UI
or °
2>l
15 0
o
(O
<
^
o
a
c
0
o
CD
JZ
C
0
CD
E
CO
CD
CO
CO
C
_o
ro
09
CD
CD
>
c
£
CO
m
Q.
O
o
J_
03
CD
)
n_
T3
CO
CO
CO
CO
^~
J3
U
£
CD
CO
CD
cx.
*
CO O ** O O O
in CD o o
O) CM <* «- i- CM
T- «"
T- r- O
T^
"-
^J. 0)
\
CO O
*- T- in CM
CD
^3 Q-
2 -o '§
o c .
1 1 1 1 H
Q C3 O. > O m
CO
2 Q- ^
to ,^ io
Q O CD
CC
X
en
CO
CM
t_
"-
CO
«-
co
t
TOTALS
X
X
O
c^
O
o
"§
0
o
£3
CD
CD
O
^
CO
10
PRODUCER
ACCURACY
CO
10
O
<
CC
13
CJ
u
oc
01
74
-------
accuracy," or measure of omission error, indicated that 69 percent and 58 percent of the
pinyon/juniper and desertscrub landcover types, respectively, were correctly classified by the GAP
data. None of the grassland, however, was correctly classified. An evaluation of the "user's
accuracy," or measure of commission error, yielded similar statistical results for the primary
landcover types of interest. The effect of an insufficient sample size becomes obvious when
evaluating the grassland classification results; with only three sample sites, the results of the
grassland assessment were not meaningful.
The preliminary vegetation assessments conducted during the pilot study were performed at
sites less than 1 hectare in area. To address the issue of plant community inclusions within
sampled vegetation, larger area measurements need to be made to more precisely determine the
referenced sample units relative to the Landsat-derived vegetation categories. To perform a valid
accuracy assessment of the remote sensing GAP data set, it will be important to increase the
number of samples per category evaluated in the field. Just as critical to the accuracy assessment,
however, is the development of a sampling design that addresses inclusions and ensures a
meaningful estimation of the extent of vegetation community types.
Preliminary results appear to indicate that plant communities distributed widely over the
landscape, with thin individual canopies, are difficult to accurately classify with Landsat TM data
alone. As the Landsat satellite instrument detects reflected energy from the Earth's surface, a
significant portion of the spectral signal reflected from arid lands is background soil. Utah State
University integrated elevation and slope data into the data set used in this evaluation, but the
construction did not include a soil layer. Incorporating a soils data base into a classification model
of the pilot study area would greatly aid in the discrimination of the desertscrub, grassland, and
woodland resource classes. Utah State University is presently integrating a soils data base from
the Soil Conservation Service (SCS) into its second generation Landsat TM-derived vegetation map
of Utah. This addition should greatly improve the accuracy of the data base in the arid vegetation
resource classes, and thus enhance the utility of the data for selecting frame materials for future
pilots and demonstrations.
To determine whether the GAP program and EMAP-Arid definitions of specific resource
classes concur, it is imperative that the EMAP-Arid group has well-developed conceptual and
operational definitions of not only resource classes but biogeographic provinces as well.
The EMAP-Arid researchers identified the following biogeographic provinces for reporting purposes
within Arid Ecosystems. Figure 1-1 shows the distribution of the Biogeographic Provinces of North
America for the EMAP-Arid resource group. These provinces'are Great Basin, Plains, Mohave,
Sonoran, Chihuahuan, Colorado Plateau, Californian, and Arctic. The EMAP-Arid team has also
identified the following resource classes for reporting purposes. These resource classes are
desertscrub, grassland, scrubland, woodland, tundra, riparian forest, riparian scrub, and strandland.
.The FWS GAP definitions of plant communities and thus resource classes in the EMAP-Arid
pilot study area concur, for the most part, with EMAP-Arid program definitions. The riparian
resource classes, however, are not adequately defined by the GAP data. A more critical issue
involving the use of the preliminary data base is its accuracy, which to date has not been adequate
to characterize arid ecosystems, specifically rangeland resource classes.
75
-------
Although definitions of provinces and resource classes have been identified by EMAP-Arid,
the 1992 pilot study was not designed to address questions of arid resource extent. For example,
definitions need to be developed to state what constitutes a pinyon juniper woodland and how
inclusions found within a woodland community are to be characterized.
The 1992 pilot study did not specifically define how area extent estimates of resources can
be carried out, although efforts have been made to determine the most appropriate method of
sampling extent. Three basic methods of extent estimation-point samples, area samples, and
"cookie cutter" samples-have been evaluated and some suggestions made as to the best choice for
EMAP-Arid. The "cookie cutter" area sample appears to be the best method from both the
ecological standpoint of landscape characterization and the statistical standpoint of providing a
smaller variance estimate.
6.2 RECOMMENDATIONS
The EMAP-Arid group definitions of vegetation resource classes, as well as biogeographic
provinces have been established. A method, or set of methods, to estimate extent of arid
resources needs to be developed, with the idea that an area sampling technique will provide a
better estimate of extent.
The preliminary evaluation of the GAP .data was inconclusive. If the EMAP-Arid group
wants to consider using GAP data in the future to select frame materials and to provide data for
extent estimation of arid ecosystems, then a scientifically valid assessment of the accuracy of the
GAP data must be performed. This assessment should be done on the second generation (or the
most recent version) of the GAP data and must include a sufficient sample number for each lancj
cover type.
76
-------
SECTION 7
QUALITY ASSURANCE, INFORMATION MANAGEMENT AND LOGISTICS
Quality assurance (QA), information management, and logistics are integral components of
EMAP field activities. In a program the magnitude of EMAP, overlooking or ignoring even
apparently minor issues or details may eventually jeopardize the success of the program. Planning
and documenting QA, information management, and logistics activities are essential. This section
documents these activities for the 1992 pilot study.
7.1 QUALITY ASSURANCE
Quality assurance plays a critical role in EMAP-Arid research efforts. As part of the EPA
Office of Research and Development, EMAP participates in the Agency's mandatory QA program
(Stanley and Verner 1985). Quality assurance activities are integrally connected with study
objectives, statistical design, logistics, field and laboratory measurements, information
management, data analysis, and product and report generation.
7.1.1 QA Program Overview
The general objective of the EMAP-Arid QA program is to maximize the probability that data
and statistical products collected by the EMAP-Arid Resource Group are of known, documented,
and adequate quality to meet and satisfy the needs of the data users. The philosophy of the QA
program for EMAP-Arid is one of guidance and assistance rather than enforcement. The
responsibility for the quality of the data rests with all project personnel, not just the QA personnel.
A specific pilot study objective related to QA was to determine whether data quality objectives
could be established for each indicator tested.
Data quality objectives (DQOs) are specific statements of the level of uncertainty a data
user is willing to accept in a body of environmental data with respect to the kind of scientific or
policy question that motivated the data collection activity. Data quality objectives are definitive,
qualitative or quantitative statements developed jointly by data users {e.g., scientists, policy
makers, interest groups) in conjunction with the QA staff. Data quality objectives will eventually be
developed on several levels so that acceptable levels of data quality will be specifically established
for individual measurements (measurement level), for indicators of ecosystem condition (indicator
level), and for the EMAP monitoring program (program level, i.e., status and trends). These levels
differ in that the higher levels contain more error sources and set constraints for the lower levels.
Data quality objectives at the program level include all sources of error (e.g., design,
sampling, measurement, indicator) that can accumulate and affect the interpretation of EMAP data
for determining status and trends. Program-level data quality is defined in terms,of the ability to
meet the EMAP program objectives with desired certainty. These DQOs will be used as
77
-------
performance criteria to assess data quality for its adequacy in determining status and trends for full
Implementation (Kirkland, in preparation). The following EMAP program-level target DQO for trend
detection has been drafted by EMAP management:
Over a decade each indicator of condition of a resource class on a regional scale should on
average detect a linear trend of 2 percent (absolute per year); that is, a 20 percent change
for a decade in the percentage of the resource class in a degraded condition. The test for a
trend will have a maximum significance level of 0.2 and a minimum power of 0.7.
Indicator-level DQOs are derived from aggregated parameter data for ecological indicators.
These DQOs may focus on the uncertainty associated with the data aggregation procedures used to
assimilate measurement-level data that provide assessment information, e.g., an index. Data
quality objectives should be developed to assure that data collected in pilot projects are of adequate
quality to support the research question (e.g., does an indicator require further development or
should it be included in the core group or be discarded). These DQOs must eventually be
established at a level such that the program-level DQOs will be satisfied.
Measurement-level DQOs (MQOs) are established for specific field and laboratory
measurement parameters. They are usually based on data user requirements and are often
estimated using existing or initial baseline data. These MQOs may define acceptance criteria for
detectability, precision, accuracy, representativeness, completeness, and comparability in field and
laboratory measurement data. The importance of MQOs at the early phase of resource group
activities stems from the fact that MQOs describe the most basic data level that will be used to
validate methods. The measurement quality objectives, listed in Tables 7-1 and 7-2 were
established for the 1992 pilot. MQOs were not established for spectral measurements.
One of the goals of the EMAP-Arfd field QA Program was to ensure the comparability or
between-crew precision of data produced by different field crews. Between-crew precision is
particularly crucial for some types of data collected, such as vegetation measurements and spectral
reflectance, because the data quality of these measurements cannot be assessed with QA samples.
The two field crews collecting data throughout the summer participated in several
between-crew remeasurements as part of the QA Program. These were of two different types-site
remeasurements and plot remeasurements. No within-crew remeasurements were performed and
therefore no assessment was made of within-crew precision.
Sfte remeasurements were performed when one. crew located, established, and measured a
site that had already been measured by the other crew. In these cases, the crew was instructed to
locate the permanent marker placed by the other crew and re-establish the transects. These
exercises did not provide a true estimation of crew variability because the crews were not sampling
exactly the same transects, subplots, and soil pits. They could, however, provide some information
about the amount of measurement plus spatial variability to be expected when sites are revisited.
They also provide fnformation on the crew ability to relocate a site previously sampled by another
crew. Site remeasurements were performed by the botanists and the soil scientists at four
locations.
78
-------
TABLE 7-1. MEASUREMENT QUALITY OBJECTIVES FOR VEGETATION
Variable
Units
Quality objective
Cover class (5 elements)
Herbaceous plants
Species
Cover clas's
Small shrub
Species
Cover class
Large shrubs and trees
Species
Widest crown diameter
Perpendicular diameter
Height
Basal trunk diameter
7 classes
7 classes
7 classes
50 cm
50 cm
50 cm
25 cm
90% agreement
90% agreement
90% agreement
95% agreement
90% agreement
98% agreement
90% within 1 m,
90% within 1 m
90% within 1m
90% within 50 cm
Plots were remeasured when both field crews visited the same previously established plot
on the same day. Two plot remeasurement sites (one pinyon-juniper and one desertscrub) were
performed at the end of the field season on August 11,1992. For this exercise, two transects and
two tree subplots were laid out and left in place until both botanists were finished with the
measurements. The botanist from each crew measured the vegetation on each transect and on
each subplot independently. Although the variance between the two botanists includes the
variance associated with placement of the sampling frames on the transect, these data can be used
to estimate variability between the two botanists.
The soil crews also participated in the plot remeasurement exercise with each crew ,
excavating and describing a soil pit. After each pit was excavated and described, the crews
switched places and described the other soil pit.
7.1.2 Results
The following sections provide an assessment of data quality and MQOs for vegetative,
spectral, and soil measurements collected during the 1992 pilot study.
7.1.2.1 Vegetation Measurements--
Two crew comparability tests (one in pinyon-juniper woodland and one in desertscrub) were
conducted using exactly the same transect lines and subplot centers. Twenty herbaceous/shrub
quadrats at each plot and two tree subplots at the pinyon-juniper site were used to compare
measurements between crews. No within-in crew precision or crew accuracy assessments were
performed.
79
-------
TABLE 7-2. MEASUREMENT QUALITY OBJECTIVES FOR SOIL ANALYSES
Precision limits
Parameter
Particle size
Sand
Silt
Clay
Organic carbon
Total carbon
Cation exchange
capacity
Soluble salts
Total nitrogen
Clod bulk^
density*
pH
Sodium absorption
ratio
Water retention
Reporting
units
% wt
% wt
% wt
% wt
% wt
meq/1 OOg
mmho/cm
% wt
g/cm3
pH units
meq/L
% wt
Detection
limits
2
2
2
0.03
0.2
0.09
0.05
0.06
0.30
0.05
0.09
1.0
Standard Coefficient
deviation of variation
1.2
1.2
1.2
0.02
0.12
1.20
0.09
0.004
0.31
0.15
1.70
0.52
3
3
3
1
1
4
2
3
15
4
4
* Was not established prior to 1992 field activities.
"Table 7-3 summarizes vegetation measurement data and calculated cover values for
selected attributes at each of the remeasurement plots. Each cover class value recorded by the
botanist was converted to a midpoint value. An average midpoint value was then calculated for
each attribute for each crew. Attributes were selected for a precision assessment based on their
relative importance to indicator development. Ground cover attributes {total vascular plant cover,
interspace rock fragments, and interspace litter) will be used in the calculation of an erosion index.
Surface type and cryptogamic crust (cyanobacteria) are proposed indicators of temporal soil surface
characteristics. The dominant shrub cover in the pinyon-juniper transect and dominant herbaceous
cover and dominant shrub cover in the desertscrub transect were selected because of their
importance in determining community composition.
80
-------
TABLE 7-3. COVER CLASS COMPARABILITY
Attribute
Total vascular plant cover
Gravel (interspace value)
(2-75 mm)
Cobble (interspace value)
(75-250 mm)
Stones (interspace value)
(> 250mm)
Litter (interspace value)
Surface Type 1 (total)
Surface Type II (total)
Surface Type III (total)
Cyanobacteria
Blackbrush
Total Surface Cover
IJTotal vascular plant cover
Gravel (interspace value)
(2-75 mm)
Cobble (interspace value)
(75-250 mm)
Litter (interspace value)
Surface Type 1 (total)
Surface Type II (total)
Surface Type III (total)
Cyanobacteria
Green Mormon Tea
Russian Thistle
Total Surface Cover
Site
Pinyon
Juniper
Pinyon
Juniper
Pinyon
Juniper
Pinyon
Juniper
Pinyon
Juniper
Pinyon
Juniper
Pinvon
Juniper
Pinyon
Juniper
Pinvon
Juniper
Pinyon
Juniper
Pinvon
Junjper
Desertscrub
Desertscrub
Desertscrub
Desertscrub
Desertscrub
Desertscrub
Desertscrub
Desertscrub
Desertscrub
Desertscrub
Desertscrub
Crew
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
Average Transect Cover Value*
(%)
18.5
18
19.2
24
3.9
2.7
3.8
2.6
12.2
16.4
3.8
8.5
60.3
25.6
32.1
40
63
68
15.7
14.5
61.4
66 3
46.5
57.1
4.2
7.8
0.9
0
27.4
16.1
0
21.8
89.1
12.1
7.3
45.8
47.5
46.3
19.3
15.3
12.5
14.6
79
81 ||
Cover class values were converted to midpoint of class (i.e., cover class 1 midpoint = 0.5; cover class 2 midpoint
= 3; cover class 3 midpoint = 15; cover class 4 midpoint = 37.5; cover class 5 midpoint = 62.5; cover class 6
midpoint = 85; cover class 7 midpoint = 97.5) and then averaged for 20 quadrats.
81
-------
Original MQOs required a 90 percent agreement between crews for individual cover classes
(I.e., agreement at the quadrat level). Only two attributes, total vascular plant cover and
blackbrush in the pinyon-juniper transects achieved this MQO. Agreement at the quadrat level is
probably an unrealistic objective because there may be slight variations in quadrat placement even
along the same transect and in decisions where actual cover is borderline between two cover
classes (i.e., one botanist could report a cover class of 3 and the other a cover class of 4 and both
botanists could be technically correct). Transect theory assumes that, if an adequate sample is
taken, the difference between measurements of the same population will not be significant. A
recommendation for a more realistic MQO would be to require that the average of the transect
cover values recorded by the botanists be within ±5 percent of each other (i.e., if one botanist
reported an average cover value of 80 percent, then the value reported by the other botanist should
be within the range of 75 to 85 percent).
With the exception of surface types, average ground cover attributes for the pinyon-juniper
transect were within ±5 percent of each other for the 20-quadrat comparison. Ground cover
attributes for the desertscrub transect were within ±5 percent of each other with the exception of
total vascular plant cover, litter, and surface types. The values for total vascular plant cover and
litter (both showed an 11 percent difference) were reciprocal between the crews, possibly
indicating a difference determining live versus dead portions of plants. The sum of ground cover
values that would be used in calculating an erosion index were within ± 5 percent of each other for
both the pinyon-juniper and the desertscrub transects. Between-crew precision for dominant
species cover was entirely acceptable.
There was very poor comparability between crews for measurements of surface types. As
much as a 77 percent difference was observed between the average values for the two crews.
Crews had trouble early in training distinguishing between surface types because of qualitative and
subjective class breaks. Quantifying class breaks for many different soil types would probably be a
formidable task requiring extensive data collection and analysis. Surface types are also associated
to a great extent with cryptogamic development on and within the soil surface. The cryptogamic
components appear to be more readily identified and quantified. Therefore, it is recommended that
surface types be dropped from further consideration as an indicator measurement.
Measurements for each tree and large woody species in a subplot included number (count)
of seedlings and saplings less than 1 meter high, longest crown diameter and largest diameter
perpendicular to the longest axis, height, and stem diameter. All MQOs listed in Table 7-1 were
met; however, analyses of the data indicated that the MQOs were much too broad and probably
inappropriate. A 90 percent comparability within a meter on crown diameters and height may be
appropriate for very large trees but extremely broad for most pinyon pine and juniper trees.
Agreement within 5 cm would have been more realistic than the 50-cm MQO listed in the QA plan.
Stem diameters above 50 cm are infrequent on many of the woodland species. In all cases, a
relative {%) comparability would be more appropriate because allowable variation would be
proportional to size. Table 7-4 gives the tree measurements recorded by each crew.
Two species were encountered on the two subplots with 100 percent agreement between
crews on species identification. Total counts for saplings were the same between crews.
However, one crew recorded five junipers and one pinyon while the other recorded four junipers
and two pinyon trees. Species identification error is highly unlikely between these two species
82
-------
TABLE 7-4. EMAP-ARID CREW COMPARABILITY FOR TREE MEASUREMENTS8
Sub- Sapl CD1
plot Spp # m
A1 JUOS 2 3.5
B1 JUQS 3 4.7
3.0
3.3
PIED 1 1.4
CREW 1
CD2
m
3.9
3.6
2.9
3.6
1.5
HT
m
2.5
3.1
1.7
2.8
1.6
SD Sapl CD1
cm #
20 2 3:7
11
15 2 4.5
42
13
21
46 2.6
6 2 1.6
CREW 2
CD2 HT
m m
3.0 2.4
3.9 3.4
2.6 2.7
1.6 1.8
-
SD
cm
20
12
16
43
15
25
47
7
a Spp: JUOS = Utah juniper, PIED = pinyon pine
Sapl: saplings <1 m high
CD1, CD2: crown diameter in meters; longest axis and longest diameter perpendicular to longest
axis.
HT = height in meters.
SD = stem diameter in centimeters (cm); SD preceded by blank cells indicates multiple stems of
the same tree.
even for very young plants. Therefore, the error is probably either a recording error or each crew
missed counting a sapling of each species. In either case, these types of errors can be minimized
with better procedural guidance and emphasis during training.
Crown diameters, CD1 and CD2, were converted to a single value, the geometric mean, for
comparison. The geometric mean was calculated as the square root of the product of CD1 and
CD2. Two junipers and the one pinyon pine were within ±5 percent for the between-crew mean
value. One juniper was only within ±15 percent. An error of this magnitude probably indicates a
recording error; however, it could also have been due to a hurried measurement. A significant error
occurred in Subplot B1 by one crew recording one tree with four stems while the other crew
recorded the same stems as two trees, one with three stems and one with a single stem. Basal
area calculations would not be affected but total crown area is greatly increased by the addition of
separate crown dimensions for another tree. Overall, the crown measurements are unacceptable
and indicate a need for increased emphasis during training or a reevaluation of procedures. Only
one of the four common height measurements exceeded ± 5 percent of the between-crew mean
and that was by ±6 percent.
Three of the eight stem diameters exceeded ±5 percent of the between-crew mean and all
were within ±10 percent; ±5 percent is desirable but difficult with woodland species. Branched
83
-------
stems near measurement points and the loose (compressible) bark of Utah juniper can impede
consistent measurement. Additional discussion with woodland experts, is advised to determine
appropriate measurement quality standards as well as additional emphasis during training.
A total of 25 plant species were identified between both crews on the desertscrub
transects. Crew 1 had a 96 percent agreement to the total plant list established by both crews,
while crew 2 had a 92 percent agreement. All plants missed by one or the other crew were "rare"
in abundance, having a transect cover value of one fourth of one percent or less.
There were 11 species identified on the pinyon-juniper transects; each crew identifying only
9 of them. Again the species missed were rare accounting for cover values of three quarters of one
percent or less. The limited number of quadrats available for comparison may be a limiting factor
for establishing a true comparison, especially considering the few species identified and the rarity of
the species missed. Minor differences in quadrat placement could easily account for "hit or miss"
differences on rare plants. Comparability relative to missing one or two rare plants is exaggerated
by the low number of species present.
7.1.2.2 Spectra] Reflectance Measurements-
Spectral data were collected along the vegetation transects and in vegetation subplots.
Each of these measurement units has a unique file, identified by site number and subplot or
transect designation. Transect measurements were made in six groups each comprising three
spectral measurements within a 2-m quadrat. These were taken at 3.5, 4.0, and 4.5 meters along
the tape, then repeated at 9.5, 10.0, and 10.5 meters; 15.5, 16.0, and 16.5 meters; 21.5, 22.0,
and 22.5 meters; 27.5, 28.0, and 28.5 meters; and finally at 33.5, 34.0, and 34.5 meters. A
meter stick was used to ensure that all measurements were taken 50 cm from the tape and at 1 m
above the ground. In the case of circular subplots, a square grid was laid out around the center of
the subplot and spectral measurements were taken in a 4 by 4 matrix with each point separated by
3.0 m. A protocol was established to ensure that the pin flags used to mark sampling locations
were removed prior to acquisition of spectra and that no shadows from the technicians or
instrument fell in the field of view of the spectrometer.
A number of variables affect spectral reflectance and the precision and accuracy of
measurements. Atmospheric haze increases reflectance in short wavebands, particularly 400 to
500 nm, and clouds affect the whole range of the visible spectra from 400 to 700 nm (Lillesand
and Kiefer, 1987). Spectral measurements were not taken at -all if cloud cover was estimated to be
greater than 50 percent. If cloud cover was up to 50 percent but no clouds obscured the sun then
measurements were taken as normal. Atmospheric clarity and the amount of cloud cover and wind
speed and direction were recorded in the comments section of the header for each transect and
subplot data file on the computer. In addition, weather conditions, soil moisture, and the
vegetation and soils comprising each spectra were recorded on field forms on a spectrum by
spectrum basis. Sun angle and shadowing also affect spectral measurements and, in an effort to
maintain consistency in the data, spectra were taken between 10:00 am and 3:00 pm when the
sun angle was highest and shadows minimal. Atmospheric and illumination conditions may cause a
change in the reference spectra at a site during the day. As this occurs, the changes can be
captured within the calibration.
84
-------
A stringent methodology for the calibration of the spectrometer must be followed prior to
beginning each group of spectral measurements. The instrument was calibrated at least thirteen
times at each site. Although the instrument was calibrated according to procedures established for
the study, a misunderstanding of the calibration procedures led to data quality problems with all of
the spectral reflectance data collected during the study. Although the operators performed the
calibration to the white reference standard as instructed, they did not toggle the necessary switch
and as a result the white reference standard information was not saved. When the instrument
executed the automatic internal calibration it was doinp so to a non-existent reference standard.
As a result there is no way to calibrate the spectral data for atmospheric conditions. The indicator
lead believes that the spectral data are biased low but this bias is neither quantifiable nor consistent
across the data. It is also believed that the variability introduced by incorrect calibration procedures
is low.
Only one spectrometer was available for use and one technician operated the instrument
from June 29 through July 26 and again from August 18 through 20, 1992. A second technician
worked from July 27 through August 17, 1993. On July 27 both technicians worked together
with the second technician taking the measurements under the direction of the first to ensure
consistency in methodology because measurements for between-crew variability for other
indicators were made on a day when only one spectral technician was present. No assessment of
the between-crew precision of spectral measurements can be made.
Both the spectrometer and the computer ceased functioning during fieldwork and had to be
replaced with equipment of the same specification. Both spectrometers measure reflectance and
are calibrated against the same standards. Tests and corrections can be made for any electronic
variability between instruments. No spectral measurements were recorded on three days while an
alternate spectrometer was acquired. These problems could be avoided in the future if two
spectrometers and computers were available for use. In the case of inexperienced crew members,
a training period of 1 week before the official field training session begins would be necessary.
Familiarity with computers is a prerequisite for operation of the spectrometer.
Spectral data were acquired from 13 sites. At 6 of the 13 sites all 220 measurements were
taken. For the other 6 sites, 2 had measurements from 10 plots, 2 from 7 plots, 1 from 3 plots
and 1 from only 1 plot. Weather conditions were largely responsible for the partially completed
sites. Three sites targeted for spectral measurement were not measured due to instrument failure
and cloudy conditions were responsible for lack of data from two other target sites. This record
could be improved by conducting field work earlier in the year before the thunderstorm season.
Overall, cloudy conditions and instrument failure were each responsible for a 20 percent loss of
potential spectral data.
7.1.2.3 Soil Measurements--
As described earlier, site remeasurements and plot remeasurements were performed several
times during the study. No MQOs for field measurements were established prior to the study. The
field soil descriptions for two sites were evaluated for consistency. The results of the comparison
between the two field crews are given in Table 7-5 for some of the more important parameters.
Similar map units, soil series, permeability, slope, soil column depth, moist A horizon, color and
texture were reported by both crews. One inconsistency noted was the recording by one soil
85
-------
TABLE 7-5. BETWEEN-CREW COMPARISONS FOR FIELD MEASUREMENTS
Parameter
Site 1 Soil series classification
Horizon A
Color dry
Color moist
Texture
Sand
Silt
Clay
Very fine sand
Permeability
Slope length
Slope %
Aspect
Site 2 Soil series classification
Horizon A
Color dry
Color moist
Texture
Sand
Silt
Clay
Very fine sand
Permeability
Slope length
Slope %
Aspect
Crew 1
Fine-loamy mixed mesic
ustic calciorthid
0-6 cm depth
5yr 6/4
5yr4/6
FSL
74
15
11
20
Moderate
3
3
SW
Fine-loamy mixed mesic
ustollic calciorthid
0-10 cm depth
5yr 6/4
5yr 4/4
FSL
73
15
12
25
Moderately rapid
3
3
SE
Crew 2
Fine-loamy mixed mesic
ustollic calciorthid
0-2 cm depth
5yr 6/6
5yr 5/6
FSL
65
25
10
20
3
3
3
SW
Fine-loamy mixed mesic
ustollic calciorthid
0-3 cm depth
7.5yr 5/6
7.5yr 4/6
FSL
56
30
14
20
2
3
3-4
SSE
86
-------
scientist of permeability as a numerical value (e.g., 2 in/hr) while the other crew recorded this value
as a descriptive term (e.g., moderately rapid). There were significant differences between the
crews for depth of horizon A, percent sand, silt, and clay content. These are, however, coarse
field measurements and are repeated in the laboratory. There were also significant differences
between the two crews for depth of A horizon. Although there was no MQO established prior to
the study, it is generally felt that the agreement between the two crews should have been better
for this parameter.
All bulk soil samples were sent to the U.S. Soil Conservation Service Soil Survey Laboratory
at the National Soil Survey Center in Lincoln, Nebraska. Analyses included particle size, organic
and total carbon, cation exchange capacity, soluble salts, nitrogen, sodium absorption ratio, pH,
and water retention. The results given in this section can be compared to the MQOs given in Table
7-2. The MQOs established for the pilot were probably unrealistic and were too rigid for the
purposes of this study. These MQOs should be reevaluated for further pilot activities.
Reference soil was obtained from a soil pit dug during training exercises. The soil was
removed from the pit, air dried, and mixed as thoroughly as possible by hand. One bag of reference
soil was included in each batch of field samples sent to the laboratory. The nine reference samples
provided a check of the consistency of the laboratory over the course of the study. The samples
do not provide an accuracy check as there were no known values for the samples. Table 7-6
summarizes the analytical data from the nine reference samples.
It should be noted that some of the variability apparent in the results may have resulted
from hand mixing the samples in the field and not from analytical variability. The results show that
most of the analyses were within or close to the MQOs listed in Table 7-2. Analyses for which no
MQOs were developed also were within limits generally accepted for these types of laboratory soil
analyses. Although the standard deviations and the coefficients of variation '(CV) were high for
carbonate and rock fragment percentages, this result is probably due partly to the field
homogenization process and partly to the very low rock fragment content of the samples.
To assess laboratory precision, seven samples were split at the analytical laboratory to
obtain three replicates of each split sample. Table 7-7 shows the range of the means, the standard
deviations, and the coefficients of variation (e.g., the lowest mean of the replicate analyses of the
seven samples for percent total clay was 2.3 and the highest mean was 28.7) for seven of the
most important analyses. All results were within acceptable limits with the exception of organic
carbon. Three of the seven replicate analyses for organic carbon resulted in high CVs (20, 22, and
115%). The 115% CV was obtained because the concentration of organic carbon present in the
sample was below the detection limit. The other two high CVs are possibly indicative of an
analytical problem.
Field duplicates were collected from the deep soil pit at each Group A site that the soil
scientists visited. The crew collected the sample, mixed it by hand, and split it into two sample
bags. The data from these samples can be used to assess system precision (i.e., the variability
associated with these results includes variability associated with the sample collection process and
with laboratory analysis).
87
-------
TABLE 7-6. SUMMARY STATISTICS FOR REFERENCE SAMPLES
Parameter
Total Clay {%)
Total Silt (%)
Total Sand {%)
Fine Silt {%)
Very Fine Sand (%}
Fine Sand (%)
Medium Sand (%)
Walktey-BIack Organic
Carbon
NH40AC Catfon
Exchange Capacity
15 Bar Water on Air
Dry Soil-Weight %
pH, 1:t Soil-Water
Suspension
pH, 1:2Soil-CACL2
Suspension
Carbonate, < 2 mm
Fraction
2-5 mm Weight Per-
centage of < 75 mm
5-20 mm Weight Per-
centage of < 75 mm
20-75 mm Weight Per-
centage of < 75 mm
Field Water Content
of Bulk Sample
Na
g
g
g
g
g
g
g
g
g
g
g
9
g
g
g
g
g
Mean
15.73
30.24
54.02
10.74
33.41
12.g7
4.87
0.27
8.46
e.go
8.29
7.73
6.30
1.00
0.44
o.?a
6.51
Maximum
17.20
33.10
55.60
11.60
35.30
14.30
5.30
0.2&
8.70
7.50
8.40
7.80
7.50
2.0O
1.00
7.0O
e.go
Minimum
14.50
28.go
51.70
id. 1.0
31.40
12.00
4.70
0.25
8.20
6.20;
8.20
7.70
3.40
0.00
O.OO
0.00
6.30
Standard
deviation
0.95
1.38
1.11
0.56
1.41
0.95
0.20
0.01
0.17
0.56
0.08
0.05
1.16
0.50
0.53
2.33
0.20
cvb
6.02
4.57
2.05
5.21
4.22
7.30
4.11
3.70
1.97
8.13
18.43
50.00
1 18.59-
300.00
3.02
8 N = number. b CV = coefficient of variation.
Table 7-8 shows the range of the means, the standard deviations and the CVs for each
crew for seven of the most important analyses. One crew collected twelve duplicate pairs while
the other crew collected seven duplicate pairs. Overall the results of the analyses of the samples
collected by Crew 1 tended to show more variabilfty than those collected by Crew 2. This may
indicate that Crew 1 did not mix the samples in the field as well as Crew 2 did. This is evident
most notably with the results of percent total clay and organic carbon. For total clay, five of the
twelve pairs of samples collected by Crew 1 demonstrated CVs above 10 percent (10.8, 10.1,
13.2, 12.g, and 33.7 percent). Most of the results were, however, close to the detection limit of
the method. Only one of the pairs collected by Crew 2 demonstrated a CV above 10 percent (17.4
percent) for total clay content. This one sample also had a very low clay content. For organic
carbon, seven of the twelve pairs of samples collected by Crew 1 demonstrated CVs above 10
88
-------
TABLE 7-7. SUMMARIZED RESULTS OF LABORATORY REPLICATE ANALYSES
Parameter
Total clay {%)
Total silt (%)
Total sand (%)
Organic carbon
Cation exchange
Range of means
2.3 -
10.1 -
33.7 -
0.01 -
1.7 -
28.7
42.3
86.5
1.78
11.3
Range of standard
deviations
0
0.7 -
0.51 -
0
0.1 -
0.61
1.25
.05
1.12
1.57
Range of coefficient
of variation
0
1.8
0.9
0
0.9
8.9
- 8.5
- 2.6
- 114.6
- 11.8
capacity
PH
1 5 bar water
7.4
2.2
- 8.8
- 10.8
0.06 -
0.12 -
0.26
0.64
0.7 -
5.1
TABLE 7-8. SUMMARY OF RESULTS OF FIELD DUPLICATE ANALYSES
Range of means
Range of standard Range of coefficient of
deviation variation
Parameter
Total clay (%)
Total silt (%)
Total sand (%)
Organic carbon
Cation exchange
capacity
PH
1 5 bar water
Crew 1
1 .6-34.4
3.9-32.9
36.4-
93.3
0.13-1.3
1.4-22.4
6.60-
8.75
1.9-12.1
Crew 2
2.8-16.6
3.5-41.6
45.3-
93.7
0.06-
0.97
3.5-10.7
7.15-
8.60
2.5-6.2
Crew 1
0.07-
1.84
0.0-2.3
0.21-
4.03
0.01-
0.47
0.0-0.42
0-0.35
0.07-
0.49
Crew 2
0.07-
1.13
0.14-
4.81
0.07-
5.02
0.0-0.08
0.14-
0.78
0.0-0.7
0.07-
0.42
Crew 1
2.0-33.7
0.0-20.2
0.2-6,1
0.0-42.8
0.0-7.4
1.6-17.0
Crew 2
0.7-17.4
1.2-11.0
0.1-11.1
0.0-18.1
1.6-8.9
2.3-9.9
89
-------
percent {19.2, 16.6, 20.0, 32.6, 30.6, 10.9, and 42.8 percent). All samples collected by Crew 1
had concentrations above the detection limit. Only one of the pairs collected by Crew 2
demonstrated a CV above 10 percent (18.1 percent).
7.1.3 Recommendations
The QA Program for EMAP-Arid should be expanded. One individual should be committed
to the position of Quality Assurance Manager for the EMAP-Arid Resource Group. This individual
should coordinate QA issues between the indicator leads and assume responsibility for follow-up on
QA issues. This individual should also serve as the QA group representative in interactions with the
EMAP-wide QA Group and the EMAP QA Director located at EPA Headquarters.
The training program should be expanded and include more dry runs. Indicator leads must
be fn attendance throughout the dry runs to ensure that the crew is performing adequately and
within the guidelines of the training. A proficiency check of the crew should be one of the last
activities during training. Crew members should be tested for all duties they will be expected to
perform during the field study. The data collected during proficiency exercises must be thoroughly
analyzed and studied. Indicator leads should be responsible for evaluating members of the field
crew and certifying that they are able to perform the necessary tasks.
An increased effort should be placed on completing forms during training. Forms must be
designed to minimize confusion and not be ambiguous. Units should either be hard-coded on the
form if appropriate or the unit should be a required field for completion. One area of confusion on
the forms was the entry for slope/aspect. Sometimes the value was recorded by the crew as an
alphabetical value (e.g., S-SW) and sometimes it was recorded in degrees. Soil permeability was
another area of confusion. One crew recorded this value as a numerical value (e.g., 2 in/hr) while
the other crew recorded this value as a descriptive term (e.g., moderately rapid) indicating a range
of values.
An audit schedule should be developed for future field studies. This schedule should
Include audits of the field crews, laboratory audits, and audits of the information management
system. Audits should be coordinated by the QA Manager, well-documented, and performed by
persons with expertise in the indicator area being audited.
In addition to the between-crew precision remeasurements by the field crews, within-crew
remeasurements should also be performed to assess a crew's ability to reproduce results.
Vegetation, soil profile, and spectral measurements should be performed by experts to provide a
reference or a "known" value. This would allow an accuracy assessment to be performed on the
field measurements.
In the future, vegetation measurement frames should be placed and left in place until all QA
remeasurements have taken place. This would eliminate the variance associated with placement of
the frames and would, therefore, lead to a truer estimate of crew variability.
Accuracy and precision measurements should take place both during training and early in a
study to ensure that the field crew is technically competent and fully understands the standard
operating procedures. An accuracy and precision assessment of all field measurements should be
90
-------
performed. Increased effort must be placed on evaluating QA data as soon as possible after data
collection to allow for early detection and correction of any problems which may affect data
quality. Electronic data capture would enhance our ability to evaluate QA data real-time.
Standardized methods should be developed for assessing precision and accuracy from
remeasurement data.
Reference samples with known concentrations should be used to monitor analytical
laboratory accuracy.
Increased effort should be place on development of MQOs. Realistic MQOs that address
the needs of the EMAP-Arid resource group should be developed.
At this time, quantitative DQOs cannot be established for each indicator tested. Analytical
paradigms for interpretation of measurements relative to societal values and goals have not been
established, nor have the measurement parameter requirements for each indicator or procedures to
correlate, combine, and index them into a DQO been completed. From information gathered during
the 1992 pilot, data quality objectives for future measurement activities will be refined and
expanded.
7.2 INFORMATION MANAGEMENT
The activities and functions performed by information management (IM) are integrally
connected with the study objectives, statistical design, logistics, field and laboratory
measurements, quality assurance, data analysis, and product and report generation. Properly
designed and implemented, IM functions as a cohesive and consistent thread which ties together
each of these components of the research effort and provides the infrastructure necessary for
turning scientific concepts into defensible results and products. For the 1992 pilot study the
following IM functions were planned for testing and evaluation:
- use of field forms in the data collection effort,
- use of Portable Data Recorders (PDRs) in selected data collection efforts,
- transfer of data between the field and EMAP-Arid IM central office,
- specifications of initial requirements for the EMAP-Arid Information Management system,
- use of external data sets,
- hardware and software requirements for the field,
- coordinating requirements between IM and the design, quality assurance, logistics, and
statistics elements, and
- availability of data to evaluate the results of the pilot study.
91
-------
7.2.1 Results
A summary of information management efforts related to field forms data collection
software, coding scheme development and training for the 1992 Pilot Study areas follows.
Field FormsThe development of the paper forms for data collection and the instructions for
completing those forms was a combined effort of the information manager, the indicator leads, and
the statistician. For each site an individual packet was prepared containing all required forms and
the assigned site and soil sample numbers. A close-to-final version of the field forms was available
for the training session but the forms did not go through a full-fledged field trial to determine their
usability until the beginning days of the pilot study. Based on the use of the field forms in the first
few days of the pilot study, changes were made to the forms concerned with site information and
vegetation transects. Changes to the latter form were cosmetic to ease the appropriate placement
of the data. Changes to the site information form allowed additional information to be captured.
The new forms were available to the field crews by mid-July.
Data Collection Software--A software program for portable data recorders (PDRs) and PCs
was developed to handle data entry in the field and data editing in post-field activities. A
documentation manual described the use of the program, care of the PDR, and the transferring,
printing, and editing of the data. Software for the PDR was developed using the C language and
C-Scape, a collection of C routines for easy screen generation. This software was not ready for
implementation prior to the field effort. Because of time limitations for software development, a
complete program to collect the vegetation data was not available until the end of July. Although
minor changes were made to the program based on a few days of field testing (in conjunction with
completing the paper field forms), the availability of the program so late in the field season
precluded adequate testing. Consequently, PDRs were not used routinely for data collection for any
of the field measurements. The PDR programs were used at the central office as one of the tools
for entering the data captured on the field forms.
Coding SchemesIdentification and coding schemes were developed to track film rolls and
site photography; PC disks; data forms; soil samples and voucher specimens of unidentified plant
species; sites, transects, pits, plots and points; and data file names. The coding scheme identified
sites, vegetation transects, circular plots and quadrats, soil pits, film rolls, and data files. This
scheme included an obvious reference in the code to the EMAP site number. The code for the
vegetation samples also included a reference to the transect and quadrat of the sample. Codes for
the soil samples reflected the format used by the Soil Conservation Service since these data would
initially be entered into that data base.
Training-Training activities covered the use of the PDR for data collection and for
transferring information between the PDR and a PC; requirements for filling out the field forms; the
importance of QA and documentation in data collection; the coding schemes and labelling
protocols; and the post-field activities for data collection. As only a partial data collection program
for the PDR was available by the time of training, use of the software was not covered during
training.
The information management system provided a reasonably effective, but not efficient
movement of data from the field to the analysis stage. Lessons were learned about the design and
92
-------
use of forms to capture the necessary data and about clarifying the directions for their use.
Lessons were also learned about the development of software on the PDR for use in the field.
Data from the field were moved to the analysis stage but with more effort and with lower
quality than had been anticipated. The prime reasons were the lack of lead time to prepare and test
the electronic data collection system for use in the field and the lack of available time in the field or
at the base office to follow quality assurance procedures for the data. Thus, the computerization of
the data took more effort; there were questions about the data for which assumptions had to be
made; and data were not transferred to the central office in a timely manner to allow for feedback
to the field crews.
The information management and data analysis systems were not reasonably adequate to
enable reporting within the required time period, partly due to the loss of information management
support for 6 months due to changes in EPA procurement approaches. Verifying the vegetation
data took a lot more effort than was anticipated and the IM systems were not efficient enough to
enable timely reporting. The efforts from this pilot study did provide a base of experience for the
next pilot.
7.2.2 Recommendations
One of the most critical learning experiences gleaned from the 1992 pilot study was
awareness of the complex process of moving data from the field to the central office. Greater
emphasis must be placed on electronic data capture in the field and on real-time QA checks of the
data. Emphasis also should be placed on coordinating the IM efforts and needs for all indicator
categories (much of the IM 1992 effort was related to the vegetation indicator) to ensure that all
data flow smoothly from the field to the data analysis stage and to ensure that all of the data
receive an adequate and timely QA review. The use of external data sets and incorporation of
these data bases into a centralized EMAP-Arid system were beyond the scope of this pilot. Future
pilots and IM activities need to address this issue.
7.3 LOGISTICS
The 1992 EMAP-Arid pilot study was conducted in part to obtain important information
related to areas such as logistical requirements. Logistics activities include site selection and
reconnaissance, training, staffing, mobilization, and field implementation. A detailed description of
these logistics planning activities is presented in Baker and Merritt (1991). Identifying logistical
constraints on field implementation and assessing the requirements for assembling and deploying
multiagency sampling teams were questions addressed in the 1992 pilot study.
7.3.1 Results
As a result of the pilot study a number of lessons were learned and logistical guidelines
were reinforced for EMAP-Arid. These were compiled in an unpublished report entitled "EMAP-Arid
Ecosystems 1992 Colorado Plateau Pilot Study Field Operations Report." This document
summarized logistic activities from the initial planning process through the final debriefing of the
field events. Documentation of these logistics activities in itself is a valuable exercise for the
planning of future EMAP-Arid field activities.
93
-------
A summary of the logistics guidelines reinforced are as follows:
- Detailed logistics planning with adequate lead time is key to a successful field program.
- Detailed protocols documented in a field operations and training manual are critical to
the performance of the field events.
- The spectral, vegetation, and soil samples could be routinely collected at a site using
4-wheel-drive vehicles, backpacking, or helicopter for very remote locations with limited
access.
- Multiagency sampling crews were effective and provided local expertise and working
facilities.
Specific lessons learned are as follows:
- Issues related to the National Environmental Policy Act and archaeological clearance
need to be identified and resolved early in the logistics planning efforts.
- A 1 to 2 week training program is required to fully train the field crews and a full
"practice run" should be included in the program to evaluate the field crews so
additional training or adjustments can be made before routine sampling starts.
- Global positioning systems (GPS) can be used by the field crews with limited training to
find the general site location, but permanent marks may have to be used in the long
term to accurately relocate a site.
7.3.2 Recommendations
The overall success of the 1992 field season sampling effort was strongly attributed to the
field personnel and how well they worked together as a team. However, a number of changes to
the 1992 pilot study operations could facilitate future activities. One such change is the addition of
an Implementation Coordinator position to the EMAP-Arid team. This key position would provide
implementation guidance, direction, and logistical support for the monitoring activities conducted by
multi-agency sampling crews. Duties would include forming communication linkages among and
between field crews, team indicator leaders, information management, and quality assurance
personnel, managers, and participating federal agencies, contractors, and cooperators.
In addition, other general guidelines could be adopted by the EMAP-Arid team to improve
future studies. These include:
Initiating anticipatory planning in the preliminary stages to help direct all components
of the projects (regardless of whether they are major or minor).
Increasing efforts to ensure clear, accurate, and timely communications within and
beyond the EMAP-Arid team. Maintaining an accurate picture of the status and
condition of the project would avoid potential problems and define progress.
94
-------
Briefing and debriefing each agency accurately on the field study. Documentation
describing the field study freely available from the inception.of the project would.also
enhance the communications networks.
95
-------
SECTION 8
CONCLUSIONS, RECOMMENDATIONS, AND ADDITIONAL STUDIES
The 1992 pilot study was the first EMAP-Arid field study and addressing the objectives
developed for this study is essential to full implementation of the program. Questions related to
these objectives will continue to be important elements in planning for future pilot studies. During
the 1992 pilot study, the EMAP-Arid team was successful in partially addressing these objectives
but, more importantly, the planning and implementation of this study uncovered issues that were
not fully understood or perceived in initial design efforts. For example, the original assumptions
that the EMAP-Forest design would be applicable to EMAP-Arid indicator measurements or that
independent site condition assessments from land management agencies could readily be used to
evaluate sensitivity of indicators were inaccurate." Only as the 1992 field work progressed were
these difficulties pinpointed. This section summarizes the major conclusions and recommendations
drawn from the 1992 pilot study, clarifies some of the issues not initially perceived, and draws
implications from the study results. Activities, conducted in a 1993 pilot study and activities
planned for future studies relative to these conclusions, issues, and implications are described
where appropriate.
Plot Design-Evaluating the EMAP-Forest sample plot design as a common design that could
also be used by the EMAP-Arid group was of interest to EMAP in the effort to help facilitate and
integrate assessment across resource groups. However, after initial adoption of this common plot
design, questions arose during the 1992 field study as to its appropriateness for arid ecosystems.
These questions were raised because of the difference in the vegetative structure and soils found in
forested and arid resource classes. The issues raised were subsequently substantiated by
preliminary analysis of the data in early 1993 and by the data analyses presented in this report. As
reported in Section 4, the variance assessment of EMAP-Arid indicator measurements obtained
using the modified EMAP-Forest plot design shows considerable discrepancy among the variables
and indicates that additional plot designs should be evaluated. These considerations led to the
development of a plot design study conducted in 1993.
The 1993 pilot study was designed to establish the sampling support area and optimum plot
size required for the selected indicator measurements. The approach was to oversample an area to
characterize the local site variability and assess an integrated plot.design that best quantifies
measures for vegetation, soils, and spectral properties for EMAP-Arid program requirements (Kepner
et al., 1993). These oversized plots or macroplots were 180 by 180 m with 36 sampling units for
soils and 120 by 120 m with 100 sampling units for vegetation and spectral properties. Five
macroplot locations were selected in the Colorado Plateau within desertscrub, grassland, and
conifer woodland resource classes. Costs, time efficiency, and site impact were also part of the
1993 pilot evaluation. Results from the 1993 pilot study will be reported in the fall of 1994 and
used to plan and develop a pilot study in 1995 at sites with known conditions.
96
-------
Indicator Sensitivity-An evaluation of indicator sensitivity against sites of "known"
condition was not possible with the results from the 1992 pilot study. The primary assumption of
the EMAP-Arid researchers that site conditions could be obtained from existing assessments made
by the various land management agencies was unfounded. There were significant differences in
agency descriptions of the condition of a site and in several cases no consistent rating of a site
could be established; this finding is similar to conclusions reported more recently by the NRC
(1994). Results for both vegetation and soils indicate that additional indicator sensitivity studies
should be conducted. Vegetation and soil indicator development will continue as part of the 1993
pilot. Results from the spectral data show a strong correlation between Landsat TM and PSII NDVI
and indicate that the overall concept of remote sensing as an indicator of ecosystem condition is
valid. The NDVI may not be the most sensitive indicator and other indices such as reflectance
(albedo) or SAVI need to be evaluated. These and other indices will be evaluated from data
collected during the 1993 pilot study.
A gradient study, evaluating a range of different environmental conditions, is planned in
1995 to assess current indicator measurements and possible additional faunal indicator
measurements evaluated in a 1994 pilot study. The study should be conducted in areas with
known levels of condition, for example, sites in different successional stages such as Long-term
Environmental Research sites (Friedel 1991) or degree of stress (e.g., different grazing intensity).
Conducting this type of study independent of the EMAP probability sampling grid would ensure that
enough data are collected to make statements on the sensitivity of indicators. By using long-term
data sets and comparing EMAP indicator data with existing data (e.g., for species composition,
erosion, and soil quality), it should be possible to determine how the indicators perform over time
and how well they meet the EMAP program-level DQOs. Specific recommendations in Section 5 for
indicator measurements should be incorporated into the 1995 Pilot Study.
Frame Material-Based on a limited assessment, preliminary data bases from the GAP and
Biotic Communities Map of North America appear quite variable with respect to identifying arid land
resource classes and appear marginal in their ability to calculate the amount or extent of individual
arid resource group types. This variability appears to result from differences in definitions and
differences in scale of point samples versus satellite-derived coverage. However, the study
included only a limited number of observations in each land cover type. If the EMAP-Arid group
wants to consider using GAP data in the future to select frame materials and to provide data for
extent estimation of arid ecosystems, then a scientifically valid assessment of the accuracy of the
GAP data must be performed. This assessment should be done on the second generation (or the
most recent version) of the GAP data and must include a sufficient sample number for each land
cover type. If improvements in accuracy are obtained an oversample strategy should be developed
to account for unresolved error. If, for example, the GAP data are determined to be accurate 85
percent of the time, then the sampling strategy must include sampling enough alternate sites to
resolve the 15 percent difference. This sample-based accuracy assessment study should be
conducted prior to any future assessment of EMAP-Arid resources using these techniques.
Quality Assurance-The 1992 pilot study demonstrated that a comprehensive QA program
encompassing logistics and all aspects of a data collection activity is an essential component of
field studies such as those conducted by EMAP. Major conclusions drawn from the 1992 pilot
study include the following items. The training program must be expanded to include more dry
runs and a formal proficiency check of all crew members. Precision and accuracy assessments
97
-------
must be made very early during field activities and then again periodically to ensure that the field
crew is technically competent and is following standard operating procedures. Audits should be
performed for field and laboratory operations to ensure that standard operating procedures are
being followed and to acquire feedback from field and laboratory personnel. Data recording and
entry procedures must be developed in such a way as to maximize recording accuracy and to allow
for real-time examination and analysis of the data.
The above findings from the 1992 pilot study were incorporated into the plot design study
design conducted in 1993 (Kepner, 1993). In that study, the training program was expanded to
include more dry runs and a proficiency check of the crew was performed during training. An
increased effort was placed on data recording procedures with the vegetation data entered
electronically on portable data recorders rather than on paper forms. Between-crew precision
remeasurements and within-crew remeasurements were made during the first week of field
operations and then again weekly to ensure that the field crew was technically competent and fully
understood the standard operating procedures. Further refinements of the QA program based on
the results from the 1993 pilot are being developed for future activities.
98
-------
Analytical Spectral Devices, Inc.
86 pp.
REFERENCES
1993. Personal Spectrometer II Procedures Reference Manual.
Anderson, B. W., and R. D. Ohmart. 1986. Vegetation. Chapter ]m Inventory and Monitoring of
Wildlife Habitat, A. Y. Cooperrider, R. J. Boyd, and H. R. Stuart (eds.), U.S. Department of
the Interior, Bureau of Land Management Service, Denver, Colorado.
Anderson, G. L, J. D. Hanson, and R. H. Haas. 1993. Evaluating Landsat Thematic Mapper
derived vegetation indices for estimating aboveground biomass on semiarid rangelands.
Remote Sensing of Environment, 45(2):165-176.
Baily, A. W., and C. E. Poulton. 1968. Plant communities and environmental relationships in a
portion of the Tillamook burn. Northwestern Oregon. Ecology, 49:1-13.
Baker, J. R., and G. L. Merritt. 1991. Environmental Monitoring and Assessment Program:
. Guidelines for Preparing Logistics Plans. EPA/600/4-91/001. U.S. Environmental
Protection Agency, Las Vegas, Nevada.
Baumer, O. W. 1989. SCS soil parameters preparation program for DRAINMOD DMSOIL user's
manual. Draft.
Breckenridge, R. P., W. G. Kepner, and D. A. Mouat. 1993. Multidimensional Monitoring for
Assessing Sustainability of Arid Ecosystems. Proceedings of the 25th International
Symposium on Remote Sensing and Global Environmental Change, April 4-8, 1993, Graz,
Austria, pp. I-583 - I-594.
Breckenridge, R. P., J. R. Williams, and J. F. Keck. 1991. Characterizing soils for hazardous waste
site assessment. EPA/540/4-91/003. U.S. Environmental Protection Agency,
Environmental Monitoring Systems Laboratory, Las Vegas, Nevada.
Brower, J. E., J. H. Zar, and C. N. von Ende. 1989. Field and Laboratory Methods for General
Ecology, William C. Brown Publishers, Dubuque, Iowa.
Brown, D. E., C. H. Lowe, and C. P. Pase. 1979. A digitized classification system for the biotic
communities of North America, with community (series) and association examples for the
Southwest. Journal of the Arizona-Nevada Academy of Science, Vol. 14, Supplement 1.
Bryant, N. A., L. F. Johnson, A. J. Brazel, R. C. Balling, C. F. Hutchinson, and L. R. Beck. 1990.
Measuring the effect of overgrazing in the Sonoran Desert. Climatic Change, 17:243-264.
Chavez, P. S., Jr. 1989. Radiometric calibration of Landsat Thematic Mapper multispectral
images. Photogrammetric Engineering and Remote Sensing, 55(9):1285-1294.
99
-------
Congalton, R. G. 1991. A Review of Assessing the Accuracy of Classifications of Remotely
Sensed Data. Remote Sensing of Environment, Vol. 37, pp. 35 - 46.
Daubenmire, R. 1968. Plant communities: a textbook of plant synecology. Harper and Row
Publishers, New York, New York.
Eckert, R. E., Jr., F. F. Peterson, and S. T. Belton. 1986. Relationship between ecological-range
condition and proportion of soil-surface types. J. Range. Manage. 39(5): 409-414.
Fuls, E. R. 1992. A technique for objective habitat condition assessment in rangelands. J. Of Arid
Environ., 22:195-198.
Franson, S. (ed). 1993. Environmental Monitoring and Assessment Program: EMAP-Arid Colorado
Plateau Pilot Study - 1992 Implementation Plan. EPA/620/R-93/001. U.S. Environmental
Protection Agency, Environmental Monitoring Systems Laboratory, Las Vegas, Nevada.
Frost, T. M., S. R. Carpenter, and T. K. Kratz. 1992. Choosing ecological indicators: effect of
taxonomic aggression on sensitivity to stress and natural variability. Pages 215-228, in
D.H. McKenzie, D.E. Hyatt, and V.J. McDonald (eds). Ecological indicators. Elsevier Appl.
Sci., London and New York.
Grossmat, R. B., and F. B. Pringle. 1987. Describing surface soil propertiestheir seasonal
changes and implications for management. In: W. C. Reybold and G. W. Peterson (eds.),
Soil Survey Techniques, Spec. Pub. 20 SSSA, Madison, Wisconsin.
Hay, A. M. 1979. Sampling Designs to Test Land-Use Map Accuracy, Photogrammetric
Engineering and Remote Sensing, Vol. 45, No. 4, pp. 529 - 533.
Hobbs, R. J., and H. A. Mooney (eds) 1990. Remote Sensing of Biosphere Functioning.
Springer-Verlag, New York, 312 pp.
Holben, B. N. 1986. Characteristics of maximum-value composite images from temporal
AVHRR data. InternationalJournal of Remote Sensing, 7:1395-1416.
Holben, B. N., and R. S. Fraser. 1984. Red and near-infrared response to off-nadir viewing.
International Journal of Remote Sensing, 5:145-160.
Holben, B. N., C. J. Tucker, and C. J. Fan. 1980. Assessing soybean leaf area and leaf biomass
with spectral data. Photogrammetric Engineering and Remote Sensing, 26:651-656.
Holmgren, G. G. S. 1988. The Point Representation of Soil, Soil Science Society of American
Journal: 52:3, pp 712-716.
Huete, A. R., and R. D. Jackson. 1987. Suitability of spectral indices for evaluating vegetation
characteristics on arid rangelands. Remote Sensing of Environment, 23:213-232.
Hunsaker, C., D. Carpenter, and J. Messer. 1990. Ecological indicators for regional monitoring.
Bull. Ecol. Soc. Amer., 71:165-172.
Justice, C. O., J. R. G. Townshend, B. N. Holben, and C. J. Tucker. 1985. Monitoring the
grasslands of the Sahel using NOAA AVHRR data: Niger 1983. International Journal of
Remote Sensing, 7:1475-1497.
100
-------
Kepner, W. G., and C.A. Fox (eds). 1991. Environmental Monitoring and Assessment Program.
Arid ecosystems strategic monitoring plan. EPA 600/4-91/018. U.S. Environmental
Protection Agency, Washington, D.C.
Kirkland, L. L. In preparation. Quality Assurance Management Plan. EPA 600/XXX/XXX. U.S.
Environmental Protection Agency, Washington, D.C.
Kucera, R. C., and B. E. Martin (eds). 1991. FY 91 Indicator Evaluation Field Study for
Environmental Monitoring and Assessment Program - Forests (EMAP-F), U.S. Environmental
Protection Agency, Atmospheric Research and Exposure Assessment Laboratory, Research
Triangle Park, North Carolina.
Lane, L. J., and M. A. Nearing (eds). 1989. USDA-Water Erosion Prediction Project: Hillslope
Profile Version, NSERL Report No. 2, West Lafayette, Indiana. 265 pp.
Larson, W. E., F. J. Pierce, and R. H. Dowdy. 1983. The threat of soil erosion to long-term crop
production. Science, 219:458-465.
Lillesand, T. M., and R. W. Kiefer. 1987, Remote Sensing and Image Interpretation, 2nd Edition.
John Wiley & Sons, New York. 721 pp.
Ludwig, J.A., and J.F. Reynolds. 1988. Statistical ecology. John Wiley and Sons, New York.
Lugo, A. E., and J. F. McCormick. 1981. Influence of Environmental Stressors upon Energy Flow
in a Natural Terrestrial Ecosystem. In: Stress Effects on Natural Ecosystems, G. W. Barrett
and R. Rosen-(eds.), John Wiley and Sons, New York, New York.
Meeuwig, R. O. 1979. Growth characteristics of pinyon-juniper stands in the western Great Basin.
USDA Forest Service, Intermountain Research Station research paper INT, 238. 22 pp.
Mouat, D. A., C. A. Fox, and M. R. Rose. 1992. Ecological Indicator Strategies for Monitoring
Arid Ecosystems. In: Ecological Indicators (Vol. I and II), D. H. McKenzie, D. E. Hyatt, and
V. J. McDonald (eds). Elsevier Applied Science, New York.
Nemani, R., and S. W. Running. 1989. Testing a theoretical climate-soil-leaf area hydrologic
equilibrium of forests using satellite data and ecosystem simulation. J. Appl. Meteorol.,
44:245-260.
NRC (National Research Council) Committee on Rangeland Classification Board on Agriculture.
1994. Rangeland health: new methods to classify, inventory, and monitor rangelands. Nat.
Acad. Press, Washington, D.C. 180 pp.
NRC (National Research Council). 1993. Soil and water quality: an agenda for agriculture.
Committee on Long-Range Soil and Water Conservation, S. S. Batie (chair). Board on
Agriculture, National Research Council. National Academy Press, Washington, D.C.
NRC (National Research Council). 1990. Managing Troubled Waters: The role of marine
environmental monitoring. National Academy Press, Washington, D.C.
Overton, W. S., D. White, and D. L. Stevens, Jr. 1990. Design report for EMAP Environmental
Monitoring and Assessment Program. U.S. Environmental Protection Agency Environmental
Research Laboratory, Corvallis, Oregon. EPA/600/3-91/053.
101
-------
Perry, C. R., Jr., and L. F. Lautenschlager. 1984. Functional equivalence of spectral vegetation
indices. Remote Sensing of Environment, 14:169-182.
Peters, A. J., B. C. Reed, M. D. Eve, and K. M. Havstad. 1993. Satellite assessment of drought
impact on native plant communities of southeast New Mexico, U.S.A. Journal of Arid
Environments, 24:305-319.
Reichenbacher, F. W., and D. E. Brown. 1992. Biotic communities of North America, Central
America, and the Islands of the Caribbean Sea. Map 1:8,000,000. U.S. Environmental
Protection Agency, Las Vegas, Nevada.
Renard, K. G., and V. A. Ferreira. 1993. RUSLE Model Description and Database Sensitivity, J.
Environ. Quality, Vol. 22, pp 458-466.
Renard, K. G., F. Asce, and J. R. Simanton. 1991. Application of RUSLE to Rangelands. 1m
Watershed Planning and Analysis in Action, R. E. Riggins, E. B. Jones, R. Singh, and P. A.
Rechard (eds.), American Society of Civil Engineers, New York, New York.
Riitters, K., M. Papp, D. Cassell, and J. Hazard. 1991. Forest health monitoring plot design and
logistics study. EPA/600/ - / . U.S. Environmental Protection Agency, Office of Research
and Development, Research Triangle Park, North. Carolina.
Running, S. W. 1990. Estimating terrestrial primary productivity by combining remote sensing and
ecosystem simulation. In: Hobbs, R. J., and H. A. Mooney (eds). Remote Sensing of
Biosphere Functioning. Springer-Verlag, New York. 312 pp.
Sabins, F. F., Jr. 1987. Remote Sensing Principles and Interpretation. W. H. Freeman and
Company, New York. 449 pp.
Schlesinger, William H., James F. Reynolds, Gary L. Cunningham, Laura F. Huenneke, Welsey M.
Jarrell, Ross A. Virginia, and Walter G. Whitford. 1990. Biological feedback in global
desertification. Science, 247:1043-1047.
Snedecor, G. W., and W. G. Cochran. 1967. Statistical Methods. 6th ed. Iowa State University
Press. Ames, Iowa.
Stanley, T. W., and S. S. Verner. 1985. The U.S. Environmental Protection Agency's Quality
Assurance Program. In: Quality Assurance for Environmental Measurements. ASTM STP
867. Pages 12-19. American Society for Testing and. Materials, Philadelphia, Pennsylvania.
Tausch, R. J., and R. S. Nowak. 1991. Prediction of annual evapotranspiration from leaf biomass
on a burned sagebrush site in western Nevada. Abs. Eco. Soc. Amer., August 1991.
Tausch, R. J., and P. T. Tueller. 1990. Foliage biomass and cover relationships between tree- and
shrub-dominated communities in pinyon-juniper woodlands. Great Basin Nat., 50:121-134.
Thornton, K. W., D. E. Hyatt, and C. B. Chapman, (eds). 1993. Environmental Monitoring and
Assessment Program Guide. EPA/620/R-93/012, Research Triangle Park, North Carolina:
U.S. Environmental Protection Agency, Office of Research and Development, Environmental
Monitoring and Assessment Program, EMAP Research and Assessment Center.
102
-------
Townshend, J. R. G., and C. 0. Justice. 1986. Analysis of the dynamics of African vegetation
using the normalized difference vegetation index. International Journal of Remote Sensing,
7:1435-1446.
Tucker, C. J., J. R. G. Townshend, and S. D. Prince. 1986. Monitoring the grasslands of the
Sahel 1984-1985. International Journal of Remote Sensing, 7:1571-1582.
USDA {United States Department of Agriculture)-ARS. 1991. Predicting soil erosion by water - A
guide to conservation planning with the Revised Universal Soil Loss Equation (RUSLE).
K. G. Renard, G. R. Foster, G. A. Weesies, and D. K. McCool, coordinators. Draft.
USDA {United States Department of Agriculture). 1983. National Soils Handbook.
Conservation Service, Washington, D.C. 619 pp.
Soil
USDA (United States Department of Agriculture). 1982. National List of Scientific Plant Names.
Volumes I and II. Washington, D.C. U.S. Department of Agriculture, Soil Conservation
Service. Government Printing Office. SCS-TP-159.
USDA SCS {United States Department of Agriculture Soil Conservation Service). 1992. Soil
Survey Laboratory Methods Manual. Soil Survey Investigations Report No. 42.
Version 2.0. U.S. Government Printing Office, Washington, D.C.
USDA SCS (United States Department of Agriculture Soil Conservation Service). 1981. Land
resource regions and major land resource areas of the United States. Agr. Handbook 296.
U.S. Government Printing Office, Washington, D.C.
USDI (United States Department of Interior). 1985. Rangeland monitoring trend studies. Technical
Reference 4400-4. USDI, Bureau of Land Management, Service Center, Denver, Colorado.
130pp.
West, N. E. 1990. Structure and function of microphytic soil crusts in wildland ecosystems of arid
to semiarid regions. Advances in Ecological Research. 20:179-223.
White, D., A. J. Kimerling, and W. S. Qverton. 1992. Cartographic and geometric components of
a global sampling design for environmental monitoring. Cartography and Geographic
Information Systems, 19(1 ):5-22.
Whitford, W. G. (ed). 1986. Pattern and process in desert ecosystems. University of New Mexico
Press, Albuquerque, NM.
Wood, M. K. 1988. Rangeland vegetation-hydrological interactions. In: Tueller, P. T. (ed).
Handbook of Vegetation Science, Volume 14, Vegetation Science Applications for
Rangeland Analysis and Management. Kluwer Academic Publishers, pp. 469-491.
103
-------
APPENDIX A
SOIL CLASSIFICATION
104
-------
[^1
MM
-JUOS/ARARA
a
LU
a.
MMl
^^^^^^
a
O
<
CC
1
o
3
-LOAMY.MIXED.MESIC USTO
A2-ATCO/BRTE-HIJA-SAKA
LU CO
Z-3
LuCD
^?
«£
Lf
cc
1
1
CO .
in a
o a
CM tn
i
MMMMBMB
Q
a
CC
S
a.
X
g
_i
R
CO
2
0
X
13
W-SKELETAL.MIXED.MESIC
A2-ATCO/BRTE-SAKA-HIJA
< CO
3i
2
<
LU
O
§
a
>^o
5
i
CO
MHM
6*
a
cc
QC
^
0
d
1
0
CO
LU
a"
LU
x
i
Ul
z
U-
MBM
^e
^f
£
cc
^
.
E--
105
-------
BBBBBBI
da:
u» ui
gQ
**
5^
i
5
^^
Lu
Ft
SOIL CLASSIRCATION: DOMINANT VEC
O2
CO UJ
CO
-SB
U- *^
Gr ^J
CL<
LU Ul
Mi
BBBBBBI
o?
CO
M/A: PIED-JUOS
a.
±£ O
° t
O ^3
*_
< CO
J^
en a
0 O
CM r
%
BBBBBBi
°_
CO
CO
O
i
a
Ul
ol
MIXED,(NON-ACID),MESIC LITHIC USTIPSAMMENT:
STRELLVAR.
*o
5»
^BBB
BBBBBBK
s*
CM
MI^^^H*
CO
O
-3
a
UJ
Q.
j^»
Ul
LOAMY.MIXED(NON-ACID).MESIC LITHIC USTORTH
PODOVAR.
e
io
S*
oS
^Bi
JBBBBB1
s2
CM
CM
BBBBBBBBBBBBB.
a
E
E
Q
o
^r
o
LOAMY-SKELETAL.MIXED.MESIC LITHIC USTOLLIC
PIED-JUOS/ARTRT-GUSA2
SHALAKO
VAR.
-8
3r
CO
CO .
a in
SCO
CO
CM CO
^BH^
BBBBBBi
s?
10
BBBBBBBBBBBBB
Q
LOAMY.MIXED.MESIC USTOLLIC LITHIC CALCIOTHI
JUOS-PIED/ARART-GUSA2
SHALAKO |
<£
CO
CO
m
^^^
BBBBBBi
&
BBBBBBMBBBBBBJ
a
x
E
COARSE-LOAMY.MIXED.MESIC USTOLLIC CALCIOf
PIED-JUOS/ARTRT-GUSA2
ANASAZI
#
eo
CO
5
^1
KBBB1
OJ
HWO^HV"
a
X
COARSE-LOAMY.MIXED.MESIC USTOLLIC CALCIOI
PIED-JUOS/ARTEM
ANASAZI
S8
xf
CO
CM
<
^^^^^
^Km
BBBBBBI
CO
^^MMHH
d
£
tc
o
2
^
t3
LOAMY-SKELETAL.MIXED.MESIC LITHIC USTOLLIC
JUOS/ARTEM
SHALAKO
VAR.
<£
s
*
Cl
m
^^^^^^
iBH^^
BBBBBBI
^
CM
BBBBBBBBBBBBB
,
a
?
cc
|COARSE-LOAMY,MIXED,MESIC USTOLLIC CALCIO
| PIED/ARTEM
| ANASAZI
3?
CO
CM
O
^M
BBBB
SS
J2
BBBBB
CC
cc
CO
o
3
-3
|FINE-LOAMY.MIXED,MESIC USTOLLIC HAPLARGID
| BARXVAR.
5?
CO
<
<
^B
en
a
CM
M
BBBB
^
M^MM
| SOIL NOT SAMPLED: JUOS/ARTRV
[(ARTIFACTS)
5
BBi
CM
BBBBBBBI
| SOIL NOT SAMPLED: JUOS/ARTRV
1
O
^MH
BBBBBBI
CO
SANDY.MIXED.MESIC TYPIC TORRIORTHENT:
CHRYS9-EPVI/AGST2-HYFI-CHAEN-ORHY
Ul
Jo
*^i ^^
< CM
CO -
" CO
en co
CM in
BBBBBBi
&
CO
SANDY.MIXED.MESIC TYPIC TORRIORTHENT:
EPVI-GUSA2/SAKA-ORHY-CHAEN
1
v0
_ I O^
m CM
^M
BBBBBi
CO
SANDY.MIXED.MESIC TYPIC TORRIORTHENT:
EPVI-CHRYS9/AGST2-HYFI-ORHY
BBBBBBBBBBH
^
OJ f*».
O CM
BBBHMHBHI
<^
T~ 2
go
CMO
106
-------
u
O
1
(9
UJ
DOMINA
<
o
E
CO
I
u
_J
o
CO
O E
en uJ
CO
Ul UJ
"
QJ
Ul
1
O
CC
cc
g
g
i
o
o
CO
Ul
g,
a
UJ
x
i
ui O
^{T
^Si
2<
-------
acc
Ul Ul
o
I
o
Ul
<
i
a
ASSIFICATI
SOIL
-1
0£
CO uI
CO
ss
n.
Mi
S>'
ui
CM
<
I
Ul
a
a
a.
s
<
co
I
_j
O
CO
in .
co a
0 0
CC
EPT
LE
S
CO
I
CO
CC
O
<:
-3
1
fc
IU
4
cc
8
§
Q.
OIL NOT SAW
CO
MMMM1
MBMNM
CM
o
1MH
MM
Ul
>
3
al
ICTORRI
a.
O
CO
LU
55
§3
~§
8?
XCM
INE-SILTY, Ml
AVE4/BRASS
U. CO
MMMMMMB
3
s
cc
0
CO
CO
CM
O .
T- O
CM "T
MM
^^^^m
in
Ul
^>
""j
B!
ICTORRI
a.
£
o
i
is
Q$
x^
>< m
INE-SILTY, M
AVE4/HAGL-I
U. CO
^^B
5
i
cc
-2
S*
CO
CO
CM
m
iM^
^
tn
^a
LU
>
D
uf
ICTORRI
a.
^
o
CO
Ul
1
0
a
Ul
£
2
S"
CO
Ul
^H
3
I
^o
^f
CO
s
^IM^
^i
2
CM
CCALCIOR
ESIC USTOI
ILIA
Y, MIXED,
PO-EPVI
?s
uiO
COCJ
EC CO
II
1
1
§
<
t
CM
ROCK
OUTCROP
RO
OUTCROP
Ul
I
o
cc
cc
e
g
i
g
x
Zj
g
co
Ul
I
Q
Ul
x
*
g
I
o
i
o
o
I
g
co
ui
Q<
UJd
x<«j
52
sr a.
c; «
ARSE-LO
S/CORA
AS
ft'
OCK
UTC
ROCK
OUTCROP
O
cc
cc
e
o
I
o
MIXED (CALC.), MESIC LI
OA
M
CC
CM
FIREBA
FAM.
108
-------
OC
LULU
5S
e*o
1
E^
2
1
o
: DOMINANT VEi
SOIL CLASSIFICATION:
«
OS
CO uj
CO
«»
|i
i^MMMMM"
^^
t_f I
---^
-.....
^*
CM
BM
Q
E
TYPIC CALCIOR
LOAMY-SKELETAL, MIXED, MESIC
ATCO-CORA/ORHY-HIJA
JLACKSTON
LU
CM O
m co
^^1^
Mi
I
O^
CM
Q
TYPIC CALCIOR
LOAMY-SKELETAL, MIXED, MESIC
ATCO-CORA/HIJA-ORHY
JLACKSTON
^M
«
Oco
^MB
.-.I
WM1
MBMMBl
ccT
UJ
^^
^^^^
^i
CXI
CM
Z
UJ
_l
E
§
1, MESIC TYPIC 1
IY
SANDY-SKELETAL, MIXED (CALC.)
JUOS/ATCA2-ARFR4/AMBRO-ORH
1.
DCC
LU <
CC>
CM
«"
5* O
S!S
CM M
^M
CD
CM
..
t
LU
I
J~
O
cc
cc
R
g
=j
}, MESIC LITHIC
1 LOAMY-SKELETAL, MIXED (CALC.
ARFR4/AQSP
RIZNOVAR.
£
CM
CD
BM"
^.-MB-B
.
£
z
LU
>
i
DC
CC
0
), MESIC TYPIC 1
SANDY-SKELETAL, MIXED (CALC.
FRAN2-AMUT/AGSP-AMBRO
REDBANK
VAR.
CM
5
OBBMBBi
..M.-.-1
.
MMV
MMMB
£"
UJ
1
^^^
CO
E
^^MM
^^a
co
to
i^^^p^^M
9
o
JSTOLLIC CALCI
COARSE-LOAMY, MIXED, MESIC I
PIED-JUOS/CEM02
ANASAZI
t
S
MMHWB
<
JM a
CM 1^-
r^
CO
N^M^^MM
l"~
-9
LU
O
CC
FHICUSTICTOR
LOAMY, MIXED (CALC.), MESIC LI!
PIED/CEMO2-AMUT
O
1
&
CM
O
M
...-*
..--»-
«
«««
a
UJ
a.
a
i
0
USTOLLIC CALC
COARSE-LOAMY, MIXED, MESIC 1
MIVIDA
HIQHRAIN
h-
CM
CM
O
^^^
I^^^BIV
C^
VJ
^
MM*HHMJ
CO
o
^
^
UJ
CL
9
OLLIC HAPLARG
1
o
CO
UJ
Q"
2
I
LU
Z
U.
CQ
O
CM
in .
r O
CM O
CM l*»
«
^
^^M^H
CO
O
7
a
LU
Q.
Q
Q
CUSTOLLICCAI
LOAMY-SKELETAL, MIXED, MESI
CO
LU
CO
fi>
CM
CM
ffl
^
CO
o
ID
^
a
UJ
a.
a
C3
cc
1 USTOLLIC HAP
LOAMY, MIXED, MESIC SHALLOW
^
SOUTHFOR
FAM.
CM
CM
O
<,_:
CMg
cc
UJ
g
CM'
Q
i
^
UJ
ITHICTORRIOR1
LOAMY, MIXED (CALC.), MESIC L
CORA-SSSS/ORHY
LU
MOENKOPI
CO
S!
£.
CM O>
CM «
^l
109
-------
SS
**
1
DOMINANTVEGETA
4>
Z
IRCATIC
SOILCLASSI
jft
DE
CO u!
CO
II
HWB^HMB
Ul Ul
gn
1
CTORRIORTHENT:
i^S
£-
U
CO
ULI
LOAMY, MIXED (CALC.), 1
CORA-SSSS/CRYPT
UJ
a.
UJ
O
"S.
^IM.HMHB.
CO
CM
m
ja
i
CTORRIORTHENT:
X
H;
O
CO
Ul
S
LOAMY, MIXED (CALC.), 1
CORA-SSSS/SAKA
lit
pj «;
IMIIBBBBB
CM
^l^^HM
5:
ICTORRIORTHENT:
-i*
£
O
CO
Ul
2
LOAMY, MIXED (CALC.),
ATCO-EPTO/HIJA
MOENKOPIE 1
'-
^M^MB>»
CO
^.«M>*«
T» M
-------
%VEG 1
COVERd 1
EGETATION6
?&
SOIL CLASSIFICATION: DOMINANT'
o E
CO ui
CO
O"» Wl
0 "*
a **
5*
Ul HI
*
»
1C CALCIORTHID:
-*
LOAMY-SKELETAL, MIXED, MESIC LITHIC USTOl
PIED-JUOS/GUSA2
UELLENTHIh
^5
eo
CM
O
MMM
m^m
c
UJ
1
Ir
8
m^m
mmm^
«
JUOS/CORA
SANDY, MIXED. MESIC USTOLLIC CALCIORTHID:
^
PIRODELFA
«
5
< _
CO O
co in
CM TT
^^^H
^^^H
to
IF"
RIORTHENT:
cc
LOAMY, MIXED (CALC.), MESIC LITHIC USTIC TO
JUOS/CORA
RIZNOVAR.
C^
co
5
1MBB
BH
RIORTHENT:
cc
LOAMY, MIXED (CALC.), MESIC LITHIC USTIC TO
JUOS/CORA
RIZNOVAR.
^^
O5
O
l^MIHB
^
co
6*
o
cc
LOAMY, MIXED, MESIC LITHIC USTOLLIC HAPLA
JUOS/SHRO-CHRYS9-GUSA2
a
1
*
<£
in .
CO CM
CM in
I^^MI
«
LIC CALCIORTHID:
-i
LOAMY-SKELETAL, MIXED, MESIC LITHIC USTO
JUOS-PIED/SYLO-EPVI
z.
MELLENTHI
fr
CM 5>
CD T-
MMi
iHBBB
Z
Ul
! USTIC TORRIORTI
«_/
LOAMY-SKELETAL, MIXED (CALC.), MESIC LITH
JUOS/SHRO-COME5-GUSA2
CO
§
CO
3?
o°
^MM
MMM
MN^MMM
0*
N
5s
ox
CCO
3?
Q m
^^^MH^Hi
^^i
I^^^M
i
^^^MMMI
TORRIORTHENT:
ft
LOAMY, MIXED (CALC.), MESIC SHALLOW USTK
CORA-COMES
SKOSVAR.
^
OJ
<*
<
^
CO .
CO CO
CM in
^^^M
^^i^^
TORRIORTHENT:
f^
LOAMY, MIXED (CALC.), MESIC SHALLOW USTII
JUOS/COME5-CORA
SKOSVAR.
-5
CM
*
en
MH^
s
ENT:CORA-GUMI
>
LOAMY, MIXED (CALC.), MESIC LITHIC USTIFLU
a,
CO
^^
oo
O
^
co z
CM 8
wmm
DUSTIFLUVENT:
| SANDY-SKELETAL, MIXED (CALC.), MESIC LITh
o
CO
0
o^
^
O
RRIORTHENT:
IY
o *-
LOAMY, MIXED (CALC.), MESIC LITHIC USTIC Tl
JUOS-PIED/CEM02-CHVIL4-ARARN/POSE-OF
1
TRAVESSII
ss
a
CM
<
Q
a
§'I~
CO
111
-------
TJ
dec
U Ul
1
5
1J
s
>
5
4
z.
S
O
a
ASSIRCATION:
o
1
A
,co
== Ul
OE
co S
CO
^0
SS
§5
li
o
i
E
Ul
r
f
D
r
r
a
MB
io
a
1
.C.),MES!CSHAI
-ARARN/HEMI2
2: ^*
Y, MIXED
-PIED/CH
1°
Si
u_
£0
DC
LU
CD
i
i
^p
fe
g
Ul
DC
0
cn
DC
o
H-
co
^ 01
II
ll
CO 4-
p
3: 0>
S8
IY, MIXED
-JUOS/CE
li
Ida.
H-
QC
Ul
a
-2
CM
O
-------
APPENDIX B
SOIL DESCRIPTION
113
-------
>oo go
000
OOO
3 $
II
ss
yog g
u
PP~
Is!
ooo
SS
is is
833
§
8
33 S
I I S? $33333 m m S 833 333333
WWW WWW QQQ QQQ QQQ QQQQQQ
.00 en m m BO oo oo « £2 rn
""pa 52 en m »o
-------
I
s
gg
ggg
333
S!
i I
a g
s o
"" 5
Ov tf
HH
55 En
O 1
oo
Igl
sa
IO O OOO
gg
ggj
ggi
|
O O
o
O CJ
88
oo
00.
OOO
S 83S 133 333
QQQ
OO 00 OO
DDD QDQ
00 00 '
Sen i
\o <
itscs
» \O \O
SRR P
vimtn tnmm ovos
.i««nrimcS«e*j «*»«o«n
QQQQQQQQ QQQ QQQ QO
SSS
S?55f
8- H
QQQ
m <*i i
QQQ
**
vi
55SSSSSSS S;
ll
! O
r- S
5 * !
5 § ca < H !
< B 2 oa M i
^ gii
iii
QQQ 000 QOO 00000000 OQQ OOQ OOO OOO QQQ
-------
a a aaaa a a aaa aa;
u
§
to
3
a aa
W to to
i g
gg
pop o o oo 500
to to O CO 03 M ES M
a a 333 a D D
O O 300 000
§ 1 3333 3 3 l3S 333 3!
S3
en en en en en en enenenen enenen enenen enenen enenenen & *r <3r
OOOQOQ OOOO QOQ OOQ QQO OOOO WWW
3333 333 333 333 3333 555
aaaa
SSSS 955 SS3S KSK S22S S fS f2
HH
g,
o!
I H H
QQQOOO 0000 000 000 000 0000 000
ua
»o
«-i en ^* '
ON o\ o% <
oo oo oo i
333 III |s| m| ||S
asSSSS.S||g aaa ass sag aagi
116
U.S. ROVERNMENT PRINTING OFFICE: 550-001/80401
------- |