United States
Environmental Protection
Agency
Environmental Research
Laboratory
Athens GA 30605
EPA 600 3-80-035
February 1980
Research and Development
Probability
Sampling to
Measure Pollution
from Rural
Land Runoff
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ECOLOGICAL RESEARCH series. This series
describes research on the effects of pollution on humans, plant and animal spe-
cies, and materials. Problems are assessed for their long- and short-term influ-
ences. Investigations include formation, transport, and pathway studies to deter-
mine the fate of pollutants and their effects. This work provides the technical basis
for setting standards to minimize undesirable changes in living organisms in the
aquatic, terrestrial, and atmospheric environments.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/3-80-035
February 1980
PROBABILITY SAMPLING TO MEASURE
POLLUTION FROM RURAL LAND RUNOFF
by
F.J. Humenik, D.W. Hayne, M.R. Overcash
J.W. Gilliam, A.M. Witherspoon, M.S. Caller
and D.H. Howe!Is
North Carolina State University
Raleigh, North Carolina 27650
Grant No. R803328
Project Officer
Ray R. Lassiter
Environmental Systems Branch
Environmental Research Laboratory
Athens, Georgia 30605
ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ATHENS, GEORGIA 30605
-------
DISCLAIMER
This report has been reviewed by the Environmental Research Laboratory,
U.S. Environmental Protection Agency, Athens, Georgia, and approved for
publication. Approval does not signify that the contents necessarily
reflect the views and policies of the U.S. Environmental Protection Agency,
nor does mention of trades names or commercial products constitute endorse-
ment or recommendation for use.
-------
FOREWORD
Environmental protection efforts are increasingly directed towards
preventing adverse health and ecological effects associated with specific
compounds of natural or human origin. As part of this Laboratory's research
on the occurrence, movement, transformation, impact, and control of envi-
ronmental contaminants, the Environmental Systems Branch studies complexes
of environmental processes that control the transport, transformation,
degradation, and impact of pollutants or other materials in soil and water
and assesses environmental factors that affect water quality.
Nonpoint sources contribute significantly to water pollution problems
in many areas of the United States. Because of this, more information is
needed on rural nonpoint sources to.guide the development of pollution control
strategies that may be required. Sound information, however, rests on the
development of accurate, cost-effective sampling methods. This report in-
vestigates the use of probability sampling, a technique that has been widely
applied to many scientific and social problems, to describe rural water
quality and evaluates the costs of this sampling method.
David W. Duttweiler
Director
Environmental Research Laboratory
Athens, Georgia
m
-------
ABSTRACT
Our objectives were to investigate the feasibility of probability sam-
pling in describing quality of rural water not affected by point sources
and to study the substantive results. We defined a rural nonpoint source
as the output of a small measurable subbasin that did not contain a point
source.
The study site was in a portion of the Chowan River Basin in Virginia
and.North Carolina. Sampling was in space and in time. In space, four
convenient subareas were used; within each, primary sites were chosen at
random (total 15), four also had automated samplers and stage recorders.
Two sampling times were selected at random from each four-week time stratum.
Flow was measured on all visits along with dissolved oxygen, temper-
ature, conductivity, and pH. Water samples for chemical analysis were iced
for grab samples and acidified for samples from the stage-activated auto-
mated samplers. Samples for kinetic studies were iced and those for algal
studies were kept at ambient temperature.
All samples were analyzed for nitrate plus nitrite nitrogen, total
Kjeldahl nitrogen, total phosphate and chloride (these data subject to
more intensive statistical examination). Chemical oxygen demand, total
organic carbon, orthophosphate, ammonia, and suspended solids were ana-
lyzed on a reduced schedule.
Costs of a one-year sampling.study were estimated (in 1975 dollars)
for grab sampling as: constant (overhead) costs, $14,211.25; per-site
costs, $153.18; and per-visit costs, $79.47. For automated sampling the
constant cost was $14,438.09; the per-site cost was either $5,667.38 or
$7,277.30, depending upon whether weekly service trips were accounted as
per-visit or per-site; per-visit cost was correspondingly $149.01 or
$32.89.
A statistical error of estimation was synthesized; to stabilize the
variance and to provide additivity of effects, analyses were under loga-
rithmic transformation with precision of estimation stated as proportional
standard error. Comparisons were on the basis of relative precision
under a constant budget.
Flow was highly variable in space and time; concentration was less so.
Stating precision as proportional standard error for a $55,000 grab
sampling study with optimal allocation, concentrations were measurable with
fair precision (0.08-0.17); flow and transport were.not (0.50-0.95) al-
though adjusting for area (to yield) somewhat improved the precision
IV
-------
(0.41-0.88). Precision was increased (variance reduced) with larger sample
size (and greater cost) more in proportion to the square root of the increase
rather than directly. Thus, improved precision is costly.
For estimating basin-wide mean values with modest budgets, grab sam-
pling was more flexible and provided better precision of estimation than did
automated sampling under the same budget. With high-budget studies (several
hundred thousand 1975 dollars or more) this relationship may be reversed.
This comparison is stated.on the restrictive basis of estimating mean values;
automated sampling may have other advantages with other kinds of studies.
Optimal allocation, while improving precision, differed with parameter.
An empirical rule-of-thumb based on this study suggests that the number of
sites sampled may be made approximately equal to the number of sampling
visits made annually to each site.
The algal community, of many species, was usually low in biomass, with
fluctuations to be explained by changes in algal input processes rather
than by growth. Growth was limited by high flushing rates and poor light,
but not by nutrients. Potential growth limitation was most frequently by
nitrogen.
The grab sampling data revealed no clear relationships between water
quality parameters and macro land-use factors. Channelized streams had
elevated concentrations of nitrate plus nitrite nitrogen. Measured con-
centrations did not display consistent functional relationships to flow.
Water yields and associated nutrient yields were greater during winter
and spring. The potential that daily weather predictions might be used
for stratification in time to reduce the effect of high flow variability
was presented.
Laboratory oxygen kinetic studies indicated that organic material
entering these streams is relatively unavailable to indigenous organisms,
a fact consistent with high in-stream.levels of dissolved oxygen. Average
rural stream water quality was relatively uniform for the 18-month period
and generally good compared to proposed standards. However, elevated con-
centrations of total phosphate even in the forested Piedmont areas demon-
strated the need for basing water quality assessments on local background
or natural levels.
This report was submitted in fulfillment of Grant #R803328-03-4 by
North Carolina State University, Biological and Agricultural Engineering,
under the partial sponsorship of the U.S. Environmental Protection Agency.
This report covers a period from July 1, 1974 to December 31, 1978, and
work was completed as of October 1, 1979.
-------
CONTENTS
Foreword i i i
Abstract iv
Figures »•> ^x
Tables X1
Acknowledgment XV11
1. Conclusions 1
2, Recommendations • • • 6
3. Study System 8
Project Scope 3
Field Equipment and Procedures 1°
Laboratory Equipment and Procedures 19
Study System Definition 24
Technology Transfer 38
4. Statistical Procedures and Results 41
Introduction 41
Conclusions 42
Recommendations 43
The Use of Sampling 4<
Statistical Model and Optimal Allocation 4b
Cost Analysis 50
Methods of Procedure ^3
Results 64
Discussion 80
5. Associated Water Quality Interpretations <&
Grab Sampling Data and Analyses 33
Subbasin Data Summary 38
Channelization Effects 90
Seasonal Fluctuations 91
Concentration Versus Water Yield 95
vii
-------
Compari sons Among Areas • 98
Point Source Impact 1°3
Headwater Versus Mainstream 1°5
Summary 1 °8
Comparison of Monitoring Techniques 109
Stream Monitoring 110
Computations m
Results and Discussion H3
Summary and Conclusions 121
6. Algal Populations in Two Small Streams 124
Introduction 124
Conclusions 124
Recommendations 125
Sites of Study and Methods of Procedure 125
Results I33
Discussion 169
7. Biochemical Oxygen Demand Studies 174
Testing Procedures 174
References 179
Publications Associated With Project Results 184
Appendix
Physico-Chemical Parameter Values For All 30 Stations Sampled. 186
vm
-------
FIGURES
Number Pa9e
1 Sampler activation switch 16
2 Typical record of a runoff event 17
3 Event recorder circuit 1£
4 Chemical concentration variation with time for various preser-
vation techniques 20
5 Schematic of Chowan River Basin 25
6 Sampling sites in point and nonpoint source study area 33
7 Water yield; precision attainable at different budgets, showing
total variance and contributions of components for grab sam-
pling (solid lines) and total variance for automated sampling
(dashed line) 77
8 Total phosphorus concentration; precision attainable at different
budgets, showing total variance and contributions of components
for grab sampling (solid lines) and total variance for automated
sampling (dashed line) 78
9 Nitrate nitrogen concentration; precision attainable at different
budgets, showing total variance and contributions of components
for grab sampling (solid lines) and total variance for automated
sampling (dashed line) 79
10 Site arithmetic data summary versus land use for grab sampling
(June 1975 to November 1976) 90
11 Site flow weighted concentrations versus land use for grab sam-
pling (June 1975 to November 1976) 91
12 Area] seasonal water yield for grab sampling (November 1974 to
November 1976) 93
13 Areal.seasonal flow weighted concentrations for grab sampling
(November 1974 to November 1976) 94
-------
Number Page
14 N03-N versus water yield at a small (0.5 km2) well-drained
Coastal Plain site for grab sampling (November 1974 to
November 1 976) .................................................. 97
15 Example of rainfall forecasts to predict sampling events .......... 122
16 Winkler azide modification vs. YSI membrane electrode ............. 175
-------
TABLES
Number
1 Laboratory-Chemical Analyses Performed on Grab and Instru-
mented Samples 22
2 Designation and Approximate Area of Sampling Sites 29
3 Land Use 31
4 Example of Field Sampling Schedule (Period 27, Visits 53 and
54) Showing Order of Visit to Sites 58
5 Average Values for Mean Squares and Example (Log-Transformed
Water Yield Data From Poorly-Drained Coastal Plain),
Showing Estimation of Variance Components 62
6 Example Showing Variance Components as Determined From Each of
the 4 Geoclimatic Strata, and Mean Values, for Water Yield
(Data Log-Transformed) 62
7 Small Drainage Basins of the Chowan River System, Listed by
Tributary, Size and Whether Rural (R) Urban (U) 65
8 Precision of Measurement (As Proportional Standard Error to be
Expected With Grab Sampling With and Without Stratification
in Time (Periods), in Each Case With Optimal Allocation of
Sampling Effort, Shown as (No. of Sites/Total No. of Visits
Per Site) and Under 3 Budget Sizes 66
9 Empirical Exploration of Effect on Variance Components of
Changing From 2 Visits (4 Weeks) to 5 Visits (10 Weeks) Per
Period for the Same Set of Data 67
10 Components of Variance and Proportional Standard Error of
Selected Parameters of Water Quality in the Chowan River
Basin; Components Are Averages Based on 15 Sites, Each Visited
52 Times in the Well-Drained Coastal Plain (4 Sites), the
Poorly-Drained Coastal Plain (4 Sites), the Silvicultural
Piedmont (4 Sites) and 38 Times in the Agricultural Piedmont
(3 Sites), Data Log-Transformed 68
-------
Number Page
11 Personnel Costs for a One-Year Grab-Sampling Study of Water
Quality of a River Basin Based on a Probability Sampling
of Sites and Visits; Costs Estimated From Experience in
the Chowan River Rural Runoff Study 70
12 Total Costs (Including Personnel) for a One-Year Grab-Sampling
Study of Water Quality of a River Basin Based on a Probability
Sampling of Sites and Visits; Costs Estimated From Experience
in the Chowan River Rural Runoff Study 71
13 Personnel Costs for a One-Year Automated Sampling Study of Water
Quality of a River Basin Based on a Probability Sampling of
Sites; Costs Estimated From Experience in the Chowan River
Rural Runoff Study 72
14 Total Costs (Including Personnel) for a One-Year Automated
Sampling Study of Water Quality on a- River Basin Based on
a Probability Sampling of Sites; Costs Estimated From Experi-
ence in the Chowan River Rural Runoff Study 73
15 Precision of Measurement (As Proportional Standard Error) to be
Expected With Grab Sampling Using Optimal Allocation of Sam-
pling Effort Under Various Budget Sizes, Showing in Each
Case the Allocation of Effort as (No. of Sites/Total No. of
Visits Per Site) 75
16 Precision of Measurement (As Proportional Standard Error) Attain-
able With a Range of Budgets, Using Optimal Allocation of
Effort With Grab Sampling (Grab) and a Standardized Schedule
of Weekly Visits to Automated Samplers (Auto.) 76
17 Channelization Effects Data Summary, December 1974-December
1976 92
18 Constituent Yield Models Incorporating Land Use and Season 95
19 Constituent Flow Weighted Average Concentration Models Incorpor-
ating Land Use and Season 96
20 Arithmetic Data Summary for Grab Sampling, June 1975-November
1976 99
21 Analysis of Average Stream Conditions for Grab Sampling, June
1975-November 1976 100
22 Flow Weighted Concentration Summary for Grab Sampling Data -
June 1975 to November 1976 101
-------
Page
Analyses of Flow Weighted Average Concentrations for Grab
Sampling Data, June 1975-November 1976 .......................... 102
24 Mean Values of Stream Reach Substudy .............................. 104
25 Comparison of Headwater and Main River Concentrations ............. 106
26 Statistical Survey Measurement Summary for Grab Sampling, June
1975-November 1976 ..................................... . ........ 107
27 Automated Data Record Summary (March 21, 1976-March 19, 1977) ..... 113
28 Comparison of Grab and Automated Samples at Five Sites ............ 114
29 Experimental Study of Automated Sampler at Four Sites ........... .. 115
30 Comparison of Grab and Automated Flow Weighted Concentrations at
Four Sites (March 1976-March 1977) .............................. 116
31 Ratios of Time Average to Flow Weighted Concentrations at Four
Sites (March 1976-March 1977) ............. . ..................... 117
32 Comparison of Grab and Automated Water Yield Values at Four Sites
1976-March 1977) .......................................... 118
33 Grab Sampling Water Yield and Skewness Values (July 1975-November
1976) [[[ 119
34 Daily Precipitation Probability Versus Daily Precipitation at
Edenton, N.C., October 1971-October 1976 ........................ 120
35 Taxonomic Distribution of Algal Species According to Division
and. Class, As Recorded in 211 Collections in Streams F-3 and
P-10, Showing for Each Species the Number of Collections in
Which Recorded, the Mean Cell count and Mean Biomass Per
Col lection [[[ 134
36 Taxonomic Distribution of Algal Species Collected in this Study,
Showing the Proportion Each Class Constitutes of the Total
Numbers and Biomass ............................................. 142
37 Relative Concentration of Algal Biomass Among the Divisions
Chrysophyta, Cryptophyta and All Other Divisions for Stream
P-10 Throughout the Study ....................................... 143
38 The 5 Algal Species Most Abundant by Cell Number and the 5 Most
Abundant by Biomass in this Study, Showing Proportion of Total
-------
Number Page
39 Distribution of 227 Algal Species According to Mean Cell Volume,
by Quarti le and by Order of Magnitude of Cell Volume, Showing
Also the Proportional Distribution of Total Cell Number and
Total Biomass [[[
40 Spatial Variation of Flow and Algal Parameters in Piedmont Stream
F-3; Arithmetic and Geometric Mean Values and Results of Tests
of the Statistical Significance of Subsite Differences for
Sel ected Vari ab les .............................................. ] 48
41 Spatial Variation of Flow and Algal Parameters in Coastal Plain
Stream P-10; Arithmetic and Geometric Mean Values and Results
of Tests of the Statistical Significance of Subsite Differences
for Selected Variables .......................................... I49
42 Spatial Variation of Chemical Parameters of Piedmont Stream F-3;
Arithmetic and Geometric Mean Values and Results of Tests of
the Statistical Significance of Subsite Differences for
Selected Variables .............................................. 150
43 Spatial Variation of Chemical Parameters in Coastal Plain Stream
P-10; Arithmetic and Geometric Mean Values and Results of
Tests of the Statistical Significance of Subsite Differences
for Se lected Var i ab les .......................................... 151
44 Temporal Variation of Algal Numbers, Biomass, and Diversity in
Piedmont Stream F-3; Arithmetic and Geometric Means for 13
Visits, Each to 4 Subsites ...................................... 152
45 Temporal Variation of Flow and Concentration of Nitrate Nitrogen
and Chloride in Piedmont Stream F-3; Arithmetic and Geometric
Means for 13 Visits, Each to 4 Subsites ......................... 153
46 Temporal Variation of Algal Numbers, Biomass, and Diversity in
Coastal Plain Stream P-10; Arithmetic and Geometric Means
for 13 Visits; Each to 4 Subsites ............................... 154
47 Temporal Variation of Flow and Concentration of Nitrate Nitrogen
and Chloride in Coastal Plain Stream P-10; Arithmetic and
Geometric Means for 13 Visits, Each to 4 Subsites ............... 154
48 Algae-Carried Nutrients; Calculated Content of Algal Biomass as
Mean and Maximum Percentage of the Total Organic Carbon, Total
Nitrogen, and Total Phosphorus Determined by Chemical Analysis
-------
Number Page
49 Tests for Regression of Algal Parameters (Cell Number, Biomass,
Diversity, Evenness) on Selected Environmental Parameters
(Calculated Independently at 11 Subsites on Two Streams)
Listing Subsites Where Regression was Statistically Signifi-
cant (p £ 0.05) 157
50 Relationship of Flow, Algal, and Chemical Parameters to Distance
Downstream for Piedmont Stream F-3, Based Upon Covariance
Analysis of Data From 4 Subsites and 19 Visits (Except For
Ammonia With 7 Visits); Data Log-Transformed (Except for
Diversity and Evenness) 158
51 Relationship of Flow, Algal, and Chemical Parameters to Distance
Downstream for Coastal Plain Stream P-10, Based Upon Covariance
Analysis of Data From 4 Subsites With 13 Visits (Except for
Ammonia With 5 Visits); Data Log-Transformed (Except for
Diversity and Evenness) 159
52 Algal Assay -- Results of Statistical Analysis for Maximum
Standing Crop for Piedmont Stream F-3; N (or P) Represents
the Mean Effect of Nitrogen (or Phosphorus) Addition, NP-P (or
NP-N) the Mean Added Effect of Nitrogen and Phosphorus Together
Over Phosphorus (or Nitrogen) Alone, and INT Represents Inter-
action, or the Excess Effect of Adding Nitrogen and Phosphorus
Together Over the Sum of the Separate Effects 162
53 Algal Assay -- Results of Statistical Analysis for Maximum Stand-
ing Crop for Coastal Plains Stream P-10; N (or P) Represents
the Mean Effect of Nitrogen (or Phosphorus) Addition, NP-P
(or NP-N) the Mean Added Effect of Nitrogen and Phosphorus
Together Over Phosphorus (or Nitrogen) Alone, and INT Represents
Interaction, or the Excess Effect of Adding Nitrigen and Phos-
phorus Together Over the Sum of the Separate Effects 163
54 Algal Assay -- Results of Statistical Analysis for Maximum
Specific Growth Rate for Piedmont Stream F-3; N (or P) Repre-
sents the Mean Effect of Nitrogen (or Phosphorus) Addition,
NP-P (or NP-N) the Mean Added Effect of Nitrogen and Phosphorus
Together Over Phosphorus (or Nitrogen) Alone, and INT Repre-
sents Interaction, or.the Excess Effect of Adding Nitrogen and
Phosphorus Together Over the Sum of the Separate Effects 165
55 Algal Assay -- Results of Statistical Analysis for Maximum Spe-
cific Growth Rate for Coastal Plains Stream P-10; N (or P)
Represents the Mean Effect of Nitrogen (or Phosphorus) Addition,
NP-P (or NP-N) the Mean Added Effect of Nitrogen and Phosphorus
Together Over Phosphorus (or Nitrogen) Alone, and INT Repre-
sents Interaction, or the Excess Effect of Adding Nitrogen and
Phosphorus Together Over the Sum of the Separate Effects 166
xv
-------
Number
56 Algal Assay — Selected Examples From Streams F-3 and P-10 With
Statistically Significant (p $ 0.01) Interactions, Showing
Means for the 10 Treatments for Maximum Standing Crop and
Maximum Specific Growth Rate 167
57 Relationship of BOD to COD and BOD to TOC 178
58 Analysis of BOD Variation 178
xvi
-------
ACKNOWLEDGMENTS
The direction, encouragement, and perseverance of Dr. Ray Lassiter,
Project Officer, and associated EPA staff have been extremely helpful in the
completion of this project.
The continuous support of participating Schools and Departments at
North Carolina State University, including the North Carolina Agricultural
Research Service; the North Carolina Agricultural Extension Service; Depart-
ments of Statistics, Zoology, Botany, and Civil Engineering; as well as
the Water Resources Research Institute, were most important.
Special appreciation is due D. H. Howells, Past-Director of the Water
Resources Research Institute of North Carolina, for his leadership in both
the conception and execution of this interdisciplinary project in an in-
creasingly important area. Special appreciation is most appropriate for the
staff of the Water Resources Research Institute which handled administrative
and budgetary responsibilities associated with this project, especially
Linda Kiger, Administrative Assistant; Rose Wilson, Accounting; and N. S.
Grigg, Director.
A.major portion of the field and laboratory responsibilities were
supervised and handled by the following project assistants. From the
Departments of Statistics and Zoology, Paul Geissler for work on statis-
tical sampling; R. M. Burr for data management; T. R. Fisher for algal
analyses. From the Botany Department, 0. Boody, R. Pearce, and P. Boody
for algal analyses. From the Civil Engineering Department, S. G. Wardak
and D. W. Blackburn for the oxygen uptake and kinetic studies. From the
Biological and Agricultural Engineering Department, F. A. Koehler, field
operations; June Preston, laboratory testing supervision; and L. F.
Bliven, water quality analyses.
Special credit was well earned by the many technicians who demonstrated an
outstanding commitment to the multiplicity of field and laboratory respon-
sibilities associated with this project.
xvi i
-------
SECTION 1
CONCLUSIONS
STATISTICAL PROCEDURES AND RESULTS
Conclusions of this study must be drawn within the constraints of our basic
assumptions: that we are to determine water quality for the output of small
drainage subbasins as river-basin averages of such quantities as concentrations
of chemical parameters, or rates of flow or transport or yield of water or
chemicals; that reference is to a one-year study; and that the basic objective
is the most precise information for the available budget. In these terms,
these conclusions follow from our study of a portion of the Chowan River Basin.
When identified as the output of small drainage subbasins, rural runoff
can be measured by use of probability sampling in space and time. Concentra-
tions were measurable with good precision. Flow and other measures, such as
transport or yield of chemicals which included the flow as a component, were
not measurable with good precision, although adjusting flow or transport to
yield by taking account of basin area improved precision somewhat. These
conclusions hold whether grab sampling or automated is to be used.
Precision was difficult to attain for flow or related measures for two
reasons. First, there was high inherent variability both in space among small
drainage subbasins and in time at any site. Second, and this is true also of
measuring concentrations, sampling was required in two dimensions; and because
both the space and time components of variance were important, neither was
reduced in direct proportion to increased total sampling effort. Rather, each
and their sum was reduced in rough proportion to the square root of the total
sampling effort. Thus, for flow not only was the variability inherently high,
but precision responded slowly to increased sampling effort. For concentra-
tions, the variability was not so high and better general precision could be
attained even though the same requirement for two-dimensional sampling exist-
ed as an impediment when attempting to increase precision. These problems
concerning variability exist by the nature of sampling river basins; they do
not arise because probability rather than judgment sampling is used.
We found that grab sampling promised better precision than automated
sampling for the same cost at modest budgets. Grab sampling is more flexible
and adjusts easily to the need for more sampling in space. In comparison,
automated sampling is costly for sampling in space but provides low-cost
sampling in time at any site where installed. Speculation suggested that
automated sampling would generally become more cost effective than grab sam-
pling at high budgets (several hundred thousand 1975 dollars or higher).
-------
We found stratification in time generally worth the effort in terms of
increased precision, but it is possible to over-stratify and reduce efficiency
by planning too-frequent visits to too-few sites.
Optimal allocation of sampling effort is important to efficiency and may
be calculated given information on costs and variances. We found, however,
that optimal allocation differs with the parameter. Because a survey can
rarely be planned for a single parameter, this means that the sampling effort
must be allocated to be of median efficiency on the average but probably near
optimal for only a few parameters. Based on our experience, such a compromise
design might have the number of sites sampled approximately equal the number
of visits per year to each site, although this entirely empirical suggestion
must be re-examined when more information is available.
ALGAL POPULATIONS IN TWO SMALL STREAMS
In two small streams the algal community included many species with
only a few dominant forms.
Algal biomass was usually low but reached high levels occasionally. Ex-
planation for fluctuations in biomass must be sought in processes bringing
algae into the flowing water from the bottom or from nearby water or land,
rather than in growth of the ambient plankton itself.
Algal population levels in these headwater streams are limited by fluc-
tuations in these unknown input processes, by high flushing rates, and possi-
bly by poor light, not by nutrients; ambient populations are only a fraction
of the potential levels realized in laboratory assay and possibly in lower
parts of the same stream.
Algal assays indicated potential growth limitation most frequently by
nitrogen, next by factors other than nitrogen or phosphorus, rarely by phos-
phorus alone. Such potential limitations apply in reality only at some
hypothetical downstream location and condition, not to populations present
in these small streams.
Within-stream fluctuations in potential nutrient limitation suggest a
within-stream heterogeneity in nutrient balance.
GRAB SAMPLING WATER QUALITY INTERPRETATIONS
Grab sample data from a two-year, statistical sampling study in arbi-
trarily selected areas of the forested and agricultural Piedmont, plus well-
and poorly drained Coastal Plain, all in the Chowan River Basin, were ana-
lyzed to permit a clearer understanding of the nature of rural runoff on an
areawide basis. These analyses provided the following conclusions:
Neither in-stream (arithmetic average) nor net export (flow
weighted) concentration data presented any clear relationships
between water quality and macro land-use factors. This result
points out the need for caution when employing model predictions
-------
to specific cases. Macro land-use factors do not account for varying
agricultural cropping and management practices, annual weather con-
ditions, stream border buffer systems, or other factors which can
minimize the impact of agricultural activities on water quality.
The impact of channelizing Coastal Plain streams was most pro-
nounced with respect to elevated N03-N concentrations in these streams
as compared to unchannelized streams which have natural swampy flood
plains and channels which increase N03-N attenuation by denitrification
and biological uptake.
Analyses for seasonal trends indicated that water yield and the
associated nutrient yields were greater during the winter and spring
seasons than during the summer and fall seasons reflecting rainfall
and evapotranspiration cycles. The model to evaluate relationships
between flow weighted concentration and both season and land use
showed that TKN and TP concentrations were related to season and N03
concentrations to both season and flow but associated r2 values were
low.
Measured concentrations did not display any consistent functional
relationship to flow (water yield) levels. However, data showed that
N03-N concentrations were elevated during flow conditions at a small
(0.50 km2 or 0.2 mi2) site but not at larger (20 km2 or 8 mi2) sites
in the well-drained Coastal Plain with similar land use. The N03-N
attenuation was judged to be the result of in-stream dynamics.
Comparisons of geoclimatic areas demonstrated that the dominant
variation was between the Piedmont and Coastal Plain with only rel-
atively minor variations occurring within these two physiographic
regions. The differences between the Piedmont and the Coastal Plain
were judged to be due to naturally occurring physiographic variations
in (a) basin characteristics, such as vegetation, soil type, and
steam hydraulics; and (b) ocean proximity.
The heterogeneous land-use mix and vegetative borders around
receiving streams characteristic of the Southeast in conjunction
with different geoclimatic conditions for studied regions are major
factors in obscuring clearer relationships between land use and water
quality for the investigation intensity employed.
Results from laboratory oxygen kinetic studies indicated that
organic material entering these streams from rural inputs is relatively
unavailable to indigenous organisms, a fact further verified by the
high in-stream mean dissolved oxygen levels.
Assessment of point and nonpoint source impacts in one small
basin verified classic point source concentrations spikes with sub-
sequent decline to intermediate levels for all investigated con-
stituents except chloride and nitrate. Therefore, for the stream
reach studied, nitrogen and phosphorus inputs which appear to come
-------
from treatment plant effluents are reduced to headwater background
levels as long as the stream assimilatory capacity is not overwhelmed
or natural inputs do not result in changed background water quality.
Point-in-time comparisons between headwater and downstream chemi-
cal parameters showed small differences.
Comparison of annual.water yield estimates from the 15 statistical
survey sites to historic values for the study region verified that
simple time stratified (STS) grab sampling provided an unbiased esti-
mate of areawide annual water yield; however, the precision of the
individual estimates was low. Sampling theory indicates that increased
estimate precision of heterogeneous populations, such as measurements
of water yield, can often be achieved by stratification. Thus, the
potential of employing daily weather predictions as a means of time
stratification to improve water yield estimate precision at a given
sampling intensity was reviewed.
Eighteen-month average rural stream water quality for the four
geoclimatic areas was relatively uniform. The quality was generally
good compared to proposed standards, but elevated TP concentrations
even in the forested Piedmont area demonstrated the need for basing
water quality assessments on measured local conditions, especially
background or natural levels. Stream sample concentrations usually
displayed large variations indicating that rural nonpoint sources are
highly variable in both space and time.
WATER QUALITY DATA COMPARISONS FOR GRAB AND AUTOMATED SAMPLING
Statistical sampling methods can be employed to measure the mean area-
wide, contribution of chemical species from rural nonpoint sources as an alter-
native to the more difficult and often impractical complete monitoring
approach. Grab sampling and automated sampling are two common methods of
assessing stream water quality that can be employed in a statistically de-
signed sampling program. Automated data was collected at 5 of the 15 grab
sites. Long-term grab data were obtained from November 24, 1974 to March
19, 1977, and automated data from May 18, 1975 to March 19, 1977. All
automated and grab sampling sites were operational from March 21, 1976 to
March 19, 1977. During these periods substudies were conducted to inves-
tigate particular sampling methodologies. The resulting data comparisons
for grab and instrument sampling support the following conclusions:
Point-in-time comparisons of manually depth integrated and fixed
point automated samples were conducted for COD, TOC, TP, TKN, N03-N
and Cl. One substudy showed that manually depth integrated samples
provides lower estimates of COD, TOC, TP, TKN, and Cl concentrations
(36, 39, 28, 18 and 8 percent, respectively) than the first sample ob-
tained from fixed point automated samplers in routine stage-activated
operation when both samples were preserved by refrigeration until
laboratory analysis.
-------
Data from a study to determine the relative contribution of sample
procurement, preservation, and storage on automated sample parameter
concentrations indicated that the sample-flush sampling cycle produced
a "first" sample effect; that is, the first sample obtained from a
sampling which was inactive for approximately a week had about 60 per-
cent higher concentrations of COD, TOC, TP, and TKN than samples taken
immediately thereafter. About 20 percent lower concentrations of COD
and TOC were observed for manually depth integrated samples preserved
by cooling compared to fixed point automated sample preserved by acid
fixation. Finally, acid-fixed automated samples which were stored in
the field for a week had about 20 percent lower COD and TOC concentra-
tions than acid-fixed automated samples returned immediately to the
laboratory for analysis. These data indicate that COD and TOC were
the parameters impacted most when recommended procedures were employed
in routine data acquisition. Furthermore, the sample-flush cycle caused
a first sample effect that had a significant impact on several parameter
concentrations.
Annual volume average concentration and water yield estimates were
obtained at four sites by routine operation of both stage-activated
automated samplers with stage recorders and simple time stratified grab
sampling. Results indicated that about 50 percent lower COD, TOC, and
TP estimates were obtained by the grab sampling method. Although rather
large differences were observed for the water yield estimates, the data
did not indicate any statistically significant difference between the
two methods. Due to confounding factors associated with the two sam-
pling methodologies, it was not possible to specifically define the
source of the concentration differences.
-------
SECTION 2
RECOMMENDATIONS
STATISTICAL PROCEDURES AND RESULTS
Probability sampling in space and time should be employed when determining
river basin average values, in particular, averages of the contribution of
small headwater subbasins to a river system.
For short-term (ca. one year) studies under modest budgets (less than
100,000 1975 dollars), the use of grab sampling methods should be given seri-
ous consideration on the basis of cost effectiveness. The longer and more
costly the study, the more the balance shifts toward favoring automated sam-
pling, especially in studies of flow, transport, and yield.
A purely empirical suggestion, based upon current information, is that in
such a one-year study based on grab sampling, the sample number of sites
should be approximately equal to the number of annual visits made to each
site,, This rule-of-thumb must be reexamined with a broader data base.
Time strata should be delineated in such a one-year survey design after
the number of visits has been determined, with two visits per time stratum.
Relatively good precision may be expected of sampling studies of concentra-
tions.
ALGAL POPULATIONS IN TWO SMALL STREAMS
The origin of planktonic algae in small headwater streams should be fur-
ther studied to determine the relative contribution from the periphyton,
from bodies of standing water, and from the soil surface.
WATER QUALITY INTERPRETATIONS
Statistical sampling methods can be employed to measure the mean area-
wide contribution of chemical species from rural nonpoint sources as an
alternative to the more difficult and often impractical complete monitoring
approach. Grab sampling and automated sampling are two common methods of
assessing stream water quality which can be employed in a statistically de-
signed sampling program. Water quality based results of the grab and auto-
mated sampling comparisons provide information which can be utilized to design
sampling plans that minimize the variance attributable to sampling methodology.
-------
Interpretation of data and experiments to compare grab and automated
sampling procedures led to the following recommendations to reduce noted re-
sult differences:
A flush-sample cycle rather than a sample-flush cycle should be
employed with automated sampling so that material that may have
settled or grown in the sampler intake can be expelled before a
sample is obtained. This should minimize the elevated first sample
TOC, TP, and TKN concentrations that were observed with the sample-
flush operational mode. However, data are needed to determine the
extent to which a flush-sample cycle may reduce this bias.
Chemical preservation techniques should be considered when
refrigeration is not feasible. If acid-fixation is employed, field
storage time should be as short as possible in order to minimize
chemical oxidation of the organics that can reduce TOC and COD con-
centrations.
The relative impact and potential interactions of a fixed intake
for automated samplers versus depth integrated grab samples, as well
as different storage and preservation techniques upon both soluble
and filterable type constituent measurements, needs further examination.
The effect of sampling methodologies on water quality measurements in
no way lessens the need to establish adequate data bases in order to provide
a foundation for sound scientific assessment of rural nonpoint sources. The
information presented should aid future investigators and monitoring agencies
in better understanding stream monitoring data and designing improved data
acquisition systems.
The relatively similar arithmetic average concentrations for sampling
sites receiving rural inputs from different land use and geoclimatic regions,
and in the main river draining these sites, indicate that more data on back-
ground conditions and relative impact of nonpoint sources are needed before
widespread implementation of best practices is required, particularly in areas
with heterogeneous land use. It also seems most important that agencies de-
veloping regulatory criteria be responsive to ambient conditions and not
entertain standards requiring better than background water quality, parti-
cularly for relatively undisturbed or pristine areas such as the forested
region of this study basin.
-------
SECTION 3
STUDY SYSTEM
PROJECT SCOPE
We undertook this study in response to the growing demand for information
on the influence of rural nonpoint sources on water quality. Policy makers
need this information to define the status of rural nonpoint sources, to guide
control strategies (if required), and to provide data for model development.
Further, a method for cost analysis of various methods and designs for samp-
ling is required to guide efficient measurement.
The objectives of our study were to investigate the feasibility of using
probability sampling in describing rural water quality not affected by point
sources on a river basin-wide basis and to examine the substantive results of
this sampling effort for insight on currently important questions. These ob-
jectives required that we define a rural nonpoint source; in this study we
have taken this to be the output of a small measurable subbasin that does not
contain a point source.
Appraisal of water quality in these very small drainage units of a com-
plete river basin requires that sampling be used, both in time and in space.
The cost of the continuous coverage of every such small basin in even one
river drainage would be prohibitive; partial (sampling) information must be
used. Sampling may be carried out either as judgment sampling or as prob-
ability sampling; our objective was to explore use of the last.
Historically, judgment sampling has dominated the study of water quality.
Sites and times for sampling are selected by professional judgment as typical
of the kind of stream being studied. The process is characterized by sub-
jective judgment in selection of the study material and in this fact lies
both the strength and the weakness of the method.
The advantages of judgment sampling are that the sites and times chosen
for sampling study can be selected by the investigator according to his ex-
perience and judgment as those that he considers best suited for the program
needs, technical requirements, and best use of sampling resources. The dis-
advantages of the system are the introduction of personal bias in the selec-
tion process, the difficulty of defining a sampling error, the fact that the
selection process is not reproducible by another investigator because of the
personal element, and the fact that the results cannot be extrapolated for
lack of a defined universe. Thus, such studies must remain a series of spe-
cial cases.
-------
Probability (or random) sampling has not been much used in the study of
water quality. Here, after a rigorous definition of the scope of the problem,
the sites and times of sampling are selected with known probability from a
list (frame) of all possible. Often in defining the larger problem it may be
subdivided (stratified) into smaller nonoverlapping but all-inclusive parts,
each of which is sampled independently.
Probability sampling has been widely applied to many scientific and so-
cial problems. Its advantages are that unbiased estimates may be derived by
methods that are reproducible; that inference may be made from the sampling
results to the defined universe; that statistical sampling error can be esti-
mated; and that with certain assumptions about statistical distribution, con-
fidence limits may be set about the estimates. The ability to estimate samp-
ling error provides a rational basis for considering the optimal allocation
of sampling effort.
Disadvantages of probability sampling are that without effective stra-
tification, too much of the field effort for maximum efficiency may be spent
at. the more numerous sites and times of generally low flow. Some randomly
selected sites may seem "nonrepresentative" to the field man (but every site
is unique in some way). Some randomly selected sites cannot be sampled (but
such sites should not have been included in the sampling frame if they cannot
be studied). Semi-synchronous sampling will generate correlations among site
records (but this may be true of judgment sampling as well).
Either an automated installation or grab (hand) methods may be used to
take water samples and measure flow. Either method may be used with judgment
sampling or with probability sampling. An automated installation is costly,
but it can provide a continuous record of stream flow and a programmed series
of water samples. Grab sampling costs less per site sampled and is therefore
a more flexible method when covering many sites, but it provides information
about flow and water quality only for the time of the technician's visit. We
investigated both methods to determine which was the more cost effective for
use with probability sampling under conditions of our study.
The cost of information determines the scope of any appraisal of water
quality. Demand for the information has risen faster than the budgets for
obtaining it, and therefore the most cost-effective sampling plan should be
used for the measurements whether they are made by automated or by grab
methods. To determine cost effectiveness some criterion of effectiveness is
needed; we have used precision of the estimate, a concept widely applied in
sampling surveys. By examining the variance components, we may judge whether
better precision can be attained; for example, by using more sites and making
fewer measurements at each, or by making more measurements at fewer sites,
with the same budget in either case. The approach required that we make an
analysis of all the costs of an investigation of water quality using prob-
ability sampling.
The data gathered in this study have also been examined for the infor-
mation they provide on water quality in the selected portions of this single
river basin. Thus the field data, which were necessary for the feasibility
study, served a dual role in also providing insight to rural water quality.
-------
This analysis included the information from studies of algal populations and
on other project substudies, as well as on water quality and land use.
FIELD EQUIPMENT AND PROCEDURES
Grab
Efforts were focused in this study on obtaining the most reliable physi-
cal and chemical data possible within given time and manpower restraints.
However, because this was a feasibility study of a methodology which might
be adopted by other agencies rather than a pure research effort to character-
ize a system, extensive site preparations or modifications for sampling were
not undertaken. Additionally, extensive physical modification of a natural
stream, especially an elaborate control structure, might well cause changes
in the stream's flow regime and quality, such as ponding in a weir pool with
a concomitant settling of particulates and absorbed nutrients.
Grab sampling was initiated on November 24, 1974. Prior to this time
procedures for making each measurement were established after which field
crews were trained. Routine grab sampling was terminated on November 27, 1976.
This afforded a full two-year data record for all sites. The overall stream
side procedure is separated into three steps: (1) Make water quality measure-
ments, (2) obtain water quality sample, and (3) measure stream flow. This
order is important, particularly in slow flowing Coastal Plain streams be-
cause dislodging bottom sediments by walking in the stream bed would cause
spurious results. The water quality measurements were made from the bank on
small streams or from a bridge on large streams.
Quality control was given considerable attention during the course of
this study, including field procedures. All technicians were given identical
training following a detailed written outline to ensure that all measurements
were made in the same manner. Additionally, retraining sessions were held
bimonthly to ensure that all personnel were using consistent methods.
All meters and instruments were subject to routine maintenance. The
schedules and steps for each piece of apparatus were written out and each
technician was assigned specific tasks to perform. A checklist aided in
establishing timely attention. This was of great benefit in minimizing
breakdowns and loss of data.
Flow--
Because stream flow was felt to be the least precise measurement made,
attention was focused on obtaining the best estimates possible. Two gen-
eral stream types were present, Piedmont and Coastal Plain. The Piedmont
streams in general have well-defined channels and velocities in a range
suitable for current meter measurement. The Coastal Plain streams were in
general sluggish and had poorly defined channels except in those areas where
channelization had occurred. These streams were extremely difficult to
measure.
10
-------
For the most part, methods recommended by the United States Geological
Survey (USGS) were employed in making flow measurements. Extensive use of
sand bags to construct restricted measuring sections was employed where
velocities were below detention limits, as well as permanent modifications at
some stations. These modifications are described in the section on site
description.
Price Meter--A Price meter manufactured by the Weathermeasure Corpora-
tion was employed during the course of this study. This meter has the advan-
tage of electronic readout directly in meters per second rather than the
method of generating clicks and amplifying them to a head phone while simul-
taneously recording time intervals with a stop watch. The meter was routinely
checked during the course of the study; both upper and lower bearings, as well
as a cup wheel, were replaced about midway through the data collection. Fol-
lowing this the calibration was checked against two other meters. No devia-
tion from the original rating curve was noted, within the precision obtain-
able. The Price meter had a somewhat higher starting velocity than the Ott
meter, and therefore was used only when the stream could not safely be waded.
In general, the meter performed in a satisfactory manner.
Ott Meter--The Ott meter is a horizontal shaft instrument manufactured
in Germany and has proven ideal for measuring the small, slow flowing streams
in this study. The meter was equipped with the optional component propeller
which automatically accounts for angularity of flow. In addition, this
propeller has a lower limit of about 3 cm/sec for reliable measurement. The
meter is used with an electromechanical counter in conjunction with a stop
watch and greatly reduces possibility of human error in comparison with the
method of manually counting clicks.
Pigmy Pr i ce Mete r--The Pigmy Price meter had a much higher starting speed
than the Ott meter and in addition, because of low velocities, it was extrem-
ely difficult to determine the angle of flow. This meter used the clicks-
count method for velocity measurements with attendant higher chance of human
error. It was felt that this meter was much less suitable than the Ott meter
under identical conditions, and therefore the meter was used only in a back-
up role.
Float Measurements—On many occasions, velocities, even in restricted
sections, were too low to measure by current meter methods. In addition, on
several occasions the level of water was too low to submerge the meter in
the stream. Under these conditions float measurements were made, generally
with small wooden sticks over a 2 to 3 m (5 to 10 ft) section.
Volumetric Measuremerit--Volumetric measurements were sometimes possible
using a calibrated bucket and stop watch. This technique is perhaps the
most precise used; but its application is limited since there must be a place
to position the bucket to catch the flow, and the upper limit of flow per
measuring point is about 10 liters per second. Several measurements can be
made across a flow drop, but this substantially reduces the precision of the
measurement.
11
-------
Estimation Method—When conditions precluded any of the preceding pro-
cedures, estimation methods were employed. This was generally true under
extremely low flow conditions. The technique employed was to create a small
shallow channel and then to visualize the filling of an appropriate sized
bottle over a one-minute period. Field crew training was accomplished by
utilizing troughs in the chemical laboratory and after some practice flows
could be estimated to within 10-15 percent.
Stage—
At the time the stream flow measurement was taken, stage was recorded for
development of a stage-discharge relationship. This served two functions.
First it provided an alternate low cost stream flow measuring technique after
the stage-discharge relation was developed, and second it established a his-
toric data base pursuant to translation of stage recorder data to stream flow
estimates at the automated sites.
Physio-Chemical Measurements —
Dissplved Oxygen--All dissolved oxygen (DO) measurements were made using a
commercial polaragraphic type instrument, calibrated before each measurement
with air saturated water. Distilled water was used for the calibration and
was kept in the back of the field van in order to maintain the calibration
temperature with 5 °C (9 °F) of the stream temperature. Where this was not
the case, the sealed calibration bottle was placed in the stream and allowed
to equilibrate until it was within 5 °C (9 °F) of the stream temperature.
Previous experience in our laboratory and many others has shown close agree-
ment with the Winkler-Azide method, and therefore no further comparisons
were made during the present study. The measurement was made at approximate
mid-depth in the center of flow of the stream. Experiments showed that
occasionally in eddys at the side of the stream low DO measurements were
recorded. Therefore, this situation was always avoided. As with all in-
struments used in the study, routine maintenance was accomplished by follow-
ing a written schedule. One problem experienced was fouling of the membrane
by dissolved organic constituents in Coastal Plain streams. Apparently some
constituents crossed the membrane and over a period of time poisoned the
cathode; subsequently a new probe was put into service.
Temperature—Temperature was measured both as a parameter and for
calibrating various instruments. Initially all meters used were compared to
carefully calibrated laboratory thermometers. Discrepancies were noted and
a correction was made for each meter. Periodically this correction was re-
checked against the laboratory thermometer.
Conductivity—Measurements were made using a commercial field model of
the standard Wheatstone bridge meter with the cell encapsulated in plastic.
The measurement was made at approximately the same stream position as the
dissolved oxygen measurement. Before each field trip the instrument was
calibrated with 0.1 normal potassium chloride solution at 25 °C following the
manufacturer's specifications. As with the dissolved oxygen meter, the cell
apparently was fouled by the"high level of dissolved organic materials in
the Coastal Plain streams; and therefore, the cell was periodically cleaned
using the manufacturer's recommended method.
12
-------
£H—Measurements were made using a commercial field pH meter. The meter
was calibrated before each use against buffers of pH 4 and 7. Once again,
dissolved organic materials appeared to cause interferences with the measure-
ment and assumedly were fouling the glass membrance of the pH electrode.
Therefore, whenever proper calibration against the two buffers was not possible,
the electrode was cleaned following the manufacturer's recommendations and then
recalibrated. All measurements were trade at mid-depth at the center of flow.
Sample Procurement—
Chemical Samples—Grab samples were obtained using a manually-depth inte-
grated method at the center of flow. An early experiment had shown that cool-
ing to a temperature of below 4 °C was an acceptable preservation method for
all parameters measured. All samples were collected and stored in preaged,
acid-washed plastic bottles. Immediately after collection all samples were
placed in ice and kept cold until analyzed.in the laboratory. An early ex-
periment had shown that there was no observable difference in values for tests
to determine nitrogen, phosphorus, and oxygen demand between manually depth
integrated sampling and using a more sophisticated depth integrating sampler.
Therefore, the manual depth integrating technique was used throughout the
study.
Sediment Samples—As with the chemical samples, the sediment samples
were taken using a manually depth integrated technique. Calibration with a
USGS DH-49 depth integrating sampler had shown no significant differences.
Because of the size of sampled streams, samples were generally taken at only
one vertical in the center of flow. These samples were also iced and re-
frigerated until analysis.
Algal Samples--Two types of algal samples were collected. One type was
for a" productivity study and the second type was for characterization of
standing populations. Those collected for productivity studies were treated
in the same manner as the sediment and chemical samples. Those collected
for characterization of the standing populations were fixed immediately after
collection by adding 30 milliliters of Lugol's solution to a 200 milliliter
water sample in an opaque brown bottle. These samples were not iced but
kept at ambient temperatures.
Kinetic Samples—These samples were designated to be run for biochemical
oxygen demand (BOD) and oxygen uptake studies as well as nutrient recycling
studies. Because of this, care was taken to ensure that the bottles were
completely filled with water and all air was excluded. As above, all samples
were iced and refrigerated until analyses.
Automated Sampling
One of the prime reasons for installing the automated sites was to make
a comparison of continuous stage recording versus grab.sampling. A second
was to study transport characteristics during runoff events. It was also
hoped that these sites might serve as "field truth stations" to assess the
overall effectiveness of the grab sampling scheme. All instrumented stations
13
-------
were operational on July 19, 1975, and operation was ceased on March 19, 1976,
affording an 18-month data record. Grab sampling was continued at these sites
after November 17, 1976, to have comparable data records.
Stage Recording--
Continuous stage was recorded on an analog recorder and later transformed
to digital computer input form. A stilling well and instrument house were
installed with design assistance from local USGS personnel. Initially, natural
controls were used at all five sites. However, stream bed shift problems were
experienced early in the study, and at three of the five sites low height arti-
ficial controls were installed to better define the stage discharge relation
under base flow conditions. Because flow was measured at each grab sampling
visit, there were as many as 52 observations for establishing a stage discharge
relation. As personnel schedules allowed, additional visits were made during
storm events to secure high flow data.
At one Piedmont site there was an active beaver colony. The dams were
removed in an effort to discourage the colony; however, they were soon rebuilt.
Virginia officials cooperated in trapping these beavers and once again the
dams were removed. However, soon thereafter a new colony of beavers began
establishing dams below the instrument station. Eventually this site had to
be abandoned because of beaver activities and the resulting continuous change
in the stage discharge relationship. This might well be a frequent occurrence
on small rural streams; and if beaver activity is noted in a proposed monitor-
ing watershed, serious consideration should be given to selecting an alterna-
tive site.
Rainfall Measurement—
Rainfall intensity data was gathered using tipping bucket rain gages
which were serviced weekly and these performed quite adequately. Initially
attempts were made to gather quality data by catching rainfall in airlock
sample containers. However, almost invariably it was discovered that foreign
materials found their way into the sample. Therefore, these attempt's to
establish rainfall quality were abandoned early in the study. The tipping
bucket rain gages were installed following recommendations of the National
Oceanographic and Atmospheric Administration.
Automated Samplers--
Automated samples were obtained using a commercial sampler which collected
up to 28 discrete samples. The sampler was activated by stage change with a
subsample being taken at each 76 mm (3 in.) or fall in stream stage. Samples
were collected in 500 milliliter plastic bottles. The bottles were pre-
charged with sufficient sulfuric acid to bring the sample pH below 2. Early
experiments had shown that acidification was an acceptable preservation tech-
nique for the parameters measured in this study. The sample time was re-
corded directly on the hydrograph by a subsidiary electronic circuit de-
scribed in the following section, "Activation and Notation System."
Placement of the intake was a prime consideration. To avoid mechanical
breakdowns which might be experienced with a moveable intake, a fixed intake
was chosen. The intake body was located just beneath the low flow level in
the stream. This assured that the intake would be submerged at the start
14
-------
of a stream rise, but also meant that the intake was near the bottom of the
water column during high flow events.
The intake figuration was a cylinder about 150 mm (6 in.) long and 38 mm
(1-1/4 in.) in diameter with about 30, 6.4 mm (1/4 in.) holes in the body.
This cylinder was placed in longitudinal alignment with stream flow and was
wrapped with fiberglass screen to prevent line clogging. Unfortunately, this
screen proved to be a good substrate for algal growth and thus was cleaned
weekly as part of the maintenance routine.
Activation and Notation System--
Because runoff events were deemed more important than baseflow conditions,
samplers were activated on the basis of stage rather than time, and discrete
rather than composite samples were collected to facilitate mechanistic inter-
pretations. However, to obtain a complete as possible record, no minimum
point for sample initiation was selected; and the primary interval for sample
activation was made as small as practical (76 mm [3 in.]). Thus, a sample
would be obtained during a period of baseflow recession when the stage dropped
76 mm (3 in.).
The sampler selected (ISCO Model 1392, Instrumentation Specialities Com-
pany, Lincoln, Nebraska) may be activated on a time or flow basis and the
actual sampling interval can be a multiple (up to 12) of the primary inter-
val. The volume of sample obtained can be selected from nine incremental
steps up to a maximum of approximately 500 ml. The sampler has a capacity
for 28 discrete samples, however, with an available option (a multiplexer),
up to four subsamples can be composited into one bottle allowing a maximum of
112 distinct sampling events. A continuous analog stage record was obtained
using a Freiz Model AU stage recorder.
Two items were needed to complete the design system: a means of activa-
ting the sampler, and a method to record sampling time. The sample activa-
tion switch was relatively straightforward. The ISCO sampler will initiate
a sampling cycle when two contacts (A and C in Figure 1) in the flow meter
connection are momentarily bridged. To accomplish this, four magnets were
cemented at equal distances around the 30.5-cm periphery of the stage recorder
float wheel and a reed switch was mounted on a sheet metal bracket such that
when the float wheel roated, the reed switch opened and closed with the
passage of the magnets (Figure 1). Thus, the primary interval for a sampling
impulse was a 76-mm (3-in.) stage change. Due to the flexibility of the
sampler, many sampling permutations were available. For example, by setting
the multiplexer switch on Position 2 and the time interval switch on Position
1, two 76-mm (3-in) subsamples were composited to obtain a sample represent-
ing a 152-mm (6-in.) stage change. A second method of obtaining a 152-mm
(6-in.) stage change sample would be to set multiplexer on Position 1 and ,
the time interval switch on Position 2. The first method potentially yields
more information. The second method avoids composting subsamples from
different regimes; i.e., the first subsample from the descending limb of a
hydrograph, and the second subsample from a succeeding runoff event.
The second part of this system is an event recorder to indicate the time
of the sample collection. An external event recorder might have been used,
15
-------
but this could present problems in relating the sample time to the hydro-
graph position, especially if the two clocks were not well synchronized.
A much simpler and less costly method of recording the time for each sampling
impulse is by marking directly on the hydrograph itself. This apparatus con-
sisted of a pulse circuit, a 12-volt coil, and a marking pen. The pen was
mounted on the coil so that opening and closing this coil would cause the pen
to move. The coil and pulse circuit were mounted on a sheet metal bracket
and positioned on the stage recorder so that the pen rested on the hydrograph
margin. When the sampler activating switch was closed, a momentary pulse was
generated by the pulse circuit, which closed the coil and caused the pen to
move. This created a "job mark" on the hydrograph margin. Care should be
taken that the pen and coil mounting do not restrict movement of the stage
recorder pen carriage at the extreme end of travel. Occasionally, the marking
pen would fail to write; however, this difficulty was minimized by monthly
replacement of the pen. A copy of the typical record is shown in Figure 2.
The pulse circuit and associated wiring are shown in Figure 3. It is
a low drain circuit, suitable for battery operation, and generated only one
pulse per switch closing. The system was powered by a 12-volt automotive
FLOAT WHEEL
(30.5 cm
CIRCUMFERENCE)
REED
SWITCH
A
C
^ > ^- ^ ^ ^^ -v -v -v -v -v ^-*~
MAGNET
SHEET METAL
MOUNTING
BRACKET
Figure 1. Sampler activation switch.
16
-------
storage battery that was recharged once a month. Bench testing demonstrated
the proper spacing between the reed switch and magnets necessary so that
only one pulse was generated per magnet cycle and that minor fluctuations in
stage did not cause multiple sampling impulses to be generated. This varied
slightly due to magnet strength but was generally about 6 mm (0.25 in.). To
ensure that momentary fluctuations in.stage did not cause multiple trigger-
ing of the sampler, a 35-y Farad, 15-volt capacitor was added to the exist-
ing circuitry for sample initiation.
Including sample losses in transit and laboratory analyses, less than
10 percent of all possible sampling data were lost over a two-year study
period. Thus as a whole, this simple sampling activation and notation package
that costs only about $25 has performed very satisfactorily on a continuous
field basis. It is well adapted to situations where line power is not avail-
able or sophisticated digital recording equipment is not used.
Stage Discharge Relationships--
Stage discharge relations were developed using regression techniques.
In three of the four cases, two curves were used. The break point was gen-
erally located just above base flow and was visually apparent from a stage
SAMPLING EVENTS
^n^
I I I I I I I I I
UJ
H
in
HYDROGRAPH
TIME
Figure 2. Typical record of a runoff event.
17
-------
discharge plot. The relation for low flow was a straight line; higher flows
were predicted using a second order equation. This provided a better fit
than a log-log relation. The high flow relations had r2 values of .95 to
.99 while the low flow r2 values ranged from .76 to .98. The lesser r2
values for low flows at two of the sites were judged to result from slight
stream-bed shifts caused by high flow scouring.
Maintenance--
Maintenance was designed into the initial sampling scheme as it was antic-
ipated that problems would occur if a regular routine was not followed. In
addition, the instrument houses were built of heavy gage plate steel rather
than aluminum as a protection against damage from firearms since these sta-
tions were in highly vunerable rural areas where hunting was taking place.
Several bullet marks on two of the station houses proved the wisdom of this
REED
SWITCH
4
^
220 K
COIL
IOK
-fc IN4004
COIL PLOTTER BRUMFIELD
*KAIIDY 12V DC
GED40C4
DARLINGTON TRANSISTOR
Figure 3. Event recorder circuit.
18
-------
approach. On one occasion the lock securing one of the instrument house doors
had been jimmied free resulting in a broken hasp. As a result the recommenda-
tion would be made that a hood be built over the lock and hasp to afford added
protection.
Record Processing--
Both the rain gages and the stage recorders.were analog devices. In
order to facilitate analyses and to manage the vast amount of data generated,
it was necessary to translate all analog data to computer acceptable digital
format. This aspect of the study was very time consuming and highly prone
to human error. The rainfall data was manually coded onto computer forms
and then keypunched. The stage data was converted to digital paper tape out-
put using an analog to digital converter developed in the Biological and Agri-
cultural Engineering Department at North Carolina State University. Our
recommendation would be that future studies of this type employ equipment
which.would produce computer acceptable format to greatly reduce the error
involved and the amount of time necessary for record processing.
LABORATORY EQUIPMENT AND PROCEDURES
Sampling Handling
Sample Preservation--
Sample integrity after collection by either grab or instrumented pro-
cedures was a major project concern. Preservation techniques were examined
to determine the most feasible procedure for the samples collected in these
rural watersheds. Three preservations were used: mercuric chloride, ice,
and acid addition to maintain sample pH<2. Comparisons were made with re-
plicate samples kept at room temperature.
A single large sample was taken in the field, returned within 15 minutes
to the laboratory and split into 12 bottles. Three bottles were preserved
with each of the three preservation methods and three were used as control
at room temperature (25 °C). Three aliquots of the original sample were
analyzed immediately for chemical oxygen demand (COD), total Kjeldahl nitrogen
(TKN), nitrate plus nitrite (N03-N), and chloride (Cl). Then at one week
intervals the bottles were sampled and analyzed, Figure 4. There was varying
results from week to week for undefined reasons. No clearly superior method
emerged from these tests; hence, the preservation was selected for convenience
of the analytical tests.
The mercuric chloride method was not used because of interference with
the nitrate and chloride analyses. The addition of acid was selected for
the instrumented samplers because extended periods could elapse before the
sample was returned to the laboratory, and thus ice or refrigeration were not
feasible. For the grab samples, storage in ice and then laboratory refriger-
ation was selected. The use of acid-fixing required considerable time for
pH correction prior to analysis and hence was not used on the larger volume
of grab samples.
19
-------
o
o
o
30
25
20
15
10
UNFIXED
--- --O ----
•RACKETS INDICATE ONE STANDARD D—
DEVIATION i
I 30
- 1.20
O
1.10
100
ro
o
2.5 -
2.0 -
28
f 26
o
24
22
TIME, WEEKS
I 2
TIME, WEEKS
Figure 4. Chemical concentration variation with time for various preservation techniques,
-------
Sample Management
Grab Samples for Chemical Parameters and Sediment--
After collection in the field, the samples were packed in ice and main-
tained until reaching the laboratory. At that time the plastic bottles were
transferred to a 4 °C refrigerator until analyzed.
Instrumented Samples--
When the bottles with acid-fixed samples were taken to the laboratory,
each was agitated and transferred to polyethylene storage bottles for refrig-
eration at 4 °C. Prior to analyses for nitrate plus nitrite nitrogen (N03-N)
and total organic carbon (TOC), pH was adjusted to a range of 6-8 with a 50
percent solution of sodium hydroxide. Preliminary tests showed no differences
in chemical oxygen demand (COD), total Kjeldahl nitrogen (TKN), total phosphate
(TP), orthophosphate (OP), chloride (Cl), ammonia (WK-N) before and after pH
neutralization, so sample pH was not adjusted for these tests.
Distribution of Chemical Analyses--
Every grab sample collected was analyzed for TKN, N03-N, TP, and Cl. A
probability assignment was made such that approximately 50 percent of the
samples were analyzed for COD, TOC, OP, and suspended solids (SS), and 25 per-
cent for NH^-N.
Instrumented samples were apportioned for analysis such that every sample
was tested for TKN, N03-N, TP, and Cl; every other sample for COD, TOC, OP,
and SS; and every fourth sample for NH4-N.
Sampling and Analysis--
Complete agitation of refrigerated samples was accomplished with magnetic
stirrers. An aliquot was taken for each analysis and the bottle returned to
the refrigerator for other analyses. Based on the previously described se-
lection technique, the average percentage of grab and instrumented samples
analyzed for respective chemical parameters are given in Table 1. Replicated
analyses of samples was not done on a routine basis.
Sample Analyses
The COD was determined according to Standard Methods (APHA, 1971) with
the shortened digestion period of 35 minutes verified for these samples
(Humenik, 1975; Overcash, 1975). The lower detection limit is about 5 mg/1.
The carbon content of all samples was obtained with a Beckman Model 915
TOC Analyzer. The amount of total carbon was determined by injecting a micro-
sample (20 yl) with a syringe into a 950 °C catalytic (Colbalt-impregnated,
asbestos packing) combustion tube. The sample was then vaporized and the
carbonaceous material completely oxidized to carbon dioxide and water in the
presence of a cobalt catalyst. Zero grade carrier gas transported the gen-
erated carbon dioxide to the infrared analyzer for measurement. The carbon
dioxide detected was directly proportional to the total carbon of the sample.
The actual concentration was determined from the peak recorded on a strip
chart, which was compared to a calibration curve.
21
-------
TABLE 1. LABORATORY-CHEMICAL ANALYSES PERFORMED ON GRAB AND INSTRUMENTED
SAMPLES
Percent of total samples collected
100% 50% 25%
Total Kjeldahl Nitrogen Chemical Oxygen Demand, Ammonia,
TKN COD NH..-N
Nitrate + Nitrite Total Organic Carbon,
NO3 TOC
Total Phosphate, TP Orthophosphate, OP
Chloride, Cl Nonfilterable Residue, SS
Inorganic carbon was determined by injecting an identical microsample
into a 150 °C combustion tube that contained quartz chips wetted with 85 per-
cent phosphoric acid. This temperature was below the value at which organic
matter oxidizes. All inorganic carbon was converted to carbon dioxide which
was measured by the infrared analyzer. Inorganic carbon values were then
determined by comparing the peak recorded on a strip chart to a calibration
curve. The sample TOC was then the difference between the total carbon and
total inorganic carbon. The inorganic carbon content of the rural runoff
samples was in the 1-4 mg/1 range.
Analysis of the OP from the start of the project until 4/20/75 was con-
ducted with the stannous chloride procedure (APHA, 1971). The lower level
of analytical sensitivity was estimated at 0.01 mg OP/1. With the develop-
ment of the low level OP automated analysis, all remaining samples were run
as an adaption of the Technicon Method (#327-74 W). The determination of
phosphorus is based on a colorimetric method in which a blue color is formed
by the reaction of OP, molybdate ion and antimony ion followed by reduction
with ascorbic acid at an acidic pH (<1). The phosphomolybdenum complex is
read at 660 nm.
In order to account for sample color, a separate analysis of background
color was made in order to determine the actual sample concentration. From
10/25/75 a dialyser was installed which automatically removed background
color. OP concentrations were measured with dialysis and without dialysis
(but substracting background level) and found to be the same. With the auto-
mated analysis and dialysis procedure, the lower detection limit was esti-
mated to be 0.01 mg OP/1.
Ammonium determinations from project initiation until 10/25/75 were
by a manual distillation as described in Standard Methods (APHA, 1971).
Utilizing this technique, an estimated lower detection limit of 1 mg NH^-N/1
was achieved. On 10/25/75 the NhU-N analysis was converted to an adaption
22
-------
of the Technicon method (Number 325-74 W) in which dialysis was used to im-
prove analytical efficiency. The determination of ammonia is based on a
colorimetric method in which an emerald green color is formed by the reaction
of ammonia, sodium salicylate, sodium nitroprusside, and sodium hypochlorite
(chlorine source) in a buffered alkaline medium at a pH of 12.8-13,0. The
ammonia-sal icylate complex is read at 660 nm. With the automated technique
adapted for low level concentrations, the lower level of analysis was esti-
mated to be 0.01 mg NHi»-N/l.
The TKN and TP analysis involved sample digestion to NhU-N and OP,
respectively as specified in Standard Methods (APHA, 1971) with the TKN re-
agents modified to use 1 g mercuric oxide/1 instead of 2 g/1 of digestion
reagents. These reagents have much lower safety risk and cost than the re-
commended automated digestion reagents (Technicon). From that stage the
procedure was the same as described for these parameters.
The block digester (Technicon BD-40) was used with a 20 ml sample and
5 ml of digestion reagent. The digestion sequences were the same as used
in Technicon procedures. After digestion and cooling, the tubes were
brought to 75 ml volume while being agitated on a vortex mixer. Next each
tube was inverted 10 times to assure complete mixing and then poured into
sample cups for the automated analyzer.
. All TP analyses were by the automated technique with dialyses. How-
ever, only the TKN samples collected after 8/31/75 were analyzed in this
manner. Prior to that date the Standard Methods (APHA, 1971) technique
for TKN was used.
The NOa-N and Cl analyses were made using adaptions of the Technicon
procedures (Number 100-70 W and 99-70 W, respectively). This automated
procedure for the determination of nitrate plus nitrite utilized the pro-
cedure whereby nitrate is reduced to nitrite by a copper-cadmium reductor
column. The nitrite ion then reacts with sulfanilamide under acidic con-
ditions to form diazo compound. This compound then couples with N-1-naphthyl-
ethelenedia-mine dihydrochloride to form a reddish purple azo dye measured at
520 nm. The automated procedure for Cl depends on the liberation of thiocy-
anate ion from mercuric thiocyanate by the formation of unionized but soluble
mercuric chloride. In the presence of ferric ion, the liberated thiocyanate
forms a highly colored ferric thiocyanate proporational to the original
chloride concentration measured at 480 nm. During July to September 1975
these procedures were modified to include dialysis in order to simplify the
analyzer operation. This conversion yielded the same results as the un-
dialyzed procedure. With the automated procedure incorporating dialysis, the
lower levels of analysis were estimated as 0.01 mg NOs-N/l and 0.50 mg Cl/1.
On a number of samples taken from the studied streams, there was an
apparent OP to TP ratio in excess of unity. An extensive analysis of the
analytical procedures indicated that all standards were determined correct-
ly, and that with known additions of OP and TP, complete recovery of each
was achieved. Preliminary tests have identified a possible positive
23
-------
interference with the OP values which did not occur after sample digestion.
The nature and magnitude of the interference have not been determined; thus
the OP must be viewed with some doubt.
Total nonf i Iterable residues were analyzed according to the procedures in
Standard Methods (APHA, 1971). This determination was made in triplicate and
was one of the more variable measurements in this study. The estimated lower
detection limit was 0.5 mg/1.
STUDY SYSTEM DEFINITION
Basin Description
The U.S. Army Corp of Engineers, as part of their water resources develop-
ment studies, made a detailed investigation of the Chowan River Basin. Results
of this study were published in a draft feasibility report (Corp of Engineers,
1975). The material in this section is taken directly from that report.
The Chowan River Basin is about 209 km (130 miles) long and drains an area
of 12,802 sq. km (4,943 sq. miles) in southeastern Virginia and northeastern
North Carolina. The Chowan River proper is located entirely in the state of
North Carolina and flows in a southerly and southeasterly direction from the
confluence of the Nottaway and Blackwater Rivers at the state line for a dis-
tance of 84 km (52 miles) to the entrance of the Albemarle Sound. The three
principal tributaries to the Chowan River are the Meherrin, Nottaway, and
Blackwater Rivers.
The upper 5180 sq. km (2000 sq. miles) of the basin lie on the Piedmont
Plateau in Virginia which is characterized by gently rolling hills. The
remaining 7700 sq. km (3000 sq. miles) of the basin lie on the flat Coastal
Plain of Virginia and North Carolina. These topographic regions are sepa-
rated by the fall line running in a general north-south direction roughly
through Petersburg, Emporia, and Roanoke Rapids as shown in Figure 5.
The Meherrin River rises in Lunenburg County, Virginia, flows toward
the AtlanticCoast in a southeasterly direction crossing the Virginia-North
Carolina state line five times, and empties into the Chowan River about
58 km (36 miles) above the Albemarle Sound. The drainage area lies between
the Nottaway and Roanoke River Basins. It is about 24-40 km (15-25 miles)
wide, about 153 km (95 miles) long, and is comprised of about 4195 sq. km
(1620 sq. miles), of which about 2590 sq. km (1000 sq. miles) lie in Virginia
and the remainder in North Carolina. The Nottaway River also arises in
Lunenburg County. Draining an area of 4403 sq. km (1700 sq. miles) in the
center of the Chowan River Basin, it flows southeasterly for 250 km (155
miles) to its confluence with the Blackwater River. While on the Piedmont
Plateau, the Nottaway is generally a fast flowing, clear stream. Then, unlike
the Meherrin, it also retains its clearness through the Coastal Plain and grad-
ually broadens as it passes through the wooded lands toward its mouth.
The Blackwater river originates on the Coastal Plain just below the city
of Petersburg and flows in a southeasterly and southerly direction for 170 km
24
-------
PETERSBURG ,Va.
ro
en
PIEDMONT COASTAL
PLAIN SEPARATION
LINE \ /
ROANOKEj-
RAPIDS /
A -
F -
w-
P-
AGRICULTURAL PIEDMONT
FORESTED PIEDMONT
WELL-DRAINED COASTAL PLAIN
POORLY-DRAINED COASTAL PLAIN
Figure 5. Schematic of Chowan River Basin.
-------
(105 miles) to its confluence with the Nottaway River to form the Chowan
River. This river drains 1920 sq. km (740 sq. miles) of gently sloping and
relatively flat terrain on the Coastal Plain of Virginia. It received its
name from its color, which is a result of tannic acids the stream receives as
it passes through wooded swamplands along its course.
Rainfall over the basin averages about 115 cm (45 in.) per year. The
snowfall varies from about 25 cm (10 in.) on the Coastal Plain to 30 cm (12
in.) on the Piedmont Plateau, Runoff from the Chowan River watershed has
averaged about 30 cm (12 in.) per year, which is about 27 percent of the rain-
fall and equivalent to about 0.0098 m3/sec/km3 (0.9 ft3/sec/mi2) of the drain-
age area. The total population of the Chowan River Basin in 1970 was 259,608
of which 66 percent resided in Virginia and 34 percent in North Carolina. The
total population for the basin represents a 1 percent decrease from the U.S.
census estimate for 1969, or in absolute figures, a decreasing population of
2294, This is primarily the result of a movement of rural agricultural work-
ers and their families to urban areas located outside the basin.
The Chowan River Basin, for the most part, has always had a rural economy.
In the 1970's the rural population in the basin was 205,973 or 79 percent of
the total population. This compares to 37 percent for Virginia, 55 percent
for North Carolina, and 23 percent of the nation. Thus in perspective, the
Chowan River is a highly rural basin. However, this is gradually changing as
in 1960 the rural population was 89 percent of the total while in 1970 the
rural population had dropped to 79 percent of the total. In Virginia the
population of the major urban centers in the basin are Franklin, 6800; Emporia,
5300; and Dinwiddy County adjacent to Petersburg, 12,435. In North Carolina,
the populations of the major urban centers in the basin are Edenton, 4766;
Ahoskie, 5105; and Murfreesboro, 3108. The remainder of the urban population
is located in scattered small towns throughout the basin.
Study Area Stratification
There are only two major physiographic regions located in this basin--
Coastal Plain and Piedmont. Therefore, two small conveniently located areas
(one Piedmont, one Coastal Plain) were arbitrarily selected for study.
Within the Coastal Plain region, previous research (Gambrell et _a_T_.,
1974) has indicated some differences in water quality parameters in drainage
water from poorly-drained soils as compared to moderately, well-drained soils.
The poorly-drained soils have a higher content of soil organic matter, and
water leaving these soils is higher in TOC. However, saturated, anaerobic
conditions within the poorly-drained soils frequently favor denitrificatibn
so less nitrate-nitrogen is removed frorv. the poorly-drained soils as compared
to well-drained soils. Well-drained soils usually support a more intensive
agricultural program and normally have greater slopes as compared to poorly-
drained soilSo Thus, the sediment losses from well-drained Coastal Plain
soils would be expected to be greater.
For these reasons, the area selected for study in the Coastal Plain was
stratified for this study into regions where most of the soils are well
drained and into regions where most of the soils are poorly drained. The
26
-------
actual subbasins within each of these stratifications were chosen at random
but from a restricted portion of the Chowan River Basin; that is, on the
North Carolina side of the river within practical working distance of Raleigh.
Within the Piedmont region, there is no simple system for stratification
based on soil resources. Nearly all of the soils are Hapludults, and major
differences in nonpoint sources of contaminates are likely to be a result of
land-use differences. Thus, a random selection of sites.was made from within
a restricted area to make travel costs reasonable. However, the area chosen
was primarily forested; therefore, another Piedmont area with more agricultural
activity was then selected to obtain additional sampling sites resulting in the
two Piedmont areas—forested and agriculture.
Study Unit Definition
The characteristics of the area sampling plan used were established by two
decisions. The first decision was that rural runoff can best be studied as
the output of small drainage subbasins that do not have point sources of
pollution. The second decision was that small drainage subbasins can best be
identified, delineated, and enumerated on some chosen topographic map. The
maps used were USGS 1:250,000 maps, AMS series V501 NJ18-10, NJ17-12, NJ17-9,
NJ18-7.
A precise but practical definition of a small drainage basin was a basic
requirement of this study because of the decision to study rural runoff by
measuring constituent concentrations and nutrient transport by random sam-
pling of small watersheds with no point sources.
The first definition of a sampling unit (a small basin) specified sampling
a second-order stream below the confluence of the two first-order streams that
generated it. The sampling point was to be as far down the stream as possible
without receiving the contribution of another tributary. The complexity of
this definition resulted from the desire to provide an opportunity to measure
stream reach dynamics, as well as nutrient contribution.
Investigators singly or in pairs made several exploratory field visits
to each of the randomly selected sites, then all the investigators visited
all sites as a team. This exploration of sites proved to be an important
step, both in contributing information and in revealing difficulties. In
consequence, certain adjustments had to be made in the original plans.
The definition initially set forth in the research plan proved to be
impractical at some points in the Coastal Plain, and accordingly was modified.
In certain Coastal Plain sites, the confluence of two first-order streams,
marked clearly on the map, was difficult to define in the field because of
extensive swamp areas, braided channels, and meandering flow. Under these
conditions, access was difficult and flow measurement practically impossible
on a routine basis. As a necessary accommodation, sampling at these partic-
ular sites was transferred downstream to the next road (or rail) crossing.
Under these conditions, the original intent to study stream dynamics and
determine confluence mass balances could not be realized, though information
27
-------
on concentration and transport variances and costs could be more easily ob-
tained for this feasibility study.
It was also observed on these exploratory trips that small basins formed
by the confluence of two first-order streams may vary greatly in drainage area
and consequently in the size of the resulting second-order stream. Further it
became clear that the distinction between point and nonpoint sources may be
difficult to determine. The first-order streams tend to be small at the point
where they cross the first road.above the confluence and at that point may be
highly influenced by human activity concentrated along roads.
The modified definition of a small drainage basin is still based on one
selected edition of topographic maps that cover the study area. Now, however,
first-order streams are selected and each is traced down to the first road
crossing. During this transit, tributaries may be acquired so the order of the
stream is no longer constant. The basin is now defined as all drainage above
this road crossing. If the resulting basin is considered to be too small, then
the next road crossing is designated, assembling in this manner a list of ba-
sins of a more uniform area. Selection of a road crossing has the further
practical advantage of total stream confinement for field measurement.
Sampling Site Selection
Four sample subbasins were randomly chosen from the poorly-drained and
well-drained Coastal Plain, and forested Piedmont; three from the agricultural
Piedmont. These same sampling sites were used for the duration of the feasi-
bility study. One selected subbasin was rejected because an impounded lake
obscured the upstream confluence point so that the subbasin did not have the
configuration required by the original sampling unit definition; the sample
site randomly selected next in order was substituted.
The designation and approximate area of the selected sampling sites are
shown in Table 2.
Field Sampling Schedule
Field sampling in this study was in two dimensions, space and time. Sam-
pling in space consisted of the selection of field sampling sites; sampling in
time produced a sampling schedule according to which the sites are being vis-
ited and the measurement made.
Necessary constraints were imposed upon the simple random sampling plan
outlined above. First, the study was confined to four subareas (well-drained
Coastal Plain, poorly-drained Coastal Plain, forested Piedmont, and agricul-
tural Piedmont) of the Chowan River Basin. These subareas were chosen arbi-
trarily to reduce travel costs and yet be representative of all major soil
types and rural land uses in the total basin. A second constraint was imposed
by limiting the number of sample sites and visiting each site on a number of
occasions, rather than selecting a site-time sample completely at random from
the two-dimensional frame. Third, certain time-space constraints were imposed
to reduce travel time as a matter of practical field operations; the sampling
sites were associated as work groups to be visited on one field trip, in a
28
-------
TABLE 2. DESIGNATION
Stratum
Poorly-Drained Coastal Plain
Well-Drained Coastal Plain
Forested Piedmont
Agricultural Piedmont
AND APPROXIMATE
Site
P-8
P-10
P-ll
P-13
W-3
W-4
W-8
W-10
F-l
F-2
F-3
F-7
A-l
A-4
A-8
AREA OF SAMPLING SITES
Area
sq. mi .
4.51
3.74
4.90
38.04
6.27
0.20
3.31
6.37
5.21
6.14
14.06
6.04
5.57
4.28
1.75
sq. km
11.7
9.69
11.7
98.52
16.2
0.52
8.57
16.5
13.5
15.9
36.42
15.6
14.4
11.1
4.53
limited series of sequences, with visits spaced apart in time and with no vis-
its at night. However, weekends and holidays were included in the time sam-
pling frame.
Subbasin Land Use
Because land use is generally accepted to have an effect on water quality,
and because many predictive models employ land use as an important variable in
generating water quality estimates, it was decided to define both the water-
shed boundaries and the current land uses within each watershed. Watershed
areas were also necessary to calculate area! yields.
The exact boundaries of each of the drainage subbasins in the Coastal
Plain used in this study were determined with help from the U.S.D.A. Soil Con-
servation Service. Topographic maps of the Piedmont sites were used for de-
termination of drainage areas. These boundaries were refined by field checks
to resolve uncertainties in surface divides. No attempt was made to determine
the pattern of groundwater flow; the assumption was made that ground and sur-
face water divides coincided.
The land-use types selected for characterization were forest (with re-
cently logged areas counted separately), crop, pasture, residential, and water
(lakes, ponds). The land use in each drainage area was determined From aerial
photographs obtained from the U.S.D.A. Agricultural Conservation and Stabili-
zation Service, dated 1970 and 1971, with information updated and supplemented
29
-------
by on-site visits. Individual land-use areas, as well as total watershed area,
were planimetered directly on these revised photographs. The sum of various
land-use areas and the total watershed area agreed within one percent in all
15 cases. This land-use information is presented in Table 3.
Sampling Site Descriptions
There were three levels of sampling employed in the study. In each of
the 15 subbasins, one station either below the confluence or at the instru-
mented sampling point was used in the statistical analyses. These stations
were designated as "Prime Sites." In addition, three subbasins had sampling
points above the confluence to measure constituent conversions and mass bal-
ances. Finally, two subbasins were designated for intensive study. This in-
cluded studies on oxygen demand kinetics and algal populations and growth po-
tential, as well as chemical constituent interconversions. For these two sites
six stations, ranging along the stream system, were selected for study.
In addition to the above stations, sampling points to assess main river
quality and the relative contribution of point and nonpoint sources in a
watershed were established. For the comparison of water quality between the
sample subbasins and rivers draining these areas, two stations were selected,
one on the Meherrin River below the Piedmont sites and one on the Chowan River
below the Coastal Plain sites. For comparison of point and nonpoint sources,
six additional stations were selected above and below one of the statistically
selected sites to study inputs and interconversions over a stream reach.
Poorly-Drained Coastal Plain (P Sites)--
This entire area is flat, with little or no relief. The cropped areas
are extensively ditched to facilitate drainage. Some areas are tile drained
to facilitate crop production in these poorly-drained soils. The major crops
are corn, soybeans and tobacco with some small grains grown' as a winter crop.
SITE P-10 was an intensive site and thus had six stations in the water-
shed.St'ation 1 was located on the west tributary about 0.8 km (0.5 mile)
above the confluence. The streambed was a channelized ditch 1.2 m (4 ft)
wide by 0.6 m (2 ft) deep and was generally dry except during and shortly
after rainfall events, Station 2 was on the north tributary and was sampled
about 6.1 m (20 ft) above the confluence. This channelized ditch was 2.4 m
(8 ft) wide by 1.8 m (1 ft) deep and generally had flowing water. Station 3
was 6.1 m (20 ft) above the confluence on the west tributary and the stream
was the same size as Station 2 and was generally flowing. At Station 4, 6.1
m (20 ft) below the confluence, the stream was also 2.4 m (8 ft) by 1.8 m
(6 ft) and flowing. Station 6 (Prime Site), an instrumented site, was loca-
ted about 3.2 km (2 miles) below the confluence. The channel was about 3 m
(10 ft) wide by 2.1 m (7 ft) deep and flowed continuously. The bridge pool
area was about 12.2 m (4 ft) wide; however, the channel was cut below the
pool area and under all but flood conditions flow was within the channel.
When the gaging station was installed, a small concrete block control was
emplaced; this did not substantially alter the existing flow pattern. Station
29 was 2.4 km (1.5 miles) below Station 6. At this station only chemical^
"algal, and oxygen demand kinetics samples were taken; no yield measurements
were made. This afforded a total stream reach of about 6.4 km (4 miles) for
30
-------
TABLE 3. LAND USE
Drainage area Forest Crop Pasture Developed Logged Ponds
Subbasin
sq mi
sq km
%
%
Poorly-Drained
P-8
P-10
P-ll
P-13
W-3
W-4
W-8
W-10
F-l
F-2
F-3
F-7
4.51
3.74
4.90
38.04
6.27
0.20
3.31
6.37
5.21
6.14
14.06
6.04
11.68
9.69
12.69
98.52
16.24
0.52
8.57
16.50
13.49
15.90
36.42
15.64
77.1
72.1
69.0
66.2
Well
44.7
52.6
56.2
48.5
72.1
91.9
90.3
82.8
17.4
25.9
23.0
26.4
-Drained
53.5
46.2
41.9
43.3
Forested
15.0
3.7
7.0
10.6
%
Coastal Plain
4.2
2.0
3.0
4.7
Coastal Plain
1.3
1.2
0.4
7.0
Piedmont
8.1
2.8
1.0
3.0
%
1.0
0.0
4.0
1.5
0.3
0.0
1.5
0.9
0.0
0.2
0.1
2.9
%
0.3
0.0
1.0
1.1
0.2
0.0
0.0
0.3
4.8
1.4
1.6
0.7
%
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Agricultural Piedmont
A-l
A-4
A-8
5.57
4.28
1.75
14.43
11.09
4.53
63.6
55.6
38.1
27.0
29.9
20.6
7.3
14.3
32.5
0.2
0.2
1.0
1.7
0.0
7.6
0.2
0.0
0.2
-------
algal and oxygen kinetic studies. Station 29 was at a broad sluggish swampy
section in contrast to the upstream sites.
SITE P-ll had three stations originally intended to be clustered about
the confluence, but field exploration revealed that measurements were not
possible in this area because of difficult conditions. Therefore, all three
stations were located at flow constrictions. Station 7 was situated on the
north tributary about 0.8 km (0.5 mile) upstream of the confluence. The
stream was channelized with a ditch 1.8 m (6 ft) by 1.2 m (4 ft) and flowed
most of the time. There was a dog pen (20 dogs) located about 30.5 m (100 ft)
upstream and 6.1 m (20 ft) from the bank. Runoff from this area during rain-
fall or cleaning-flushing events had a visible impact on receiving streams.
Additionally, there were about 10 houses located along the road crossing the
stream and it is suspected that septic tank drainage also entered the stream.
A large amount of filamentous algae (predominantly Anabaena) could be seen
in the pool upstream of the culvert. Station 8 was located on the west trib-
utary, about 0.8 km (0.5 miles) upstream of the confluence and roughly 0.8 km
(0.5 mile) south of Station 7. This stream was channelized and was the same
size as Station 7. There were about 20 houses along the road and the stream
also drained about 1 sq. km (0.4 sq. mile) of urban and developed area in
the town of Ahoskie. Station 9 (Prime Site) was roughly 0.8 km (0.5 mile)
downstream from the confluence. The sampling point was at a railroad cross-
ing; three, 1C8 m (6 ft) corrugated iron pipes carried the flow under the
embankment. There was always flow through one pipe, but the other two carried
water only under very high flow conditions. The streambed proper was about
18.3 m (60 ft) wide and had been channelized in the past but was slowly re-
verting to a natural swampy state with a thin sheet of water about 9.1 m (30
ft) wide slowly coursing through a thick vegetative mat in addition to flow
in the 1.2 m (4 ft) by 0.61 m (2 ft) channel.
SITE P-13 was by far the largest watershed in the study with a drainage
area of 98 sq. km (38 sq. miles). This site is the Ahoskie Creek watershed
which is an active Soil Conservation Service channelization project. There
was only Station 10 (Prime Site) on this stream in the main study, but a
substudy was done along the course of this watershed. (See below.) Original
information from state and local officials indicated that sewage effluent
from a small town in the watershed did not drain into the basin. However,
analysis of the stream reach study data indicated a high impact upstream of
Station 10, and contacts with state regulatory officials (A. C. Turnage,
North Carolina.Department of Natural Resources and Community Development,
Division of Environmental Management, Personal Communication, September 1976),
as well as field inspection, revealed that effluent from an 1100-person septic
tank was entering the catchment. This is discussed more fully in the section
on the stream reach study. At Station 10 the streambed was 18.3-21.3 m (60-
70 ft) wide and crossed.the road under a 30.5-m (IQQ-ft) wide bridge. There
were three additional overflow culverts but these seldom had any flow. This
entire area was extensively ditched and channelized.
There were 7 stations in the stream reach substudy, designated AH-1 to
AH-7, shown in Figure 6. Station AH-1 was in the headwaters of the stream
about 10 km (6 miles) from the origin. The entire catchment was flat and was
32
-------
extensively ditched and channelized. At the sampling point the man-made
channel was 9.1 m (30 ft) wide; some sedimentation had occurred and under low
flow conditions the channel was 2.5-3 m (8-10 ft) wide and 0.6 m (2 ft) deep.
Station AH-2 was Station 10 in the main study, described above. This station
was 5.6 km 13.5 miles) downstream of Station AH-1. Station AH-3 was located
8 km (5 miles) below Station AH-2. Again, the entire area draining to this
station was ditched and channelized. The stream channel was about 15 m (50
ft) wide and little siltation or revegetation had occurred; thus under all but
the lowest flows, the entire channel bottom carried water. Station AH-4 was
5 km (3 miles) downstream from Station AH-3. The channel at this point was
21 m (70 ft) wide and the bottom was relatively smooth and free of vegetation;
under baseflow conditions the water was 0.6 m (2 ft) deep. Station AH-5 was
on a tributary that drained a large, swampy area to the south of the main
stream. This watershed was approximately one-fifth the size of the main water-
shed. The stream was wide, sluggish, and meandering with a characteristic
dark brown coloration of the water. At the sampling point, which was a bridge
crossing, the swamp was confined to a channel 15 m (50 ft) wide by 2 m (6 ft)
deep. The stream flow was sluggish, with barely detectable velocities. Sta-
tion AH-6 was located 0.8 km (0.5 mile) downstream of the confluence of the
main stream and the swamp. This road crossing was the effective end of the
Trickling Filter
System
Septic Tank
System
Figure 6. Sampling sites in point and nonpoint source study
area.
33
-------
channelized section of the main stream; below this point the stream was
essentially a natural meandering Coastal Plain swamp. At the sampling point
the man-made channel was about 24 m (80 ft) wide and 0.9 m (3 ft) deep.
This channel was clean and level and generally held water across the entire
bottom. The final sampling point, Station AH-7, was 5 km (3 miles) down-
stream. At this point the stream was broad and flat with a winding channel
surrounded by low-lying swampy areas. Velocities were low; under high-flow
conditions the adjacent floodplain was under water. The main channel was
21 m (70 ft) wide at the road crossing where samples were taken. The stream
was generally less than 1.2 m (4 ft) deep and much vegetation was present in
the stream bottom.
SITE P-8 had two stations assigned to the watershed. Station 11 was
located at the south tributary, at a road crossing approximately 2.4 km (1.5
miles) from the confluence. The stream was very small at this point, about
0.61 m (2 ft) by 0.3 m (1 ft), and had an extremely small flow except during
and after rains when flow increased. Sixty-five houses were located along
the road and a roadside ditch fed this tributary. The ditch apparently re-
ceived input from septic tank drain fields and contained a whitish water
with a sudsy scum on the surface. Initial samples exhibited high levels of
organics, nitrogen and phosphorus, and a detailed field exploration revealed
that a house about 0.4 km (0.25 mile) upstream had a direct, untreated waste-
water flow through a pipe into the streambed. Following this discovery, sam-
ples were taken above this point; at this location the stream was generally
dry. The data summary (Appendix A) includes the early samples with attend-
ant high concentrations, as well as later samples. Station 12 (Prime Site)
was located downstream of the confluence at a road crossing.The area above
this station was an extensive swamp and the stream was sluggish with a typi-
cal dark coloration. The bridge was 15 m (50 ft) wide, the bridge pool was
the width of the bridge and about 1.2 m (4 ft) deep at the center. This was
also an automated site. The bridge pool never went dry, but during dry
periods it was difficult to discern any flow out of the swamp above the
station or into the swamp below the road crossing. Frequently, velocities
were too low to measure unless the channel was restricted by placing sand bags
along the banks.
The main river sampling point, used for comparison to the poorly-drained
sites, was designated CHO-1. The sampling point was the U.S. Highway 13 bridge
crossing the main stream of the Chowan River. At this point the river is 0.4
km (0.25 mile) wide. The area adjacent to the river on both sides was low-
lying and swampy. Under high-flow conditions this area was underwater. The
river was sampled from the bridge at the centroid of flow.
Well-Drained Coastal Plain (W Sites) —
In contrast to the poorly-drained area, these watersheds, for the most
part, are cropped on the uplands and an extensive buffer zone of woods and
swamp extended along stream banks. Three of the four watersheds in this area
have extensive swamps representative of the cypress sweet gum biome. Impor-
tant crops, in addition to corn, soybeans, and tobacco, include peanuts and
cotton. While there may be differences in pesticide levels due to the pre-
sence of these two crops that are treated with high levels of herbicides,
fungicides and insecticides, the level of nutrient application is not greatly
different than that for other row crops.
34
-------
SITE W-3 had Station 13 (Prime Site) located at a road crossing about
9.5 km (0.3 mile) below the confluence. The area above and below the road
crossing was swampy; however, a channel had been cut on the downstream side
and most measurements were made at this point. A plank walkway across the
stream was installed here to facilitate flow measurements under runoff con-
ditions. The bridge pool was wide, deep and always held water; but on a
few occasions there was no measurable flow. On one trip during July 1975
this site experienced two days of violent thundershowers which caused a flood
that was judged by local observers as roughly a ten-year high.
SITE W-4 had a very well-defined stream channel. The sampling point,
Station 14 (Prime Site), was located about 0.3 km (0.2 mi) below the conflu-
ence and 0.8 km (0.5 mile) from the nearest road. This was the smallest
subbasin studied and was dry on many sampling visits. The channel at the
sampling point was about 1.8 m (6 ft) wide and 0.61 m (2 ft) deep. Under
high flow conditions the water was out of the banks and spread across the
wooded flood plain making flow measurements extremely difficult.
SITE W-8 sampling, Station 15 (Prime Site), was about 0.3 km (0.2 mile)
below the confluence at a road crossing. On either side of the bridge the
stream was extremely swampy with a braided channel. At the road crossing,
the channel was about 15 m (50 ft) by 1.2 m (4 ft) and velocities were so
low that current meter measurements were impossible except under high flow
conditions. Therefore, a channel restriction was erected, using block and
concrete to narrow the channel to 1 m (3 ft) wide and block was laid across
the stream bottom to give the best possible measuring section. Under high
flow conditions measurements were made from the bridge.
SITE W-10 was almost an exact replica of Site W-8. Sampling Station 16
(Prime Site) was 0.3 km (0.2 mile) below the confluence and was located at a
road bridge. The stream opened out into a swamp on both sides of the road.
Because the velocities were below detection limits at most times, a channel
was built, as Site W-8 except that it was 3 m (10 ft) wide. Site W-10 had a
large beaver dam (80-100 ft wide by 3 ft high) on the upstream side of the
bridge. Several times during the two-year measurement period the local land-
owner cut holes in the dam which the beavers quickly repaired. The dam created
a two-acre swamp of standing water.
Forested Piedmont (F Sites)--
The area selected is in Virginia and lies west of the fall line and is
an intergrading of Coastal Plain and Piedmont in the east to the true Pied-
mont of the west. The area is predominately forest with small plots of corn,
soybeans and small grains present. There is very little tobacco grown in this
area. There is very little livestock production in the area, with only a few
small swine and pastured beef operations present.
SITE F-2 was selected as a mass balance watershed with three stations.
Station "17 was on the northwest branch of the main stream. Access was by
logging roads and then a short hike, as the site was approximately 0.3 km
(0.5 mile) from the nearest road. The sampling-measurement site was
about 6 to 9 m (20 to 30 ft) above the confluence in a well-defined channel
about 2.7 m (9 ft) wide by 0.9 m (3 ft) deep. The stream was dry several
35
-------
months of the year and v:as still measurable by wading during high flows. Sta-
tion 18 was on the west branch and access was as described for Station 17.
The sampling-measuring site was about 6 to 15 m (20 to 50 ft) above the con-
fluence in a well-defined channel about 3.7 m (12 ft) wide by 1.2 m (4 ft)
deep. This stream was dry during drought season, but less frequently than
at Station 17. During high flow this stream became difficult to measure by
wading and there was no bridge for using the Price meter. A wooden plank
walkway was laid across the stream and secured with ground anchors, but a
flood washed it away between visits. Thereafter, surface flow estimates were
made when the stream overflowed its banks.
Station 19 (Prime Site) was 6 to 9 m (20 to 30 ft) below the confluence
of the streams with Stations 17 and 18, and access was as for Station 17.
This stream ranged from dry during drought periods to a flood stage where it
was measured about 183 m (600 ft) downstream at the crossing of the logging
road used to gain access. The channel was well defined with rock outcroppings
and gravel to sandy banks. At the sample-measurement point the channel was
about 4.6 m (15 ft) wide by 1.5 m (5 ft) deep.
SITE F-3 was an intensive site and had six stations in the watershed.
Five of these stations were normal measuring sites, the sixth site was sampled
for chemical, algal, and kinetic study but flow was not measured. The area
was forested and logging had begun in the watershed near the end of the study
period. A small amount of farming was noted with corn, soybeans, and home
gardening predominating. Also some cattle were being grown on pasture. There
were numerous rock outcroppings present in the stream channel, but generally
the stream bottom was sandy with stretches of gravel. The stream was typical
uplands forest stream and the areas around all six stations were heavily
wooded. Station 20 was located on the north branch at a road crossing about 3
km (2 miles) upstream from the confluence of the two tributaries which formed
the main creek. Thus Station 20 was the furthest upstream.and the channel was
small; about 2 m (6 to 8 ft) wide by 1 m (3 ft) deep. Above the road, the
creek flowed in a well-defined channel with a broad flood plain through gently
rolling hills. Flow measurements -were generally taken in this channel. Below
the road, there was a large pool about 1 m (3 to 4 ft) deep and 2 m2 (20 sq.
ft) in area. Low flow measurements were taken in a shallow riffle area below
the pool. During droughts the stream dried up except for water standing in
the pool, and even during high flows the stream stayed within its banks. Sta-
tions 22, 23, and 24 were at the confluence of the two major tributaries which
formed the main creek, and access was on foot from a logging road 0.8.km (0.5
mile) away. Station 22 was on the north branch about 6 m (20 ft) above the
confluence, and the stream channel was well defined and about 3.7 m (12 ft)
wide by 1 m (3 ft) deep. Station 23 was on the southeast branch about 15 m
(50 ft) above the confluence"!! Here, a log restriction was put in place to
narrow the stream through a sandy area in the channel which had widened at
this point to 4 m (15 ft) by 1 m (3 ft) deep. Station 24 was located about 6
m (20 ft) below the confluence. Here, the channel was 6 m (20 ft) wide by
2 m (6 ft) deep. Stations 20, 22, 23, and 24 went completely dry during the
drought in the fall of 1976. At times the stream overflowed its banks during
flash floods, becoming impossible to wade. Because of the remote location, no
alternate high flow measurement points were available, so surface estimates of
high flows were made. Station 25 (Prime Site) was an automated, as well as a
36
-------
grab station, and was located on a logging road about 1,6 km (1 mile) off the
nearest hard-paved road. This station was approximately 0.8 km (0.5 mile)
below the confluence. The bridge was a wooden culvert 4 m (12 ft) wide by
4 m (12 ft) deep. The stream channel above the bridge was well defined and
about 3 to 3.6 m (12 to 12 ft) wide by 0.6 m (2 ft) deep. Below the bridge
there was a large pool about 1.5 m (4 to 5 ft) deep by 9.3 sq. m (100 sq. ft)
in area. A rock dam was constructed below the bridge and above the pool to
insure a steady channel around the area of the stilling well. Flow measure-
ments were taken in a fast-flowing, straight, stable section below the pool.
The channel here was 1.2 m (4 ft) wide by 0.6 m (2 ft) deep. This stream
rarely dried up; high flow measurements were taken with a Price meter off the
bridge. Station 28 was a chemical, algal and kinetic sampling point; no flow
measurements were taken. The channel was 6.1 m (20 ft) wide by 1.8 m (6 ft)
deep.
SITE F-l, Station 26 (Prime Site), was located about 0.8 km (0.5 mile)
below the confluence in a wide flood plain on an unpaved, raised roadway with
9.1 m (30 ft) embankments. The bridge was a double culvert with each section
being 2.4 m (8 ft) wide by 2.4 m (8 ft) high. Below the road was a large
pool about 1.8 m (6 ft) deep and 14 sq. m (150 sq. ft) in area. Flow measure-
ments were taken in a stable section below the pool on the downstream side of
the bridge, except that.extremely low flow measurements were taken in a sandy
area just below the culverts and above the pool. This stream stopped flowing
during the drought, but the.pool never completely dried up. The stream
channel itself was about 3.0 m to 3.7 m (10 to 12 ft) wide by 1 m (3 to 4 ft)
deep. High flow measurements were taken with the Price meter from atop the
culvert.
SITE F-7, Station 27 (Prime.Site), was an automated as well as a grab
statKJrK ITTwas located on a paved road about 457 m (500 yd) below the con-
fluence of the two small tributaries which formed the main stream. The road-
way here was raised with 9.1 m (30 ft) embankments; and on the upstream side
of the road, the flood plain was narrow with steep hills rising sharply on both
sides of the creek. Below the road there was a large pool and the flood plain
widened considerably. The channel was 7.6 m (25 ft) wide by 1 m (3 ft) deep.
Flow measurements were taken about 60 m (200 ft) upstream in an area free of
debris because beaver activity resulted in downed trees impeding the flow near
the culverts. The bridge was a double culvert 2.4 m (8 ft) wide by 3.7 m (12
ft) high, and the instrument package was attached to a wingwall on the up-
stream side. During low flows, measurements were taken about 152 m (500 ft)
upstream and during high flows measurements were taken from atop the culverts
with a.Price meter. At the start of this study, the watershed was depopulated
of beavers. However, shortly thereafter the subbasin was recolonized. The
attendant dam building frustrated efforts to establish a stage-discharge re-
lation; the automated station was subsequently deactivated.
The Meherrin River station, MER-1, was downstream of the F-sites; data
from this station was used for comparison purposes. The river was about 120
m (400 ft) wide at this point with a rocky, gravely bottom. It was fast flow-
ing with a pronounced riffle directly upstream of the sampling point.
37
-------
Agricultural Piedmont (A Sites)--
These selected watersheds were well into the Piedmont and had both forest
and agricultural activities. The farms were larger than in the forested
Piedmont area with corn, soybeans, and some fire-cured tobacco predominating.
Also, there were cattle and some sheep on pasture.
SITE A-8 also had only Station 30 (Prime Site). It was the westernmost
site and thus was furthest into the Piedmont. It was located on a paved rural
road about .1.6 km (1 mile) above the stream's confluence with the Meherrin
River. The stream was a single branch in a well-defined.channel about 3.7 m
(12 ft) wide by 1 m (3 ft) deep. The stream was slow moving above the bridge,
but the section below the bridge had a gravel bottom and about 0.7 (2 ft) of
fall over a 6 m (20 ft) reach. Low flow measurements were taken in this
section. High flow measurements were taken from the bridge with a Price meter.
The flood plain was not wide; rolling hills sloped gently down to the stream.
SITE A-4 had only Station 31 (Prime Site), located on a single reach of
stream about 2.4 km (1.5 miles) long above the station. It was located on a
state-maintained unpaved road. The channel above and below the road crossing
was well defined and about 3.0 m (10 ft) wide by 0.9 m (3 ft) deep. Flow
measurements were made under the bridge during normal conditions. Under high
flow conditions, measurements were made from the bridge; also, the stream
stopped flowing only during very dry weather. The flood plain was fairly
typical of the Piedmont area: fairly wide and bounded by gently sloping hills.
SITE A-1 had only Station 32 (Prime Site) which was an automated, as well
as a grab, station located at a bridge crossing of a paved rural road. The
bridge was about 9 m (30 ft) wide and the stream pooled somewhat under it.
The main channel downstream was 4.6 m (15 ft) wide by 2 m (6 ft) deep and the
stream was slow moving. Upstream from the bridge the stream was joined by two
small tributaries. Flow measurements were taken downstream except during high
flow conditions when measurements were taken off the bridge. The instrument
package was placed on the bank of the stream on the downstream side of the
bridge. Because the stream had a sandy bottom and a meandering channel, a
block structure was installed as a control to stabilize the stream and insure
that the intake pipes were in the channel.
TECHNOLOGY TRANSFER
To acquaint water quality researchers and planners with the feasibility
of using probability ("random") sampling to estimate areawide water quality
parameters, three workshops were held in Raleigh, North Carolina; Atlanta,
Georgia; and Washington, D.C. The workshop programs were developed in con-
junction with the Water Resources Research Institute of the University of
North Carolina which along with the North Carolina Agricultural Experiment
Station, the North Carolina Agricultural Extension Service, and the U.S.
Environmental Protection Protection Agency sponsored these presentations.
Over 150 persons attended these workshops during September and October
1977 and received a 134-page workbook presenting basic principles, project
procedures, sample data forms, calculations, and project results. Requests
33
-------
for this workbook were received from many who were not able to attend the
workshops. Attendance at the Raleigh workshop was primarily from in-state
technical and service agencies whereas a broad spectrum of national repre-
sentatives attended the Atlanta and Washington sessions.
These workshops allowed immediate dissemination of project results on
the feasibility of statistical sampling. Further, the extensive data collect-
ed during this feasibility study provided base for making water quality
judgments and assessing ongoing or proposed water quality programs.
The topics covered the first day of a workshop were project scope; pro-
bability sampling; drawing up a sampling plan: work session; field and lab-
oratory procedures and problems. Topics presented on the second day included
statistical results: estimates, precision and cost; substantive results: con-
clusions, recommendations and discussion. An extended time was set aside for
discussion and questions following the presentations.
The need for these workshops and their nature is described in the advance
notice:
Workers in water quality have seen an explosion in the
scope of their responsibilities in recent years. Nonpoint
sources and the 208 planning and implementation process have
become of primary concern because traditional technology is
not suitable to point source contributions on an areawide
basis, the total number of sites (hundreds) is too large for
complete coverage, and therefore some method of sampling must
be used. Similar sampling needs arise in evaluating urban
runoff, air pollution impact and initial assessment of
possible pollutant or hazardous material problems in
agriculture and forestry.
This workshop is based on a three-year study of the
feasibility of using probability ("random") sampling in
measuring rural runoff, carried out by an interdiscipli-
nary team at North Carolina State University. Members
of this group will present the statistical, chemical, and
operational methods they have developed and will discuss
how the results can be applied by others.
This sampling technology is just now being applied to
the water quality area. This approach allows sampling
with various budget levels, precision estimates, and asses-
ment of problems on an areawide basis. This workshop will
demonstrate the effectiveness of this sampling tool and its
transferability to other problems and areas.
We continue to receive inquiries about the potential of statistical sam-
pling for evaluation of nonpoint sources and areawide water quality because
of these workshops and results presented through other professional contacts
and publications. In fact, watershed sampling following the rainfall pro-
39
-------
ability sampling plan developed in the initial portion of this project was
initiated in May 1978 by the North Carolina Department of Natural Resources
and Community Development, Division of Environmental Management, in con-
junction with the agricultural portion of the North Carolina 208 program.
This field testing of sampling stratification on the basis of rainfall
probability is being executed by this state regulatory agency under super-
vision of University faculty involved in this EPA grant. This opportunity
to work with a state regulatory agency in testing the feasibility of statis-
tical sampling provides mutual benefits and allows a realistic test of a
service agency's ability to conduct this type of field monitoring. It also
demonstrates that the rainfall probability concept developed during this
project is acceptable to a state water quality agency and can be implemented
when circumstances suggest that it may be the most cost-effective sampling
strategy.
The User Manual being developed to demonstrate the actual utilization of
the developed methodology to statistically assess areawide inputs from rural,
urban, municipal and industrial sources throughout an actual watershed repre-
sents an additional technology transfer opportunity associated with this
research project.
40
-------
SECTION 4
STATISTICAL PROCEDURES AND RESULTS
by
Don W. Hayne
INTRODUCTION
Recent demand for information on the quality of the waters of the
nation has presented workers with an unprecedented problem of coverage.
This study explores the feasibility of using probability sampling to
derive estimates of average water quality in a river basin and in par-
ticular to study the small headwater drainages to which the primary
contribution is rural runoff.
Determining water quality in all small drainage subbasins of a river
system is one of many problems where it is impossible to achieve com-
plete and continuous recording. Were such complete recording possible,
it would, of course, be superior to any system of sampling. Considera-
tion of sampling is enforced here, and elsewhere, by the impossibility
of complete coverage; for this same reason, sampling is a pervasive
element of our scientific, commercial and personal life.
There are two broad objectives of any sampling study. First, it
must provide the right kind of information, and second, it must provide
information of the best quality possible under the budget. Often the
first time a clear definition of exact objectives is required is in
planning a sampling study when decisions must be made about exactly
what to measure and how. As to the second objective, quality is defined
in terms of accuracy and precision, two terms of different meaning.
Accuracy has to do with the reduction or elimination of bias, while
precision describes the repeatability of a measurement or an estimate.
Both accuracy and precision may be improved through survey design.
Accuracy is attained by using probability (or random) sampling; preci-
sion is increased by changing sample size and by use of a survey design
which is adapted to the universe under study and which allows the avail-
able sampling effort to be employed with the greatest efficiency.
A unique feature of sampling the small subbasins of a river basin
is that it must be in two dimensions, in space among subbasins, and in
41
-------
time throughout the year. The larger the number of subbasins sampled,
the less frequently each may be visited under any constant budget.
In the present study we made a probability sampling study of water
quality of small drainage subbasins in a portion of the Chowan River
basin of North Carolina and Virginia, gathered data for two years,
carried out an analysis of costs and variance components, and based on
these findings, made estimates of the precision attainable under optimal
allocation of sampling effort for a range of possible budgets. Using
the same information and methods, we made comparisons of the cost-
effectiveness of grab sampling (use of hand methods at field sites)
compared with automated (instrumented) sampling.
CONCLUSIONS
Conclusions of this study must be drawn within the constraints of
our basic assumptions: that we are to determine water quality for the
output of small drainage subbasins as river-basin averages of such
quantities as concentrations of chemical parameters, or rates of flow,or
transport or yield of water or chemicals; that reference is to a one-
year study; and that the basic objective is the most precise information
for the available budget. In these terms, these conclusions follow from
our study of a portion of the Chowan River basin.
When identified as the output of small drainage subbasins, rural
runoff can be measured by use of probability sampling in space and time.
Concentrations were measurable with good orecision. Flow, and other
measures such as transport or yield of chemicals, which included the
flow as a component, were not measurable with good precision, although
adjusting flow or transport to yield by taking account of basin area
improved precision somewhat. These conclusions hold whether grab
sampling or automated is to be used.
Precision was difficult to attain for flow or related measures
for two reasons. First, there was high inherent variability, both in
space among small drainage subbasins, and in time at any site. Second,
and this is true also of measuring concentrations, sampling was
required in two dimensions, and because both the space and time components
of variance were important, neither was reduced in direct proportion to
increased total sampling effort. Rather, each, and their sum, was
reduced in rough proportion to the square root of the total sampling
effort. Thus, for flow, not only was the variability inherently high,
but precision responded slowly to increased sampling effort. For con-
centrations, the variability was not so high and better general precision
could be attained, even though the same requirement for two-dimensional
sampling existed as an impediment when attempting to increase precision.
These problems concerning variability exist by the nature of sampling
river basins; they do not arise because probability rather than judgement
sampling is used.
42
-------
We found that grab sampling promised better precision than auto-
mated sampling for the same cost at modest budgets. Grab sampling is
more flexible and adjusts easily to the need for more sampling in space.
In comparison, automated sampling is costly for sampling in space but
provides low-cost sampling in time at any site where installed. Specu-
lation suggested that automated sampling would generally become more
cost-effective than grab sampling at high budgets (several hundred
thousand 1975 dollars or higher).
We found stratification in time generally worth the effort in terms
of increased precision, but it is possible to over-stratify and reduce
efficiency by planning too-frequent visits to too-few sites.
Optimal allocation of sampling effort is important to efficiency
and may be calculated, given information on costs and variances. We
found, however, that optimal allocation differs with the parameter.
Because a survey can rarely be planned for a single parameter, this
means that the sampling effort must be allocated to be of median effi-
ciency on the average, but probably near optimal for only a few
parameters. Based on our experience, such a compromise design might
have the number of sites sampled approximately equal the number of
visits per year to each site, although this entirely empirical sug-
gestion must be reexamined when more information is available.
RECOMMENDATIONS
Probability sampling in space and time should be employed when
determining river basin average values, in particular, averages of the
contribution of small headwater subbasins to a river system.
For short-term (ca. one year) studies under modest budgets (less
than 100,000 1975 dollars) the use of grab-sampling methods should be
given serious consideration on the basis of cost-effectiveness. The
longer and more costly the study, the more the balance shifts toward
favoring automated sampling, especially in studies of flow, transport,
and yield.
A purely empirical suggestion, based upon current information, is
that in such a one-year study based on grab sampling, the sample number
of sites should be approximately equal to the number of annual visits
made to each site. This rule-of-thumb must be reexamined with a broader
data base.
Time strata should be delineated in such a one-year survey design
after the number of visits has been determined, with two visits per
time stratum. Relatively good precision may be expected of sampling
studies of concentrations.
43
-------
THE USE OF SAMPLING
Why should sampling be used; why not study the whole universe of
interest instead of only a part? The primary reason is that even if
the whole universe could in principle be studied, it usually costs too
much to do so. Even if the whole universe could be studied, it often
happens that only a superficial examination could be made, with study
in greater depth possible only for a sample. Sometimes complete study
is impossible because measurements are destructive; chemical and bio-
logical samples must often be destroyed to be identified. In choice of
sampling method, most factors relate to the balance between cost and
precision.
A sampling study must be possible to carry out. Further, it should
be reproducible and yield unbiased results, or if results are biased,
then there should be an understanding of the bias. Finally, a sampling
study should provide some measure of precision of the estimate to guide
the user of the information.
Use of Probability Sampling
Why use probability sampling? Up to this point, discussion of
sampling has been general without specification of whether the sample
was random or one selected by judgement. The most-frequently asked
question is how a random number table or coin toss can be instrumental
in selecting a better sample than one derived through the judgement
of an experienced investigator. The answer rests upon the definition
of better and this reflects the advantages of probability sampling.
Probability sampling has several advantages. First, it prevents
bias from the unconscious action of personal selection. Studies in
several areas have shown that when samples are selected by judgement
they tend to be biased in that they reflect the investigator's personal
views of the distribution of values for items in the universe. Although
the mean value for some fraction of the possible random samples may lie
farther from the mean of the universe than any judgement sampl-e, the
probability sample has the advantage of being ultimately objective and
providing assurance against personal bias in selection.
A second advantage of probability sampling is that it allows use of
statistical methods for making tests and inferences. These statistical
methods are based upon mathematical reasoning which assumes random
sampling and random events. Judgement sampling depends upon human
behavior which is infinitely variable and subject to the required mathe-
matical analysis only under very restrictive assumptions about its
nature. For the purpose of being able to use the large body of work in
mathematical statistics that supports the battery of applied methods,
the assumption is often made that judgement sampling is effectively
random. Yet when this assumption is tested empirically, it is usually
found to be in error.
44
-------
A third advantage of probability sampling is that the work is
absolutely repeatable as to method. One worker can follow the same
sampling method at another time and place just as well as the original
investigator could. In contrast, a judgement sample is doubtful of
comparable repetition even by the same worker at another time or place
and impossible of repetition by anyone else because it is based upon a
subtle combination of quantitative information, hunch, and a subjec-
tive distillation of experience.
Probability sampling methods are widely known to be objective and
to negate the potential criticism of subjectivity of choice of illus-
trative examples. At the same time, to simply state that samples are
taken randomly is no guarantee of objectivity; acceptance of any
scientific knowledge must rest upon faith in the integrity of the
investigator.
Stratified Random Sampling
Any probability sampling design should be planned to take maximum
advantage of professional judgement and experience in setting up the
sampling design. The difference is that judgement is taken account
of in the form of the design, rather than in the ultimate selection
of the sample from the universe. There are several methods of accom-
plishing this purpose; the only one to be discussed here is stratified
random sampling. Here the universe is divided into parts which in
principle include the whole universe and are non-overlapping. Each
of these parts is then studied separately, often with a different
sampling intensity. The primary statistical objective in dividing
the universe into parts is to create units within which the vari-
ability is less than in the whole. For example, if the flow of water
is known to be strongly seasonal in pattern, it is only good sense
to make separate studies of the different seasons and sum the results
for an annual total rather than ignore seasons. Stratified random
sampling also allows the allocation of increased sampling effort to
the more important parts of the universe. Other advantages are that
it allows separate estimates to be made for the separate parts of the
universe, subject to the fact that only part of the total sample is
available for this purpose. Stratified random sampling may also be
used simply to assure that there is a relatively even distribution of
sampling effort over all parts of the universe.
Stratified random sampling is discussed in numerous works on
sampling techniques. A good exposition at the user level is Chapter
17 in Snedecor and Cochran (1967) while more technical discussions
will be found in Cochran (1977), Yates (1960) and Hald (1952).
Although the mathematical background is presented by Riggs (1967),
Chow (1964) and Haan (1977), none of these references deals with
stratified random sampling.
45
-------
Design of Present Study
The sampling design of the present study reflects the physical
nature of a river basin and its component small drainages. There are
many small drainages, of which only a sample may be selected for
study. These drainages are changing in time and may be measured only
at sample instants in time. Time is divided into periods or strata,
within each of which two sample times are selected. Thus we are
sampling in two dimensions, space and time. Ideally, we should
sample at random in the space-time matrix, but this is physically
impossible; we must select certain sites, and visit all of these
within some short interval, chosen at random from the available time.
It is not practical to select a new sample of sites at each visit; a
certain amount of exploration and site preparation is required even
for grab sampling, and automated sampling requires established sites.
STATISTICAL MODEL AND OPTIMAL ALLOCATION
In this study we are sampling in two dimensions, space and time.
Measurements are made at a number of small drainage basins, randomly
selected from a complete listing of such basins, and at each of these
selected sample drainages, measurements are made at times selected in
a restricted random fashion within time periods. These periods con-
stitute a stratification in time, and were set at 4 weeks in length
for this study, with 13 periods per year. Throughout the following
discussion of the statistical model, the figure of 13 periods per
year is treated as a constant; the discussion, however, is more
general and applies to periods of any length. In fact, the empirical
results suggest that we should have used longer periods. The element
of stratification in space is not considered in this discussion of
the statistical model. The model applies independently within strata
in space, without considering the problem of allocation of sampling
effort among strata.
An additive linear model is assumed in the following discussion,
but in fact, because the data were log-transformed for analysis, the
model is actually multiplicative and variances assumed to be propor-
tional rather than absolute.
The Statistical Model
The problem, then, is of sites randomly selected in space, and
visited at randomly selected times within each successive time period.
Under these conditions each measurement is assumed to conform to the
following linear model:
(1)
where X... /.\ = the single measurement made at the il" site in the
J u; period and the kth visit within the jtn period
j
46
-------
p = general mean
S. = effect of i site (random)
n. = effect of j period (fixed)
u
(Sn)..
= effect of interaction, ith site and j period (random)
V = effect of k visit within j period; this also
^' includes any possible batch effect in chemical
analysis (random)
e... ,.\ = random measurement error, which includes any inter-
1J (3' action of site with visit within period, the
variability due to field sampling, and that due to
variance of chemical analysis within batch, or to
other sources of random error.
The variance of this single observation will be:
varXijk(j)=0S+0?Sn)+aV+ae
where a| = variance due to differences among sites
O
0. ,2 = variance due to site x period interactions
ou2 = variance due to differences among visits within sites
a 2 = variance due to random error.
e
This variance is considered within time strata (periods) and therefore
the variance among periods drops out as a result of stratification in
time. The number of periods to be used is a constant element, set
before the number of sites and visits is decided.
The variance of the mean value for s sites each sampled at pv
visits, is:
var x = (o§/s) + (afsn)/sp) + (aj/pv) + (o'/spv) (2)
where elements are as described previously and:
s = number of sites
p = number of periods
47
-------
v = number of visits per period.
This variance must be synthesized by estimating variance components
(CT'S) from the field data and substituting in the relationship, using
the denominator values actually used.
When estimating variance components from field data a small nega-
tive value may sometimes occur. In such case, the component was
assumed to equal zero because a variance, being a squared quantity,
cannot carry a negative sign. In the case where the component a?<- \
had a negative value (and was assumed to be zero) the components were
recalculated, dropping the element (Sn).. in formula (1), and using
' \j
the following model for the variance of the mean:
var x = (
-------
change with more or less sampling (although they may fluctuate randomly
by sampling error). Therefore the only way to reduce the variance of
the mean (with the same sampling design) is to increase sampling effort,
thus increasing the denominator values.
Inspection of formula (2) shows, however, that increasing the
number of sites sampled will decrease some of the terms while increasing
the number of visits will decrease others. Under a constant budget,
an increase in number of sites will require a reduction in the number
of visits that may be made to each site. The problem of optimal allo-
cation is to find the combination of number of sites and number of
visits that will minimize the variance of the mean under a constant
budget.
Optimal Allocation
Optimal allocation (sometimes called Neyman allocation) here follows
closely the classical method of J. Neyman (as expounded in numerous
references, e.g., Cochran 1977). First, a cost function is set up as:
C = s(cs + pvcv)
where C = total available budget (not including the fixed costs of
the survey)
c = cost of establishing a site exclusive of cost of visiting
and sampling
c = average cost of visiting and sampling an already established
v site.
Then:' s = G/(CS + pvcy) (4)
If formula (2) for the variance of the mean is rewritten in these terms,
then:
var x = (a|(cs + pvcv)/C) + (ofsn)(cs + pvcJ/pC) + (aj/ov)
+ (a|(cs + pvcv)/pvC.
Next, take the derivative of var x with respect to v, equate to zero
and solve for v:
V = [(Ca + Ca2)/PC(pa2 + a)]* (5)
The solution for v gives the optimum number of visits per period,
that is, the number that will produce the minimum variance. This value
49
-------
must be rounded to an integer: with the present survey design, there
must be 2 or more visits per period to provide an estimate of variance
within each time stratum.
Given a value for v, the number of visits per period, the available
budget determines the number of sites that can be covered, according
to formula (4). This value of s, the number of sites, must also be
rounded to an integer. This practical requirement for integral values
of both visits and sites prevents a precise matching of available bud-
get and estimated cost.
This solution differs from the classical result with hierarchal
sampling where the optimal number of the secondary sampling unit
depends only on the variances and costs of the primary and secondary
sampling units, and not on the budget. In the present case there is
no distinction of primary and secondary units even though there is a
superficial resemblance in that sites are established and then visited
repeatedly. The solution here requires a knowledge of the budget.
Once the optimal numbers of sites and visits are calculated for a
given budget and set of variance components, the best precision
attainable can be estimated, using formula (2). This device is used
later to explore the potential precision when using either grab or
automated sampling, and to compare the possibilities of the two methods
in a standardized manner.
COST ANALYSIS
Information on cost is basic to planning any investigation. Cost-
efficiency is the primary argument for use of sampling of any type, be
it probability sampling or judgement sampling; if there were no cost
considerations every parameter would be studied continuously and at
every point. But in fact, the availability of information is always
limited by the cost of obtaining it and some kind of cost analysis is
made before undertaking any investigation.
Cost is especially important in the type of investigation described
here and in considering the optimal allocation of sampling effort among
sites and visits. Not only must the relative cost of establishing a
site in comparison to the cost of a visit to an established site be
known with fairly good precision, but it is important that these costs
be accurate in relation to the total budget. For this reason, we under-
took as complete an analysis of the costs of probability sampling as was
within our capability. On the other hand, there is no doubt that we
would have benefited from the help of a professional in this area.
Because dollar costs of goods and services rose during the course
of our study, and have risen considerably since it ended, we have
stated all costs in 1975 dollars. Thus we reduced some costs to prices
paid during that calendar year for inclusion in our analysis. Anyone
50
-------
attempting to use our cost information for general guidance must use
some economic index to inflate our costs to current levels.
To make this cost analysis, ws found it necessary to define the
kind of investigation as a standard of reference. We assumed the
objective to be for an agency report based upon a study of one-year
duration and with no research aspects. This qualification meant that
we could not transfer the costs of our investigation directly to our
analysis because we had carried on numerous larger or smaller research
studies, and because our investigation was to explore feasibility and
not designed to report on the basin-wide status of water quality.
This postulated one-year study would cover a river basin similar to
the Chowan River and would determine average flow, yield of water per
unit area of drainage basin, concentration of chemicals and transport
and yield of chemicals. Algal studies were not included in the cost
analysis.
Cost Functi on
In making the analysis of costs, we broke each down into either a
fixed cost, a per-site cost, or a per-visit cost, in conformity with
the following cost function:
Total Cost = c-j + scg + spvcv
where c-, = constant cost
c = per-site cost (proportional to number of sites)
c = per-visit cost (proportional to total number of visits to
v sites)
s = number of sites
pv = number of visits per site (product of number of periods and
number of visits per period).
In this relationship, constant costs are the same regardless of the
magnitude of the study and constitute the overhead required to under-
take a study, no matter how small. The per-site costs are the share
of the single site in those costs that are proportional to the number
of sites. This is the cost of adding another site no matter how many
or few visits are to be made and includes the planning, exploration
and location of the site, site preparation and instrumentation if any,
or construction if it is an automated site. Costs of the minimum
number of service trips required to operate an automated sampler should
be listed as per-site costs. Per-visit costs include a pro-rata share
of all costs that are proportional to volume of day-to-day operations,
such as travel, subsistence, supplies, laboratory analyses, data manage-
51
-------
ment and physical and administrative costs that are proportional to
volume.
Basis of Cost Estimation
The activity to carry out the study was assumed to be ready to
function, staffed with knowledgeable investigators who would be avail-
able at any level of employment, thus avoiding the problem of planning
a study to fit full employment of personnel. The availability of basic
facilities and services was assumed, with this cost being covered by over-
head charges which we calculated as 50 percent of personnel costs. Appro-
priate motor-pool vehicles were assumed available at a charge of $0.15
per mile and field subsistence was calculated as $23.00 per day. Our
own field costs per sample were doubled to allow for covering an entire
river basin rather than only the conveniently located study areas which
we used to minimize travel. We assumed that there would be no added
costs for taking more water samples after the first one was taken by
the team when it reached a site on a particular trip.
Personnel were assumed to be available, trained in the required
skills, and employable by the hour at any fractional level up to full
time. The costs of accomplishing tasks were mostly counted in hours
and then evaluated on the basis of a work year being 2,080 hours with
annual reimbursement of personnel (including fringe benefits) being
set at:
Administrator $25,000
Investigator $16,000
Trained technician $ 9,000.
All equipment was assumed to be totally depreciated at the end of
a one-year study.
For field construction at sites to have an automated sampler and
recording gage, the work was assumed to be contracted but with planning,
supervision, and inspection remaining a cost to the investigating agency.
This field construction would cover a stilling well and a vandal-proof
instrument housing. Use of a digitalized recording gage was assumed.
To calculate the costs of automated sampling it was necessary to make
assumptions about the number of service visits and the number of water
samples to be taken and stored between these visits. The figures of
weekly visits and 3.3 samples per visit were used, drawn from our
experience with a stage-activated system.
The per-site and per-visit costs for using an automated sampler
were calculated two ways (which yielded the same overall cost of an
automated site). In the first calculation, the cost of weekly service
visits was allowed as a per-visit cost, to which was added the cost of
chemical and statistical analysis of the average 3.3 samples taken per
52
-------
week. The second method of calculation reasoned that weekly service
trips to the sampler were an absolute requirement of any use of the
instrument, and therefore that the cost of 52 service trips should be
added to the annual per-site cost. This view reduced the per-visit
costs to those of chemical and statistical analyses of a single water
sample. This second method of calculation is necessary if one is to
speculate on the optimal allocation of automated sampling resources
between number of sites and number of samples per site.
The costs of chemical analyses were assumed to be contracted to
a commercial laboratory with the costs of supervision, record keeping,
and quality control remaining as further costs to the project. To
establish the costs of the chemical analyses we obtained from several
commercial laboratories their charges for handling a range of water
analyses at several levels between 250 and 3,000 per year. The
resulting information showed general use of a volume discount with the
costs fitting fairly well the following linear function:
total cost = $2,200 + $30.10 (number of analyses).
This function fitted well only if more than 250 analyses were planned
and yielded results that were somewhat high for 3,000 analyses. It
did, however, provide us with information in the form of our cost
function. The planned determinations included in each analysis of a
water sample included: chemical oxygen demand, Kjeldahl nitrogen,
nitrate, total phosphorus and chloride for every sample; biological
oxygen demand (5-day), total solids, and ammonia for one-quarter of
the samples.
For data management, we assumed that general facilities would be
available, including a high-capacity computer, general software, and
competent personnel, for the use of which standard computer charges
would be paid. We estimated the cost of developing our specific soft-
ware such as programs, procedures and storage and have listed this as
a separate item. We have attempted to estimate the reduction in costs
if some other agency were to use our computer programs instead of
developing their own; such use of our work would have to be based upon
familiarity with SAS (Barr et al. 1976).
METHODS OF PROCEDURE
The primary objective of this study was to explore the feasibility
of applying probability sampling to the characterization of rural run-
off as represented by the output of small drainage basins. A secondary
objective was to compare the cost-effectiveness of grab sampling and
automated sampling in making this characterization. To accomplish
these objectives we went through all the steps of a probability sampling
survey with the only departure from a real survey being that we restricted
our activities to a relatively small portion of the Chowan River basin.
By not covering the entire basin we were able to devote more of our
resources to field sampling and less to travel.
53
-------
Any sampling survey proceeds according to six steps. First, there
is defining the problem, then second, building a sampling frame, third,
selecting a sample, fourth, gathering the data, fifth, making an estimate
of some characteristics of the universe under study, and sixth, describing
the precision of the estimate by calculating a variance or standard
error.
Definition of the Problem
Definition of the problem includes a statement of its general form,
and identification of the appropriate statistical model (as outlined
previously). It also includes a precise statement of the universe
under study, and a delineation of what parts of the universe are to be
excluded from study.
Sampling in this problem was in two dimensions, spatial in the
selection of sites for study, and temporal in the selection of times
to visit these sites, with time stratified into 4-week periods. The
decision to use 4-week periods was made arbitrarily at the start,
based upon the need to have two visits to each site within each period
(to provide a basis for calculating a variance), and the number of
sites thought practical to cover. In the actual study, there was also
an element of stratification in space in that 4 study areas were arbi-
trarily chosen as easily accessible work sites, each represented a
recognizable geoclimatic type. Within each of these study areas all
sites were listed and a total of 15 chosen at random, with 4 sites in
the well-drained coastal plain, 4 sites in the poorly-drained coastal
plain, 4 sites in the siIvicultural piedmont and 3 sites in the agri-
cultural piedmont. This selection of different geoclimatic types
assured a good test of feasibility of field operations and provided a
broad base of observation, but the sample sizes were too small to allow
separate estimates for these strata. Also, because of this arbitrary
limitation of our study, our results cannot be used to characterize
the Chowan River basin, but apply only to the portions within which we
worked.
If often happens, as it did here, that it is impossible or imprac-
tical to extend probability sampling over the entire spatial or
temporal universe. For example, our definition of small drainage
basins (below) requires a road crossing as a sampling point. The spa-
tial sampling frame covering the small drainages in the Chowan River
basin as defined included 61 percent of the total basin area. Those
small drainages excluded include those that, before crossing a road,
join tributaries that have already been listed in their headwaters.
Also excluded are any areas draining directly into such tributaries
or into the main stream, as well as the surface area of the tributaries
and the main stream. In addition, certain sites chosen on a map may
not be useful in the field either because they are inaccessible or for
other reasons. Finally, there may be mistakes in the map and unrecorded
drainages may exist. Thus the scale of the map and the policies of the
cartographer become an element in the specification of the problem.
54
-------
For this reason, it would be important to use the same kind of map for
comparative studies.
Certain exclusions were also made in the time sampling frame. For
example, following an almost universal practice we confined all of our
sampling activities to the daylight hours, even though it is known that
dissolved oxygen and some biological parameters may differ from day to
night. We included weekends and holidays in the sampling schedule,
even though many studies do not.
In principle, this problem of possible exclusions from the universe
under study may be met in two ways. First, and probably most desirable,
is to subdivide, or stratify the universe, and cover the more difficult
portions with different methods and perhaps lower sampling intensity.
For example, it would be possible, though expensive and difficult, to
sample at night, and probably the excluded small drainages could be
covered by using different methods. But to make such a complete study
would take resources from the rest of the study. The second, and more
common method of meeting the problem, is to clearly define the universe
to include only the parts studied. Then, after the study is completed,
the investigator may state his opinion as to how far the findings
should be extended, making it clear that any extension beyond the areas
sampled is by judgement unsupported by data.
There are advantages to a complete coverage of a universe. There is
no need then to qualify the results concerning portions omitted or to
justify the extension of the conclusions to the entire universe unsup-
ported by anything but opinion. There may in fact be real differences
between the more accessible and the less accessible portions of a river
basin and a complete study may make these known. On the other hand, a
complete study is often costly, wearing on personnel, and may take
resources away from areas of primary interest. Further, if studies of
other workers have covered only part of the universe, one must be care-
ful to stratify a more complete study so that partial results may be
compared directly.
Sampling Frames
Separate frames were drawn up for sampling in space and in time.
Sampling in space was carried out on four U.S.G.S. 1:250,000 topo-
graphic maps (NJ17-9 Roanoke 1971; NJ17-12 Greensboro 1966; NJ18-7
Richmond 1974; NJ18-10 Norfolk 1973). On each map all of the small
drainage basins in the Chowan River basin were outlined according to
our definition which was:
"Start at origin of stream, trace down to road crossing
that defines the single largest basin in the range 1.3-
90.6 square kilometers (0.5-35 square miles); do not
count as a road crossing the situation where a bridge
crosses an impoundment but do count as a road crossing
a road on top of a dam; omit a basin if the only streams
55
-------
are marked 'dry' or 'intermittent1; trace basin boundary
by topography, otherwise by roads or railroads, otherwise
divide equally between streams."
One could eliminate any small basins that are primarily urban or that
receive major point discharge if the emphasis is to be on rural runoff.
Alternatively, all small basins as defined might be listed, as here,
where urban was distinguished on the basis that a substantial portion
of the basin fell within city limits as indicated on the map.
Each such small basin was outlined on the map and identified with
a serial number; the size as determined by planimeter was recorded as
well as the identity of the tributary system and whether rural or urban.
At this stage it would be possible to divide the river basin into
several subareas or sampling strata. The objective of such stratifica-
tion might be to reduce the variability of the streams being sampled,
to derive estimates for subareas, or to exercise some control over the
distribution of sample sites. This question did not arise in the
present feasibility study which was confined to four minor, arbitrarily
selected portions of the Chowan River basin.
The frame for sampling in time is a calendar. Most environmental
parameters have a strong seasonal element and therefore stratification
in time is desirable. In planning this study it was reasoned that the
shorter the periods used as time strata, the more they would remove of
any seasonal variation. For this reason, time strata were planned to
be no longer than necessary to provide two visits per period to the
number of sites to be covered; this required 4 weeks. The field schedule
was planned to provide for certain ability to cover all sites twice
during each period. Although 3 trips of 3 days each were thought ade-
quate to cover all sites and return samples, a further day was allowed
each trip for emergency use if needed. Six such 4-day intervals
allowed far visiting each site twice and a seventh equal interval
allowed time for repair of equipment and time off for personnel. Also,
7 periods of 4 days resulted in each period starting on a different
day of the week, thus balancing any day-of-the-week effect.
Drawing the Sample and Planning the Field Schedule
A sample of sites is drawn from the serially numbered list of small
basins. In our case, the sites were drawn from a complete list of all
sites in the chosen work areas but in principle, the sample would be
drawn from the entire river basin, or a designated number would be
selected from each of the strata if these were set up.
In drawing such a sample, a table of random numbers is used. With
a relatively small universe (here 397 possible sites) the items of the
universe may be listed by serial number and numbers of sufficient digits
drawn from a random number table to identify the sample sites, ignoring
all numbers greater than the largest serial number and any random
numbers that are repeated. For a larger universe it may be more
56
-------
convenient to assign a random number to each member of the list and then
re-sort the list in order of random number and accept the first so many
as a designated sample. This last operation is most easily carried out
by computer.
Once a sample of sites has been chosen, there is a question of the
order in which these are to be visited on any trip into the field. If
the sites are always covered in the same order, then each site may always
be visited at about the same time of day, thus confounding any time of
day differences with site differences. Further, the same set of sites
will be always visited on the same day, thus increasing correlations among
sites. For statistical purposes the best method would be to visit the
sites in a completely random order, rerandomizing this order for each
visit. But with sites spread over an entire river basin such a plan
would be impractical. As a compromise, we grouped the 35 sites to be
covered into 3 "routes" each of which constituted a practical unit of
field effort, requiring three work days to leave Raleigh, visit the sites,
and return. For any one route, several practical orders of visiting the
sites were set up and one of these chosen at random for any one visit.
In laying out the field schedule (Table 4) the interval for repair
and rest was allocated first at random to 1 of the 7 possible positions
within the 28-day period. Then, of the remaining six 4-day intervals,
the first 3 were allocated to the first visit and the rest to the second
visit. The 3 routes were assigned at random to each set of three 4-day
intervals (using a random permutation). Then, within each route an
order of visiting the sites was designated.
Use of Logarithmic Transformation
Among the problems of data management is the fact that the logarithmic
transformation was used throughout the analyses of these data. The
logarithmic transformation was used for several reasons. First, variation
in a number of measurements and especially in the measurement of flow can
more reasonably be thought of as proportional or multiplicative rather
than as additive in an absolute sense. Statistical methods used here are
based upon additive linear models, and multiplicative models become
additive when the data are log-transformed. The distributions of many
of the data were strongly skewed with relatively few large values and
many small values; use of the logarithmic transformation tends to bring
such distribution closer to the normal and thus more tractable of statis-
tical manipulation. Further, standard statistical tests such as the
analysis of variance assume that the standard deviation is independent
of the mean while with many of these measurements the standard deviation
was obviously correlated with the mean when subsets of the data were
examined. Use of the logarithmic transformation makes the standard
deviation independent of the mean value under these conditions.
Use of the logarithmic transformation creates some difficulties in
the interpretation of results. When a transformation is used, tests of
hypotheses and statements of confidence intervals must be carried out
57
-------
TABLE 4. EXAMPLE OF FIELD SAMPLING SCHEDULE (PERIOD 27, VISITS 53 AND
54) SHOWING ORDER OF VISIT TO SITES
Year
Month
Visit
Number
1976
Nov
53
Dec
53
53
54
54
54
Day Alternate Period 27
Route or NH3 (sites to be sampled from left to right)
Group sample
28
29
30
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
C
III A NH3
B
B
I A A
C
A
II B A
C
off
C
I A NH3
B
A
II B
C
C
III A A
B
M2-17A; M2-18K; M2-19W
M3-20K; M3-28Y; M3-22M; M3-23A; M3-24VI;
M7-27W; M1-26W
W10-16W; W8-15W
P10-6Y; P10-1A; P10-2B; P10-3K; P10-4W;
W3-13W
U1-32W; U8-30W; U4-31W
P11-9W; P11-7A; P11-8K; W4-14W
P13-10W; P8-11K; P8-12W
W3-13W
P10-29Z; P10-1A; P10-2B; P10-3K; P10-4W
W10-16W; W8-15W
U4-31W; U8-30W; U1-32W
P11-7A; P11-8K; P11-9W; W4-14W
P13-10W; P8-11K; P8-12W
M2-17A; M2-18K; M2-19W
M3-28X; M3-22M; M3-23A; M3-24W; M3-25X;
M7-27W; M1-26W
M3-25X
P10-29Z
; P10-6Y
M3-20K
Note: Geoclimatic codes changed in manuscript: U is now A, M is now F.
58
-------
with the transformed data. When the mean value of a series of log-
transformed data is detransformed, the result is the geometric mean
rather than the arithmetic mean. Most readers are accustomed to
reasoning from the arithmetic mean; the geometric mean is always smaller
than the arithmetic mean by a factor that depends upon the variability
of the data (if the data are truly log-normal). Therefore, to allow
comparison between different studies and allow interpretation 'of the
data, it is desirable to state both the arithmetic mean and the geometric
mean, and not the mean in logarithmic units because this is difficult to
translate to the proper scale. Further, detransformed values of the
standard deviation, standard error, or confidence interval must be read
as a factor to be multiplied by or divided into the geometric mean
because in the log-transformed form such statistics are understood to be
added or subtracted. Thus, in this way the standard error is stated as
a multiple of the (geometric) mean value.
The proportional standard error, that is, the standard error divided
by the mean, is often used as a statement of relative precision of an
estimate. This quantity is also sometimes called the coefficient of
variation, although that usage is not followed here because it generates
confusion with an identically named quantity, the standard deviation
divided by the mean. When using logarithmically transformed data, an
equivalent to the proportional standard error may be derived by taking
the anti-logarithm of the standard error, and subtracting unity.
Extensive use of this quantity is made in comparing relative precision
of different sampling plans later in this study. For a feasibility study
such as this, this value of the proportional standard error has the great
advantage that it does not require the direct calculation of the mean
value but may be derived directly from the component variances.
Use of the logarithmic transformation raises a further problem, that
of what to do about zero values. At some times during this study, the
measurement of flow rate was actually zero either because it was too
low to be measured, or because the stream bed was dry, at least in
sections. The conventional advice under these conditions is to add unity
to each quantity before taking the logarithm, but this seems highly inap-
propriate when the measurements concerned may be small decimal values.
The practice adopted was to substitute for the zero values one-tenth of
the lowest recorded non-zero value in the data series.
Missing Data
There were a certain number of cases where data were not recorded.
These fell into four categories, the first of which was when the sampling
plan called for making the measurement on only part of the visits. For
example, ammonia was determined for only one visit per period instead'
of the usual two. In this case, calculations would have to follow the
special design that resulted from this change.
The second category of missing information resulted from no concen-
trations being measured when there was zero flow. The desired mean
59
-------
concentration was that when there was a flow of water; with no flow, no
measurement was made. This constitutes a definition limiting the scope
of the study and corresponds to what Cochran (1977) describes as domains
of interest. The overall mean value of concentration was calculated
without including any value when there was no flow of water. However,
when calculating a variance of the mean, a value of zero was substituted
for the missing value, following the advice in this reference.
The third category of missing information was when there was a mea-
sured non-zero flow but no concentration was measured because of accident,
as when the sample was lost. Here a synthetic missing value was substi-
tuted, calculated by regression on values at other sites in the same
geoclimatic stratum.
Finally, there was the case where flow was not measured for various
reasons, most commonly either because flow was too low to be measured,
or when flood conditions prevented a measurement. Concentrations
measured under these conditions were valid. A flow value was synthesized
and substituted, based upon regression on values at other sites.
To synthesize substitute values where data points were missing we used
a multiple regression on values for other sites of the same geoclimatic
stratum. The regression equation was set up on the basis of available
data, and then the missing value was calculated as that expected from the
values observed at the other sites. Such missing Values constituted a
very small percentage of the total number.
Making an Estimate
To calculate the mean value for a study of, say, 1-year duration,
first the mean for each of the periods should be calculated and then the
annual value calculated as the mean of the period means. Within each
period there should be a value, real or synthetic (and possibly zero)
for each site visit for flow or for transport but not necessarily a
value for concentration.
To estimate a variance and a standard error for such a mean value
it is necessary with this sampling design to synthesize a value following
the reasoning of the statistical model. The sampling design is based
upon repeated use of several randomly selected sites which are sampled
at random in time and one cannot calculate a value for the variance in
the usual manner, assuming that all measurements are independent. In
other words, it is necessary to allow for the fact that the sample size
for variability among sites is limited to the number of sites selected,
no matter how many times each of these may be visited. The synthesizing
of such an error term is carried out in two steps, first, the estimation
of the variance components from the field data, and second, the calcula-
tion of the standard error, which is presented here mostly as the
proportional standard error.
60
-------
Components of Variance
Components of variance are probably best understood in the context
of an analysis of variance and are discussed in a number of textbooks
including that of Steel and Torrie (1960; see p. 214 for a mixed model
similar to this design). We are concerned with a mixed model where the
effect for period is fixed; there is no question of sampling periods,
each is represented. All other effects, including sites, visits within
periods, the interaction of sites and periods, and the error term, are
random effects. This design is laid out in the usual form in Table 5
which shows the average values expected for the mean squares for each
of the effects. Variance components are estimated by subtracting the
correct mean square values and dividing by the appropriate coefficients
(Table 5 includes a numerical example).
Variance components were estimated separately for each of the geo-
climatic strata, resulting in four different values for each water quality
parameter, one for each stratum, as illustrated for yield of water in
Table 6. As shown in this table, these values varied considerably from
stratum to stratum, each being based upon a very small sample number of
sites. For further work, the arithmetic mean of these four values
was used. In principle it would be possible to proceed with the infor-
mation from each stratum separately and then combine these into a
weighted average. We chose to average the components and then to discuss
the problem as though it were one of simple random sampling, unstratified
in space. This step was taken because of the high variability, the
small sample sizes within each stratum, and for lack of good evidence
that the components differed among strata (pair-wise F-test of variance
ratio).
While it is quite possible to calculate variance components by hand
as illustrated in Table 5, it is more practical with large amounts of
data to use a computer program as we did. The SAS program used calculated
the components as though all effects including periods were completely
random, and not as a mixed model, with the result that variance components
for sites had to be recalculated by hand where the interaction component
was non-zero.
Estimating Standard Error and Comparing Precision
Estimates of the standard error of the estimate for a particular
study may now be calculated, given the values for the components of
variance. These values are substituted in formula (2), which is:
var x
= (02/s) + (a*S]I)/sp) + (aj/pv) + (o^/spv) (2)
The constants in the denominators depend upon the survey design; in our
study there were 15 sites (s), 13 periods (p), and 2 visits per period
(v). Thus using the information on water yield from Table 6 the vari-
ance of an estimate for a 1-year study would be:
61
-------
TABLE 5. AVERAGE VALUES FOR MEAN SQUARES AND EXAMPLE (LOG-TRANSFORMED
WATER YIELD DATA FROM POORLY-DRAINED COASTAL PLAIN),
SHOWING ESTIMATION OF VARIANCE COMPONENTS
Source of
variability
Site
Period
Site x Period
Visit within Period
Error
Examjj
Degrees
of
freedom
3
25
75
78
207
le
Mean
square
17.9169
5.3020
0.9583
1.0051
0.2961
Average value of
square
9 i ..9
aa + PVCTC
°e + vatsn) + SOV
ae Scrv
a2
mean
+ svzn2
FT
Estimates:
Variance component for Site = (17.9169-0.2961)/52 = 0.3389
Variance component for Period = (5.3020-0.9583-1.0051+0.2961 )/8 = 0.4543
Variance component for Site x Period = (0.9583-0.2961)72 = 0.3311
Variance component for Visit within Period = (1.0051-0.2961)/4 = 0.1772
Note: Data are for s = 4 sites, p = 26 periods of 4 weeks (2 years), and
v = 2 visits per period; recall that Period is a fixed effect.
TABLE 6. EXAMPLE SHOWING VARIANCE COMPONENTS AS DETERMINED FROM.EACH
OF THE 4 GEOCLIMATIC STRATA, AND MEAN VALUES, FOR WATER
YIELD (DATA LOG-TRANSFORMED)
Components of
Land-use stratum
Well -drained coastal plain
Poorly-drained coastal plain
Silvicultural piedmont
Agricultural piedmont
Number
of
sites
4
4
4
3
Site
0.2007
0.3389
0.0600
0.2142
Site
x
Period
0.2163
0.3311
0.0900
0.2709
variance for:
Visit
within
Period
0.6465
0.1772
0.2636
0.1782
Error
1.2521
0.2961
0.3771
0.1691
Mean 15 0.2034 0.2271 0.3164 0.5236
62
-------
var x = (0.2034/15) + (0.2271/195) + (0.3164/26) + (0.5236/390)
= 0.02824.
The standard error of the estimate would be the square root of this
quantity, or 0.1680. The anti-logarithm of 0.1680 is 1.472 and sub-
tracting 1, the proportional standard error is 0.472. Similar
calculations were made for other parameters.
The primary objective of this study, however, was not to calculate
estimates and standard errors for this actual study, but rather to
compare the precision obtainable under optimal allocation of effort
with various budgets. For example, the budget available for allocation
of sampling effort for a design of the kind we used would be $33,291
(as calculated from formula (3); recall that this is in 1975 dollars
and that $14,211 must be added as a constant or overhead cost). By
coincidence, it happened that for water yield the allocation of effort
in the design we used was exactly that to be calculated as optimal for
this budget from formulas (4) and (5), that is, two visits per period
and 15 sites. Now imagine that we asked what would happen to the
precision if we doubled the budget to a total of $95,004 (with $80,793
available for sampling). Here the optimal allocation would be 3
visits per period and 25 sites. Recalculating the variance of the
estimate:
var x = (0.2034/25) + (0.2271/325) + (0.3164/39) + (0.5236/975)
= 0.01748.
Here the standard error would be 0.1322 and the proportional standard
error 0.356. Note, incidentally, that the ratio of the standard errors
(0.1322/0.1680) is about 0.79 instead about 0.71, the value expected
when doubling sampling effort under simple random sampling. Neither
the number of sites nor the number of visits could be doubled, rather,
each was increased one-half or two-thirds.
Similar calculations were made for the water quality parameters
studied here at a range of budgets between about $28,000 and $269,000;
in one comparison the upper level was $497,000. The particular budget
sizes to be compared were chosen in anticipation of also calculating
precision from use of automated sampling, where the budget would have
to be increased in the large increment of $13,415.90, the cost of
adding a single automated site. Thus the lowest value, $27,854, was
the sum of the constant cost for automated sampling ($14,438) plus the
cost of a single automated site. Statistically speaking, such a design
would be impractical for any study where an estimate of variance was
needed because there would be only a single value for site, but this
lowest budget could be used with grab sampling and it provided a con-
venient starting point for the range of budgets to be examined.
63
-------
The precision attainable was calculated under the present design at
different budget levels and also under designs without stratification
or where the length of periods was changed. In each case the procedure
was the same as above. Using the components of variance calculated from
our field experience, the optimal allocation as to number of sites and
number of visits was calculated using formulas (4) and (5) and then a
proportional standard error was synthesized. This allowed a comparison
to be made under optimal conditions at the same budget levels.
RESULTS
Results are reported here for drawing up the sampling frame, for
some incidental investigations on the usefulness of stratification in
time and the effect of changing length of periods, then on the calcula-
tion of components of variance and on analysis of costs. This information
allows the exploration next of the relative precision attainable at
different budgets, and a comparison of the cost-efficiency of grab sam-
pling as compared to automated sampling.
The Samp1 ing Frame
The sampling frame was drawn up for the entire Chowan River basin
by delineating each of 397 small drainage basins on U.S.G.S. 1:250,000
topographic maps, determining the area of each basin by planimeter,
numbering and listing the basins according to major tributary of the
river. This frame was constructed to confirm the feasibility of this
part of the sampling approach and determine the cost of such an opera-
tion even though the frame itself was not used in our study. Constructing
such a frame is not inexpensive; the 100 hours of technician time required
made up by far the largest single item in the cost of planning the study.
Table 7 shows the distribution of these small drainage basins according
to tributary, whether rural or urban, and by size class (with limits
set arbitrarily to divide the total small-basin area about equally).
The table also repeats the definition according to which the basins were
outlined on the map. Such small basins accounted for 61 percent of the
total area of the Chowan River basin.
Effect of Stratification in Time
The effect of stratification on precision was considered in two ways,
first, by contrasting precision attained when using stratification (as in
our design) with results using no stratification, and second, considering
the effect of using time strata of differing length.
The relative precision of grab sampling with and without stratifica-
tion in time is shown in Table 8 as proportional standard error. Optimal
allocation of effort was assumed whether with or without stratification;
this distribution of effort is recorded for 3 budget levels. In most
cases, stratification in time seemed to provide appreciable advantage.
For example, with flow measurements this stratification reduced propor-
tional standard error by 15 percent-at the lowest budget and by 24 percent
64
-------
TABLE 7. SMALL DRAINAGE BASINS OF THE CHOWAN RIVER SYSTEM, LISTED BY TRIBUTARY, SIZE AND WHETHER RURAL
(R). URBAN (U)
CTl
Ol
Small BVackwater Nottoway Sbmerton Meherrin
basin River River Creek River
area
km2
(mi2) R U R U R U R U
<2.6 16 - 31 - - 15 1
5.2-20.7 27 5 77 2 4 54 5
(2-8)
23.3-38.8 12 - 13 4 4 20 2
(9-15)
41.4-57.0 43 13 1 11 63
(16-22)
59.6-90.6 31 10 2 - 73
(23-35)
Total 62 9 144 9 91 102 14
Area in
small R 1072 2504 215 2010
^linS U 259 350 57 469
Wiccacon Other Total Percent of
Creek total Chowan
basin areat
R U R U R U Sum
7 - 6 - 75 1 76
9 - 11 1 182 13 195
21 3 - 54 7 61
1 2 - 27 8 35
11 2 - 23 7 30
20 2 24 1 361 36 397
Remainder of Chowan basin
306 470 6577
93 18 1246
Total 7823
1.4
15.6
14.7
13.3
16.1
61.1
38.9
* Data from planimeter measurement of small basins outlined on U.S.G.S. 1:250,000 topographic map. Rule
for defining a small basin: "Start at origin of stream, trace down to the road crossing that defines
the single largest basin in the range 1.3-90.6 square kilometers (0.5-35.0 square miles); do not count
a bridge crossing an impoundment as a road but do count a road on a dam; omit if the only streams are
marked "dry" or "intermittent"; trace basin boundary by topography, otherwise by roads or railroads,
otherwise divide equally." Urban was differentiated from rural on the basis that a substantial portion
of the basin fell within city limits as indicated on the map.
t Total area of Chowan basin -- 12,802 km3 (4,943 mis).
-------
TABLE 8. PRECISION OF MEASUREMENT (AS PROPORTIONAL STANDARD ERROR o BE
EXPECTED WITH GRAB SAMPLING WITH AND WITHOUT STRATIFICATION IN
TIME (PERIODS), IN EACH CASE WITH OPTIMAL ALLOCATION OF SAMPLING
EFFORT, SHOWN AS (NO. OF SITES/TOTAL NO. OF VISITS PER SITE) AND
UNDER 3 BUDGET SIZES
StratifiedApproximate budget for a one-year study
into time
periods? $28,000 $108,000 $269,000
Flow
Yield
of
of
Water
Water
yes
no
yes
no
0
1
0
0
.906(6/26)
.060(7/22)
.694(6/26)
.837(6/29)
0
0
0
0
.412(42/26)
.538(20/57)
.340(29/39)
.454(15/75)
0.
0.
0.
0.
299(60/52)
394(34/93)
254(48/65)
332(26/123)
Concentration of:
Nitrate Nitrogen
Kjeldahl Nitrogen
Total Phosphorus
Chloride
yes
no
yes
no
yes
no
yes
no
0.304(6/26)
0.187(25/5)
0.115(4/39)
0.119(3/55)
0.148(6/26)
0.152(7/23)
0.104(6/26)
0.106(8/21)
0.110(42/26)
0.104(79/13)
0.064(11/104)
0.067(8/143)
0.080(29/39)
0.087(19/60)
0.054(29/39)
0.062(21/55)
0.069(115/26)
0.079(134/22)
0.048(20/156)
0.050(14/236)
0.061(40/78)
0.066(32/98)
0.042(48/65)
0.047(35/90)
* Log-transformed data; proportional standard error is expressed here as
antilog standard error, minus 1. The smaller this quantity, the better
the precision.
at the highest. For nitrate nitrogen, however, the situation was reversed
with the non-stratified study providing a potential of considerably better
precision at the lowest level of budget. This anomaly arose because of
the relative predominance of the variance component for site which, as is
pointed out later, was not well accomodated by the particular sampling
design used. At the highest budget considered for the same material, the
ratio was reversed with the stratified design having a proportional
standard error 13 percent less than the unstratified, thus resembling the
balance with the other parameters.
The effect on the variance components of changing the length of period
from four weeks to 10 weeks was explored empirically with the data of this
study. This was possible because our design called for one visit to each
66
-------
site within each interval of about two weeks and while the original plan
was for two visits per four-week period, it was possible to redraw the
lines defining periods and include successively more visits, up to five,
within a period. This manipulation of the data was carried out with only
three of the variables, flow of water and concentration of nitrate nitrogen
and Kjeldahl nitrogen, as shown in Table 9. The proportional standard
error was calculated on the basis of a one-year study assuming 15 sites
and a total of 26 visits per site, divided into periods of varying lengths.
This table also lists the equivalent component for periods which is not
listed in other tables because it is not of itself a contributing element
to the variance of the mean; in fact, the object of stratification in
time is to eliminate this element. As period length increased, with
Kjeldahl nitrogen concentration there was the expected decrease in the
component for periods with a corresponding increase in the component for
visit within periods; there was no consistent change with flow of water
or concentration of nitrate nitrogen.
TABLE 9. EMPIRICAL EXPLORATION OF EFFECT ON VARIANCE COMPONENTS OF CHANGING
FROM 2 VISITS (4 WEEKS) TO 5 VISITS (10 WEEKS) PER PERIOD FOR THE
SAME SET OF DATA
Length
of period
in visits
Flow of
Water
Nitrate
Ni trogen
Concentratio
2
3
4
5
2
8
4
5
Kjeldahl 2
Ni trogen -.
Concentration^
4
5
Period*
0.6258
0.6814
0.6066
0.6903
0.0082
0.0082
0.0058
0.0070
0.0201
0.0178
0.0029
0.0072
Variance
Site
0.3573
0.3611
0.3574
0.3726
0.0766
0.0762
0.0767
0.0758
0.0032
0.0032
0.0032
0.0033
components for:
Site
X
Period
0.1955
0.1172
0,1446
0.1649
0.0137
0.0087
0.0095
0.0108
0.0100
0.0057
0.0052
0.0146
Visit
within
Period
0.3394
0.3208
0.3877
0.3368
0.0043
0.0035
0.0058
0.0056
0.0312
0.0306
0.0490
0.0420
Pro
Error s
0.5356
0.6220
0.5950
0.5952
0.0256
0.0314
0.0303
0.0304
0.0686
0.0747
0.0741
0.0647
portional
tandard
errorf
0.578
0.575
0.601
0.598
0.184
0.184
0.186
0.186
0.098
0.097
0.118
0.114
* Period is a fixed effect.
t Using design of present study, i.e., 15 sites with a total of 26 visits
per site, for 1 year. Log-transformed data; proportional standard
error is expressed here as antilog standard error, minus 1. The
smaller this quantity, the better the relative precision.
67
-------
Components of Variance
Components of variance based upon the present study are listed in
Table 10, which also shows the proportional standard error for a one-year
TABLE 10. COMPONENTS OF VARIANCE AND PROPORTIONAL STANDARD ERROR FOR
SELECTED PARAMETERS OF WATER QUALITY IN THE CHOWAN RIVER
BASIN; COMPONENTS ARE AVERAGES BASED ON 15 SITES, EACH
VISITED 52 TIMES IN THE WELL-DRAINED COASTAL PLAIN (4 SITES),
THE POORLY-DRAINED COASTAL PLAIN (4 SITES), THE SILVICULTURAL
PIEDMONT (4 SITES) AND 38 TIMES IN THE AGRICULTURAL PIEDMONT
(3 SITES), DATA LOG-TRANSFORMED
Variance components for: Pr
Flow of Water
Yield of Water
Concentration of:
Nitrate Nitrogen
Kjeldahl Nitrogen
Total Phosphorus
Chloride
Transport of:
Nitrate Nitrogen
Kjeldahl Nitrogen
Total Phosphorus
Chloride
Yield of:
Nitrate Nitrogen
Kjeldahl Nitrogen
Total Phosphorus
Chloride
Site
0.3573
0.2034
0.0766
0.0032
0.0126
0.0071
0.9507
0.3481
0.4103
0.3274
0.8282
0.1649
0.1870
0.1926
Site
X
Period
0.1955
0.2271
0.0137
0.0100
0.0219
0.0104
0.3259
0.0490
0.2144
0.1808
0.3467
0.0547
0.2321
0.1984
Visit
within
Period
0.3394
0.3164
0.0043
0.0312
0.0226
0.0092
0.7271
0.4553
0.4044
0.2941
0.6570
0.4070
0.3721
0.2642
Error
0.5356
0.5236
0.0256
0.0686
0.0555
0.0258
1.0553
1.0869
0.5820
0.4724
0.9595
1.0366
0.5521
0.4462
•oportional
standard
error
0.578
0.472
0.184
0.098
0.107
0.074
1.039
0.619
0.634
0.541
0.955
0.486
0.484
0.441
* Using the design of present study, i.e., 15 sites, 13 periods, each
with 2 visits per site for 1 year. Log-transformed data; proportional
standard error is expressed here as antilog standard error, minus 1.
The smaller this quantity, the better the relative precision.
68
-------
survey based on these components and using the present design, i.e., 15
sites, 13 periods of 4 weeks each, each with 2 visits per site. These
components of variance are averages of values calculated for each of
the four geoclimatic study areas and the proportional standard error is
calculated from these average values as though there were no stratifica-
tion in space. Components of variance were calculated here for a selected
set of parameters, including flow of water, yield of water per unit area
of drainage, and concentration, transport and yield of nitrate nitrogen,
Kjeldahl nitrogen, total phosphorus and chloride. This set of variables
was then examined in greater detail with reference to planning an effi-
cient study at each of a range of budget levels.
These components of variance provide part of the raw materials for
calculating the optimal allocation of sampling effort. Information on
cost is also required; this is presented next.
Costs
Information on cost is presented in four tables, two each for grab
and for automated sampling. Tables 11 and 12 present the information
on grab sampling and Tables 13 and 14 the data on automated sampling;
in each pair the first table presents only the personnel costs and the
second table the total costs including personnel so that the non-
personnel costs may be obtained by subtraction. Each table assumes a
one-year study of water quality based upon probability sampling. Costs
are presented under the three headings of constant (overhead) cost, the
cost per site established, and the cost per single visit to an estab-
lished site. All costs are stated here in 1975 dollars; these values
must be expanded using an appropriate multiplier if guidance is to be
sought here in planning other studies.
The information on costs that is used in the further work with
optimal allocation may be summarized as the following values, which
are treated hereinafter as constants:
Grab sampling Automated sampling
Fixed cost (c-,) $14,211.25 $14,438.09
Per-site cost (c } 153.18 (1) 5,667.38
5 (2) 7,277.30
Per-visit cost (c ) 79.47 (1) 149.01
v (2) 32.89
For automated sampling, the per-site and per-visit costs were cal-
culated two ways, (1), to fit the sampling-schedule used in our study,
and (2), counting field sampling costs of 52 weekly service visits as
per-site costs, reducing the "per-visit" costs to those of chemical
and statistical analysis of a single water sample.
69
-------
Fixed costs were approximately the same for both types of sampling.
The per-site cost was much greater for the automated sampling because of
the instruments and construction required at each site. Depending upon
how service trips were accounted, the per-visit cost for automated
sampling"was either about twice as great or about half as much as that
for grab sampling.
TABLE 11. PERSONNEL COSTS FOR A ONE-YEAR GRAB-SAMPLING STUDY OF WATER
QUALITY OF A RIVER BASIN BASED ON A PROBABILITY SAMPLING OF
SITES AND VISITS; COSTS* ESTIMATED FROM EXPERIENCE IN THE
CHOWAN RIVER RURAL RUNOFF STUDY
Item
Administration
Planning
Chemical analysis (contracted)
Field exploration, construction
Field sampling
Data management:
using our computer programs
not using our computer programs
Subtotal :
using our computer programs
not using our computer programs
Indirect costs: Add 50 percent of
Constant
cost
$1,919.94
664.76
384.50
0
615.20
856.16
2,417.10
$4,440.56
6,001.50
total personnel
Per site
cost
$ 2.94
0
0
62.49
7.69
0
0
$73.12
7.3.12
cost.
Per visit
cost
$ 0
.08
.46
0
18.86
2.05
2.05
$21.45
21.45
Example: Total personnel cost if our computer programs are not used
Total without indirect costs: $6,001.50 $73.12 $21.45
Total including indirect costs: $9,002.25 $109.68 $32.18
* Costs in 1975 dollars.
70
-------
TABLE 12. TOTAL COSTS (INCLUDING PERSONNEL) FOR A ONE-YEAR GRAB-SAMPLING
STUDY OF WATER QUALITY OF A RIVER BASIN BASED ON A PROBABILITY
SAMPLING OF SITES AND VISITS; COSTS* ESTIMATED FROM EXPERIENCE
IN THE CHOWAN RIVER RURAL RUNOFF STUDY
Item Constant
cost
Administration $
Planning
Chemical analysis (contracted)
Field exploration, construction
Field sampling
Data management:
using our computer programs
not using our computer programs
Subtotal :
using our computer programs $
not using our computer programs 1
Indirect costs: Add 50 percent of total
Example: Total cost (including personnel
not used
Total without indirect costs: $1
1,919.94
864.76
2,584.50
50.00
2,649.20
1,371.16
3,142.10
9,439.56
1,210.50
personnel
) if our
1,210.50
Total including indirect costs: $14,211.25
Per site
cost
$ 2.94
0
0
101.49
12.19
0
0
$116.62
116.62
cost.
Per visit
cost
$ 0
.08
30.56
.75
30.96
6.39
6.39
$68.74
68.74
computer programs are
$116.62
$153.18
$68.74
$79.47
* Costs in 1975 dollars.
71
-------
TABLE 13. PERSONNEL COSTS FOR A ONE-YEAR AUTOMATED SAMPLING STUDY OF
WATER QUALITY OF A RIVER BASIN BASED ON A PROBABILITY
SAMPLING OF SITES; COSTS* ESTIMATED FROM EXPERIENCE IN
THE CHOWAN RIVER RURAL RUNOFF STUDY
Item Constant
cost
Administration $1,
Planning
Chemical analysis (contracted)
Field exploration, construction
Field sampling
Data management:
using our computer program
not using our computer program 2,
Subtotal :
using our computer programs $4,
not using our computer programs 6,
Overhead: Add 50 percent of total personnel
919.94
664.76
384.50
184.56
615.20
856.16
417.10
625.12
186.06
cost.
Example: Total cost if our computer programs are not
Total without overhead: $6,
Total including overhead: $9,
186.06
279.09
Per site
cost
$ 2.94
0
0
193.29
7.69
0
0
$203.92
203.92
used
$203.92
$305.88
Per visit
cost
$ 0
.08
.46
0
18.86
2.05
2.05
$21.45
21.45
$21.45
$32.18
* Costs in 1975 dollars.
72
-------
TABLE 14. TOTAL COSTS (INCLUDING PERSONNEL) FOR A ONE-YEAR AUTOMATED
SAMPLING STUDY OF WATER QUALITY ON A RIVER BASIN BASED ON
A PROBABILITY SAMPLING OF SITES; COSTS* ESTIMATED FROM
EXPERIENCE IN THE CHOWAN RIVER RURAL RUNOFF STUDY
Item
Administration
Planning
Chemical analysis (contracted)
Field exploration, construction
Field sampling
Data management:
using our computer programs
not using our computer programs
Subtotal :
using our computer programs
not using our computer programs
Overhead: Add 50 percent of total
Constant
cost
$ 1,919.94 $
864.76
2,584.50
184.56
2,649.20
1,371.16
3,142.10
$ 9,574.12 $
11,345.06
personnel cost.
Per site
cost
2.94
0
0
5,550.29*
12.19
0
0
5,565.42
5,565.42
Per visit
cost
$ 0
.08
100.85f
0
30.96*
6.39
6.39
$138.28
138.28
Example: Total cost if our computer programs are not used
Total without overhead: $11,345.06 $5,565.42 $138.28
Total including overhead: $14,438.09 $ 5,667.38 $149.01
Annual cost per site, assuming weekly visits $13,415.90
* Costs in 1975 dollars.
t On the basis of 3.3 water samples per weekly visit.
* The total per-site cost of an automated sampler must include not only
this cost of the instrument and construction needed for its installation
but also the cost of servicing (number of annual service trips multi-
plied by per-visit cost of field sampling, above).
73
-------
Precision Attainable under Different Budgets
In considering the question of relative precision at different bud-
getary levels, budgets were set according to costs of automated sampling,
with the increment of increase being the cost of installing and operating
one automated site ($13,415.90). The lowest budget considered, $27,854,
was thus the sum of the constant cost for automated sampling ($14,438)
and the one-year cost of a single automated site. Budgets set in this
fashion and examined as to the potential precision of measurement both by
grab sampling and by automated sampling ranged over almost a ten-fold
span from $28,000 to $269,000 (and in one instance, to $497,000).
For each such budget, the optimal allocation of grab sampling sites
and visits was estimated and then the attainable precision of determina-
tion was calculated as the proportional standard error, using the variance
components already presented. Results of this process are presented in
Table 15 which shows for each of the parameters under each of five budgets
the proportional standard error and the optimal allocation of effort into
sites and visits, for a one-year study. It is clear from this table that
while respectable precision may be obtained with a low budget for concen-
tration of Kjeldahl nitrogen, total phosphorus or chloride, that a higher
budget would be required for the same precision with nitrate nitrogen
concentration, and that even with the highest budget considered here,
precision would not be good with either flow or yield of water (though
yield could be measured with better precision than flow, as should be
expected). Equally poor precision characterizes either transport or
yield of the chemical entities because these values are so strongly influ-
enced by flow. It is also clear that for any given budget, the optimal
allocation of sampling effort differs somewhat depending upon the param-
eter being considered.
Comparing Grab and Automated Sampling
A comparison of the attainable precision of determination when using
automated samplers as against grab sampling under equal budget, is
presented in Table 16. This table repeats for grab sampling some of the
values from the previous table, presents values derived for automated
sampling, and extends the range of budgets to about a half-mi 11 ion dollars
to demonstrate the convergence of precision which was apparent in the
trends of certain data. The variance components used for the automated
sampler were those derived for grab sampling, as described in the methods
section. Comparisons of grab and automated sampling are made here only
for flow and yield of water and the concentrations of the four chemical
parameters. For transport and yield of chemicals we lacked a model for
combining the variability in flow and concentration but reasoned that
such a model should follow closely the trend of the corresponding mea-
surement of water. For yield of water, concentration of nitrate nitrogen,
and of total phosphorus these results are shown in graphical form in
Figures 7, 8, and 9. From the figures and tables it is clear that
superior precision is obtained by grab sampling at lower budgetary levels.
While the trends of increasing precision with greater budgets suggests a
74
-------
TABLE 15. PRECISION OF MEASUREMENT (AS PROPORTIONAL STANDARD ERROR*) TO BE
EXPECTED WITH GRAB SAMPLING USING OPTIMAL ALLOCATION OF SAMPLING
EFFORT UNDER VARIOUS BUDGET SIZES, SHOWING IN EACH CASE THE
ALLOCATION OF EFFORT AS (NO. OF SITES/TOTAL NO. OF VISITS
PER SITE)
Approximate budget for a one-year study •
$28,000$55,000$108,000$175,000$269,000
FWate°f 0.906(6/26) 0.537(18/26) 0.412(42/26) 0.343(50/39) 0.299(60/52)
Y,i,e!d of 0.694(6/26) 0.445(18/26) 0.340(29/39) 0.289(38/52) 0.254(48/65)
Wo Cc i
Concentration of:
Nitrogen °-304(6/26) 0.168(18/26) 0.110(42/26) 0.084(73/26) 0.069(115/26)
Nitrogen °-115(4/39) 0.079(8/65) 0.064(11/104)0.055(15/130)0.048(20/156)
Phosphorus °-148(6/26) 0.102(18/26) 0.080(29/39) 0.069(38/52) 0.061(40/78)
Chloride 0.104(6/26) 0.069(18/26) 0.054(29/39) 0.047(38/52) 0.042(48/65)
Transport of:
Nitrogen ]-781(6/26) 0.952(18/26) 0.692(42/26) 0.573(50/39) 0.496(60/52)
Nitrogen 0-942(6/26) 0.579(18/26) 0.437(29/39) 0.368(38/52) 0.322(48/65)
Phosphorus °-998(6/26) 0.589(18/26) 0.445(42/26) 0.376(50/39) 0.327(60/52)
Chloride 0.849(6/26) 0.503(18/26) 0.383(29/39) 0.321(50/39) 0.280(60/52)
Yield of:
Nitrfen 1-612(6/26) 0.877(18/26) 0.643(42/26) 0.533(50/39) 0.462(60/52)
Nitrogen °-678(6/26) 0.457(12/39) 0.343(22/52) 0.293(30/65) 0.256(34/91)
Pho^horus °-69°(6/26) 0.458(18/26) 0.350(22/52) 0.298(30.65) 0.260(40/78)
Chloride 0.653(6/26) 0.414(18/26) 0.317(29/39) 0.270(38/52) 0.237(48/65)
* Log-transformed data; proportional standard error is expressed here as
antilog standard error, minus 1. The smaller this quantity, the better
the relative precision.
75
-------
possible convergence for the two types of sampling for some of the param-
eters, there was only one, yield of water per unit area, where an actual
crossing of the lines was observed and this was at a very high budgetary
level (Figure 7). These figures also illustrate the relative contribu-
tion made by each component of variance to the overall estimate of
precision, the variance of the mean.
TABLE 16. PRECISION OF MEASUREMENT (AS PROPORTIONAL STANDARD ERROR*)
ATTAINABLE WITH A RANGE OF BUDGETS, USING OPTIMAL ALLOCATION
OF EFFORT WITH GRAB SAMPLING (GRAB) AND A STANDARDIZED
SCHEDULE OF WEEKLY VISITS TO AUTOMATED SAMPLERS (AUTO.)*
Approximate budget for a one-year study
Flow of
Water
Yield of
Water
Concentrati
Nitrate
Nitrogen
Kjeldahl
Ni trogen
Total
Phosphorus
Chloride
grab
auto.
grab
auto.
on of:
grab
auto.
grab
auto.
grab
auto.
grab
auto.
$28,000
0.906
2.896
0.694
1.755
0.304
0.897
0.115
0.161
0.148
0.313
0.104
0.224
$55,000 $108,000 $175,000 $269,000 $497,000
0.537
1.193
0.445
0.795
0.168
0.447
0.079
0.094
0.102
0.172
0.069
0.125
0.412
0.672
0.340
0.467
0.110
0.274
0.064
0.065
0.080
0.112
0.054
0.081
0.343
0.481
0.289
0.340
0.084
0.203
0.055
0.054
0.069
0.086
0.047
0.062
0.299
0.366
0.254
0.262
0.069
0.159
0.048
0.047
0.061
0.070
0.042
0.050
0.249
0.254
0.212
0.184
0.054
0.113
0.040
0.040
0.052
0.054
0.035
0.038
* Log-transformed data; proportional standard error is expressed here as
antilog standard error, minus 1. The smaller this quantity, the better
the relative precision.
t Assuming an average of 3.3 water samples taken per week for chemical
analysis.
76
-------
0.500.
a
-p
"a 0.200.
o
rH
% 0.100.
S 0.050.
o
0)
o
£J
•? 0.020,
?H
cd
O
-P
O
•H
-P
0.010_
5 0.005.
O
O
0.002
0.001
Water
Yield
site
visit
rror
inter-
action
T
200
I T
2.00
i.oo
0.50
a
Oj
0.30
O
•H
0.15
0.10
I I I I I 1 I I
20 50 100 200 500
Budget in Thousands of 1975 Dollars
Fiqure 7 Water yield; precision attainable at different budgets,
showing total variance and contribution of components
for grab sampling (solid lines) and total variance for
automated sampling (dashed line).
77
-------
w
-p
•H
no
o
o
G
aJ
•H
O
-p
o
•H
-P
3
§
0.0200-
0.0100-
0.0050 _
0.0020-
0.0010-
0.0005-
0.0002
0.0001
Total
Phosphorus
Concentration
0.03
Figure 8.
20 50 100 200
Budget in Thousands of 1975 Dollars
Total phosphorus concentration; precision attainable at
different budgets, showing total variance and contributions
of components for grab sampling (solid lines) and total
variance for automated sampling (dashed line).
78
-------
0.1000.
0.0500-
'£ 0.0200-
o
° 0.0100-
o
03
0.0050.
0.0020.
^ 0.0010-
g
-H
-P
I 0.0005-
o
o
0.0000 -
0.0001.
v Automated Nitrate
\ Total Concentration
\
\
\
\
site
\
\
\
\
\ visit
interaction
LI.00
-0.50
.0.30
0.20 TJ
I
.0.15 I
cd
-0.10 o
-P
O,
_0.05
_0.03
20
i r i i i j i
50 100 200
I i
500
Budget in Thousands of 1975 Dollars
Figure 9. Nitrate nitrogen concentration; precision attainable at
different budgets, showing total variance and contributions
of components for grab sampling (solid lines and total
variance for automated sampling (dashed line).
79
-------
DISCUSSION
The Sampling Frame
The geographic universe for a study of small drainage basins of the
Chowan River system is represented by the enumeration of these basins
in Table 7. By definition, any such study would be limited to these
basins and would not include certain other important sections of the
river basin, for example, small drainages that join larger tributaries
before crossing a road, where these tributaries have already been listed
in their headwaters. Also excluded are any areas draining directly into
such tributaries or into the main stream. Although our definition
included the majority of the small drainage basins (it covered 61 percent
of the entire basin) there is no doubt that we excluded an important
fraction of the total small drainages. If we need to know about these
excluded basins then we must devise some other method pf study and treat
this other part of the river drainage basin as a separate stratum. The
methods of study we have used here restrict attention to those small
basins that can be defined, and then sampled and studied under the
definition.
Effect of Stratification in Time
Stratification in time was used in the design for this study to
provide partial control over the seasonal variability expected in water
measurements, especially in those of flow. If there are seasonal
fluctuations with periods that are relatively long compared to the length
of the time stratum, then stratification can improve the precision of
estimation by eliminating from the variance the average differences among
strata. Such a move does nothing about the short term or random fluctua-
tions which still occur within the time strata. Thus, it would seem
that the shorter the time strata can be made, the more of the long-term
variability that should be eliminated from the error term.
By and large our findings confirmed this dictum (Table 8). The
interesting case of the nitrate nitrogen, however, points out that this
rule, "the shorter the time strata the better," must be modified to
take account of the type of stratified design to be used. With nitrate
the predominant variability was between sites and therefore the best
design would emphasize sampling a larger number of sites and this could
be done only by reducing the number of visits per period. Our design
had fixed 4-week periods and two visits per period (26 per year) were
required as the minimum to provide an estimate of sampling error. Under
this constraint, only 6 sites could be visited at the lowest ($28,000)
budget. In contrast, an unstratified design could provide much better
precision at this lowest budget by visiting 25 sites 5 times a year each.
This result does not show that stratification of itself would fail here
in its usual objective but only that the particular stratified design
used was not suited to the parameter being measured. For this parameter,
80
-------
using periods of greater length would probably increase the efficiency
of the operation whether or not it resulted in any changes in the vari-
ance components.
The empirical exploration of the effect of increasing length of
periods (Table 9) failed to show any appreciable change in the variance
components for flow and nitrate nitrogen concentration, but for Kjeldahl
nitrogen concentration the changes were in the expected direction, that
is, a decrease in the variance among periods with increased length of
period, accompanied by a corresponding increase in the variance among
visits within periods. It was not possible to extend the same type of
examination over longer periods with this set of data; one could expect
that at some point in increasing the length of periods there would be a
stronger effect on the variance components than was found. After all,
inclusion of stratification generally yielded better precision than no
stratification. The effect of changing length of period is tied up with
the structure of the time series for the particular water parameter;
apparently Kjeldahl nitrogen concentration fluctuated on a short enough
cycle that a change of period length from four to ten weeks altered
measurably the distribution of variability between and within periods.
Components of Variance
The components of variance (Table 10) provide insight into the
inherent differences in precision among measurements. In a component
by component comparison, values associated with the measurement of
flow are in the neighborhood of 10 times as great as those for concen-
trations of chemicals. Adjusting flow to account for the area of the
drainage basin reduced the variance component for site by almost half
but did not change the others importantly. The components for either
transport or yield of a chemical, which combine the variability of
flow and concentration (and are dominated by the large values associated
with flow) indicate generally poorer overall precision than even the
measurement of flow.
Considering the anatomy of the variance of an estimated mean value
(Formula 2), for the parameters considered here, the variance components
with the greatest influence on the final value are those for sites and
for visits within periods. This is true not so much because of numerical
value, but because their divisors increase less rapidly with increase in
the budget. Further, the balance between components for sites and for
visits within period helps determine the optimal allocation of sampling
effort between sites and visits. The greater the relative importance of
the component for sites, the greater the number of sites that should be
covered for efficiency. For example, with nitrate nitrogen the component
for sites is over 15 times as great as that for visits within periods and
correspondingly Table 8 shows an emphasis on number of sites for nitrate
at high budgets with the stratified sampling design; this fact also
resulted in the unstratified design being the more efficient at lower
budget levels. For Kjeldahl nitrogen the component for visits exceeded
that for sites. For phosphorus an-d for chloride the two components were of
31
-------
about the same magnitude; the same was true for flow and yield of water
and consequently, for all transport or yield of chemicals.
Costs
The information on costs of all phases of our study is of central
importance to the further analysis of cost-effectiveness. The high cost
of establishing one automated site means that fewer will be set up but
that more samples will be taken from those that are established. The
opposite trend will govern grab sampling where a site is relatively
inexpensive to establish but where the proportional cost of a set of
chemical measurements is higher. The contrast is well illustrated by
the ratio of the cost of a set of chemical measurements to that of
establishing a site. With automated sampling, it costs only 0.4-0.8
percent of the cost of establishing a site to obtain and analyze one
set of measurements, while with grab sampling the figure is 53 percent.
Comparison by the same ratio must be even more disproportionate with the
measurement of flow, could the comparison be clearly defined. Thus it
is clear that use of an automated sampler argues for taking many obser-
vations, once a site is established, but against the establishment of
many sites, while use of grab sampling presents the opposite view.
We have attempted to list all costs, especially those of personnel
and overhead. In trying to use our cost figures, anyone must make
allowance for this fact. There sometimes seems to be a tendency for
an agency to reason that once workers are employed and laboratories
established, then the only costs are for field expenses and supplies.
From this point of view, our costs will seem too high, or perhaps
discourage investigation. In fact, however, many of the costs we list
are already being paid by the agency.
Precision Attainable
The method used here to determine the rate at which proportional
standard error decreases as the budget increases was to consider the
same general sampling design over a range of budgets, each time
assuming optimal allocation of effort within the constraints of the
survey design. Results of such comparisons are shown in Table 15, and
in Figures 7, 8, and 9 for some parameters.
These results apply conceptually to an entire river basin, even
though derived from restricted parts of a basin, under the assumption
that the variance components would not change in moving to the larger
geographic unit. It is possible, of course, that the components for
sites and for interaction of sites and periods in particular, might be
greater if remote parts of a basin were included, but for discussion
here we assume no such change.
A first conclusion is that the determinations of water quality
parameters were of disappointingly poor precision. Exactly what con-
stitutes useful precision is a matter of subjective, not statistical,
82
-------
judgement. Workers in some fields may feel uncomfortable about propor-
tional standard errors greater than 0.05. For ecological studies I have
sometimes offered the purely subjective personal judgement that a pro-
portional standard error of 0.15 or less is probably useful
administratively, though perhaps not in research. Even this permissive
criterion is not met here by flow or yield of water, or by any determina-
tions of transport or yield of chemicals, even at the highest budgets
considered. Therefore, we may conclude, first, that while concentrations
may be measurable with precision within usually accepted standards, it is
likely that flow and the related measures of yield and transport may not
be so measurable. Second, however, perhaps we must learn to use informa-
tion of such poor precision, because probably it is the best we are going
to have.
The relatively poor precision of these estimates arises from the
inherent variability of the data; it is not a result of using probability
sampling. Sampling by any other method will be subject to similar errors,
although the imprecision may not be apprehended because it is not quanti-
fied. In contrast, when using probability sampling we can estimate the
precision. Of course, it is quite possible that certain special plans of
judgement sampling could be subject to less variability as, for example,
if only moderate or high flows were to be measured, but then the universe
under study could not be defined exactly.
A second conclusion is that precision increases more slowly with
increased budget than should be expected, for example, with simple random
sampling. With the latter, one may expect the proportional standard
error to decrease as the reciprocal square root of the sampling effort.
The budget is a good measure of sampling effort if adjusted for the
constant (overhead) cost. In Table 15 changing from an available budget
of $94,000 ($108,000 minus $14,000) to one of $225,000 ($269,000 minus
$14,000) should reduce standard error (which must be calculated here as
log [proportional standard error plus 1]) to about 61 percent of its
value at the lower budget if the square root rule applied (and to 78
percent with a fourth root rule). Actually, all but one of the values
at the higher budget ranged between 75 and 78 percent of those at the
lower budget. The single exception and lowest value was for nitrate
nitrogen at 64 percent, but this was a special case.
The reasons for departure from the square root rule are apparent in
examining the basic relationship (Formula 2) and Figures 7, 8, and 9.
The components of variance for sites and visits are most important in
determining the precision, and these are divided by either number of
sites or number of visits, but not by both. Any increase in budget must
be spread over increases in both number of sites and number of visits;
therefore neither can be increased in direct proportion to the budget
(except in the special case so well represented by nitrate nitrogen in
this study). The divisor for the variance component for error, in
contrast, includes the product of number of sites and number of visits,
and thus error follows the square root rule. But it is of little
relative importance in determining the proportional standard error
83
-------
because its divisor is so large and its value decreases so rapidly with
increased sampling.
The contributions made by the different components of variance are
illustrated in Figures 7, 8, and 9. Each figure shows the value for the
total variance (and proportional standard error) for both automated and
grab samples over the range of budgets considered. The curve for auto-
mated sampling would be a straight line were it not for the effect of
the overhead costs; in each figure its general slope approximately illus-
trates the square root rule because the number of sites is proportional
to the available budget. With yield of water (Figure 7) and total
phosphorus concentration (Figure 8) the line representing total variance
of the estimate declines at a lesser rate reflecting the generally
parallel course of contributions from sites and visits. The contribu-
tion from error declines at the rate expected under the square root rule.
The interaction also seems to decline at this rate for the brief range
of budgets for which it is visible on these graphs, but this appearance
results from the fact that in all three graphs the number of sites
increases in direct proportion to the available budget in this portion
of the curve. This happens because at these lower budgets the number
of visits (fixed here as the minimum number required to calculate a
variance) already exceeds the calculated optimal value, therefore is not
increased, allowing sites to increase in direct proportion to the budget.
This last effect also governs the special case of nitrate nitrogen
(Figure 9). Here the requirement of 2 visits per period so far exceeds
the optimal value that the contribution from the component for visits
remains constant and relatively unimportant throughout the entire range
of budgets considered. In consequence, the contribution from the compo-
nent for sites declines throughout this range at the rate expected by
the square root rule, and the total variance declines at a similar but
decreasing rate.
This information on attainable precision and optimal allocation of
effort (Table 15) is more important for guidance in future studies than
as a postmortem of the one we have completed. The figures for optimal
allocation at lower budget levels are suspect; the requirement of 2
visits per period (imposed so that a variance could be calculated) had
the result that with the minimum number of 26 visits per year, only a
few sites could be covered before exhausting the budget. At these
lower budgets, the use of longer periods, fewer visits per site and
more sites would have been better for all parameters except Kjeldahl
nitrogen. The figures at the higher budget levels, however, better
illustrate the differences among parameters with reference to optimal
division of sampling effort among sites and visits. The greatest
difference is between nitrate nitrogen with its high demand for number
of sites and Kjeldahl nitrogen with its demand for number of visits.
84
-------
Planning Other Studies
In designing a survey of another river basin, one would not have
the information we have in retrospect with the Chowan. Further, even
if one does have such information, the chances are that the optimal
allocation of sampling effort for one variable may be inefficient for
another (e.g., as here with nitrate nitrogen and Kjeldahl nitrogen).
This is a common problem with surveys involving numerous variables,
and compromise is required because almost never may a survey be based
on a single variable.
One may design the survey to best suit the most important variable
and let the rest be determined as may happen. Or, one may select some
median status of variables and design a survey for this. In the study
of water quality of a river basin of the type discussed here, the
difficulty of obtaining any reasonable precision in the measurement of
either flow or yield of water and the closely related values of transport
and yield of the chemical constituents may urge that primary attention
be given to the measurement of flow. It also happens that in this study
flow and water yield occupy median positions as to optimal distribution
of sampling effort. Determination of flow calls for a number of sites
somewhat greater than the number of visits per year, but when the flow
of each small basin is adjusted for its area, the balance shifts and
the number of sites required is somewhat less than the number of visits.
As a rule of thumb, then, we propose that the number of sites be
made approximately equal to the number of visits. If some variable
where information is needed has unusually -high spatial variability, then
the balance may favor sites, and vice versa. There is no theoretical
basis for this suggestion beyond the observation in this study that
approximate equality of variance components holds a median position among
the parameters investigated. The suggestion may well be discarded after
further work, but for now it is suggested as a conservative plan.
Ordinarily, the best known dimension of a pending study is the
budget. Rewriting formula (4), and setting s (number of sites) equal to
pv (number of visits per year), the number of sites (or annual visits)
may be calculated as:
(-cs + (C2 + 4cvC)1'2)/2cv
where C is the available budget. Substituting the values for costs
observed here for grab sampling, this becomes:
-0.964 + (0.929 + 0.0126C)
This suggests about 13 sites visited 12 times per year for an available
budget within $15,000 in 1975. This would break conveniently into 6
85
-------
periods of about 2 months each, each with 2 visits per site. For an
available budget of $100,000 in 1975, 35 sites should be used, with 34
visits to each; this would provide 17 time strata of about 21 days each.
Grab Sampling Compared with Automated Sampling
The basis of comparing grab sampling with automated was the attain-
able precision (expressed as proportional standard error) of estimates
produced by each method under equal budget. Grab sampling was assumed
to allocate the sampling effort in an optimal manner but within our
sampling design, this constraint resulted in less than maximum efficiency
at low budgets. Automated sampling was assumed to be on a fixed schedule
of weekly service visits at which an average of 3.3 water samples would
be retrieved; these samples were considered to be random in time. This
fixed schedule also resulted in less than possible efficiency; in
principle the schedule for an automated sampler might be changed within
limits set by the design and capacity of the sampler and by sample
storage problems. Thus the present comparison is strongly dependent
upon these assumptions.
The comparison was made for the range of budgets previously considered,
extended here to demonstrate that precision could actually be equal for
the two methods if the budget were high enough. Otherwise, grab sampling
resulted in greater precision than automated sampling for all cases
examined, as may be noted in Table 16 and in greater detail in Figures
7, 8, and 9 for three of the parameters. The single exception was
observed with measurement of yield of water where automated sampling was
equally cost-effective at a very high budget and more efficient at even
higher budgets. In all other cases, and especially at modest budgets,
grab sampling was clearly more cost-effective.
This claim for superiority of grab sampling must, however, be clearly
understood to be made on the limited basis of comparison of precision of
determination of an average value under an equal budget, in a study
limited to a single year. No weight is allowed for the other functions
that an automated sampler can accomplish. For example, an automated
sampler provides a continuous record of flow and allows study of the
characteristics of different portions of the hydrograph during a runoff
episode, a capability lacking in the instantaneous measurement provided
by the grab sampling. Neither was any allowance made for salvage value
of instrumentation or continued use of an automated site after the con-
clusion of a one-year study. But for the special comparison made here,
grab sampling appears to provide generally better precision at equal
cost.
The superiority of grab sampling in this comparison arises from its
greater flexibility in covering numerous sites, and thus reducing the
numerical importance of the variance component for sites as a contribu-
tion to the standard error. That is, it is easier to maximize information
under a fixed budget by adjusting the balance between number of sites
covered and annual visits per site.
36
-------
Automated sampling, in contrast, is very effective in reducing
variance due to temporal fluctuation; in fact, it entirely eliminates
that component in the continuous recording of flow. But automated
sampling costs so much per site added that it is impossible to reduce
the effect of the spatial component very much under modest budgets and
optimal allocation. The economics of automated sampling are such that
once the instrument is established at a site, it can provide at rela-
tively low cost numerous samples for chemical determinations, and
measure flow continuously. These characteristics make the apparatus
ideal for studying what happens at a single site. But the high cost of
covering more sites prohibits any large spatial sample.
In fact, the ease with which more samples can be taken for chemical
analysis may be misleading. Each of these samples must be analyzed
chemically, and this adds to the cost of the study. A trial cost analy-
sis on the basis of including- the weekly service trips in the per-site
costs and adjusting the number of water samples to be analyzed, with
the option of more sites sampled, suggested that only with Kjeldahl
nitrogen would anything approaching our average of 13 water samples per
4-week period be justified on the basis of precision; optimal strategy
usually called for fewer analyses and more automated sampling sites.
This last comparison is not really appropriate in our study because the
stage-activated sampling schedule we used was designed for other pur-
poses. But had the automated sampling schedule been designed primarily
for better overall precision, the cost-effectiveness could have been
improved somewhat, though probably not enough to match the grab sampling
efficiency.
87
-------
SECTION 5
ASSOCIATED WATER QUALITY INTERPRETATIONS
GRAB SAMPLING DATA AND ANALYSES
The field data necessary to evaluate the feasibility of statistical sam-
pling have been analyzed to take advantage of this extensive information base
in order to better understand the characteristics of rural runoff and area-
wide water quality. These analyses are summarized in sections that present
(1) a data summary for each of the 15 statistical survey sites that include
relationships between water quality and land use, (2) comparisons of unchan-
nelized and channelized Coastal Plain stream water quality, (3) analysis of
seasonal fluctuations, (4) concentration versus water yield, (5) comparison
among geoclimatic areas,(6) a stream reach substudy, (7) comparison of study
basin versus mainstream water quality, and (8) an overall basin data summary.
Two years of data (November 1974 to November 1976) were collected at
the forested Piedmont (F); poorly-drained Coastal Plain (P) and well-drained
Coastal Plain (W) sites; and 18 months (June 1975 to November 1976) of data
were collected at the agricultural Piedmont (A) sites. The 18 months of
compatible data from all sites were employed for most analyses so comparisons
could be based upon a uniform time interval. The full two-year data set was
employed, however, for analyses of seasonal variations. The data analyses
fit into two broad classes which either relate site values to land-use ac-
tivity or investigate differences in regional water quality.
Arithmetic concentration averages provide information about in-stream
conditions. Volume average concentrations (flow weighted averages), how-
ever, provide information about the net water quality during a time interval
and thus are meaningful to downstream receivers. Therefore, several sec-
tions provide dual analyses so that the data can be reviewed with respect
to these two concepts.
SU8BASIN DATA SUMMARY
Prediction of present and future rural water quality is often based
upon models which relate water quality to macro land-use factors. These
models are probably best supported by a national water-quality land-use
study of 928 sites reported by Omernik (1977). That study found increased
nitrogen and phosphorus concentrations with increased agricultural intensity.
Water-quality land-use relationships for the 15 statistical survey sites were
investigated to determine if similar relationships were evident in this
88
-------
watershed. These analyses Investigate both in-stream conditions and exported
water quality as a function of percent forested area.
The land-use summary (Table 3) for these subbasins showed that the for-
ested area, plus agricultural area, was approximately equal to the total sub-
basin area at each site. Thus, in these subbasins decreased forested area
indicated increased human activity which was primarily agriculture. There-
fore, major trends which are observed for water-quality constituents with
respect to percent forested area would also exist (only inversely) with re-
spect to percent agriculture.
Site water quality versus land-use relationships for COD, TP, TKN, and
NOs-N concentrations during June 1975 to November 1976 are presented in Figure
10. For each of the four parameters, the grab sampling mean, mean ± one
standard deviation, maximum, and minimum values at each site are plotted as
a function of percent forested area. From a water-quality perspective, the
COD, TP, and TKN graphs indicate that no meaningful mean concentration increase
occurred as the percent forested area decreased. Indeed, even the range of
values for these parameters was relatively constant for all sites. The mean
NOa-N concentration also does not display a concentration increase with de-
creased forested area, but the range of values varied considerably from site
to site which indicates that the sites were probably differentially impacted
by the general land-use activities considered. The major conclusion obtained
by examination of these graphs was that the mean time average concentration
was relatively uniform throughout the watershed and did not increase as agri-
cultural activity increased.
The effects of land use on receiver system water quality was assessed
by analyzing flow weighted concentrations of COD, TP, TKN, and N03-N versus
percent forested area for the 15 statistical survey sites during June 1975-
November 1976. These data are displayed in Figure 11. The flow weighted
concentrations do not present any clear relationship between water quality
and percent forested area so regression models were not attempted.
It is both reasonable and often reported that nutrient levels in streams
increase as agricultural intensity increases, but often a wide range of
confidence limits exist for regression models developed for large geographic
areas. For example, one regression model (Omernik, 1977) relating mean total
phosphorus concentration to percent agriculture plus urban area for the
eastern U.S. had broad confidence limits as measured by a ratio for the ±
one sigma range to predicted mean value of 130 percent. The Chowan concen-
tration versus land-use graphs point out the need for caution when employing
model predictions to specific cases. Regression models employing macro land-
use factors do not account for varying agricultural cropping patterns and
management practices, annual weather conditions, stream border buffer systems
or other factors which can impact agricultural effects on water quality.
While direct water-quality land-use relationships for the Chowan data
were weak, differences related to geoclimatic area, season, and size were
observed. These differences are the subject of the following sections.
89
-------
CHANNELIZATION EFFECTS
The influence of subbasin soil type on stream parameter levels may have
been confounded in the Coastal Plain region of this study by the impact of
channelization. The Coastal Plain study areas were chosen on the basis of
well- and poorly-drained soils. However, parts of both the well- and poorly-
drained Coastal Plain have been channelized to expedite surface and sub-
surface drainage for greater agricultural flexibility and productivity. To
illuminate possible channelization effects, the Coastal Plain sites were re-
classified into unchannelized and channelized streams with four sites in each
class. It should be noted that this study and that by Kuenzler et al. (1977)
are not before and after channelization evaluations at given streams, so
differences attributed to channelization may be confounded with or due to
other factors.
A summary of the chemical and physical properties of the sampled un-
channelized and channelized streams is presented in Table 17. An F-test was
performed to determine whether the mean values for the two classes differed
significantly considering the variation that was observed for the mean of
each site within the two classes. The analysis showed that there was no
evidence (P>0.10) of a difference in the mean TKN, NH3-N and SS concentrations
or the mean temperature and pH values between the channelized and unchannel-
ized stream classes. However, because significant (P<0.10) differences were
noted for the remaining parameters, the unchannelized stream values were
employed as the standard in discussing possible effects of channelization.
• FP Sill Ural
• AP Sill M.on
A PDCP Siti UIIK
O WDCP Sill Mian
O
O
O
IOO:
9O
eo
TO
60
90
40
30
20
10
0
<-»Mioi>: FSTO
' i
> 1
|
|.
* 1
1
t 1
1
t
— Mo., Min
•|i
•
• FP Sill Mrai
• AP Sill Mion
a PDCP Sill Ul«l
0 WDCP Sill Mion
«-» M»n ! I STO
<—'Mom, Min
I.O
07!
05
029
i
IT
!:
]
,1;
1 I
i
i
,
1
^
i
i
I
i
<
SO 40 90 CO 70
20
1.9
10
0.9
».0| 2.9,
1
4.4 1
2.2
1 T
:
i
J
i
,4
•
4:
:J ,« .k.
JO 40 90 SO 70 80 »0 10
FORESTED AREA (%)
FORESTED AREA (%)
Figure 10. Site arithmetic data summary versus land use for grab sampling
(June 1975 to November 1976).
90
-------
The data showed that the average velocity and DO concentration were
about 12 percent greater in the channelized streams than in the unchannel-
ized streams. The channelized streams also had higher average TP (40 per-
cent) and N03-N (380 percent) concentrations, but the average TOC and COD
concentrations were about 15 percent lower. In a similar study, Kuenzler
et al. (1977) also found that the mean annual nitrate concentrations for
unchannelized streams (0.06-0.11 mg/1) was much less than for channelized
streams (0.4-1.7 mg/1) for several rural streams in the North Carolina
Coastal Plain. Thus, the influence of stream channelization appears to be
the loss of natural swampy buffer strips in which plants can extract both
nitrate and phosphate and where nitrate can also be lost through denitri-
fication in the saturated soil layers. However, the swampy buffer strips
results in increased organic inputs as measured by TOC and COD. The in-
creased DO levels in the channelized streams probably reflect both the in-
creased aeration association with increased water velocity and the lower oxy-
gen demand of the less swampy type inputs.
SEASONAL FLUCTUATIONS
Earlier studies (Stanley and Hobbie, 1976) found that seasonal timing of
nutrient concentrations and loads was important to the algal growth in the
lower Chowan River so the rural runoff data were analyzed for seasonal
trends. The seasonal fluctuations of water yields (WYD) and flow weighted
concentrations from the four geoclimatic areas are presented in Figures 12
o
8
• FP SITE MEAN
• AP SITE MEAN
& PDCP SITE MEAN
O WDCP SITE MEAN
• FP SITE MEAN
• AP SITE MEAN
a PDCP SITE MEAN
O WDCP SITE MEAN
05
,-. 04
f 03
t -
O.I
e
o
00
0
FORESTED AREA (%)
FORESTED AREA (%)
Figure 11 Site flow weighted concentrations versus land use for grab
sampling (June 1975 to November 1976).
91
-------
TABLE 17. CHANNELIZATION EFFECTS DATA SUMMARY, DECEMBER 1974-
DECEMBER 1976
Variable Units
Mean
Minimum
value
Maximum
value
C.V.
%
4 Unchannelized Streams
COD
TOC
TP
NH3
N03-N
TKN
Cl
SS
DO
VELOCITY
PH
TEMP
COD
TOC
TP
NH3
N03-N
TKN
Cl
SS
DO
VELOCITY
PH
TEMP
i
•
mg/1
i
>
m/sec
°C
•
mg/1
1
m/sec
°C
116
115
185
48
186
186
185
71
181
192
123
199
31.0
13.4
0.13
0.12
0.24
1.30
7.54
10.5
7.37
0.10
5.91
13.8
4.0
1.0
0.02
<0.01
<0.01
<0.04
1.50
0.5
0.20
0.0
4.9
0.1
71.0
36.0
1.65
1.30
1.46
11.1
51.4
167.0
14.2
0.57
7.6
26.0
47
54
109
223
121
83
57
193
41
114
13
48
4 Channelized Streams
121
122
197
52
198
198
198
77
183
191
121
198
25.5
11.3
0.18
0.14
0.92
1.20
<4.0
<1 .0
0.02
<0.01
<0.01
<0.04
9.40
13.3
8.25
0.16
6.02
14.5
1.50
0.56
1.70
0.0
4.80
2.0
91.0
39.0
1.36
1.15
8.96
5.25
20.60
75.0
14.4
0.75
7.20
26.0
61
63
106
198
142
64
33
103
30
83
8
43
92
-------
and 13. Water yields tend to be higher in winter and spring, but relation-
ships between area constituent concentrations in different geoclimatic areas
and season were less well defined.
The concentration data do indicate two interesting relationships. The
-N concentration graph shows that the Piedmont values were less than the
Coastal Plain values for all recorded seasons and the TP concentrations in
the agricultural Piedmont were comparable to the forested Piedmont values
except for the summer of 1975.
Next, both land use and seasonal effects on water and nutrient yields
from the 15 statistical survey watersheds were assessed by a model of the
form:
Y = Ye
o
-BF
(1)
where Y is the yield; Ys is a seasonal constant; B is an attenuation coeffi-
cient; and F is the percent forested area of the subbasin. A year was com-
posed of four, 3-month seasons representing winter, spring, summer and fall,
with winter corresponding to December, January, and February. The results of
fitting WYD, TKN, N03-N and TP transport data (15 site means x 8 seasons)
10
E
Q
_)
UJ
tr
UJ
• FP AREA MEAN
• AP AREA MEAN
A PDCP AREA MEAN
O WOCP AREA MEAN
W74-75 S75 S75 F75 W75-76 S76
SEASON
S76
F76
Figure 12. Areal s-easonal water yield for grab sampling (November 1974
to November 1976).
93
-------
to this model are summarized in Table 18. Analyses showed that the models
were not appreciably improved by letting B vary with season. This analysis
indicated that water yield was about 430 percent greater during the winter
and spring seasons than during the summer and fall seasons. The nutrient
yield models demonstrated relationships similar to the water yield model.
Finally, the land use and seasonal effects on flow weighted concentra-
tion were assessed by a model of the form:
C = Cge
-BF
(2)
where C is the concentration and Cs is a seasonal constant. The results of
fitting TKN, NOs-N and TP concentrations (15 site means x 8 seasons) to this
model are summarized in Table 19. In general, the models displayed signifi-
relationships but had lower r2 values. With this model, TKN and TP concen-
tration variations were attributable to seasonal but not land-use effects,
while NOs-N concentrations were a function of both season and land use.
Again, the models were not appreciably improved by allowing B to vary with
season.
o
o
100
eo
60
«o
20
0
• FP AREA MEAN
• AP AREA MEAN
6 POCP AREA MEAN
0 WDCP AREA MEAN
6
0 0 0
a o • ago
S . • « r
• a • •
i i i i i i i
4
o
a
• FP AREA MEAN
• AP AKEA MEAN
a POCP AREA MEAN
O WDCP AREA MEAN
I SO-
1.2*5 •
100-
07S-
OSO
02S
OOO
9 JS-
*~ 20 •
O
°^
WT4-73 STS S7S FT3 WT5-7S STS S7« FTS
STS
F7S W7S-T6 S7« S7t F 7*
SEASON
SEASON
Figure 13. Aereal seasonal flow weighted concentrations for grab
sampling (November 1974 to November 1976).
94
-------
CONCENTRATION VERSUS WATER YIELD
Models of nonpoint source constituent transport typically employ a
hydrologic model for the main program and assume some empirical relationship
between flow rate and constitute concentration to calculate constituent
loads. The rural runoff nutrient data were employed to investigate possible
relationships between concentration levels and water yield values to assist
in model development for the Chowan watershed.
TABLE 18. CONSTITUENT YIELD MODELS INCORPORATING LAND USE AND SEASON
Variable
Season
Land Use Season
WYD
(mm/day)
TKN
(kg/km2 -DAY)
N03-N
(kg/km2 -DAY)
TP
(kg/km2-DAY)
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
1.20±0.14
0.87+0.58
0.23+0.18
0.16±0.05
97±9
57±34
29±24
8±1
236±56
190±151
35±23
12+6
11+3
16±14
3±2
2±1
0.0235 0.52
0.0235
0.0235
0.0235
0.0248 0.53
0.0248
0.0248
0.0248
0.0667 0.51
0.0667
0.0667
0.0667
0.0288 0.48
0.0288
0.0288
0.0288
P<0.01** P<0.01**
P<0.01** P<0.01**
P<0.01** P<0.01**
P<0.01** P<0.01**
RF
Model: Y = Yse"Dr
**Highly Significant
95
-------
TABLE 19. CONSTITUENT FLOW WEIGHTED AVERAGE CONCENTRATION MODELS INCORPOR-
ATING LAND USE AND SEASON
Variable
Season
B
Land use
Season
TKN
(mg/1)
N03-N
(mg/1)
TP
(mg/1)
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
Winter
Spring
Summer
Fall
1.20±0.02
1.05±0.12
1.65±0.26
0.78±0.15
2.88±0.36
2.72±0.78
2.76±0.64
1.08±0.28
0.135±0.017
0.208±0.096
0.194±0.010
0.196±0.036
0.0013
0.0013
0.0013
0.0013
0.0432
0.0432
0.0432
0.0432
0.0052
0.0052
0.0052
0.0052
0.44 P>0.10 P<0.01**
0.30 P<0.01** P<0.05*
0.20 P>0.10 P<0.01**
Model: C = Cse"BF
*Significant
**Highly Significant
For the purpose of this discussion, water yield (mm/day) refers to indi-
vidual flow measurements divided by the appropriate subbasin area. These
values need to be distinguished from annual average water yield (mm/yr) values
which are computed from several water yield (mm/day) values. Water yield
values were employed rather than flow rates to allow direct comparisons among
sites.
Plots of COD, NOa-N and TP concentrations versus water yield are gener-
ated for each site. Because a review of these graphs indicated little or no
consistent relationship between the parametic concentrations and the water
yield, no empirical relationships with general application to all sites were
generated. However, reviewing the.data did produce some qualitative relation-
ships which are helpful for related 208 type activities.
The NOs-N data showed that the five highest concentration measurements,
which were all greater than 5 mg/1, were all obtained in the well-drained
Coastal Plain at the W-4 subbasin. The N03-N graph in Figure 10 shows the
high level of these values relative to those at other sites. W-4 was the
smallest watershed with an area of only 0.5 km2 (0.2 mi2) and the next
larger watershed was almost 10 times as large. The W-4 stream channel was
well defined and extended into the cropped fields of the watershed. The
NOa-N concentration at W-4 increased with decreasing water yield as shown in
Figure 14. Four of the elevated N03-N measurements were made at low water
yield conditions during the period of June to September 1976, and the fifth
96
-------
measurement during a low water yield condition in August 1975. Animal pro-
duction had not been observed during site inspections; and data also indicated
there was no relationship between either COD, TP, or Cl concentration and flow
at this site, so the increased N03-N concentrations were not attributable to
animal waste.
In plot studies of nitrogen losses from row crops in well-drained soils
of the North Carolina Coastal Plain, Gambrell and Fisher (1966) measured near-
ly constant nitrate concentrations of about 15 mg/1 from subsurface drainage.
Because the W-4 watershed was approximately 50 percent agriculture and 50
percent forested and the nitrate nitrogen concentrations during low flow con-
ditions were 6-7 mg/1, it appears that these high values resulted from the
subsurface drainage of the agricultural land. Although the other three WDCP
subbasins had similar land use, these subbasins were 15 to 30 times larger
than W-4, ar,d the maximum N03-N concentrations at these sites were all less
than 2.5 mg/1. It was hypothesized that in-stream processes were responsible
for the reduction in N03-N concentration at these sites.
In summary, measured concentrations did not.display any consistent func-
tional relationship to water yield levels. However, data from Gambrell and
Fisher (1966) and this study showed that N03-N concentrations at small
(<0.5 km2) well-drained Coastal Plain sites were elevated during low flow
conditions during the summer; however, elevated NQ3-N concentrations were
not observed at sites draining 10 km2 (2.6 mi2) watersheds.
taor-
a>
— SO
•Ti «• •• • •
• *•• •
10-
10'1
10
10 '
WATER YIELD (mm/day)
Figure 14. NOs-N versus water yield at a small (0.5 km2) well-drained
Coastal Plain site for grab sampling (November 1974 to
November 1976).
97
-------
COMPARISONS AMONG AREAS
The purpose of this section is to investigate possible differences be-
tween geoclimatic areas based upon site,average values. Analyses are pre-
sented for both in-stream (arithmetic average) and export (flow weighted
average) summaries.
Arithmetic Average Concentrations
Stream conditions for the 15 statistical survey streams are summarized
by geoclimatic areas in Table 20. The 15 statistical survey site mean
values were employed to evaluate possible physical and chemical differences
between streams in the Piedmont and Coastal Plain, agricultural Piedmont and
forested Piedmont, poorly-drained Coastal Plain and well-drained Coastal Plain.
These comparisons were performed by an analysis of variance using log-
transformed data, and the results are summarized in Table 21.
The analysis of variance indicated a significant (P<0.10) difference
between the well-drained Coastal Plain and poorly-drained Coastal Plain water
yield (WYD) values. The WYD data indicated that the mean WYD from the W
sites was about five times as great than from the P sites and that the WYD
variation within the W sites was relatively large. The elevated water yield
value of the W sites was primarily due to the measurement of high.water
yields at two of the four sites. At one of these sites, a relatively in-
frequent, very high flow event resulted in the elevated mean value. At the
second site, the high mean value was attributable to numerous moderate water
yield measurements which may reflect groundwater input from areas outside of
the basin surface drainage area. Due to the proximity of the W and P sites, it
was reasonable to assume that rainfall amounts were similar during the 18-month
study period and that the difference in mean water yields was a data set arti-
fact. However, because the water yield difference was 400 percent, it had a
major impact on nutrient yield comparisons among the four study areas.
The analysis of variance also showed significant (P<0.10) differences
existed among geoclimatic areas for some of the other physical and chemical
parameters. The mean water velocity was approximately 100 percent greater in
the Piedmont than in the Coastal Plain. Although temperature differences were
established between both the Piedmont versus Coastal Plain and the poorly-
drained Coastal Plain versus well-drained Coastal Plain areas, they were not
considered meaningful on a general water quality basis because the variations
were less than 5 percent. Thus, the 25 percent greater dissolved oxygen level
in the Piedmont compared to the Coastal Plain probably reflected increased
aeration.associated with high velocities in the Piedmont streams. The lower
conductivity (COND) level (28 percent) and chloride concentration (45 percent)
in the Piedmont compared to the Coastal Plain probably were due to decreased
atmospheric inputs with increasing distance from the ocean as recorded in pre-
cipitation studies by Gambell and Fisher (1966).
The average in-stream N03-N concentration of the Piedmont streams (0.07
mg/1) was substantially less than the Coastal Plain value (0.64 mg/1). How-
ever, the NOa-N analysis of variance model had a smaller r (0.25) because the
98
-------
TABLE 20. ARITHMETIC DATA SUMMARY FOR GRAB SAMPLING. JUNE 1975-NOVEMBER 1976
Variable
WYD
VELOCITY
TEMP
DO
PH
COND
Cl
COD
TOC
TP
N03-N
TKN
ss
WYD
VELOCITY
TEMP
DO
pH
COND
Cl
COD
TOC
TP
NOj-N
TKN
SS
Mean of
Units sites
Forested
mm/ day
m/sec
PC
mg/1
ymhos/cm
mg/1
Agricultural
mm/day
m/sec
°C
mg/1
ymhos/cm
mg/1
Piedmont
1.28
0.20
14.5
8.56
6.14
49.7
5.46
23.5
8.78
0.12
0.04
1.12
8.87
Minimum
site
average
- 4 Sites
0.85
0.17
14.2
8.40
5.99
45.4
4.23
17.1
6.74
0.10
0.02
1.02
6.15
Maximum
site
average
1.5
0.29
14.7
8.81
6.22
56.1
6.93
29.8
10.5
0.14
0.06
1.19
12.1
C.V.
.«
23
30
1
2
2
10
21
23
18
17
50
6
21
Piedmont - 3 Sites
1.58
0.17
14.9
9.08
6.27
51.2
4.20
17.8
6.98
0.10
0.11
1.00
6.82
1.32
0.13
14.5
8.93
6.17
42.2
3.51
13.9
6.05
0.10
0.08
0.92
4.26
1.75
0.20
15.7
9.21
6.37
68.0
5.08
20.7
8.32
0.11
0.17
1.13
11.0
15
24
4
2
2
29
19
20
17
4
45
11
54
Variable
WYD
VELOCITY
TEMP
DO
pH
COND
Cl
COD
TOC
TP
N03-N
TKN
SS
WYD
VELOCITY
TEMP
DO
PH
COND
Cl
COD
TOC
TP
NOi-N
TKN
SS
Units
We 11 -Drained
mm/day
m/sec
°C
mg/1
ymhos/cm
mg/1
Poorly-Drained
mm/day
m/sec
°C
mg/1
ymhos/cm
mg/1
Mean of
sites
Coastal
2.18
0.07
14.9
7.00
5.93
60.1
8.25
26.1
10.5
0.12
0.75
1.10
8.7
Coastal
0.43
0.11
15.6
7.13
6.01
80.4
9.37
26.6
10.7
0.22
0.53
1.18
14.4
Minimum
site
average
Plain - 4
0.48
0.05
14.5
5.87
5.61
54.5
6.36
16.9
7.3
0.07
0.06
0.94
3.5
Plain - 4
0.30
0.06
14.8
5.56
5.82
67.8
7.19
16.8
7.6
0.13
0.03
0.84
6.9
Maximum
Site
average
Sites
3.77
0.09
15.3
7.87
6.08
63.0
9.88
32.6
13.3
0.14
2.20
1.28
18.4
Sites
0.52
0.13
16.2
8.38
6.07
89.5
11.36
40.5
18.3
0.41
1.17
1.33
28.0
C.V.
it
84
29
2
13
4
6
18
28
27
25
131
15
75
23
27
4
19
2
14
18
37
46
59
108
19
67
-------
Coastal Plain streams had a large inherent variation in mean concentrations
(0.03 to 2.20 mg/1). The mean TP concentration in the poorly-drained Coastal
Plain was about twice the value of 0.12 mg/1 recorded for the well-drained
Coastal Plain. The COD, TOC, TKN, and SS concentrations did not show any
significant (P<0.10) variation among the four geoclimatic areas.
Flow Weighted Concentrations
Flow weighted concentrations for the statistical survey sites are summa-
rized in Table 22. This table presents a data summary for water exported
from the four geoclimatic areas as determined from the 18-month volume aver-
age concentration from each site within an area. These flow weighted con-
centrations for the 15 statistical survey sites were employed to evaluate
TABLE 21. ANALYSIS OF AVERAGE STREAM CONDITIONS FOR GRAB SAMPLING
JUNE 1975-NQVEMBER 1976
Probability of Significant
Variable
WYD
VELOCITY
TEMP
DO
PH
COND
Cl
COD
TOC
TP
N03-N
TKN
SS
Model
r2
0.37
0.70
0.49
0.59
0.54
0.71
0.75
0.23
0.24
0.40
0.25
0.17
0.19
Piedmont vs.
Coastal Plain
P>0.10
P<0.01**
0.050.10
P>0.10
P>0.10
0.050.10
P>0.10
AP vs. FP
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
Difference
PDCP vs. WDCP
0.010.10
0.050.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
0.050.10
P>0.10
P>0.10
*Significant Difference
**Highly Significant Difference
100
-------
TABLE 22. FLOW WEIGHTED CONCENTRATION SUMMARY FOR GRAB SAMPLING DATA - JUNE 1975 TO
NOVEMBER 1976
Mean of
Variable sites
Minimum
site
value
Forested Piedmont -
COD
TOC
TP
N03-r
TKN
Cl
COD
TOC
TP
N03-r
TKN
Cl
30.1
11.9
0.072
-------
possible differences between geoclimatic areas. The comparisons were per-
formed by an analysis of variance and the results are summarized in Table 23.
No differences (P<0.10) were observed for volume average COD, TOC, and TKN
concentrations among the study areas. For the 15 statistical survey sites,
the average flow weighted COD, TOC, and TKN concentrations (30, 13, and 1.3
mg/1, respectively) were the same magnitude as the arithmetic average stream
values (24, 13 and 1.1 mg/1, respectively) indicating that concentrations
were relatively uncorrelated with respect to flow.
Significant (P<0.10) differences were noted between Piedmont and Coastal
Plain NQa-N and Cl concentrations. The flow weighted average NOa-N concen-
trations were much lower in the Piedmont (0.08 mg/1) than in the Coastal Plain
(0.45 mg/1). Likewise, the Cl concentrations were about 55 percent lower in
the Piedmont than in the Coastal Plain. Finally, the analysis of variance in-
dicated that there was a difference between the agricultural Piedmont TP con-
centration value of 0.18 mg/1 and the forested Piedmont value of 0.07 mg/1.
In summary, analyses of variance employing 18-month arithmetic average
and flow weighted average concentrations for the 15 statistical survey sites
in four geoclimatic areas indicated that the dominant variation was between
the Piedmont-and Coastal Plain with relatively minor variation occurring be-
tween (1) the agricultural Piedmont and forested Piedmont and (2) the poorly-
drained and well-drained Coastal Plain. The water quality differences between
the Piedmont and Coastal Plain were judged to be primarily the result of
naturally occurring physiographic variations in (1) basin characteristics,
such as vegetation, soil type, stream hydraulics; and (2) ocean proximity.
TABLE 23. ANALYSES OF FLOW WEIGHTED AVERAGE CONCENTRATIONS FOR GRAB SAMPLING
DATAT JUNE 1975-NOVEMBER 1976
Variable
Model
**Highly Significant
Probabi lity of Significant Difference
Piedmont vs.
Coastal Plain
AP vs. FP
PDCP vs. WDCP
COO
TOC
TKN
N03-N
TP
Cl
0.18
0.25
0.06
0.42
0.41
0.68
P>0. 10
P>0.10
P>0.10
0.050.10
P<0.01**
P>0.10
P>0.10
P>0.10
P>0.10
0.050.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
P>0.10
102
-------
POINT SOURCE IMPACT
A substudy of a 32-km (20-mi) stream reach, shown schematically in Figure
6, was conducted for a 15-month period (12/19/75 to 3/12/77). This system
was viewed as a model from which inferences as to the water quality impacts
of various land uses and urban influences could be drawn. Thus, this sub-
study is separate and different from the statistically-based main study.
Arithmetic sample averages for this 32-km (20-mi) stream reach showed
variations which seem related to inputs from point sources and swampy areas
(Table 24). The COD and TOC appeared to be increased as a result of effluent
input from a 1,200 person septic tank system (Site 2), a secondary treatment
plant (Site 4), and a swamp (Site 5). The recovery zones (Sites 3 and 7) were
physically different because Site 3 was channelized and Site 7 was heavily
influenced by swampy inputs. Therefore, COD and TOC concentrations after the
Site 2 input returned closer to background levels than after the downstream
effluent (Site 4) and swamp inputs (Site 5). Total nitrogen increased as a
result of effluent discharges (Sites 2 and 4) but decreased after the swamp
input (Site 5) because swamp waters had a low nitrogen content. It was inter-
esting that while ammonia variations were large, especially after the effluent
input from secondary treatment (Site 4), that changes in both nitrate and total
organic nitrogen (ON) (calculated using average TKN minus ammonia) were less
than might have been expected throughout the total reach.
Total organic nitrogen values were probably impacted by the relationship
between nitrogen inputs and incorporation to stream biomass. While the sample
size was too small, the relatively high total organic nitrogen value at the
last confluence (Site 6) may have been related to the large upstream reduction
of ammonia. Ammonia is the nitrogen form most easily assimilated biologically
and thus converted to organic nitrogen. Phosphorus concentrations were in-
creased after both point source inputs but declined in the recovery reaches.
Suspended solids were not markedly increased as a result of the first point
source input (Site 2) because of the long stream reach prior to confluence,
but at times appeared higher as a result of the second point source input and
then decreased to a very low value at the last site compared to Site 4. This
final decrease in suspended solids from Site 4 to Site 7, as well as phosphorus
and total organic nitrogen concentrations, may have been due to settling in
this slow, meandering stream reach.
In overview, the studied stream reach had high assimilatory capacity until
inputs from point sources and swampy areas overrode this dynamic capability.
It was noteworthy that concentrations in the background zones and the first
recovery zone were very similar for most parameters (TP and N03-N were slightly
higher). However, concentrations after the second point source and swamp
input remained higher than in the background reach for all parameters but
NH3-N and ON. It was also interesting that the swamp input, while exhibiting
high COD and TOC levels, actually resulted in reduced total nitrogen and total
phosphorus values because of the dilutional effects of these very low nitrogen
and phosphorus water and coincident settling that occurred in this slow moving
stream. Therefore, for this stream reach major inputs of nitrogen and phos-
phorus, which appeared to come from treatment plant effluents, were reduced
103
-------
TABLE 24^ MEAN VALUES OF STREAM REACH SUBSTUDY
Constituent
in mg/1
COD
TOC
TP
NH -N
ON
TKN
NO -N
TN
C1
SS
Samples
17
17
27
5
5
27
27
27
27
10
Head-
waters
1
21.2
6.6
0.13
0.05
1.00
1.05
0.26
1.31
8.56
12.9
First
Point
Source
Input
2
28.8
8.0
0.40
0.16
1.22
1.38
0.57
1.95
9.66
12.6
Recovery
Reach
3
23.7
7.2
0.21
0.04
1.03
1.07
0.49
1.57
9.06
11.2
Sites
Second
Point
Source
Input
4
26.6
8.8
0.82
1.47
1.01
2.48
0.56
3.04
11.17
18.8
Swamp
Input
5
33.4
13.2
0.19
0.05
1.16
1.21
0.03
1.24
6.92
5.4
Conflu-
ence
6
27.4
9.6
0.78
0.08
1.38
2.18
0.49
2.68
12.25
11.3
5 km
Downstream
7
32.3
10.2
0.48
0.14
1.15
1.29
0.52
1.82
11.28
4.3
-------
to near background levels as long as the stream assimilatory capacity was not
overwhelmed. Moreover, differing natural inputs because of changed physio-
graphic factors may cause changes in water quality, such as possible differ-
ences in COC and TOC, due to swamp inputs as in this study. However, phos-
phors and nitrogen species concentrations, except nitrate, evidenced real
but small differences when contrasting the background reach (Site 1) and
swamp input (Site 5).
These stream reach results suggested that nonpoint sources establish
background water quality and that during nonrunoff conditions, point sources
caused concentration increases that may or may not be reduced to headwater
background levels, depending upon impact and recovery zone capabilities, as
well as relative location along the stream reach. For example, concentration
values at recovery Site 3 after the point source input (Site 2) were very
similar to background Site 1 for most parameters (TP and N03-N slightly high-
er). However, the second point source, along with naturally occurring swamp
inputs began to establish a different water quality at Site 7 as compared to
Site 1; i.e., higher organics and oxygen demand, and elevated nitrogen and
phosphorus levels.
HEADWATER VERSUS MAINSTREAM
To gain additional insight as to the relative water quality in headwater
or rural areas and the main stream of the Chowan River system, initial com-
parisons were made using project data and information obtained from a com-
patible study in the lower Chowan (G. Cook, Personal Communication, North
Carolina Department of Natural & Economic Resources, Raleigh, North Carolina,
1976). Because constituent concentrations in the small study basins and main
stream draining these sites appeared to be similar, samples were taken at main
stream locations receiving drainage from the statistical survey sites and the
related survey sites on the same day. Comparison of average concentration val-
ues for these headwater and main river regions are shown in Table 25. These
overall averages were based upon one main stream value and the average of one
to three statistical site values for each sampling date. This data verified
that water quality is very similar in the rural headwaters and the main river
draining these rural areas based upon measured constituent concentrations.
However, there were differences between regions indicating that in these pre-
dominantly rural study areas, main stream water quality reflects inputs from
rural nonpoint sources and that these inputs or background conditions differ
with geoclimatic regions.
General Basin Summary
The physical and chemical grab sample measurements for the 18-month sta-
tistical survey are summarized by geoclimatic areas in Table 26. Mean values
for the four areas were relatively uniform throughout the river basin. How-
ever, the coefficient of variation (CV), standard deviation divided by mean,
was typically greater than 30 percent for all measured parameters except pH
which was typically about 8 percent. In many cases the CV approached 100
percent which indicated enormous naturally occurring variations within each
geoclimatic area due to spatial, temporal, and random factors and demon-
strated the large variation in individual sample values.
105
-------
In order to evaluate overall rural nonpoint source impact, watershed mean
concentration values were calculated and compared to water quality criteria
proposed by the Environmental Protection Agency (EPA) and also typical values
for secondary treated municipal wastewater as listed by Rubin et al. (1978).
The 50 states have adopted individual receiving water quality criteria based
upon EPA guidelines and local natural or background conditions. Although
the North Carolina and Virginia regulations are somewhat different than the
EPA guidelines, the EPA values are used for the following discussion of rural
stream water quality. These values are presented only to provide a range
of source and quality values for comparative purposes.
Dissolved oxygen (DO) is necessary to maintain fish and other aquatic
life, so oxygen supply was measured as DO concentration and the potential for
oxygen der.iand was assessed by TOC and COD concentrations. The mean DO levels
of the four geoclimatic areas of the Chowan Basin were greater than the pro-
posed minimum level of 5 mg/1; however, infrequent violations were measured.
These violations usually occurred during low flow conditions in the warmer
months. The watershed average TOC and COD concentrations were about 10 and
25 mg/1 respectively, while the maximum recorded concentrations were about
40 and 90 mg/1, respectively. The COD concentration was typically about
2.5 times as 'great as the TOC concentration; this indicates a relatively
oxidized state. Although EPA has not established water quality criteria for
either TOC or COD, typical COD concentrations for raw and secondary treated
municipal wastewater are 450 and 50 mg/1, respectively.
Nitrogen and phosphorus are basic elements for primary productivity and
thus relative concentrations are often employed to assess the potential for
algal production. Additionally, EPA has proposed a drinking water quality
criterion for nitrate based upon human health considerations and suggested
an upper limit of 10 mg/1 as N. While the maximum measured N03-N concentra-
tion (9.0 mg/1) of the rural stream water approached this limit, almost all
samples had concentrations much less than the proposed standard. For the
study geoclimatic areas, the mean TKN and N03-N concentrations were approxi-
TABLE 25. COMPARISON OF HEADWATER AND MAIN RIVER CONCENTRATIONS
F-Sites
Meherrin River
(Below F-Sites)
P-Sites
Chowan River
(Below P-Sites)
Visits
26
26
25
25
COD
19.2
19.4
35.3
32.8
TOC
6.7
7.6
14.6
13.6
TM
my/ i
0.10
0.10
0.24
0.24
NH3-M
0.02
0.04
0.06
0.06
TKN
0.85
0.64
0.98
0.85
N03-N
0.02
0.05
0.06
0.06
Cl
3.9
5.3
9.3
13.0
106
-------
TABLE 26, STATISTICAL SURVEY MEASUREMENT SUMMARY FOR GRAB SAMPI ING .1IJNE 1Q7R-NOUFMRFR 1Q7K
Variable
WYD
VELOCITY
TEMP
DO
pH
COND
Cl
COD
TOC
TP
N03-N
TKN
NHj-N
SS
WYD
VELOCITY
TEMP
DO
pH
COND
Cl
COO
TOC
TP
NOs-N
TKN
NH3-N
SS
Units
mm/day
m/sec
°C
mg/1
umhos/cm
T
1
mg/1
mm/day
m/sec
°C
mg/1
prahos/cm
T
i
mg/1
Number Minimum
of Samples Mean Value
Forested Piedmont
144 1.28
144 0.20
144 14.5
142 8.56
70 6.14
144 49.7
144 5.46
72 23.5
72 8.78
144 0.12
144 0.04
144 1.12
50 0.03
74 8.87
Agricultural Piedmont
112 1.58
112 0.17
112 H.9
105 9.08
55 6.26
114 51.2
112 4.20
56 17.8
56 6.98
112 0.10
112 0.11
112 1.00
42 0.0?
53 6.82
0.0
0.0
1.5
1.70
5.00
10.0
1.50
<4.0
<1.0
0.02
<0.01
0.11
<0.01
0.33
0.0
0.0
1.0
5.50
4.80
19.0
2.10
<4.0
<1.0
O.OZ
<0.01
<0.04
<0.0?
0.33
Maximum
Value
31.1
0.85
24.0
13.9
7.20
110
55.3
87.0
30.0
1.15
0.24
4.19
0.29
94.3
34.1
1.11
26.0
13.6
7.60
220
29.3
75.0
31.0
0.50
0.87
3.73
0.14
30.2
C.V.*
%
280
85
39
30
8
38
109
59
65
150
100
74
167
182
278
106
42
24
8
50
73
76
84
80
100
77
200
98
Variable
Units
Number Minimum
of Samples Mean Value
Well-Drained
WYD
VELOCITY
TEMP
DO
pH
CONO
Cl
COD
TOC
TP
NOj-N
TKN
NH,-N
SS
rim/day
m/sec
°C
mg/1
Mmfios/cm
t
1
mg/1
140
140
140
129
67
140
140
70
70
140
140
140
41
56
Poorly-Drained
WYD
VELOCITY
TEMP
DO
pH
COND
Cl
COD
TOC
TP
NOj-N
TKN
NHt-N
SS
mm/day
ffl/sec
°C
mg/1
umhos/cm
mg/1
136
136
136
136
71
136
136
68
68
136
136
136
68
68
Coastal
2.18
0.07
14.9
7.00
5.93
60.1
8.25
26.1
10.5
0.12
0.75
1.10
0.07
8.68
Coastal
0.43
0.11
15.6
7.13
6.01
80.4
9.37
25.6
10.7
0,22
0.53
1.18
0.17
H.4
Plain
0.0
0.0
0.05
1.40
4.80
25.0
2,44
<4.00
<1.00
0.02
<0.01
0.30
<0.01
0.55
Plain
0.0
0.0
2.0
0.20
4.70
1.30
2.63
<4.0
<1.0
0.03
<0.01
<0.04
O.01
0.8
Maximum C.V.*
Value %
75.6 340
0.56 114
26.0 44
13.6 35
7.40 11
190 /IS
51.4 58
68.0 52
33.0 59
0.60 83
8.96 187
4.77 S2
1.03 245
167 260
7.62 44
0.50 82
26.0 41
13.1 38
7.60 10
300 50
20.6 31
91.0 71
36.0 77
1.65 109
1.69 158
5.25 62
1.30 188
75.0 97
*C.V. = Standard Deviation/Mean
-------
mately 1.2 and 0.3 mg/1, respectively, so the average total nitrogen con-
centration was about 1.5 mg/1. Secondary treated municipal wastewater
typically has a total nitrogen concentration of 25 mg/1 which is approxi-
mately 10 times as great as the rural nonpoint source value.
To prevent biological nuisances, EPA suggested water quality criteria
for TP is 0.1 mg/1 for flowing streams not discharging directly to lakes or
impoundments, 0.05 mg/1 in any stream at the point where it enters a lake
or reservoir, and 0.025 mg/1 within a lake or reservoir. Because all four
geoclimatic areas of the Chowan Basin including the forested Piedmont had
a mean TP concentration greater than 0.1 mg/1, the natural or background
phosphorus level in the Chowan Basin exceeded the suggested water quality
criteria. Nevertheless, it should be noted that the rural nonpoint source
maximum TP concentration (1.5 mg/1) was only about one-tenth the average TP
concentration of 10 mg/1 for secondary treated municipal wastewater.
The mean Cl concentration for the rural runoff was 6.5 mg/1 compared to
a mean Cl concentration of 100±75 mg/1 for secondary treated wastewater.
In summary, average rural stream water quality for the four geoclimatic
areas was relatively uniform. The water quality was generally good relative
to proposed standards, but elevated TP concentrations even in the forested
Piedmont area demonstrated the need for water quality goals and criteria
to be responsive to measured local conditions, especially natural or back-
ground levels. Stream sample concentrations usually displayed large varia-
tions with respect to mean values indicating that rural nonpoint sources
are highly variable in both space and time.
SUMMARY
Grab sample data from a two-year, statistical sampling study of forested
and agricultural Piedmont, plus well- and poorly-drained Coastal Plain areas
of the Chowan River Basin, were analyzed to permit a clearer understanding of
the nature of rural runoff on an areawide basis. These analyses provided the
following conclusions:
1. Neither in-stream (arithmetic average) nor net export (flow weighted)
concentration data presented any clear relationships between water quality
and macro land-use factors. This result points out the need for caution when
employing model predictions to specific cases. Macro land-use factors do not
account for varying agricultural cropping and management practices, annual
weather conditions, stream border buffer systems, or other factors which can
minimize the impact of agricultural activities on water quality.
2. The impact of channelizing Coastal Plain streams was most pronounced
with respect to elevated N03-N concentrations in the channelized streams as
compared to unchannelized streams which have natural swampy flood plains and
channels which increase N03-N attenuation by denitrification and biological
uptake.
3. Analyses for seasonal trends indicated that water yield and the asso-
ciated nutrient yields were greater during the winter and spring seasons than
during the summer and fall seasons reflecting rainfall and evapotranspiration
108
-------
cycles. In analysis of seasonal flow weighted average concentration data,
the models generally demonstrated significant relationships but had rather
low r2 values.
4. Measured concentrations did not display any consistent functional
relationship to flow (water yield) levels. However, data showed that NOa-N
concentrations were elevated during flow conditions at a small (0.50 km2 or
0.2 mi2) site but not at larger (20 km2 or 8 mi2) sites in the well-drained
Coastal Plain with similar land use. The N03-N attenuation was judged to be
the result of in-stream dynamics.
5. Comparisons of geoclimatic areas demonstrated that the dominant vari-
ation was between the Piedmont and Coastal Plain with only relatively minor
variations occurring within these two physiographic regions. The differences
between the Piedmont and the Coastal Plain were judged to be due to naturally
occurring physiographic variations in (a) basin characteristics, such as
vegetation, soil type, and stream hydraulics; and (b) ocean proximity.
6, Assessment of point and nonpoint source impacts in one small basin
verified classic point source concentration spikes with subsequent decline
to intermediate levels for all investigated constituents except chloride and
nitrate. Therefore, for the studied stream reach, nitrogen and phosphorus
inputs which appear to come from treatment plant effluents are reduced to
headwater background levels as long as the stream assimilatory capacity is
not overwhelmed or natural inputs result in changed background water quality.
7. Point-in-time comparisons between headwater and downstream constituent
concentrations showed small differences on a water quality basis.
8. Eighteen-month average rural stream water quality for the four geo-
climatic areas was relatively uniform. The quality was generally good com-
pared to proposed standards, but elevated TP concentrations even in the
forested Piedmont area demonstrated the need for basing water quality assess-
ments on measured local conditions, especially background or natural levels.
Stream sample concentrations usually displayed large variations with respect
to mean values indicating that rural nonpoint sources are highly variable tn
both space and time.
COMPARISON OF MONITORING TECHNIQUES
Grab and automated monitoring techniques were elevated at four of the
statistical survey sites to supplement the cost-effective evaluations of these
procedures to estimate rural nonpoint source nutrient yields and concentrations
on an areawide basis. Consistent differences (bias) between mean value esti-
mates obtained by grab and automated sampling techniques were documented al-
though determination of causative effects was beyond the study scope. There-
fore comparisons of point-in-time measurements obtained by grab and automated
sampling systems are presented with the understanding that both procedures
estimate actual field truth. Next, annual data from four sites which were
both grab sampled and automated sampled were compared to assess possible
differences in annual water quality values associated with monitoring methods.
Finally, a sampling scheme which was flow stratified by the use of daily
rainfall probability predictions was explored as a possible method to obtain
more precise water yield estimates at a given sampling level because field
data showed that subbasin water yield distributions were highly skewed.
109
-------
STREAM MONITORING
Long-term grab were obtained from November 24, 1974 to March 19, 1977,
and automated data from May 18, 1975 to March 19, 1977. All automated and grab
sampling sites were operational from March 21, 1976 to March 19, 1977. During
this period substudies were conducted to investigate particular sampling metho-
dologies. Because these substudies often followed a sampling program which was
built into the long-term schedule, the essentials of the long-term sampling
plan are briefly reviewed followed by appropriate substudy information.
Long-Term Sampling Plan
The long-term sampling plan employed a sample time stratification (STS)
strategy which has been presented in detail (Section 4). The time stratifi-
cation was employed to ensure that measurements were obtained at a uniform
rate throughout the study. Basically, the stratification was such that each
of the 15 statistical survey sites was monitored 26 times per year at the
rate of two visits (chosen by a restricted random sampling) per 28-day period.
During each grab sample site visit, the flow rate was measured by standard
USGS (Buchanan and Somers, 1969) procedures and a manually depth integrated
water sample was obtained at the midpoint of the stream flow. Grab samples
were collected and stored in preaged, acid-washed polyethylene bottles and
were stored at 4 °C until laboratory analysis.
Automated sampling systems were established at five of the 15 statistical
survey sites. Stream stage was recorded continuously by analog recorders;
however, at one site a reliable stage discharge relationship and thus flow
data could not be developed because of beaver activity. Automated samplers
employed had the capability of collecting 28 discrete, 500-ml water samples.
To retard constituent transformations during field storage, the plastic
sample bottles were precharged with sufficient sulfuric acid to bring the
sample pH below 2 as recommended by EPA Methods for Chemical Analysis of
Water and Waste (1974). Each automated sampler intake manifold was at a
fixed location in the stream such that a 250-ml subsample was taken at 75-mm
(3-in.) stage increments, and two subsamples were composited to fill a sample
bottle. The sampler was designed and operated on a sample-flush cycle as
opposed to a flush-sample cycle. The time each subsample was obtained was
recorded on the stage strip-chart by a relay activated pen so each sample bot-
tle was assigned the mid-point time between samplings.
Grab vs. Automated Sampling
In Substudy A, data were collected to test whether manually depth inte-
grated grab samples and fixed point automated samples produced similar water
quality values for samples obtained at approximately the same time. Eighty-
seven paired samples, one of each type, were obtained during routine visits
to all five automated sites. Samples were obtained in the following manner:
first, the grab sample was taken downstream from the automated sampler intake
manifold; then the automated sampler was actuated to obtain a sample. The
time interval between the paired samples was less than five minutes. All
samples were stored at 4 °C until laboratory analysis. The samplers were in
110
-------
the stage activated mode between visits; and the intake systems were not clean-
ed or purged before taking the sample for these paired observations.
Grab and Automated Sampling and Storage
Another detailed study of the two sampling techniques was undertaken in
which data were obtained at four automated sites once per week during a four-
week period. The sampling procedure was to manually activate the automated
sampler to take five successive samples (designated A1-A5 as explained below).
Next, a manually depth integrated sample was taken (Gl). The technician then
installed a new intake system such that the intake manifold was in the stream
approximately 150 mm (6 in.) upstream and 75 mm (3 in.) to the side of the
usual sampling point. After waiting a sufficient time for all streambed dis-
turbance to subside (5 to 30 minutes), the intake system was thoroughly
flushed and two automated samples were taken (Ml, N2). These represented
duplicate samples with a new intake tube. Finally, a second grab sample
was taken (G2). All automated samples were acid fixed, whether stored a
week in the sampler or immediately returned to the laboratory. All grab sam-
ples were stored at 4 °C and returned to the laboratory. The new intake
system consisted of an intake manifold and hose which has been "aged" by
flushing with tap water for 24 hours. One "new" intake system was used for
this entire study, so before leaving the site it was removed and the old in-
take manifold was cleaned (routine procedure) and reconnected. The sampler
was inactive between these weekly visits. The sample procurement and dispo-
sition are summarized as follows:
1. Al = Acid fixed, taken to the laboratory as the first sample;
2. A2 = Acid fixed, taken to the laboratory;
3. A3 = Acid fixed and left in the field sampler for one week;
4. A4 = Same as A3;
5. A5 = Same as A2;
6. Gl = Iced, taken to the laboratory;
7. Nl = Acid fixed, taken to the laboratory;
8. N2 = Same as Nl; and
9. G2 = Same as Gl .
COMPUTATIONS
Both grab and automated data sets were employed to calculate annual (1)
water and nutrient yields, (2) volume average concentrations, and (3) time
average concentrations. The computational methods employed with the grab
data set are reviewed to illustrate the use of a statistical grab sampling
data set with respect to employed data analyses. Data set manipulations to
permit computations with the continuous stream hydrograph record plus stage-
activated water sampling system are also discussed. Finally, the grab data
set had no missing values for the year analyzed, but the automated data set
did; thus computational methods employed to generate substitute values for the
missing data are presented.
Ill
-------
Grab Data
Computational methods for the grab data set are considered first. Annual
water yield (WY) values were calculated using Equation 3.
WY = z3 z uFLOW.. *DAYS..)/AREA] (3)
1=1 1 = 1 (_ 'J U J
where FLOW-jj and DAYSjj are respectively the flow measurement and duration of
the jth sampling unit of the itn 28-day period, and AREA is the subbasin area.
The variable time factor was employed because a sampling unit could be 12, 14,
or 16 days. Annual constituent yields (CY) were computed as:
CY = z2 z [(C..*FLOW..*DAYS..)/AREA] (4)
i = l j = l L 10 10 U J
j. L*
where C-jj is the grab sample constituent concentration for the j sample of
the itn period. Annual volume average (flow weighted) concentrations (VAC)
were computed by dividing the annual constituent yields by the annual water
yield. Annual time average concentrations (TAG) were computed for each site
as:
TAG = Z Z (C,-*DAYS.-/DAYS) (5)
i=l j=l 1J 1J
where DAYS represents a time interval of 364 days. Equations 3-5 reduce to a
more common format if the sampling unit duration is chosen to be constant.
Automated Data
Hourly water yield values were calculated by digitizing the hydrographs
on an hourly basis and employing the appropriate stage-discharge relationship.
Hourly constituent transports were also determined for the automated data set.
The stage activated samples did not obtain samples on an hourly basis so a
method of assigning hourly values had to be chosen.
The hydrographs had considerably steeper sections during the rising por-
tion of runoff events (more than two water samples/hr) than during the re-
cession limb (less than one sample/hr) and these events were usually separ-
ated by relatively long baseflow periods (less than one sample/day). Thus,
water sampling frequency varied across the hydrograph with samples being taken
as often as several per hour on rising limbs of runoff events to as infrequent-
ly as less than one per week during nonrunoff conditions. So that the first
sample on the rising limb of an event would not impact the possible long an-
tecedent baseflow assessment, the assignment method judged to best represent
the physical system for calculating constituent yield elements (CY. ,) was:
K , 1
112
-------
where C|< is the concentration of the kth.water sample and Yu i is the ith 5-
minute water yield during the time interval between the k and k+1 water sam-
ples. Ci was a special case in which the concentrations were extended back-
ward to time zero because no sample was obtained when the instruments were
initially activated. Five-minute water yields for these calculations were
obtained by linear interpolation between hourly values. The 5-minute con-
stituent yield values were summed on an hourly basis and the resulting values
were used in data analyses.
The automated sampling data records for each site are summarized in Table
27 which shows that the P8 system was 100 percent operational and the other
three systems were operational more than 95 percent of the year. Analytical
techniques were employed to generate substitute values for the inoperable
hours. The hourly water yields were highly correlated for sites within either
the Piedmont or Coastal Plain so linear regressions were employed on adjacent
site values to generate missing water yield values. No satisfactory regression
method could be formed to predict concentrations at a site so the year was
broken into four, 3-month seasons (November-December corresponding to winter);
and seasonal flow weighted average concentrations were employed to generate
missing hourly constituent yield values.
The measured and synthesized automated data were combined to provide a
complete hourly data record. Annual yields and volume average concentrations
were computed with respect to definitions. On an hourly basis, constituent
yields were divided by water yield to obtain volume average concentrations.
These hourly volume average concentrations were arithmetically averaged to
provide annual psuedo-time average concentrations which are simply referred to
as automated time average concentrations.
RESULTS AND DISCUSSION
Grab vs. Automated Sampling - Substudy A
Constituent mean values from the initial comparison of grab and automated
sampling at five sites and the significance of differences determined by an
analysis of variance are presented in Table 28. The analysis of variance was
performed on logarithmically transformed paired sample data because measure-
ment errors were considered to be proportional to the magnitude of the measured
TABLE 27. AUTOMATED DATA RECORD SUMMARY .(MARCH 2,1|T ,1976-MARCH, 19. 1977)
Site Percent record complete Hours missing Dates missing
F03
A01
P08
P10
97.8
97.7
100.0
98.6
192
195
0
119
January 28-February 5
July 17-July 25
None
January 10-January 15
113
-------
TABLE 28. COMPARISON OF GRAB AND AUTOMATED SAMPLES AT FIVE SITES
Parameter
COD
TOC
TP
TKN
N03-N
Cl
Grab mean
concentration
mg/1
25.10
8.10
0.13
1.22
0.09
6.19
Automated mean
concentration
mg/1
39.40
13.20
0.18
1.49
0.08
6.71
Statistical
significance
of difference
0.010.10
0.010.10) for N03-N
concentrations. The grab mean concentrations were less than the automated
mean concentrations by 36, 39, 28, 19, and 8 percent for COD, TOC, TP, TKN, and
Cl respectively.
Grab vs. Automated Sampling and Storage
This detailed experiment was performed to assess possible sources of ob-
served differences between parameter concentrations measured by manual, depth-
integrated, grab sampling and fixed-point, automated sampling. Several pos-
sible contributing factors were evaluated statistically by an analysis of var-
iance which employed logarithmically transformed data. Mean values and signif-
icance of differences are summarized in Table 29. Four single degree of free-
dom comparisons were examined. Although these are not all orthogonal, they
comprise the comparisons of interest. The error term was a composite of first
order interactions of visits and sites for the nine samples taken. For the
constituents considered, the experiment showed (P<0.10) that when a sample-
flush cycle was employed, the first automated sample (A!) concentrations of
COD, TOC, TP, and TKN from samplers which were inactive for a week were 55, 60,
71, and 34 percent greater than the concentration of samples (A2, A5) taken
immediately thereafter; i.e., a first sample effect existed. For comparisons
in which the first sample was excluded, it was found that:
114
-------
TABLE 29. EXPERIMENTAL STUDY OF AUTOMATED SAMPLER AT FOUR SITES
Effect tested
Parameter
Argument mean
concentration
mg/1
Standard mean
concentration
mg/1
Intake systems;
Nl, N2 vs. A2, A5
COD
TOC
TP
TKN
N03-N
Cl
51.100
16.300
0.220
1.020
0.098
6.310
Storage;
Statistical
significance
of difference
First
A1 vs.
s amp 1.e ;
A2, A5
COD
TOC
TP
TKN
N03-N
Cl
82.600
30.400
0.040
1.590
0.630
5.990
53.300
19.000
0.140
1.190
0.079
6.020
P<0.01**
P<0.01**
P<0.01**
0.010.10
P>0.10
53.300
19.000
0.140
1.190
0.079
6.020
P>0.10
P>0.10
P>0.10
0.050.10
P>0.10
A3, A4 vs. A2, A5
Grab vs.
routine automated
61, G2 vs. A2, A5
COD
TOC
TP
TKN
N03-M
Cl
COD
TOC
TP
TKN
N03-N
Cl
42.400
16.100
0.120
1.090
0.089
6.250
42.300
12.500
0.140
0.960
0.083
4.780
53.300
19.000
0.140
1.190
0.079
6.020
53.300
19.000
0.140
1.190
0.079
6.020
P<0.01**
0.050.10
P>0.10
P>0.10
P>0.10
0.010.10
0.01
-------
Annual Concentration and Yield Comparisons
The annual volume average concentrations obtained by simple time strati-
fication (STS) grab sampling and stage-activated automated sampling at four
sites are presented in Table 30. For a given site, relatively large differ-
ences can be observed between the grab and automated estimates. Because the
grab to automated concentration ratio varied among constituents at a site, no
consistent factor related the two estimates. Further, for each water quality
constituent the grab to automated concentration ratio varied among the sites;
thus a factor relating the two estimates which was independent of site did
not exist. The significance of difference between the grab and automated vol-
TABLE 30. COMPARISON OF GRAB AND AUTOMATED FLOW
WEIGHTED CONCENTRATIONS AT FOUR SITES,
(MARCH 1976-MARCH 1977)
Parameter
COD
TOC
TP
TKN
N03-N
Cl
Subbasin
F3
Al
P8
P10
F3
Al
PS
P10
F3
Al
PS
P10
F3
Al
P8
P10
F3
Al
P8
P10
F3
Al
P8
P10
GVAC*
mg/1
23.7
12.9
54.5
38.4
5.86
4.80
9.76
9.04
0.094
0.085
0.176
0.174
1.16
0.96
1.95
2.40
0.033
0.112
0.075
0.372
3.50
3.28
4.51
10.25
AVAC**
mg/1
42.0
41.1
56.4
81.9
12.6
13.1
14.3
21.4
0.139
0.190
0.256
0.521
1.32
1.45
1.87
2.46
0.054
0.092
0.074
0.413
4.58
4.00
6.44
9.02
GVAC/AVAC
%
56
31
96
47
47
37
68
42
68
45
69
33
88
66
104
97
61
122
101
90
76
82
70
113
*GVAC - Grab Sampling Flow Weighted Concentration
**AVAC - Automated Sampling Flow Weighted Concentration
116
-------
ume average concentration estimates was evaluated employing a two-sided t-test
utilizing logarithmic transformed paired samples (values for the two methods
at each of four sites). The test indicated (P<0.10) that annual, volume aver-
age, grab concentration estimates of COD, TOC, and TP were less than automated
sampling values by a statistically significant amount averaging about 50 per-
cent; for TKN, NOs-N and Cl grab sampling was less by an average of about 13
percent, not significant statistically.
The ratio of the annual, time average concentration to the annual, volume
average concentration obtained by both grab and automated sampling is presented
in Table 31. For some nonpoint sources such as sediment, higher concentrations
occur during high flows in which case the time average concentration (TAG) is
less than the volume average concentration (VAC). However, the values in Table
31 indicate that this is not always the case for the.measured constituents and
TABLE 31. RATIOS OF TIME AVERAGE TO FLOW WEIGHTED
CONCENTRATIONS AT FOUR SITES,
(MARCH 1976-MARCH 1977)
Parameter
COD
TOC
TP
TKN
N03-N
Cl
Subbasin
F3
Al
P8
P10
F3
Al
P8
P10
F3
Al
P8
P10
F3
Al
P8
P10
F3
Al
P8
P10
F3
Al
P8
P10
Grab
ratio
1.12
0.91
0.74
0.59
1.14
0.90
1.43
0.74
0.96
0.98
0.93
0.87
1.00
0.91
0.74
0.49
1.71
0.75
0.63
0.50
0.91
1.05
1.59
0.94
Automated
ratio
1.07
1.18
1.16
1.02
1.13
1.17
1.51
1.04
0.95
0.77
1.25
0.96
0.91
0.94
1.12
1.01
1.00
0.48
0.62
0.78
0.96
0.90
1.25
1.18
117
-------
computational methods employed. The fact that many values are approximately
1.0 implies that concentrations are not correlated with flow.. Use of a two-
sided t-test on log-transformed paired summary values found evidence of a
statistically significant difference in only 2 of the 12 data sets; for grab
sampling the time averaged mean for TP was less than the volume average, while
for automated data the time average values exceeded the volume average (in both
cases 0.05
-------
per two-week frequency did not provide a very precise estimate of the annual
water yield for the 15 statistical survey streams during the sampling period.
Further analysis of the water yield measurements for each site indicated
that the distributions were highly skewed (Table 33); i.e., many low flow
measurements and_only a few high flow measurements were made at each site.
Stratified sampling on the basis of the measured parameter may provide a more
precise estimate of the mean at a given sampling intensity compared to simple
random sampling for skewed distributions (Cochran, 1977). In practice, this
suggests a method of stratifying stream measurements based upon flow regime.
Daily rainfall probability predictions provide a mechanism for stratifying
days into two strata with either a high or low chance of rainfall and hence
runoff. As a first approximation, daily water yield should be proportional to
daily rainfall. Naturally other factors, such as soil moisture and rainfall
intensity, impact this relationship; but as an initial attempt, it seem appro-
priate to examine the probability of rainfall as an indication of flow regime.
To pursue this approach, five years of NOAA rainfall predictions (U.S. Depart-
ment of Commerce, 1971-76) and rainfall measurements (Wiser, 1975) for a pre-
cipitation recording station in the general vicinity of our study (Edenton,
North Carolina) were- compiled. The contingency table for these data, Table
34, indicating relationships between rainfall predictions and measured rainfall
TABLE 33. GRAB SAMPLING WATER YIELD AND SKEWNESS VALUES, (JULY 1975-NOVEMBER
1976)
Subbasin
mm/yr
Water yield
(in./yr)
Skewness
value
nondimensional
F-l
F-2
F-3
F-7
A-l
A-4
A-8
W-3
W-4
W-8
W-10
P-8
P-10
P-ll
P-13
Mean
376
345
348
211
417
330
437
942
940
127
175
124
132
76
102
338
(14.8)
(13.6)
(13.7)
(8.3)
(16.4)
(13.0)
(17.2)
(37.1)
(37.0)
(5.0)
(6.9)
(4.9)
(5.2)
(3.0)
(4.0)
(13.3)
5.6
3.4
3.5
2.9
4.3
5.7
4.6
4.6
1.5
4.1
5.1
4.3
3.0
119
-------
TABLE 34. DAILY PRECIPITATION PROBABILITY VERSUS DAILY PRECIPITATION AT EDENTON, N.C., OCTOBER
1971-OCTOBER 1976
ro
o
Forecast
probability
of precipitation,
in percent
0
10
20
30
40
50
60
70
80
90
100
Sum of
Probabilities
0.0-T*
(0.0-T)
25.00
14.30
11.00
7.62
4.36
2.10
2.04
0.94
0.55
0.22
0.00
68.97
Dai
0.1-12.4
(0.01-0.49)
1.99
2.37
3.64
2.82
2.65
2.10
2.71
2.10
1.27
0.22
0.11
21.98
ly precipitation
occurrence,
12.5-25.2
(0.50-0.99)
0.17
0.11
0.72
0.77
0.55
0.61
0..88
0.44
0.77
0.28
0.22
5.52
range, in mm
in percent
25.3-37.8
(1.00-1.49)
0.11
0.06
0.33
0.22
0.39
0.17
0.22
0.17
0.28
0.06
0.06
2.04
(in.);
>37.9
0.06
0.00
0.28
0.17
0.22
0.06
C.ll
0.28
0.11
0.06
0.17
1.49
Sum of Ranges
27.89
16.84
16.23
11.60
8.17
5.02
5.96
3.92
2.98
0.83
0.55
100.00
*T = Trace
-------
shows that either no rainfall or less than 13 mm (0.5 in.) of rainfall was
measured approximately 90 percent of the time and that as the daily rainfall
increased, the likelihood of occurrence decreased rapidly.
This daily rainfall distribution indicates a cause of the skewed daily
water yield estimates provided by STS grab sampling. However, for a specified
sampling level, utilization of rainfall predictions can provide an increased
proportion of high flow measurements as compared to simple random sampling. If
a field trip were made every time the precipitation probability reached or ex-
ceeded some chosen level, the chance of measuring streamflow during a signifi-
cant rainfall would increase as the rainfall probability increased. This is
illustrated in Figure 15 where the predicted number of trips per year and the
expected number of "successes" for several minimum rainfall probability levels
of field trip initiation are shown. Success was defined as a daily rainfall
of 13 mm (0.5 in.) or greater which was the amount assumed to result in eleva-
ted streamflows. As a further refinement the rainfall distribution of the
successes is shown in Figure 15. As an example, if a sampling trip were made
only when the probability of rainfall was 80 percent or greater, about 7 of
16 field trips per year would result in elevated stream flow measurements; and
the rainfall intensity would be greater than 39 mm (1.5 in.) about once per
year.
Thus, the chance of predicting significant rainfall and correspondingly
the chance of measuring elevated stream flow can be increased by employing rain-
fall probability predictions as compared to simple random sampling. Because
the distributions of daily water yield for small subbasins in the study area
of the Chowan Basin were skewed, a sampling plan stratified by the probability
of rainfall instead of a simple time stratification may have allowed increased
precision of annual water yield estimates for a given sampling intensity. Bas-
ically this method permits a monitoring agency to put a small fraction of re-
sources into measuring baseflow conditions and a larger fraction into the more
difficult measurement of high flows. As a consequence, better estimates of
runoff conditions and annual yields may be obtained on an areawide basis while
still utilizing the statistical approach.
SUMMARY AND CONCLUSIONS
Statistical sampling methods can be employed to measure the mean areawide
contribution of chemical species from rural nonpoint sources as an alternative
to the more difficult and often impractical complete monitoring approach. Grao
sampling and automated sampling are two common methods of assessing stream
water quality which can be employed in a statistically designed sampling pro-
gram Water quality based results of the grab and automated sampling compari-
sons'provide information which can be utilized to design sampling plans that
minimize the variance attributable to sampling methodology.
Point-in-time comparisons of manually depth integrated and fixed point
automated samples were conducted for COD, TOC, TP, TKN, N03-N, and Cl. One
substudy showed that manually depth integrated samples provides lower estimates
of COD TOC, TP, TKN, and Cl concentrations (36, 39, 28, 18 and 8 percent,
respectively) than the first sample obtained from fixed point automated sam-
plers in routine stage-activated operation when both samples were preserved by
121
-------
refrigeration until laboratory analysis. Data from a study to determine the
relative contribution of sample procurement, preservation, and storage on auto-
mated sample parameter concentrations indicated that the sample-flush sampling
cycle produced a "first" sample effect; that is, the first sample obtained from
an automated sampler operating in the sample-flush mode which was inactive for
approximately a week had about 60 percent higher concentrations of COD, TOC,
TP, and TKN than samples taken immediately thereafter. About 20 percent lower
concentrations of COD and TOC were observed for manually depth integrated sam-
ples preserved by cooling compared to fixed point automated sample preserved by
acid fixation. Finally, acid-fixed automated samples which were stored in the
field for a week had about 20 percent lower COD and TOC concentrations than
51.6
AVERAGE NUMBER OF
FIELD TRIPS PER YEAR
AVERAGE NUMBER OF
MEASURED RAINFALL
EVENTS PER YEAR
13- 24 mm
(0.5- 0.99 in )
25-38 mm
(1.0-1.49 in)
3t 39mm
1.5 in)
60%
70%
80%
90%
MINIMUM PROBABILITY OF RAINFALL NECESSARY
TO INITIATE A FIELD TRIP
Figure 15. Example of rainfall forecasts to predict sampling
events.
122
-------
acid-fixed automated samples returned immediately to the laboratory for anal-
ysis. These data indicate that COD and TOC were the parameters impacted most
when recommended procedures were employed in routine data acquisition. Further-
more, the sample-flush cycle caused a first sample effect that had a signifi-
cant impact on several parameter concentrations.
Annual volume average concentration and water yield estimates were obtain-
ed at four sites by routine operation of both stage-activated automated sam-
plers with stage recorders and simple time stratified grab sampling. Results
indicated that about 50 percent lower COD, TOC, and TP estimates were obtain-
ed by the grab sampling method. Although rather large differences were ob-
served for the water yield estimates, the data did not indicate any statis-
tically significant difference between the two methods. Due to confounding
factors associated with the two sampling methodologies, it was not possible
to specifically define the source of the concentration differences. However,
investigators should be alerted to the limitations of estimates using one or
the other of these sampling methods.
Comparison of annual water yield estimates from the 15 statistical survey
sites to historic values for the study region verified that simple time
stratified (STS) grab- sampling provided reasonable estimate of areawide
annual water yield; however, the precision of the individual estimates was
low. Sampling theory indicates that increased estimate precision of skewed
populations, such as recorded by the STS sampling of water yield, can often be
achieved by stratification on the population variable. Thus, the potential of
employing daily weather predictions as a means of time stratification to im-
prove water yield estimate precision at a given sampling intensity was demon-
strated.
In conclusion, the effect of sampling methodologies on water quality
measurements in no way lessens the need to establish adequate data bases in
order to provide a foundation for sound scientific assessment of rural non-
point sources. The information presented should aid future investigators and
monitoring agencies in better understanding stream monitoring data and design-
ing improved data acquisition systems.
123
-------
SECTION 6
ALGAL POPULATIONS IN TWO SMALL STREAMS
by
Augustus M. VJi therspoon and Thomas R. Fisher11
INTRODUCTION
In recent years much attention has been focused on the pollutional
burden carried in streams of the United States. The concern has been
generated in part from the objectional aesthetic deterioration of
waters receiving the discharge of all streams, large and small. Algal
blooms of increasing frequency, duration, and magnitude have played a
large part in generating social pressure to rectify man-induced environ-
mental deterioration. The resulting legislation has been directed at
monitoring water quality and detecting and correcting deleterious
changes.
Knowledge is incomplete on the contributions made to this problem
of algal growth by elements of rural runoff as it occurs in small head-
water streams of a drainage basin. The contribution of nutrients made
by point sources, and the characteristics of algal populations in
impoundments and in slow-moving large streams are better understood.
The objectives of this phase of the overall study were to appraise
the algal populations found in small drainages, to list the species
found, to measure the population biomass and to apply probability
sampling to the study of population changes. A further objective was
to appraise the usefulness of a static algal assay technique developed
and recommended at the national level by the Environmental Protection
Agency (Anon. 1971).
CONCLUSIONS
In these two small streams,
with only a few dominant forms.
the algal community included many species
Manuscript assembled and edited by Don W. Hayne.
124
-------
Algal biomass was usually low, but reached high levels occasionally.
Explanation for fluctuations in biomass must be sought in processes
bringing algae into the flowing water from the bottom or from nearby
water or land, rather than in growth of the ambient plankton itself.
Algal population levels in these headwater streams are limited by
fluctuations in these unknown input processes, by high flushing rates,
and possibly by poor light, not by nutrients; ambient populations are
only a fraction of the potential levels realized in laboratory assay
and possibly in lower parts of the same stream.
Algal assays indicated potential growth limitation most frequently
by nitrogen, next by factors other than nitrogen or phosphorus, rarely
by phosphorus alone. Such potential limitations apply in reality only
at some hypothetical downstream location and condition, not to popula-
tions present in these small streams.
With in.-stream fluctuations in potential nutrient limitation suggest
a within-stream heterogeneity in nutrient balance.
RECOMMENDATIONS
The origin of planktonic algae in small headwater streams should
be further studied to determine the relative contribution from the
periphyton, from bodies of standing water, and from the soil surface.
Further investigation should be made of the possible synchrony
between plankton pulses in small drainage basins, in principal tribu-
taries and in the main river, to determine what role the populations
in headwater streams may play in blooms in the estuary.
The bottle assay method should be refined and used to determine
by how far the nutrient concentrations in small drainages fall short
of those required to support an algal bloom.
SITES OF STUDY AND METHODS OF PROCEDURE
These algal studies were carried out in -two streams, arbitrarily
chosen for more intensive study from the set of 15 small drainage basins
chosen at random for the overall study. One of these streams lay in the
poorly drained Coastal Plain while the other was in the forested
Piedmont. Several sampling stations or subsites were established on
each stream.
The piedmont stream (Quarrel Creek in Brunswick County, Virginia,
designated F-3) lay in a rolling piedmont location entirely wooded
along the stream and for some distance away from it in both directions
with a mesic hardwood forest. Subsites from which data were used were
numbered 20, 22, 23, 24, 25 and 28 with 20, 22, 24, 25 and 28 in one
direct line of flow and 23 on a tributary just above the confluence
which lay just above 24. A few other subsites were sampled at irregular
125
-------
intervals. The stream had many looping hairpin turns, especially in
the broader part of the forested flood plain which had a thick deposit
of leaves mixed with sand from previous flooding. The banks were clay
0.5-1.5 m high and the stream bottom was sand with some forest debris.
This stream was 2-5 m wide, 0.2-0.4 m deep, and recorded velocities
were 0.16-0.31 m per sec, all measurements increasing downstream. The
length of stream above the lowest subsite used here (25) was 10.2 km.
These characteristics suggest that this stream volume was replaced in
9-18 hours.
Within the stream there were many rocks covered with algae and an
unidentified macrophyte up to 0.6 m long, carrying algal periphyton.
Blue-green algae were observed growing on sand in places where the
turbulence was not great. There were also many standing pools with
large algal populations (mostly filamentous greens and blue-greens);
some of these had a direct connection with the stream and slowly drained
into it. Thus the F-3 system was a winding natural stream with many
internal sources of algae as well as some external but the flushing rate
was such that there could not be much growth of planktonic algae within
the stream itself.
The coastal plain stream (Flat Swamp in Hertford County, North
Carolina, designated P-10) had been channelized and was virtually
without pools or bends, 1-3 m wide and 0.2-0.5 m deep. Subsites from
which data were used were numbered 1, 2, 3, 4 and 6 with 1, 3, 4 and
6 in one direct line of flow and 2 on a tributary just above the con-
fluence which lay just above 4. Several other subsites were sampled
at irregular intervals. The banks of P-10 were clay sub-soil with a
U-shaped cross section typical of channelized streams. The bottom
was of clay chunks and some sandy clay with virtually no rock or hard
substrate. The water flowed straight and fast; litter did not accu-
mulate to any extent. Estimates of water velocity ranged from 0.41-
0.70 m per sec; with a total length of 6 km above the lowest downstream
subsite (6) water was replaced in the stream every 2-4 hours. The
water was stained with dissolved organics and in addition carried a
load of fine clays and other suspensoids with the bottom not visible if
the water was deeper than 0.3 m. Although agricultural fields dominated
the drainage basin, a margin of mostly deciduous trees bordered the
entire length of the stream and provided heavy shade when leafed out.
Thus light was not plentiful during most of the year.
Possible sources of algae in the P-10 watershed were numerous.
During one visit in winter the cultivated fields had a noticeable
green cast, indicating a surface bloom of soil algae; clumps of fila-
mentous algae were observed growing on wet soil of empty fields.
Other sources of algae were nearby pocosin swamps; one of these
appeared to be fed directly by the effluent from a pig farm and to be
covered by a green layer of algae. Thus, algal concentrations in the
stream probably reflected local extrinsic events and not intrinsic
conditions.
126
-------
The basic schedule for making algal collections at these two sub-
basins was the same as that for the overall study, that is, two visits
at randomly selected times within each four-week period. Algal collec-
tions were made on this schedule from the first visit (24 November 1974)
through the seventh (7 March 1975) and then the schedule was shifted to
half this frequency because of the cost of processing the samples, and
this schedule was used through visit 34 (9 March 1976). After examining
the data from the algal assays, however, it was necessary to discard
results during the period of developing the technique prior to visit 16
(9 July 1975). Thus, information is available on the identity of the
algal species and on their numbers and biomass from November 1974 to
March 1976, but for the algal assays the only information is for the
shorter period of July 1975 to March of the following year.
Typically, at each field visit to a sampling point water samples
were taken for algal study and for chemical analysis and a flow measure-
ment was made. Water samples were collected with a bottle, manually
integrating over the water column in each stream. Samples for algal
assay were iced; those for enumeration of endemic algae were preserved
with Lugol's solution in the field. Samples were brought to the
laboratory according to the work schedule, sometimes within the same
day but never after more than three days.
The endemic algae were counted and identified within one week after
arrival of the sample in the laboratory. From each water sample, a 100
ml subsample was counted in an Utermohl settling chamber using an
inverted microscope. Fields were counted until 100 individuals of each
of the more common species had been recorded. From each sample the
volume of 10 individuals of each species was estimated using an appro-
priate solid geometric model depending upon the shape of the organism.
The number of cells per ml was calculated for each species, using a
constant unique to each sample (depending upon the number of fields
counted, the field area and the settling chamber area). The total cell
volume for each species was then estimated from the average cell volume
and the estimated number of cells per ml. Cell volume was taken to
equal biomass assuming that density was unity (Vollenweider 1974).
To compare the endemic algal population levels observed in the
streams with the potential levels as measured in the algal assays, the
latter values had to be converted to rough approximations in similar
units. For this reason the optical density measurements of maximum
standing crop for the unenriched controls were converted to cell numbers
using data from Forsythe (1973, Table 7).
Diversity and evenness were calculated here, based upon the genus
taxon. Weber (1973) suggested in reference to macroinvertebrates, that
diversity, and especially "equitability" of Lloyd and Ghelardi (1964)
might be sensitive indices to water quality. In following this sug-
gestion with algae, we have used the Shannon-Weaver form (Poole 1974,
Pielou 1975):
127
-------
H1 = c£ p. In p.
where s = number of genera
p. = proportion of the total number of individuals in the ith
1 genus
c = constant to state results as logarithms to base 2.
The Margalef evenness value has been calculated as the diversity value
divided by the value for maximum diversity possible with the observed
number of genera.
The importance of the algae in transporting carbon, nitrogen and
phosphorus was estimated for each collection. First the carbon,
nitrogen, and phosphorus contents of the algal cells were estimated
from the bioniass, assuming that cell organic carbon represents 10
percent of algal fresh weight (Vollenweider 1974), and that the atomic
C:N:P ratio is 100:16:1 (Redfield et al . 1963, Winberg 1971). These
assumptions result in an uncertainty of 3-5 in the final results
(Vollenweider 1974), i.e., the correct value lies in the interval 1/3-
1/5 to 3-5 times the calculated values. These values were then calcu-
lated as percentages of total organic carbon, total nitrogen (nitrate
plus Kjeldahl) and total phosphate as determined by chemical analysis
of the particular water sample.
The measurements of algal biomass made in these streams were converted
to approximate equivalent values of chlorophyll a^ through use of the data
of Nichols (1976).
The Environmental Protection Agency has developed an algal assay
procedure which has been recommended as useful in studies of water
quality (Weiss and Helms 1971, Fitzgerald 1972, Forsythe 1973, Maloney
et al . 1972). This procedure is a static culturing system which tests
the effect of additions of nutrient on the growth rate and maximum
standing crop of a standard test alga. The results of the static bio-
assays have been used to determine limiting nutrients, and can be used
to assess the growth potential for algae of a particular body of water.
Our algal assays used this Provisional Algal Assay Procedure Bottle
Test (Anon. 1971).
Upon receipt, water samples for algal assay were micropore (0.45
micron) filtered to remove living cells and then used to prepare a
series of 250 ml Erlenmeyer flasks with 100 ml water each. Three repli-
cated flasks (sometimes two) each for 10 treatment levels received
differing spikes of nitrogen and phosphorus according to the experi-
mental design detailed below. Sodium nitrate was added for nitrogen at
0.75, 0,225 and 0.75 mg/1, and potassium phosphate for phosphorus at
0.005, 0.015, and 0.050 mg/1. Each flask was then seeded with 1,000
123
-------
cells /ml of Selena strum capricornutum maintained and prepared for
inoculation as directed in the reference. The seeded flasks were held
in a constant temperature room (24° ± 1° C) under a continuous light
regime of 400 ft-c. With each set of flasks (which usually comprised
all assays for a single visit to a stream), three replicate flasks
were also seeded with the same cell concentration of the test organism
on the standard culture nutrient medium and held for the same period
under similar temperature and illumination conditions. These "medium
controls" served an important function; their erratic behavior before
visit 16 demonstrated that the laboratory procedure was not under control
The optical density of the algal standing crop was measured daily
using a Beckrnan spectrophotometer at 750 nm until increase was less
than 5 percent in a day. For each flask two parameters were calculated
from the set of successive daily measurements. Optical density readings
were used in these calculations rather than estimated cell density,
following the finding of Weiss and Helms (1971) that optical density is
highly correlated with algal biomass. Maximum standing crop was taken
to be the first point reached after which the daily increase in biomass
was less than 5 percent. Maximum specific growth rate was the maximum
value of the specific growth rate (or instantaneous rate of growth)
which was calculated for each day by the usual relationship as follows:
specific growth rate = •
where X2 = biomass at the end of the time interval
Xj = biomass at the beginning of the time interval
t = elapsed time in days, here 1 day.
Units of this quantity are day'1.
An argument can be made that maximum specific growth rate may be a
less useful response variable here than maximum standing crop, because
growth rate may be limited by rate of uptake of available nutrients.
Recent research on the uptake kinetics of nitrate, ammonia, and
phosphate (Eppley et al . 1969, Fuhs et al . 1972, Maclsaac and Dugdale
1969, Stanley and Hobbie 1977, Balducci 1977) have documented a
generally hyperbolic relationship between the rate of uptake of a
nutrient and its concentration. Uptake is usually expressed as a
velocity (uptake/ standing crop, both in units of N or P) which is equi-
valent to a specific growth rate in terms of either nitrogen or
phosphorus. The two parameters used to define the curve are Vmax, the
maximum or saturated uptake velocity (the asymtote), and Kt, the sub-
strate concentration at one-half the maximum uptake velocity. At
concentrations exceeding Kt, at most a doubling in uptake velocity can
occur, and at concentrations exceeding two or three times Kt difficulty
may be anticipated in detecting increased uptake velocity with increased
nutrient concentration. Although Kt varies depending on environment
129
-------
(Maclsaac and Dugdale 1969), most estimates fall into the ranges 0.01-
0.05 mg N/l and 0.001-0.005 mg P/l (Eppley et al. 1969, Maclsaac and
Dugdale 1969, Funs et al. 1972). Stanley and Hobbie (1977) and Balducci
(1977), working in the lower Chowan River, estimate Kt as 0.01-0.02 mg
N/l and 0.004-0.005 mg P/l, respectively. Considering the average
concentrations in the two streams it seems that little relationship
between concentrations and growth rate should be expected, especially
for stream system P-10. The concentrations in F-3 were closer to Kt
values than are those in P-10. Except in lakes or reservoirs where
concentrations are equivalent to or less than Kt values (e.g., Maloney
et al. 1972), maximum specific growth rate may not be as useful as
maximum standing crop in algal assays.
An incomplete factorial design was used for the algal assay procedure,
following somewhat the suggestions of the publication describing the
method (Anon. 1971). A factorial design had the advantage that certain
"contrasts" could be examined efficiently and without ambiguity by
planning certain combinations of spikes, but in this context the design
had the practical difficulty that inclusion of all treatments and
combinations of treatments at all levels would have been costly in terms
of separate treatments. In the incomplete design used here, certain
combinations of treatments were dropped to reduce labor costs, retaining
only the marginal treatments of nitrogen and phosphorus used alone along
with the diagonal set of combinations, as shown below. As a result,
with four levels each of nitrogen and phosphorus (counting in each case
the 0 level of each for unfortified natural water) the cost of each assay,
per replication, was reduced from 16 to 10 bottles (plus medium control).
The design used was as follows:
NOP3 - - N3P3
NOP2 - N2P2
NOPT N1P1
NOPO N1PO N2PO N3PO
Here, NOPO represents the zero level of addition for both nitrogen and
phosphorus, NOP1 the zero level for nitrogen and the first level for
phosphorus, etc.
The 10 combinations of treatments listed above, along with the
replicated trial of the growth medium alone allowed a testing of treat-
ments with ten degrees of freedom. A very large number of possible
contrasts could have been examined here, even though some of the tradi-
tional factorial contrasts had been confounded in the incomplete design.
It was not possible to find a set of orthogonal contrasts that included
all the comparisons of possible interest and that did not include
contrasts of no interest. Therefore, seven contrasts (not all mutually
orthogonal) were set up to examine what seemed to be the important
130
-------
biological questions. For each contrast the same error term was used,
based upon the variability among replicated flasks treated alike.
In the present context the term "contrast" refers to a comparison
between algal responses to treatments. When a contrast has a positive
numerical value it means that the first element (response) of the compari-
son exceeded the second; a negative contrast indicates that the first
element was less than the second.
The contrasts used in the analysis concentrated on detecting the
average change from addition of the three levels of nitrogen, for
example, rather than testing for a linear response to the series of
levels. This choice was made for two reasons. First, testing for a
linear response would require that the ambient level of nutrient be
determined, in each case to be added to the spike values, with confusion
from the fact that at least a small part of the measured ambient level
had been removed with the live cells. Second, if factors other than
nitrogen or phosphorus were potentially limiting, then response to nitro-
gen and phosphorus might not be linear, but rather, addition above some
certain level might only establish a plateau reflecting the other
limitations.
The following definitions of the contrasts identify the elements of
each comparison or contrast first in words, then in symbols. A positive
contrast is assumed in each case.
Medium vs. River Water: Response with medium exceeds that with
river water alone.
Medium vs. N3P3: Response with medium exceeds that with river water
plus N3P3.
N_ (N vs. River Water): Response from adding N to river water exceeds
the response with river water alone (average for 3 levels).
NP-P (N after P): Response from adding N and P to river water
exceedsTthat from adding only P to river water (average for 3 levels).
P (P vs. River Water): Response from adding P to river water exceeds
the response with river water alone (average for 3 levels).
NP-N (P after N): Response from adding P and N to river water
exceichTthat from adding only N to river water (average for 3 levels).
Interaction: Contrast when adding N and P together to river water
exceeds the sum of the contrasts when adding N and P separately to river
water (average for 3 levels) (this is a difference in contrasts and may
be described several ways with perhaps the easiest to understand being
that from the contrast when adding N and P together we subtract the sum
of the contrast when adding N separately plus the contrast when adding P
separately, where each of the three contrasts is measured as the excess
131
-------
of the response with the addition over the response with river water
alone).
Calculation of contrasts from mean values for treatments (in symbols):
Medium vs. River Hater: M - NOPO
.Medium vs. N3P3: M - N3P3
(N1PO + N2PO + N3PO - 3*NOPO)/3
(N1P1 + N2P2 + N3P3 - NOP1 - NOP2 - NOP3)/3
(NOP1 + NOP2 + NOP3 - 3*NOPO)/3
(N1P1 + N2P2 + N3P3 - N1PO - N2PO - N3PO)/3
(N1P1 + N2P2 + N3P3 - N1PO - N2PO - N3PO -
NOP1 - NOP2 - NOP3 + 3*NOPO)/3.
Statistical analyses of a number of the parameters were carried out
under logarithmic transformation both to stabilize the variance and to more
nearly normalize the data. Logarithms to base 10 were used. As one
consequence the tabulated mean values in some cases are stated both as
arithmetic and as geometric means, the first because it is a more
accustomed method of presentation for most readers and the second because
the statistical analyses based on log-transformed data are best repre-
sented by the geometric mean. As a second consequence of use of the log
transformation, the slope where a log-transformed variable is regressed
on distance downstream (as in the covariance analyses used here) may be
read in the antilog form as proportional change per unit distance.
Statistical analyses were carried out as two-way analyses of variance,
simple regression, analysis of variance of factorial contrasts for the
algal assays, and covariance analyses of a number of variables with
reference to changes up and down stream. Standard procedures were employed
for these analyses (Snedecor and Cochran 1967) being implemented through
the general linear models procedure of SAS (Barr et al. 1976).
The analysis of covariance was used to explore the dynamics of these
two streams; this analysis allowed an examination of changes both in
space (up and down stream) and in time (from visit to visit). The covari-
able used was the approximate distance in kilometers from the origin of
the stream for .each of four subsites chosen so that the almost simultane-
ous records at each could be interpreted as following the water mass as
it changed in moving down stream. The subsites were 20, 22, 24, and 25
for stream F-3 and 1, 3, 4, and 6, for stream P-10. The covariance analy-
sis consisted of fitting three linear models in succession to the data
for each stream (Snedecor and Cochran 1967). First it tested whether
there was evidence of difference in slopes (of regressions on distance)
132
-------
among the data subsets for visits, second, whether the visit mean values
differed when adjusted for the covariable, and third, whether a single
regression fitted the mass of data significantly better than the mean
value ignoring the covariable of distance. Carrying out the covariance
analyses with log-transformed values of the dependent variable allowed a
statement of slopes as proportional changes per kilometer of stream.
RESULTS
In total, 226 species of 108 genera of algae were recorded in 211
collections made at 12 subsites, 6 on each of the two streams, F-3 and
P-10. The number of genera recorded per sub site visit averaged 12 and
ranged from 3 to 27.
The taxonomic distribution of these species is shown in Table 35 along
with the cell density (cells per ml) and species total volume (cubic
microns per ml, or as biomass, 10~6 mg per liter) averaged over all
collections. Table 36 summarizes this information and shows as propor-
tions the distribution of species according to Class, with the
contribution of each Class to the average total cell number and biomass
per collection. The Cyanophyta dominated numerically (making up 57
percent of the cell number) while the Chrysophyta dominated in biomass
(65 percent). The relative dominance pattern among algal divisions
changed seasonally, however. In stream F-10, for example (Table 37)
the Chrysophyta had the highest relative cell volume during cooler
months while Cryptophyta were more dominant in the warmer months, with
tne other divisions dominant in the fall.
Most of the species were recorded only rarely. Of the 226 species
(plus a group of unidentified spores counted as one "species") 85 (37
percent) were found in only a single collection and 175 (77 percent) in
10 collections or fewer. The most common species, Cryptomonas er_osa.
was recorded in 162 collections, followed by the unidentified spores in
157 and Achnanthes sp in 155. Only 3 other species were recorded in
more than half (106) of the collections; these were Nitzschia sp (134),
Navicula sp (127) and Eunotia sp (107).
Most of the numbers and the biomass of these algal populations were
accounted for by a small list of species. There were only 16 species
each of which constituted one percent or more of cell numbers and there
were only 16 species where each had one percent or more of the total
biomass. Table 38 lists the 5 most numerous species (which constituted
50 percent of the total cell number) and the 5 species with the most
biomass (these made up 48 percent of the total biomass). Oscillator!a
subtillissima, a blue-green, had the highest total cell count, and
Frustulia spT a diatom, accounted for the largest fraction of the total
cell volume. Two species were on both lists.
In ceil volume these species ranged over 6 orders of magnitude, from
Merismopedia tenuissima, a blue-green with an average volume of 3 cubic
133
-------
TABLE 35. TAXONOMIC DISTRIBUTION OF ALGAL SPECIES ACCORDING TO DIVISION
AND CLASS, AS RECORDED IN 211 COLLECTIONS IN STREAMS F-3 AND
P-10, SHOWING FOR EACH SPECIES THE NUMBER OF COLLECTIONS IN
WHICH RECORDED, THE MEAN CELL COUNT AND MEAN BIOMASS PER
COLLECTION
Genus and Species
DIVISION CYANOPHYTA, CLASS
Agmenellum quadriduplica
Anabaena circinalis
Anabaena wi sconsinense
Anacystl s cyanaea
Anacysti s firma
Anacystis incerta
Aphanizomenon flos-aquae
Calothrix sp
Chroococcus limneticus
Coccochlon's sp
Coccochloris stagnina
Coelosphaerium naegelianum
Gomphosphaeria lacustris
Hapalosiphon sp
Marssoniella elegans
Merismopedia punctata
Merismopedia tenuissima
Oscillatori a geminata
Oscillatoria sp
Oscillatoria subtilissima
Phormidium sp
Plectonema battersii
Raphidiopsis curvata
Rhabdoderma lineare
Spirulina laxa
Unknown
Times
recorded
CYANOPHYCEAE
2
17
1
2
7
3
2
1
4
1
1
7
2
1
1
2
1
10
11
2
54
2
3
2
1
1
Mean
cell
count
(cells/ml)
1.83
4.56
1.41
5.44
16.09
19.96
1.23
7.18
0.66
0.01
18.83
3.40
1.08
0.06
1.51
4.18
1.85
4.17
5.16
202.74
43.43
0.82
1.01
0.13
0.52
0.71
Mean
biomass
(lO'6 mg/1)
5.5
455.0
305.4
91.2
289.5
359.2
1,401.0
776 ..1
5.9
0.1
339.0
681.1
22.8
5.7
16.6
87.6
5.6
451.5
557.1
21,896.2
738.5
14.0
191.3
3.3
31.9
282.1
134
-------
TABLE 35. (continued)
Genus and Species Times Mean
recorded cell
count
(cells/ml)
DIVISION CHLOROPHYTA, CLASS CHLOROPHYCEAE
Ankistrodesmus contorta
Ankistrodesmus convolutus
Ankistrodesmus falcatus
Ankistrodesmus falcatus var. mirabilis
Ankistrodesmus quarternatus
Asterococcus limneticus
Carteria fritschii
Characium rostratum
Chlamydomonas botryopara
Chlamydomonas botrys
Chlamydomonas sp
Chlorogonium spirale
Closteriopsis longissima
Closterium peracerosum
Closterium dianea
Coccomonas orbicularis
Coccomonas sp
Coelastrum microporum
Cosmarium speciosum
Cosmocladium pusillum
Crucigenia crucifera
Crucigenia quadrata
Crucigenia^ sp
Crucigenia tetrapedia
Dictyosphaerium pulchellum
Euastrum affine
Eudorina elegans
Gloeocystis botryoides
1
11
5
16
1
1
7
2
1
3
102
1
8
1
5
10
1
1
1
1
5
3
1
10
1
2
1
1
0.01
0.29
0.10
0.70
0.03
0.50
0.09
0.02
0.12
0.40
5.95
0.01
0.12
0.08
0.41
0.64
0.07
0.11
0.04
0.34
0.97
0.14
0.08
1.18
0.03
0.02
0.46
0.15
Mean
biomass
(10-6 mg/1)
0.5
13.5
1.4
13.1
2.3
173.7
118.7
5.4
7.7
208.2
3,103.1
1.5
100.0
858.8
2,758.3
263.3
29.6
33.1
83.6
140.1
12.5
4.0
2.0
28.1
1.9
8.0
32.4
39.2
135
-------
TABLE 35. (continued)
Genus and Species Times
recorded
Gloeocystis sp
Gloeocystis vesiculosa
Golenkinia radiata
Gonatozygon kinahana
Hyalotheca dissi liens
Kirchneriella bibrianum
Kirchneriella lunaris
Kirchneriella obesa
Mephrocytium agardhiamum
Monoraphidium contortum
Hougeotia sp
Nephrocytium agardhianum
Palmodictyon viride
Pandorina morum
Pediastrum tetras var. tetraodon
Scenedesmus acuminatus
Scenedesmus alternans
Scenedesmus bijuga
Scenedesmus denticulatus
Scenedesmus dimorphus
Scenedesmus ecornis
Scenedesmus quadricauda
Scenedesmus quadricauda var. alternans
Scenedesmus arcuatus
Selenastrum bibrianum
Selenastrum westii
Staurastrum orbicularis
Staurastrum cerastes
Staurastrum turgescens
1
1
}
1
2
2
4
1
1
8
2
1
3
7
1
1
1
8
3
1
2
12
2
14
1
14
6
11
1
Mean
cell
count
(cells/ml)
0.06
0.06
0.06
0.01
0.25
0.05
0.41
0.01
1.18
0.17
0.55
5.22
8.88
6.64
0.16
0.02
0.01
0.33
0.11
0.19
0.10
0.92
0.06
0.44
<0.01
0.43
1.66
0.12
<0.01
Mean
biomass
(10-6 mg/1)
14.8
14.3
41.6
38.0
1,412.3
2.6
23.0
0.6
39.8
8.3
5,325.1
177.4
266.3
1,135.0
6.4
3.4
3.4
20.2
6.8
26.0
6.0
117.4
3.2
30.7
0.3
17.0
3,355.0
464.4
88.9
136
-------
TABLE 35. (continued)
Genus and Species
Tetraedron sp
Tetrastrum glabrum
Tetrastrum staurogeniaforme
Treubaria crassispina
Ulothrix sp
Westell a botryoides
DIVISION CHRYSOPHYTA, CLASS
Achnanthes deflexa
Achnanthes lanceolata
Achnanthes lemmermanii
Achnanthes linearis
Achnanthes microcephala
Achnanthes minutissima
Achnanthes sp
Amphi pleura sp
Amphora oval is
Amphora sp
Bacillaria paradoxa
Caloneis alpestris
Caloneis lewisii
Times
recorded
1
1
3
1
1
4
BACILLARIOPHYCEAE
5
1
2
1
1
2
155
1
1
9
1
1
4
Caloneis lewisii var. inflata 1
Ceratoneis arcus
Cocconeis sp
Cyclotella glomerata
Cyclotella meneqhioriania
Cyclotella sp
Cymatopleura solea
Cymbella naviculi forma
Cymbella sp
1
16
3
1
12
1
1
46
Mean
cell
count
(cells/ml)
0.03
0.03
0.46
<0.01
0.32
0.21
0.97
<0.01
0.02
0.01
0.05
0.19
10.62
0.09
<0.01
0.20
<0.01
<0.01
0.03
0.05
0.08
0.20
0.26
0.17
0.73
0.02
0.09
0.76
Mean
biomass
(10-6 mg/1)
2.9
0.5
4.2
1.2
79.7
23.4
148.6
2.1
7.0
1.4
7.6
19.1
4,202.5
47.3
75.2
988.0
0.7
13.0
47.8
251.1
269.5
486.0
8.7
198.0
23.9
85.8
676.8
490.6
137
-------
TABLE 35.
Genus and Species
Cymbella tumida
Cymbella ventricosa
Diatoma vulgare
Diploneis oblongella
Epithemia sp
Eunotia arcus
Eunotia exigua
Eunotia pectinalis
Eunotia pectinalis var. minor
Eunotia sp
Fragilaria sp
Frustulia rhomboides
Frustulia sp
Gomphonema acuminatum var. coronata
Gomphonema gracillima
Gomphonema olivaceum
Gomphonema parvulum
Gomphonema sp
Mastigloia elliptica var. danseii
Melosira distans
Melosira distans var. alpisena
Melosira granulata
Melosira granulata var. angustissima
Helosira italica
Melosira italica var. tenuissima
iMelosira sp
Melosira varians
Meridion circulare
Meridion circulare var. constrictium
(continued)
Times
recorded
1
16
17
1
2
1
13
1
1
107
16
10
47
1
1
21
2
29
1
2
3
5
1
8
21
1
5
66
5
Mean
cell
count
(cells/ml)
0.02
0.89
1.47
0.01
0.10
0.01
1.34
0.01
0.10
6.02
0.75
0.15
0.76
0.03
0.17
1.76
0.02
1.43
0.01
0.08
0.09
0.14
0.04
0.98
1.05
0.69
0.23
3.04
0.09
Mean
biomass
(10'6 mg/1)
258.8
419.4
1,107.5
20.5
72.6
19.8
453.0
24.3
157.8
8,819.5
752.7
2,176.5
77,007.0
13.4
71.8
734.7
9.8
3,100.7
7.1
12.8
9.5
87.3
9.0
2,234.3
196.1
893.1
709.6
1,370.5
40.3
138
-------
TABLE 35.
Genus and Species
Micro si phona potomos
Navicula cryptocephala
Navicula graciloides
Navicula hambergii
Navicula pupila
Navicula pupila var. rectiosa
Navicula rhynchocephala
Navicula sp
Navicula viridula
Neidium affine
Neidium sp
Nitzschia acicularis
Nitzschia filaformis
Nitzschia ignorata
Nitzschia linearis
Nitzschia palea
Nitzschia parvula
Nitzschia recta
Nitzschia sp
Nitzschia thermal is
Nitzschia thermal is var. minor
Pinnularia sp
Pinnularia subcapitata
Rhizosolenia eriensis
Stauroneis phoeni center
Surirella ovata
Surirella sp
Synedra incisa
Synedra sp
(continued)
Times
recorded
2
1
4
1
2
2
2
127
1
9
88
13
4
4
1
39
3
3
134
10
3
49
2
2
6
2
56
2
23
Mean
cell
count
(cells/ml)
0.76
0.08
0.06
<0.01
0.06
0.05
0.18
5.13
0.01
0.34
2.89
0.46
0.04
0.04
0.03
5.82
0.09
0.14
7.19
0.20
0.07
1.98
0.18
0.02
0.26
0.01
2.01
0.02
0.55
Mean
biomass
(10~6 mg/1)
76.5
40.6
44.4
2.6
110.2
56.1
394.1
2,516.1
6.4
1,508.8
13,202.1
189.9
16.6
178.0
11.1
4,112.9
95.8
57.1
299.2
207.0
63.4
8,735.0
905.7
14.1
2,836.1
35.7
6,347.7
18.2.
1,672.0
139
-------
TABLE 35. (continued)
Genus and Species
Synedra ulna
label 1 aria fenestrata
label! aria sp
Unknown
DIVISION CHRYSOPHYTA, CLASS
Dinobryon acuminatum
Dinobryon ravaricum
Dinobryon divergens
Dinobryon sertularia
Dinobryon sp
Mallomonas sp
Mallomonas tonsurata
Mo n o c h ry s i s ap h a nas t e r
Rhizochrysis sp
Stichococcus bacillaris
Synura carol ini ana
Synura sp
Syn u ra sp ha g n i co 1 a
Synura spinosa
Synura uvella
Tribonema sp
Unknown
DIVISION EUGLENOPHYTA, CLASS
Eug_lena_ acus
Euglena ejinenbejrg_ii_
Euglena gracillima
E!ug_lena_ proxima
Euglena sp
Times
recorded
25
1
52
44
CHRYSOPHYCEAE
1
2
1
10
5
9
1
1
2
1
5
43
10
24
5
1
1
EUGLENOPHYCEAE
3
1
1
5
44
Mean
cell
count
(cells/ml)
2.54
0.01
1.61
8.22
0.01
0.15
0.01
0.27
0.06
0.12
<0.01
0.23
0.01
0.03
7.92
5.91
0.30
23.99
7.34
0.96
0.16
0.13
0.01
0.04
0.26
1.51
Mean
biomass
(10~6 mg/1)
8,904.7
15.8
1,890.4
10,028.5
5.9
70.1
1.2
130.3
25.4
34.3
2.0
158.6
7.8
2.4
3,681.4
2,736.1
234.6
53,917.7
3,415.0
2,201.0
62.8
784.5
25.1
117.5
953.7
2,626.4
140
-------
TABLE 35.
Genus and Species
Lepocinclis ovum
Lepocinclis sp
Phacus acuminatus
Phacus longicauda
Phacus pleuronectes
Phacus sp
Phacus tortus
Strobomonas sp
Trachelomonas gibberosa
Trachelomonas hispida
Trachelomonas horrida
Trachelomonas oblonga
Trachelomonas robusta
Trachelomonas sp
Trachelomonas volvocina
(continued)
Times
recorded
1
8
1
3
2
14
8
12
1
9
3
13
22
41
5
DIVISION PYRRHOPHYTA, CLASS DINOPHYCEAE
Ceratium hirundinella 1
Glenodinium penardi forme
Glenodinium sp
Gymnodinium neglectum
Peri dim' urn inconspicuum
Peri dini urn sp
1
1
1
6
28
DIVISION CRYPTOPHYTA, CLASS CRYPTOPHYCEAE
Chroomonas acuta 27
Chroomonas coerulea
Chroomonas sp
Cryptomonas erosa
Crvutomonas erosa var. reflexa
4
14
162
17
Mean
cell
count
(cells/ml)
0.11
0.41
0.01
0.08
0.13
0.26
0.20
0.57
0.05
0.36
0.08
0.74
0.43
0.98
0.11
0.01
0.01
0.04
0.05
0.20
0.43
9.49
0.82
0.26
16.57
0.93
Mean
biomass
(10-6 mg/1)
107.6
144.9
35.3
950.5
1,437.2
360.6
.1,180.5
5,762.3
301.2
1,513.3
153.4
229.2
1,543.4
2,846.2
56.2
1,572.0
233.7
662.2
212.0
355.4
7,245.1
996.5
58.8
26.6
11,585.7
644.3
141
-------
TABLE 35. (continued)
Genus and Species
Cryptomonas ovata
Cryptomonas reflexa
Cryptomonas sp
Cyanomonas americana
Times
recorded
77
7
34
16
Mean
cell
count
(cells/ml)
4.74
0.19
1.14
0.34
Mean
biomass
(10~6 mg/1)
10,684.6
135.9
513.2
228.9
DIVISION CHLOROMONADOPHYTA, CLASS CHLOROMONADOPHYCEAE
Gonyostomum depressum 1 0.01 23.3
DIVISION UNKNOWN, CLASS UNKNOWN*
Unknown 157 58.64 23,471.3
* Unidentifiable small green spheres, mostly spores, probably of numerous
species.
TABLE 36. TAXONOMIC DISTRIBUTION OF ALGAL SPECIES COLLECTED IN THIS STUDY,
SHOWING THE PROPORTION EACH CLASS CONSTITUTES OF THE TOTAL
NUMBERS AND BIOMASS
Division and Class
(common name)
Cyanophyta, Cyanophyceae
(blue-green algae)
Chlorophyta, Chlorophyceae
(green algae)
Chrysophyta, Bacillariophyceae
(diatoms)
Chrysophyta, Chrysophyceae
(yellow-green algae)
Euglenophyta, Euglenophyceae
(euglenoids)
Genera
20
36
31
8
5
Species
26
63
84
17
20
Proportion of
total by
numbers biomass
0.566 0.079
0.069 0.056
0.125 0.469
0.077 0.180
0.010 0.057
142
-------
TABLE 36. (continued)
Division and Class Genera Species
(common name)
Pyrrhophyta, Dinophyceae 4 6
(dinoflagellates)
Cryptophyta, Cryptophyceae 3 9
(naked flagellates)
Chloromonadophyta, Chloromonadophyceae 1 1
(chloromonads)
Unknown, Unknown*
Total 108 226
Proportion of
total by
numbers biomass
0.001 0.028
0.056 0.067
<0.001 <0.001
0.095 0.063
* Unidentifiable small green spheres, mostly spores, probably of numerous
species.
TABLE 37. RELATIVE CONCENTRATION OF ALGAL BIOMASS AMONG THE DIVISIONS
CHRYSOPHYTA, CRYPTOPHYTA AND ALL OTHER DIVISIONS FOR STREAM
P-10 THROUGHOUT THE STUDY
Visit no.
1
2
3
4
5
6
7
9
12
Date
3 Dec 74
19 Dec 74
6 Jan 75
15 Jan 75
31 Jan 75
20 Feb 75
7 Mar 75
27 Mar 75
14 Apr 75
Proporti
Chrysophyta
0.153
0.111
0.280
0.357
0.731
0.416
0.539
0.503
0.456
on of Biomass
Cryptophyta
0.252
0,222
0.252
0.145
0.122
0.190
0.090
0.061
0.225
by Division
Other Divisions
0.595
0.667
0.468
0.498
0.147
0.394
0.371
0.436
0.319
143
-------
TABLE 37. (continued)
Visit no.
13
16
17
19
21
23
26
28
30
32
33
Date
18 Apr 75
9 July 75
21 July 75
15 Aug 75
7 Sept 75
5 Oct 75
23 Nov 75
20 Dec 75
18 Jan 76
11 Feb 76
5 Mar 76
Proportion of Biomass
Chrysophyta Cryptophyta
0.368
0.122
0.164
0.238
0.124
0.114
0.274
0.749
0.783
0.856
0.639
0.197
0.279
0.741
0.100
0.232
0.000
0.081
0.126
0.032
0.002
0.075
by Division
Other Divisions
0.435
0.599
0.095
0.662
0.644
0.856
0.645
0.125
0.185
0.142
0.286
TABLE 38. THE 5 ALGAL SPECIES MOST ABUNDANT BY CELL NUMBER AND THE 5 MOST
ABUNDANT BY BIOMASS IN THIS STUDY, SHOWING PROPORTION OF TOTAL
CELL NUMBERS AND TOTAL BIOMASS ACCOUNTED FOR BY EACH OF THESE
SPECIES
Species
Mean per sample Species
cell proportion
number biomass of total
(cells/ml)(10~6 mg/1) cell
number biomass
Common name
Five highest by cell number
Oscillatoria subtillissima
Phormidium sp
Synura spinosa
Anacystis incerta
203
43
24
20
21 ,896
738
53,918
359
0.330
0.070
0.039
0.032
0.059
0.002
0.146
0.001
blue-green
blue-green
chrysophite
blue-green
144
-------
TABLE 38. (continued)
Species Mean per sample Species
cell proportion
number biomass of total
(cells/ml )(1(T6 mg/1) cell
number biomass
Common name
Coccochloris stagnina
Total
19
339 0.031 0.001 blue-green
0.502 0.209
Five highest by biomass
Frustulia bp
Synura spinosa
Qscillatoria subtil!is si ma
Neidium sp
Cryptomonas erosa
Total
Total of all species
col1ected
0.76 77,007
24 53,918
203 21,896
2.9 13,202
16 11,586
615 369,716
0.001 0.208 diatom
0.039 O.H6 chrysophite
0.330 0.059 blue-green
0.005 0.036 diatom
0.027 0.031 cryptophyte
0.402 0.480
An important component of the algae was a complex of unidentifiable small
green spheres, mostly spores, probably of numerous species; the entity
would rank second in numbers and third in biomass, above.
microns, to Ceratium hirundinella. a dinoflagellate with a volume estimated
as 110,657 cubic microns per cell. The distribution of species according
to cell volume is shown in Table 39 both according to quartiles of numbers
of species and by order of magnitude of cell volume. Species with small
cell volume predominated, with half being of less than 500 cubic microns.
The total cell count (all species) averaged 615 over all 211 collections
and ranged from 8 to 8,400 per ml. Total cell biomass averaged 0.37 and
ranged from 0.01 to 17 mg per liter; 11 values exceeded 1 mg per liter.
Using the data of Nichols (1976) the abundance of algae represented in
these collections corresponded to an average chlorophyll a_content of
approximately 2 with a range of 0.04 to 80 micrograms per liter.
145
-------
TABLE 39. DISTRIBUTION OF 227* ALGAL SPECIES ACCORDING TO MEAN CELL VOLUME,
BY QUARTILE AND BY ORDER OF MAGNITUDE OF CELL VOLUME, SHOWING
ALSO THE PROPORTIONAL DISTRIBUTION OF TOTAL CELL NUMBER AND
TOTAL BIOMASS
Number ProportToh VoTurne range Proportion of Proportion of
of of species (cubic microns) algal cell algal biomass
species numbers
By quarttles of species
1st
quartile
2nd
quartile
3rd
quartile
4th
quartile
By order
56
57
57
57
of magnitude
4
47
98
64
12
2
0.247
0.251
0.251
0.251
of cell
0.017
0.207
0.432
0.282
0.053
0.009
3-102
103-448
450-1 ,764
1,785-110,567
volume
1-10
10-100
100-1,000
1,000-10,000
10,000-100,000
over 100,000
0.250
0.524
0.141
0.085
0.008
0.240
0.633
0.115
0.003
0.001
0.010
0.158
0.177
0.655
<0.001
0.010
0.260
0.457
0.061
0.212
* Including as a "species" the group of unidentifiable spores.
The detailed quantitative aspects of the total algal community are
presented for each of the two streams for a subset of data which includes
all the visits when all of four selected subsites were represented on
each stream. These subsites were selected to record a continuing flow
of the same mass of water. For the piedmont stream F-3, these subsites
were numbered 1, 3, 4, 6, proceeding downstream, with a tributary entering
between 3 and 4. For coastal plain stream P-10, the subsites were
numbered 20, 22, 24 and 25, proceeding downstream, with a tributary
entering between 22 and 24.
146
-------
Variability in time and in space are shown in tables of mean values
for these two stream systems. For parameters of the algal populations
and for most chemical parameters, each subsite mean was based on 19 visits
for piedmont stream F-3 and on 13 visits for coastal plain stream P-10.
In both streams, each visit mean was based upon the 4 subsite values.
Both arithmetic and geometric means are presented.
Mean values showing spatial variation in flow and in parameters of
the algal population are presented in Table 40 for stream F-3 and Table
41 for stream P-10. A two-way analysis of variance of log-transformed
data was used to test differences between subsites in each stream for
flow, concentration of cell number and biomass, and the measurements of
diversity and evenness. In both streams, the subsites differed in flow
in a highly significant manner, increasing downstream as should be
expected. With cell number and biomass concentration, sites differed
significantly in stream P-10, but not in stream F-3. For diversity and
evenness there were no significant differences among sites in either
stream. These same tables also show values for transport of algal
numbers and biomass which generally follow those for flow, increasing
downstream.
The same type of information is presented for 5 chemical parameters
in Table 42 for stream F-3 and Table 43 for stream P-10. Again, analysis
of variance tests were made for differences between subsites with refer-
ence to concentrations but not for transport. Statistically significant
differences between subsites were found for nitrate nitrogen for both
streams, for ammonia nitrogen for stream F-3, and for chloride in stream
P-10. No significant differences were found in either stream for con-
centrations of Kjeldahl nitrogen or orthophosphate. In most but not all
cases, transport increased downstream in the general pattern of flow.
Temporal variation among visit means is shown for piedmont stream
F-3 in Tables 44 and 45 and for coastal plain stream P-10 in Tables 46
and 47 for algal numbers, biomass, and diversity, and for flow and con-
centration of nitrate nitrogen and chloride. Two-way analyses of
variance showed that for all these variables in both streams there were
statistically significant differences among mean values for visits,
indicating synchronous changes in parameter values across the 4 subsites.
The occurrence of an algal bloom is an important aspect of temporal
variation in biomass. According to one rule of thumb, a biomass of 10
cubic microns per ml (1 mg per liter, wet weight) is near the m nimum
level for a bloom. As mentioned above, in the tota of 211 collections,
there were 11 where the biomass exceeded 1 mg per liter and thus quali-
fied as a bloom by this definition. Of these 11 samples, 6 were
collected on 2 visits to stream P-10 (on 23 November 1975 and on 5 March
1976) and 5 were collected on 1 visit to stream F-3 (on 9 March 1976).
The ubsets of data for each stream where 4 subsites were represented
at each visit provide information for approximating the expected
frequency of bloom conditions in the streams Values (in og}fl units)
for mean biomass concentration and the standard error of visit mean
147
-------
values (each based upon 4 subsites) for stream F-3 are -1.0566 ± 0.3962,
and for stream P-10, -0.7232 ± 0.4703 (values derived from covariance
analysis of biomass data, detailed below). Assuming a log normal distri-
bution of the values of biomass concentration, these values for mean and
standard error allow estimation of the frequency with which visit geometric
means (each for 4 subsites) will exceed the criterion for a bloom (1 mg
per liter). In these terms, these data suggest that with stream F-3
the criterion will be exceeded about one-half of one percent of the time,
while with stream P-10 blooms will be present about 6 percent of the time.
Empirically, the subsets of data show one instance in 19 visits for
stream F-3 and two cases in 13 visits for stream P-10. Similar calcula-
tions of the expectation of single values exceeding the 1 mg per liter
criterion yield about 2 percent for stream F-3 and about 10 percent for
stream P-10. The reasoning leading to these approximations makes no
allowance for possible serial correlations and is thus based upon use of
a sampling interval similar to that used here.
TABLE 40. SPATIAL VARIATION OF FLOW AND ALGAL PARAMETERS IN PIEDMONT
STREAM F-3; ARITHMETIC AND GEOMETRIC MEAN VALUES AND
RESULTS OF TESTS OF THE STATISTICAL SIGNIFICANCE OF
SUBSITE DIFFERENCES FOR SELECTED VARIABLES
Type Means for 19 visits to subsites Statistical
of significance
Parameter mean 20 22 24 25 of subsite
differences
Distance from origin (km)
Flow (m3/sec)
arith.
geom.
Cell number
concentration arith.
(cells/ml) geom.
transport arith.
(109 cells/d) geom.
Biomass
concentration (mg/1) arith.
geom.
transport (kg/d) arith.
geom.
1.2
0.132
0.041
535
146
0.186
0.087
2.63
0.31
8.2
0.451
0.129
362
129
8.4
0.774
0.281
362
152
0.160
0.084
12.09
0.93
0.195
0.091
21.61
2.23
10.2
0.886
0.290
359
154
4,933 32,108 26,790 58,174
519 1,434 3,696 3,872
0.245
0.089
40.64
2.24
**
ns
ns
Diversity
arith. 2.40 2.56 2.30 2.29
ns
148
-------
Evenness
TABLE 40. (continued)
Type Means for 19 visits to subsites Statistical
of significance
Parameter mean 20 22 24 25 of subsite
differences
arith. 0.749 0.768 0.673 0.705
ns
** Statistical significance: p<_0.01.
ns Statistical significance: p > 0.05.
TABLE 41. SPATIAL VARIATION OF FLOW AND ALGAL PARAMETERS IN COASTAL PLAIN
STREAM P-10; ARITHMETIC AND GEOMETRIC MEAN VALUES AND RESULTS
OF TESTS OF THE STATISTICAL SIGNIFICANCE OF SUBSITE DIFFERENCES
FOR SELECTED VARIABLES
Parameter
Type Means of 13 visits to subsites Statistical
of significance
mean 1 3 4 6 of subsite
differences
Distance from origin (km)
Flow (m3/sec)
Cell number
concentration
(cells/ml)
transport
(109 cells/d)
Biomass
concentration (mg/1 )
transport (kg/d)
Diversity
Evenness
arith.
qeom.
arith.
geom.
arith.
geom.
arith.
geom.
arith.
geom.
arith.
arith.
0.2
0.010
0.002
424
222
191
49
1.65
0.33
0.39
0.073
2.75
0.766
2.8
0.013
0.005
3,033
232
848
95
0.52
0.22
0.26
0.089
2.79
0.759
3.0
0.024
0.012
272
144
333
152
0.24
0.16
0.34
0.169
2.92
0.795
6.0
0.104
0.072
204
95
983
594
0.82
0.11
3.64
0.694
2.35
0.736
**
*
**
ns
ns
* Statistical significance:
** Statistical significance:
ns Statistical significance:
0.01 < p <_ 0.05,
p <_ 0.01.
p > 0.05.
149
-------
TABLE 42. SPATIAL VARIATION OF CHEMICAL PARAMETERS OF PIEDMONT STREAM
F-3; ARITHMETIC AND GEOMETRIC MEAN VALUES AND RESULTS OF
TESTS OF THE STATISTICAL SIGNIFICANCE OF SUBSITE
DIFFERENCES FOR SELECTED VARIABLES
Parameter
Type Means of
of
mean 20
Distance from origin (km)
1
.2
19
visits to subsitest Statistical
significance
22 24 25 of subsite
differences
8.2
8.4
10.2
Nitrate nitrogen
concentration
transport
(ppm)
(kg/d)
arith.
geom.
arith.
geom.
0.
0.
0.
0.
0632
0478
57
17
0.
0.
1.
0.
0589
0359
70
41
0.
0.
3.
0.
0553
0351
10
87
0.
0.
3.
0.
0547
0356 *
02
91
Kjeldahl nitrogen
concentration
transport
(ppm)
(kg/d)
arith.
geom.
arith.
geom.
1.
0.
14.
2.
24
74
0
0
0.
0.
41.
5.
99
64
3
2
1.
0.
61.
14.
06
74
5
6
1.
0.
87.
22.
07
91 ns
8
8
Ammonia nitrogen
concentration
transport
Qrthophosphate
concentration
transport
Chloride
concentration
transport
(ppm)
(kg/d)
(ppm)
(kg/d)
(ppm)
(kg/d)
arith.
geom.
arith.
geom.
arith.
geom.
arith.
geom.
arith.
geom.
arith.
geom.
0.
0.
1.
0.
0.
0.
0.
0.
3.
3.
65
12
0800
0184
56
35
0637
0311
49
08
99
38
1
0.
0.
1.
0.
0.
0.
1.
0.
3.
3.
04
35
0143
0055
49
18
0637
0355
56
36
34
17
0.
0.
1.
0.
0.
0.
2.
0.
4.
3.
215
91
0128
0040
79
08
0842
0275
37
42
46
73
0.
0157
0.0056 *
2.
0.
0.
0.
2.
0.
5.
3.
99
79
0616
0218 ns
81
31
95
65 ns
213
92
* Statistical significance: 0.01 < p < 0.05.
ns Statistical significance: p > 0.05."
t Exception: ammonia had 9 visits.
150
-------
TABLE 43. SPATIAL VARIATION OF CHEMICAL PARAMETERS IN COASTAL PLAIN STREAM
P-10; ARITHMETIC AND GEOMETRIC MEAN VALUES AND RESULTS OF TESTS
OF THE STATISTICAL SIGNIFICANCE OF SUBSITE DIFFERENCES FOR
SELECTED VARIABLES
Type Means of 1
Parameter
of
mean
Distance from origin (km)
1
0.
2
3
3
2.
visits to subsitest
8
4
3.
0
6
6.
0
Statistical
significance
of subsite
differences
Nitrate nitrogen
concentration
transport
(ppm)
(kg/d)
arith.
geom.
arith.
geom.
0.
0.
0.
0.
293
137
32
02
0.
0.
0.
0.
430
277
36
11
0.
0.
0.
0.
450
334
74
35
0.
0.
2.
0.
174
115
11
72
*
Kjeldahl nitrogen
concentration
transport
(ppm)
(kg/d)
arith.
geom.
arith.
geom.
1.
0.
0.
0.
39
96
68
13
1.
0.
1.
0.
30
91
32
29
1.
0.
2.
1.
14
96
44
01
1.
0.
9.
3.
04
62
27
49
ns
Ammonia nitrogen
concentration
transport
Orthophosphate
concentration
transport
Chloride
concentration
transport
* Statistical
** Statistical
ns Statistical
t Exception:
(ppm)
(kg/d)
(ppm)
(kg/d)
(ppm)
(kg/d)
sigm
signi
signi
arith.
geom.
arith.
geom.
arith.
geom.
arith.
geom.
arith.
geom.
arith.
geom.
ficance:
ficance:
ficance:
ammonia had 6
0.
0.
0.
0.
0.
0.
0.
0.
11.
9.
6.
2.
180
012
002
001
106
092
06
02
,2
,7
.9
.0
0.
0.
0.
0.
0.
0.
0.
0.
268
035
096
017
108
,097
,11
,04
12.1
10.9
8.9
4.5
0.01 < p <
0.05.
0.
0.
0.
0.
0.
0.
0.
0.
188
025
073
022
132
120
30
13
11.8
10.8
16.6
11.3
i
0.
0.
1.
0.
0.
0.
1.
0.
8.
7.
64.
49.
436
067
719
340
131
097
27
61
1
,9
,2
,7
ns
ns
**
p < 0.01.
p > 0.05.
visits.
151
-------
TABLE 44. TEMPORAL VARIATION OF ALGAL NUMBERS, BIOMASS, AND DIVERSITY IN
PIEDMONT STREAM F-3; ARITHMETIC AND GEOMETRIC MEANS FOR 13
VISITS. EACH TO 4 SUBSITES
Visit
1
2
4
5
6
7
9
10
12
14
15
18
20
21
26
27
30
31
34
Date
25 Nov 74
10 Dec 74
23 Jan 75
26 Jan 75
11 Feb 75
28 Feb 75
1 Apr 75
8 Apr 75
10 May 75
11 Jun 75
27 Jun 75
25 Jul 75
4 Sep 75
15 Sep 75
27 Nov 75
9 Dec 75
14 Jan 76
26 Jan 76
9 Mar 76
Algal nos.
(cells/ml)
arith. geom.
mean mean
50
552
140
52
65
111
254
193
88
55
20
182
421
1,707
783
259
60
400
2,297
34
293
102
49
62
110
229
158
80
52
17
146
401
644
668
244
56
190
2,064
Algal biomass
(mg/1)
arith. geom.
mean mean
0.050
0.080
0.059
0.048
0.085
0.118
0.222
0.116
0.067
0.036
0.025
0.273
0.164
0.253
0.318
0.152
0.038
0.180
1.445
0.036
0.080
0.041
0.046
0.074
0.099
0.206
0.101
0.064
0.032
0.022
0.127
0.089
0.161
0.272
0.148
0.030
0.092
1.063
Algal
diversity
arith.
mean
1.49
1.77
1.81
2.79
3.04
3.22
2.88
2.50
2.86
2.69
1.82
2.45
1.90
2.25
0.48
3.42
3.19
2.05
2.75
152
-------
TABLE 45. TEMPORAL VARIATION OF FLOW AND CONCENTRATION OF NITRATE NITROGEN
AND CHLORIDE IN PIbDMONT STREAM F-3; ARITHMETIC AND GEOMETRIC
MEANS FOR 13 VISITS, EACH TO 4 SUBSITES
Visit
1
2
4
5
6
7
9
10
12
14
15
18
20
21
26
27
30
31
34
Date
25 Nov 74
10 Dec 74
23 Jan 75
26 Jan 75
11 Feb 75
28 Feb 75
1 Apr 75
8 Apr 75
10 flay 75
11 Jun 75
27 Jun 75
25 Jul 75
4 Sep 75
15 Sep 75
27 Nov 75
9 Dec 75
14 Jan 76
26 Jan 76
9 Mar 76
Flow (m3/sec)
arith. geom.
mean mean
0.05
0.22
0.34
2.00
0.31
0.25
0.34
0.15
0.1 1
0.02
0.01
0.25
0.02
0.03
0.16
0.89
0.30
3.54
1.67
0.04
0.17
0.29
1.34
0.26
U.17
0.24
0.10
0.08
0.01
0.01
0.21
0.02
0.02
0.13
0.67
0.24
2.93
1.36
Nitrate
nitrogen
concentration
(ppm)
arith. geom.
mean mean
0.002
0.035
0.065
0.052
0.058
0.020
0.018
0.012
0.052
0.148
0.238
O.U85
0.038
0.052
0.038
0.042
0.048
0.025
0.075
0.002
0.034
0.063
0.052
0.054
0.017
0.016
0.012
0.052
0.146
0.237
0.085
0.037
0.048
0.037
0.042
0.047
0.024
0.074
Chloride
concentration
(ppm)
arith. qeom.
mean mean
3.6
3.8
3.2
2.9
3.0
2.8
2.3
2.5
2.4
8.5
17.7
3.0
2.9
4.2
4.2
3.5
3.9
2.2
7.7
3.5
3.8
3.2
2.9
3.0
2.7
2.2
2.5
2.4
6.4
9.2
3.0
2.9
4.0
4.1
3.5
3.7
2.2
5.6
153
-------
TABLE 46. TEMPORAL VARIATION OF ALGAL NUMBERS, BIOMASS, AND DIVERSITY IN
COASTAL PLAIN STREAM P-10; ARITHMETIC AND GEOMETRIC MEANS FOR
13 VISITS, EACH TO 4 SUBS1TES
Visit
2
3
4
5
7
9
12
13
17
26
28
30
33
Date
19 Dec 74
6 Jan 75
15 Jan 75
31 Jan 75
7 Mar 75
27 Mar 75
14 Apr 75
18 Apr 75
21 Jul 75
23 Mov 75
20 Dec 75
18 Jan 76
5 Mar 76
Algal nos.
(cells/ml)
arith. geom.
mean mean
120
184
109
124
326
130
88
133
143
10,096
50
103
1,180
112
172
97
88
300
108
86
103
135
2,622
41
93
894
Algal biomass
(mg/D
arith. geom.
mean mean
0.134
0.203
0.096
0.184
0.321
0.089
0.131
0.229
0.159
1.830
0.125
0.106
6.887
0.118
0.197
0.088
0.124
0.292
0.080
0.118
0.136
0.152
1.287
0.081
0.098
2.650
Algal
diversity
arith.
mean
2.19
2.43
3.25
2.94
3.27
2.97
3.27
3.16
1.22
1.64
3.03
3.41
2.35
TABLE 47, TEMPORAL VARIATION OF FLOW AND CONCENTRATION OF NITRATE NITROGEN
AND CHLORIDE IN COASTAL PLAIN STREAM P-10; ARITHMETIC AND
GEOMETRIC MEANS FOR 13 VISITS, EACH TO 4 SUBSITES
Visit Date
2 19 Dec 74
3 6 Jan 75
Flow
arith
mean
0.014
0.040
(m^/sec)
. geom.
mean
0.008
0.026
Nitrate
nitrogen
concentration
(ppm)
arith. geom.
mean mean
0.462 0.384
0.582 0.582
Chloride
concentration
(ppm)
arith. geom.
mean mean
15.9 15.2
9.6 9.5
154
-------
TABLE 47. (continued)
Visit
4
5
7
9
12
13
17
26
28
30
33
Date
15
31
7
27
14
18
21
23
20
18
5
Jan
Jan
Mar
Mar
Apr
Apr
Jul
Nov
Dec
Jan
Mar
75
75
75
75
75
75
75
75
75
76
76
Flow (n
arith.
mean
0
0
0
0
0
0
.175
.045
.037
.042
.009
.007
0.037
0
0
0
0
.006
.022
.043
.015
i^/sec)
geom.
mean
0.122
0.033
0.022
0.028
0.001
0.002
0.003
0.002
0.010
0.027
0.006
Nitrate
nitrogen
concentration
(ppm)
arith. geom.
mean mean
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
245
285
412
250
110
320
285
010
660
388
368
0.242
0.273
0.370
0.219
0.083
0.190
0.252
0.010
0,565
0.362
0.062
Chloride
concentration
(ppm)
arith. geom.
mean mean
6.1
6.1
8.6
5.5
19.1
14.7
6.3
16.0
9.8
8.7
14.0
6.0
6.1
8.5
5.5
17.9
14.3
6.3
15.4
9.7
8.6
13.5
The percentage that the algae carried of the total organic carbon, total
nitrogen and total phosphorus in the two streams was approximated for a
number of samples; mean values are shown in Table 48. These results
TABLE 48. ALGAE-CARRIED NUTRIENTS; CALCULATED CONTENT OF ALGAL BIOMASS AS
MEAN AND MAXIMUM PERCENTAGE OF THE TOTAL ORGANIC CARBON, TOTAL
NITROGEN, AND TOTAL PHOSPHORUS DETERMINED BY CHEMICAL ANALYSIS
OF WATER SAMPLES FROM TWO STREAMS
Site Minimum
sample
size
Coastal plain stream P-10 56
Piedmont stream F-3 90
Algal content as percent of total
Organic Total Total
carbon nitrogen phosphorus
mean max. mean max. mean max
1.1 15 3.6 78 3.9 79
0.3 2 0.6 55 1.0 9
155
-------
indicate that on the average the algae contained only small percentages
of the nutrients available in the stream. On the other hand, there were
10 values in all exceeding 10 percent (including the maximum values in
the table); these reflect the rare occurrence of high biomass values.
The standing crop of numbers of algae observed in samples from these
two streams was compared in an approximate way with the maximum standing
crop obtained with algal assay of the same water (using a single test
species in the laboratory as described later). The comparison was made
by calculating the ratio: observed standing crop in numbers divided by
the assay maximum standing crop (calculated as numbers in the manner
described in the methods section). These ratios were extremely low;
for piedmont stream F-3 they averaged 0.0002 (maximum 0.002) and for
coastal plain stream P-10 they averaged 0.0005 (maximum 0.009). These
values indicate that in these streams under natural conditions the
standing crop lies far below the potential (usually by a factor of
several thousand) inasfar as this may be judged by such an approximate
comparison to the single laboratory test species.
Simple regression analysis was used to explore the relationship
between the algae and environmental parameters. A separate line was
fitted for each of 11 subsites for regressions of algal parameters
(log-transformed cell numbers and biomass, untransformed diversity and
evenness) on nitrate nitrogen, Kjeldahl nitrogen, total phosphorus,
chloride, temperature C, pH, day length and dissolved oxygen as deter-
mined for the water samples. An analysis was carried out for data from
each of subsites 1, 2, 3, 4, and 6 of stream P-10 and from subsites 20,
22, 23, 24. 25, 28 of stream F-3; each regression was based upon the
records of from 9 to 16 visits. Thus 344 analyses were examined (sub-
site 28 lacked complete data on pH and dissolved oxygen). Results of
these regression analyses are presented in Table 49 by listing those
subsites where a statistically significant (p £0.05) regression was
found. This table shows that there were scattered occurrences of
significant relationships (in total, 21 out of the 344; one would
expect about 17 on a basis of chance alone). All that may be said for
this regression analysis is that it provides no evidence of consistent
relationships between the algal and chemical parameters measured; one
must recall, however, that sample sizes were low and the ranges of the
independent variables were restricted.
Regression analysis was also used to test for a possible temporal
relationship in variation in algal numbers and diversity in the two
streams using the mean values listed in Tables 44 and 46. In both
streams the apparent relationship was negative, indicating a lower
diversity with higher algal population but in neither stream was the
relationship significant at the 0.05 level in spite of the coincidence
in coastal stream P-10 of very high algal numbers and a low diversity
value.
The two stream systems were compared directly on the basis of algal
numbers, biomass, diversity and evenness, using an analysis of variance
156
-------
TABLE 49. TESTS FOR REGRESSION OF ALGAL PARAMETERS (CELL NUMBER, BIOMASS,
DIVERSITY, EVENNESS) ON SELECTED ENVIRONMENTAL PARAMETERS
(CALCULATED INDEPENDENTLY AT 11 SUBSITES ON TWO STREAMS*)
LISTING SUBSITES WHERE REGRESSION WAS STATISTICALLY
SIGNIFICANT (p <. 0.05)
Independent variable
Nitrate nitrogen
Kjeldahl nitrogen
Total phosphate
Chloride
Temperature C
Day length
PH
Dissolved oxygen
Algal
cell
number
2, 6
24
25, 28
2, 25
22
Dependent variable
Algal Diversity
biomass
2, 6 2
28
2
25
22 24, 25
25
Evenness
2
25
25
* A total of 344 regressions were calculated; the 21 "significant" is
close to the 17.2 expected by chance.
of data from 4-week sampling periods when both streams were represented.
Only for biomass was the difference between the two streams statistically
significant (p < 0.001) and here there was also a highly significant
(p < 0.001) interaction of stream and period. This means that while
stream P-10 generally exceeded F-3 in biomass, the difference was of a
different order in different periods or might be reversed in some. In
all tests, period differences were highly significant indicating that
all four parameters fluctuated in time more or less in synchrony in the
two streams, subject to the significant interaction of stream and period
with numbers, biomass, and diversity.
The analysis of covariance was used to examine the changes in flow
and algal and chemical parameters in space (up and down stream) and in
time (from visit to visit). Overall mean values have been presented
above to show the spatial variation in flow and algal and chemical
parameters in Tables 40, 41, 42, and 43 with temporal variation illus-
trated for part of the variables in Tables 44, 45, 46, and 47.
Results of the covariance analysis for piedmont stream F-3 are
shown in Table 50 and for coastal plain stream P-10 are shown in Table 51.
157
-------
TABLE 50. RELATIONSHIP OF FLOW, ALGAL, AND CHEMICAL PARAMETERS TO DISTANCE
DOWNSTREAM FOR PIEDMONT STREAM F-3, BASED UPON COVARIANCE
ANALYSIS OF DATA FROM 4 SUBSITES AND 19 VISITS (EXCEPT
FOR AMMONIA WITH 7 VISITS); DATA LOG-TRANSFORMED
(EXCEPT FOR DIVERSITY AND EVENNESS)
Statistical
Parameter differences
Flow
Algal parameters
Numbers
concentration
transport
Biomass
concentration
transport
Diversity
Evenness
Chemical parameters
Nitrate nitrogen
concentration
transport
Kjeldahl nitrogen
concentration
transport
Ammonia nitrogen
concentration
transport
Orthophosphate
concentration
transport
Chloride
concentration (log)
transport (log)
slopes
ns
ns
ns
**
ns
ns
ns
**
ns
ns
ns
**
ns
ns
ns
ns
ns
significance of
among visits:
means
**
**
**
**
**
**
**
**
**
**
**
**
**
**
**
**
**
Common slope
fitted through
visit means
0.0948**
0.0007ns
0.0955**
0.0011ns
0.0959**
-0.0064ns
-0.0047ns
-0.0158**
0.0780**
0.0045ns
0.1079**
-0.0680**
0.0064ns
-0.0106ns
0.0805
0.0028ns
0.0976**
Antiloq of
common
slope
1.24
1.00
1.24
1.00
1.25
0.98
0.99
0.96
1.20
1.01
1.28
0.86
1.01
1.02
1.20
1.01
1.25
* Statistical significance: 0.01 < p <_ 0.05.
** Statistical significance: p < 0.01.
ns Statistical significance: p > 0.05.
158
-------
TABLE 51. RELATIONSHIP OF FLOW, ALGAL, AND CHEMICAL PARAMETERS TO DISTANCE
DOWNSTREAM FOR COASTAL PLAIN STREAM P-10, BASED UPON COVARIANCE
ANALYSIS OF DATA FROM 4 SUBSITES WITH 13 VISITS (EXCEPT
FOR AMMONIA WITH 5 VISITS); DATA LOG-TRANSFORMED
(EXCEPT FOR DIVERSITY AND EVENNESS)
Statistical significance of Common slope
Parameter differences among visits: fitted through
Flow
Algal parameters
Numbers
concentration
transport
Biomass
concentration
transport
Diversity
Evenness
Chemical parameters
Nitrate nitrogen
concentration
transport
Kjeldahl nitrogen
concentration
transport
Ammonia nitrogen
concentration
transport
Orthophosphate
concentration
transport
Chloride
concentration
transport
slopes
ns
ns
ns
ns
ns
ns
ns
ns
ns
*
*
**
ns
ns
*
ns
ns
means
**
**
**
**
**
**
**
**
**
**
*
**
**
**
**
**
**
visit means
0.2553**
-0.0658**
0.1895**
-0.0821**
0.1732**
-0.0710ns
-0.0052ns
-0.0175ns
0.2618**
-0.0328ns
0.2474**
0.1263*
0.4519**
0.0040ns
0.2654**
-0.0159**
0.2455**
Antilog of
common
slope
1.80
0.86
1.55
0.83
1.49
0.85
0.99
0.96
1.83
0.93
1.77
1.34
2.83
1.01
1.84
0.96
1.76
* Statistical significance: 0.01 < p <. 0.05.
** Statistical significance: p <.0.01.
ns Statistical significance: p > 0.05.
159
-------
First, for each parameter the table shows the results of testing differ-
ences among slopes for the different visits. Second, the table indicates
the statistical significance of differences among visit mean values.
Third is stated the slope for a set of lines fitted, one through each
visit mean, with common slope (a value which has meaning only if there
is no evidence of difference between slopes for the different visits)
and the indication of whether this common slope fits significantly
better than the general regression fitted to the entire set of data.
Finally, each table shows the antilog of the common slope value; this
shows in the form of a multiplier the proportional change accompanied by
moving a kilometer downstream. Again, this value has meaning only if
there is no evidence of difference among visits with regard to slope.
There were significant differences among visit means for all param-
eters examined by covariance in the two streams. In most cases there
was no evidence that the sets of data for the different means differed
in slope; exceptions to this generality were with Kjeldahl nitrogen
concentration and transport, ammonia nitrogen concentration and ortho-
phosphate transport in stream P-10 and biomass concentration, nitrate
nitrogen concentration and ammonia nitrogen concentration in stream
F-3. This means that, aside from the exceptions, the values for the
parameters varied proportionately from visit to visit, maintaining the
same relationship up and down stream.
Flow increased in going downstream, with the common slope equivalent
to a gain of 24 percent per kilometer in stream F-3 and 80 percent per
kilometer in stream P-10. In concentration of algal numbers and biomass
and the chemical constituents, the only significant changes in stream
F-3 were with nitrate nitrogen and ammonia nitrogen, both of which
decreased significantly at average rates of 4 percent and 14 percent per
kilometer respectively. In stream P-10, however, concentration of algal
numbers decreased at the rate of 14 percent per kilometer, biomass con-
centration at the rate of 17 percent per kilometer, and chloride
concentration at the rate of 4 percent per kilometer while in this stream
ammonia nitrogen increased at the rate of 34 percent per kilometer.
Other changes in both streams were nonsignificant.
Values for transport of algae and chemicals generally followed the
trend of flow, increasing downstream (in most cases significantly) but
at a rate reflecting both the changes in flow and in concentration. For
example, in stream P-10 where algal biomass concentration decreased when
going downstream at the rate of 17 percent per kilometer while the flow
was increasing at the rate of 80 percent per kilometer, the transport
increased at the rate of 49 percent. With ammonia nitrogen, in stream
F-3 the concentration decreased significantly and with an increase in
flow, resulted in a nonsignificant change in transport, while in stream
P-10 with ammonia nitrogen concentration increasing at a rate of 34
percent per kilometer and the flow increasing at the rate of 80 percent,
the combined effect was an increase of 183 percent per kilometer. But
recall that sample sizes were small with ammonia.
160
-------
Results from the algal assays comprise a large mass of data in spite
of the fact that sampling was on a reduced schedule, both in time and in
space and part of the results had to be discarded. For each assay there
were 10 treatment levels with every lot of bottles having its own medium
control. Each assay level was carried out with three (or two) replicated
bottles. Results are presented here for 83 assays; several from subsites
visited on an irregular basis were not used.
Results from the algal assays are summarized in Tables 52, 53, 54,
and 55 which show the statistical results in terms of whether the effects
were statistically significant. Mean values are not shown except for
the 10 mean values for several selected assays (Table 56) presented as
examples of statistically significant interactions.
Tables 52, 53, 54, and 55 record the statistical significance of
results of five factorial contrasts; a negative sign signals that results
of the particular contrast were negative. In these tables, N and P
indicate the mean effect of adding nitrogen or phosphorus to the water
being assayed, the contrast being between water with the three levels of
nutrient added and the natural water with no addition. Two other con-
trasts, NP-P and NP-N, test for the effect of adding one nutrient to
water already containing the other. The contrast labeled INT tests
whether addition of both nutrients together produced a greater effect
than that calculated as the sum of the effects of adding them separately.
Table 56 presents several examples of assay mean values selected to
illustrate an interaction in this context, and incidentally, cases of
negative effects. Greater detail on these contrasts has been presented
in the section on methods.
With each lot of bottles run through the laboratory assay, a repli-
cated set of bottles was used with the laboratory medium and the test
alga as a control. Two statistical comparisons were made in each
analysis, contrasting the results for the medium with those for the
unenhanced natural water, and with those ^natural water plus the
highest levels of nitrogen and phosphorus added (N3P3). These contrasts
were positive with one exception; in stream P-10 visit 16 subsite 6 the
relationship was reversed, probably indicating a doubtful laboratory
result Except for this one, most differences were highly significant
(p < 0.01); there were seven exceptions with one difference significant
at the 0.05 level and the others nonsignificant. These results confirmed
laboratory control of the process.
In the broad view, using the 0.01 level for statistical significance,
about half (86) of the 166 assay results tabulated in these tables
^ _f _C J. *" -IJ—-1--I— — *•. * M t +•
indicate a significant effect
one or more negative values obscure the interpretation. signiican
effect was associated with addition of phosphorus in four cases, in three
of which the effect had negative values. Thus, over the whole, about
161
-------
TABLE 52. ALGAL ASSAY -- RESULTS OF STATISTICAL ANALYSIS FOR MAXIMUM
STANDING CROP FOR PIEDMONT STREAM F-3; N (OR P) REPRESENTS
THE MEAN EFFECT OF NITROGEN (OR PHOSPHORUS) ADDITION, NP-P
(OR NP-N) THE MEAN ADDED EFFECT OF NITROGEN AND PHOSPHORUS
TOGETHER OVER PHOSPHORUS (OR NITROGEN) ALONE, AND INT
REPRESENTS INTERACTION, OR THE EXCESS EFFECT OF ADDING
NITROGEN AND PHOSPHORUS TOGETHER OVER THE SUM OF THE
SEPARATE EFFECTS
Subsite
Visit
Visit
Visit
Visit
18
20
22
23
24
25
28
20
20
22
23
24
25
28
21
20
22
23
24
25
28
23
20
22
23
24
N NP-P NP-N
25 July
**
**
**
**
**
**
1975
**
**
**
**
**
**
4 September
**
**
**
**
**
**
**
**
**
**
**
**
15 September
**
**
**
**
**
**
**
**
ns
**
**
**
**
**
**
**
**
*
1975
ns
*
ns
ns
_**
-ns
1975
-ns
ns
-ns
ns
_*
_**
P
-ns
-ns
-ns
-ns
-ns
ns
-ns
ns
ns
-ns
ns
-ns
**
-ns
ns
ns
ns
**
INT
**
**
**
**
**
-ns
ns
ns
-ns
ns
_*
ns
_**
*
-ns
-ns
_*
_**
14 October 1975
**
**
** _*
**
*
*
**
ns
_**
_**
-ns
ns
**
-ns
-ns
-ns
_**
_*
ns
1
Subsite
Visit 26
20
22
23
24
25
28
Visit 27
20
22
23
24
25
28
Visit 30
20
22
23
24
25
28
Visit 31
20
22
23
24
62
N NP-P
27 November
ns ns
ns *
ns ns
ns ns
ns ns
ns *
9 December
ns *
** **
** **
ns **
** **
ns *
14 January
ns ns
* **
** **
** **
* **
26 January
-ns ns
** **
** **
** **
NP-N
1975
-ns
ns
-ns
-ns
-ns
ns
1975
-ns
-ns
-ns
-ns
-ns
-ns
1976
-ns
ns
ns
ns
ns
1976
-ns
ns
-ns
-ns
P
-ns
-ns
ns
ns
-ns
-ns
-ns
-ns
ns
_•*
-ns
-ns
ns
-ns
-ns
-ns
ns
-ns
ns
ns
-ns
INT
ns
ns
-ns
-ns
ns
ns
-ns
-ns
-ns
-ns
ns
ns
-ns
ns
ns
ns
ns
ns
-ns
-ns
ns
-------
TABLE 52. (continued)
Subsi
Visit
te
25
28
34
20
22
23
24
25
28
N NP-
ns
ns
9 March
**
**
**
**
**
**
ns
1
**
**
**
**
**
P NP-N
**
_**
976
ns
ns
-ns
ns
**
P
-ns
-ns
-ns
-ns
-ns
-ns
-ns
INT Subsite
* 25
ns 28
ns
ns
-ns
ns
ns
N NP-P NP-N P INT
** ** -ns ns -ns
ns ** ns -* *
* Statistically significant, 0.01 < p < 0.05.
** Statistically significant, p <_0.01.
- Negative sign indicates a negative effect (a reduction).
TABLE 53. ALGAL ASSAY — RESULTS OF STATISTICAL ANALYSIS FOR MAXIMUM
STANDING CROP FOR COASTAL PLAINS STREAM P-10; N (OR P)
REPRESENTS THE MEAN EFFECT OF NITROGEN (OR PHOSPHORUS)
ADDITION, NP-P (OR NP-N) THE MEAN ADDED EFFECT OF
NITROGEN AND PHOSPHORUS TOGETHER OVER PHOSPHORUS
(OR NITROGEN) ALONE, AND INT REPRESENTS INTERACTION,
OR THE EXCESS EFFECT OF ADDING NITROGEN AND PHOSPHORUS
TOGETHER OVER THE SUM OF THE SEPARATE EFFECTS
Subsite
Visit
Visit
16
1
2
3
4
6
19
1
2
3
N NP-P NP-N P
9 July 1975
* ** ns ns
** ** ns ns
** ** _|-|s -**
15 August 1975
INT Subsite
Visit 26
1
-ns 2
3
ns 4
* 6
Visit 28
1
2
3
N
23
-ns
ns
ns
ns
ns
20
ns
-ns
ns
NP-P
November
**
ns
*
*
*
December
**
ns
**
NP-N
1975
*
ns
ns
ns
ns
1975
*
-ns
*
P
_**
ns
ns
-ns
ns
-ns
_*
-ns
INI
**
-ns
-ns
ns
ns
ns
ns
ns
163
-------
TABLE 53. (continued)
Subsi
Visit
Visit
Visit
te N NP-P NP-N P
4
g ** ** ** **
21 7 September 1975
1
2
3
4
6 ** ** ns -ns
23 5 October 1975
1
2
3
4
6 * ns ns -ns
33 5 March 1976
1 ** ** ns _ns
2 ** ** ns -ns
3 ** ** * ns
4 ** ** _ns _ns
6 ** ** -ns -ns
INT Subsi te
4
ns 6
Visit 30
1
2
3
4
ns 6
Visit 32
1
2
3
4
-ns 6
ns
ns
ns
ns
ns
N
-ns
**
18
ns
*
ns
ns
**
11
ns
**
**
**
**
NP-P
ns
**
January 1
**
**
*
ns
**
February
**
**
**
**
**
NP-N
ns
-ns
976
*
ns
-ns
ns
ns
1976
**
ns
-ns
-ns
_*
P
-ns
ns
-ns
ns
-ns
ns
ns
ns
ns
_*
ns
-ns
INT
ns
-ns
ns
-ns
-ns
-ns
ns
**
ns
ns
-ns
-ns
* Statistically significant, 0.01 < p < 0.05.
** Statistically significant, p <.0.01.
- Negative sign indicates a negative effect (a reduction).
164
-------
TABLE 54. ALGAL ASSAY -- RESULTS OF STATISTICAL ANALYSIS FOR MAXIMUM
SPECIFIC GROWTH RATE FOR PIEDMONT STREAM F-3; N (OR P)
REPRESENTS THE MEAN EFFECT OF NITROGEN (OR PHOSPHORUS)
ADDITION, NP-P (OR NP-N) THE MEAN ADDED EFFECT OF NITROGEN
AND PHOSPHORUS TOGETHER OVER PHOSPHORUS (OR NITROGEN) ALONE,
AND INT REPRESENTS INTERACTION, OR THE EXCESS EFFECT OF ADDING
NITROGEN AND PHOSPHORUS TOGETHER OVER THE SUM OF THE SEPARATE EFFECTS
Sub site
Visit 18
20
22
23
24
25
28
Visit 20
20
22
23
24
25
28
Visit 21
20
22
23
24
25
28
Visit 23
20
22
23
24
25
28
N
25
**
**
**
**
**
**
NP-P
July 1975
**
**
**
**
**
**
4 September
**
ns
**
**
ns
**
15
**
**
**
**
**
**
14
ns
ns
-ns
-ns
**
**
**
**
ns
**
**
**
September
**
**
**
**
**
**
October 1
*
**
**
-ns
ns
**
NP-N
-ns
*
ns
**
**
ns
1975
_**
**
-ns
-ns
ns
-ns
1975
ns
ns
_*
ns
**
ns
975
-ns
-ns
-ns
-ns
_**
_**
P
ns
ns
_*
ns
ns
ns
_**
-ns
**
_**
_**
_*
*
-ns
ns
-ns
*
-ns
-ns
_**
_**
ns
-ns
_*
INT Subsite
Visit
-ns
ns
*
ns
**
ns
Visit
ns
**
_**
**
**
ns
Visit
-ns
*
-ns
ns
-ns
ns
Visit
-ns
*
*
-ns
_**
r.s
165
26
20
22
23
24
25
28
27
20
22
23
24
25
28
30
20
22
23
24
25
28
31
20
22
23
24
25
28
N
27
**
*
*
ns
ns
**
NP-P
NP-N
P
INT
November 1975
**
**
**
**
**
ns
9 December
ns
ns
-ns
ns
ns
-ns
14
**
**
**
**
**
26
-ns
**
-ns
**
ns
ns
ns
*
**
*
**
ns
January
**
**
**
**
**
January
ns
**
ns
-ns
**
-ns
ns
-ns
-ns
ns
ns
-ns
1975
-ns
ns
ns
-ns
*
-ns
1976
**
ns
ns
_*
ns
1976
ns
-ns
ns
-ns
-ns
-ns
*
-ns
-ns
_*
_**
*
-ns
ns
_*
-ns
-ns
-ns
_**
-ns
-ns
**
_*
-ns
ns
_*
*
_*
ns
-ns
-ns
ns
*
**
_*
-ns
ns
**
ns
ns
ns
**
*
ns
_**
• *
ns
-ns
ns
_*
ns
-ns
-------
TABLE 54. (continued)
Subsite
Visit
34
20
22
23
24
25
28
N NP-P
NP-N
P
INT Subsite N NP-P NP-N P INT
9 March 1976
* -ns
** **
** **
** **
* **
-ns
-ns
-ns
ns
*
ns
-ns
-ns
-ns
-ns
_*
ns
ns
ns
*
* Statistically significant, 0.01 < p < 0.05.
** Statistically significant, p <.0.01.
- Negative sign indicates a negative effect (a reduction),
TABLE 55. ALGAL ASSAY — RESULTS OF STATISTICAL ANALYSIS FOR MAXIMUM
SPECIFIC GROWTH RATE FOR COASTAL PLAINS STREAM P-10; N (OR
P) REPRESENTS THE MEAN EFFECT OF NITROGEN (OR PHOSPHORUS)
ADDITION, NP-P (OR NP-N) THE MEAN ADDED EFFECT OF NITROGEN
AND PHOSPHORUS TOGETHER OVER PHOSPHORUS (OR NITROGEN)
ALONE, AND INT REPRESENTS INTERACTION, OR THE EXCESS
EFFECT OF ADDING NITROGEN AND PHOSPHORUS TOGETHER
OVER THE SUM OF THE SEPARATE EFFECTS
Subsite
Visit
Visit
16
1
2
3
4
6
19
1
2
3
4
6
N NP-P NP-N P
9 July 1975
-ns ns ns -ns
ns * ns ns
-* ns -ns -**
15 August 1975
** ns -ns **
INT Subsite
Visit 26
1
ns 2
3
ns 4
* 6
Visit 28
1
2
3
4
-** 6
N
23
ns
ns
ns
ns
*
20
ns
ns
-ns
-ns
ns
NP-P
November
ns
-ns
ns
**
-ns
December
**
-ns
*
ns
ns
NP-N
1975
ns
_*
-ns
-ns
_**
1975
*
*
ns
**
ns
P
ns
ns
-ns
ns
ns
-ns
ns
-ns
ns
ns
INI
-ns
-ns
-ns
ns
_*
ns
-ns
ns
ns
-ns
166
-------
TABLE 55. (continued)
>ubsite N
P
Sub site
NP-P NP-N
INT
Visit 21 7 September 1975
1
2
3
4
6 ** ** -ns
Visit 23 5 October 1975
1
2
3
4
6 ** ** -ns
Visit 33 5 March 1976
Visit 30 18 January 1976
1 ns ns ns
2 ns -ns -ns
3 ns ** ns
4 ns -ns ns
ns
*
-ns
ns
ns
_*
**
**
-ns
Visit 32 11 February 1976
-ns
ns
-ns
_*
ns
-ns
_*
1
2
2
4
6
**
*
**
**
ns
**
-ns
_**
*
**
ns
-ns
_*
-ns
**
*
ns
**
-ns
-ns
-ns
-ns
_**
-ns
*
1
2
3
4
6
*
ns
-ns
ns
**
ns
**
ns
-ns
**
_*
ns
ns
ns
ns
ns
-ns
-ns
ns
ns
-ns
ns
ns
-ns
-ns
* Statistically significant, 0.01 < p < 0.05.
** Statistically significant, p <_0.01.
- Negative sign indicates a negative effect (a reduction).
TABLE 56. ALGAL ASSAY — SELECTED EXAMPLES FROM STREAMS F-3 AND P-10 WITH
STATISTICALLY SIGNIFICANT (p <.0.01) INTERACTIONS, SHOWING
MEANS FOR THE 10 TREATMENTS FOR MAXIMUM STANDING CROP AND
MAXIMUM SPECIFIC GROWTH RATE
Form used to display mean
values in this table
NOP3 N3P3
NOP2 N2P2
NOP1 N1P1
1/Standing crop; P-10 visit 26;
subsite 1; interaction = 0.113
0.153 0.285
0.154 0.197
0.154 0.194
167
-------
TABLE 56. (continued)
NOPO N1PO N2PO N3PO
Standing crop; M-3 visit 18;
subsite 23; interaction = 0.040
0.020 0.186
0.019 0.077
0.017 0.029
0.023 0.032 0.062 0.092
Growth rate; P-10 visit 19;
subsite 6; interaction = -0.374
1.064 1.419
0.980 1.059
1.057 0.971
0.728 1.115 1.182 1.357
2/
3/
0.226 0.156 0.164 0.232
Standing crop; M-3 visit 23;
subsite 22; interaction = -0.044
0.191 0.195
0.188 0.179
0.185 0.168
0.169 0.197 0.205 0.215
Growth rate; M-3 visit 26;
subsite 25; interaction = 0.177
0.588 0.847
0.530 0.679
0.485 0.666
0.702 0.634 0.706 0.822
I/
2/
3/
F
Also illustrates negative contrast; P = -0.072.
Also illustrates negative contrast; NP-N = -0.025.
Also illustrates negative contrast; P = -0.167.
half the tests showed a clear effect from addition of nitrogen and about
a quarter of the tests no effect from either nutrient.
Negative values were obtained in a number of the assays of stream
water. Most of these were nonsignificant and thus by definition were
within the range to be expected from the variability of the measurements,
but several were highly significant (p <.0.01). The incidence of these
significant negative values is shown as follows:
Parameter
Standing crop
Growth rate
Stream Number of assays N NP-P NP-N
INT
F-3
P-10
F-3
P-10
52
31
52
31
0
0
0
0
0
0
0
1
5
0
3
1
0
2
7
1
2
0
3
2
Each of these cases was examined for unusual data such as extremely high
values for NOPO. Three instances of these significant negative contrasts
168
-------
are included in Table 56 among the five examples of significant inter-
actions. Of the 20 cases of negative effects (excluding interactions)
all but one involved the addition of phosphorus, either alone or after
nitrogen. The exception involved the addition of nitrogen after
phosphorus.
In the 166 analyses there were 22 highly significant interactions,
meaning that the effect of phosphorus and nitrogen together was either
significantly greater or significantly less than the effect expected
from action of the nutrients separately. Table 56 exhibits five examples
of interactions selected to demonstrate the effect (calculation of an
interaction is explained in the methods section).
DISCUSSION
Working so far up into the head waters of these two small streams
with waters that had been exposed to biological exploitation for so
short a time, we may ask, first, why we found as many algae as we did,
and second, why we did not find more. Generality of our conclusions
must be limited because we worked with only two streams, one in the
Piedmont and the other in the Coastal Plain.
We found a fairly large and diversified flora, dominated by a few
species in numbers and biomass. Most species were rare, with over a
third appearing in only a single collection; this implies that further
collecting and examination of collections would record many more species.
Generally, algal abundance was very low compared to the assay-tested
capability of the water to produce a standing crop but occasionally the
abundance rose into the range of a possible bloom. Algal biomass was
typically one to two orders of magnitude less than that reported by
Stanley and Hobbie (1977) for the lower Chowan River, suggesting that
realization of some of the available growth potential may be a partial
explanation for the nuisance algal blooms reported there (Bond et al.
n.d.).
In one of these streams the number of algae per volume of water
decreased as the water moved downstream but at the same time with
increasing flow the transport of algae increased, implying that there
were substantial additions to the population mass between stations even
though population density was diluted. In the other stream total algal
transport also increased but there was no evidence of change in popula-
tion density. Thus the processes of population increase differed somewhat
between streams.
An important question is the source of these population increases, a
source capable of providing the pulse for a bloom. Although this study
provides no detailed answer, it is clear that only a few sources are
possible. The source most often credited for stream phytoplankton is
washout from periphyton living on the bottom of the stream; Swanson and
Bachman (1976) found a relationship between the export of phytoplankton
and the area of streambed above the sampling point. Another source of
169
-------
phytoplankton in these small drainage basins may be drainage from standing
water in ponds in fields and woods. A third possible source is from
runoff over the land surface where algae may grow under suitable condi-
tions. A final possible source of algal increase is reproduction of
cells carried by the flowing water, although the short residence times
in these streams prevent this from being an important factor. These
observations argue against significant growth of planktonic algae; thus
the algae observed in the stream must originate from other environments.
Similar conclusions have been drawn in recent reviews of stream ecology
(Whitton 1975, Hynes 1970). The observations also suggest that consid-
erable algal growth may occur downstream when conditions become favorable.
Algal populations fluctuated in time in a manner at least consistent
with the notion that the changes were of the same proportion in the
various parts of the stream examined. Some of the chemical parameters
behaved in the same manner, others did not and ammonia differed in
behavior between the two streams. The simple idea of dilution by ground
water, at first thought to be a possible explanation (Hayne et al. 1977)
cannot explain the changes; probably the interaction of a series of
processes is involved. It is not possible to distinguish between such
influences with the present data.
Except during the rare bloom conditions, the algae contained only
small percentages of the nitrogen, phosphorus and organic carbon trans-
ported by the streams. Studies of the nearby Pamlico River support this
conclusion (Harris and Hobbie 1974, Robbie et al. 1972), and data from
Nichols (1976) and Nichols and MacCrimmon (1975) on the Holland River
also indicate that only 5-6 percent of the total nitrogen and total
phosphorus are present in algal biomass.
Conceptually, there can be a very large number of environmental
variables with an influence on growth of phytoplankton, but it has been
found that in a given situation only a few are important (e.g., Walsh
1971). We found that in both streams inorganic nitrogen and phosphorus
were usually available for algal growth, averaging 0.38 mg per liter
nitrate plus ammonia and 0.14 mg per liter inorganic phosphorus. These
values greatly exceed the concentrations generally recognized as limiting
uptake of nutrients, i.e., 0.03 mg per liter nitrogen and 0.06 mg per
liter phosphorus (Eppley et al. 1969; Maclsaac and Dugdale 1969; Fuhs et
al. 1972). Because nitrogen and phosphorus were present in concentra-
tions sufficient to saturate uptake kinetics these nutrients could not
be limiting growth. Instead, we suggest that stream flushing rate and
amount of available light may have been important.
If the flushing rate is high enough, the phytoplankters will be
washed through a system before the population can reproduce appreciably
(Lund 1965). As flushing rate decreases and residence rate increases,
more algal growth can occur in a reach of given length and thus, with
other factors equal we may expect more phytoplankton in more slowly
moving streams with the potential for visible bloom directly proportional
to the residence, time. Both of these streams had high rates of flushing.
170
-------
The availability of enough light for photosynthesis is a second
environmental variable that may be important for controlling algal
growth in streams. Some streams flow under dense forest cover; suspended
sediment and water color limit light penetration in some. In any stream,
algal cells experience exponentially decreasing light with depth and
growth will depend upon the integrated light exposure of cells as they
rise and fall in the turbulence of flowing waters. The concept of criti-
cal mixing depth has been developed to describe this situation in oceanic
systems (Yentsch 1975); this is the depth of mixing at which the net
growth is zero because integrated photosynthesis (a function of depth)
is balanced by respiration (independent of depth). Something similar
must apply also in flowing streams.
The dynamics of flow, algal populations and chemical parameters were
studied here using probability sampling in time and the analysis of
covariance. Results suggest that this is a promising technique for the
study of small streams. The linear regression model used here provides
only an approximation to the measurement of change over a distance of
several kilometers. Probably change is not a smooth process with a
continuous rate for all variables as might be implied by the linear model.
The general idea of applying probability sampling to this problem need
not be confined to the linear regression model used here; it might well
be simplified to the comparison of successive pairs of points in a more
detailed analysis of dynamics.
We cannot generalize from these results to processes in small streams
in general, except to point out that these two streams differ in some
characteristics. Therefore, if the dynamics of small streams are to be
studied by choosing a random sample of streams and studying changes in
selected reaches, the sample sizes must be large enough to deal with the
variability we have demonstrated.
The algal assays provided two kinds of evidence about the water
quality of these small drainage basins. First, the populations of the
test algae developed in the unspiked natural water provided a measure
of the potential growth possible in this water, and furnished a basis
for comparison with the ambient algal populations present in the same
water. If the growth potential of these particular waters is the only
point of interest, a much simplified bottle assay method could be
devised to use only the unfortified natural waters to approximate the
maximum standing crop for comparison with the standing crop of indigenous
algae observed in the stream. Such a technique would provide estimates
both of the standing crop achieved in the stream, and the approximate
fraction this comprises of the maximum possible.
The second kind of information from the algal assays relates to
potential growth limitation and at best has meaning for interpreting
field data as applying to the hypothesized continuation of algal growth
in exactly these same waters until nitrogen or phosphorus or both have
been so reduced that low availability of the elements inhibits further
growth. In the small drainage basins where this work was done such
171
-------
hypothetical conditions would occur far downstream, perhaps in the main
portion of the river or in the estuary. The further assumption that
these waters would be unaltered in their travels except by the uptake
of nutrients by the algae further removes the interpretation of results
from direct application to the small streams. With these limitations in
mind, let us examine the results of the algal assays made here.
In the total of 166 values (83 assays, each calculated as maximum
standing crop and the maximum specific growth rate) three patterns of
results predominate, ignoring the presence or absence of a significant
interaction. The most frequent pattern (61/166) was for the effect of
N and of NP-P to both be significant with neither of the phosphorus
effects significant; this clearly indicated potential limitation by
nitrogen. The second pattern in frequency (44/166) was non-significance
in all four effects (N, NP-P, NP-N, P) indicating no potential limitation
by any of these elements and thus probably limitation by some other
entity. The third most frequent pattern (21/166) was for a non-
significant effect of N, a significant effect of NP-P, and lack of
significance for the two phosphorus effects. This may be classified
as limitation by nitrogen and near limitation by phosphorus with the
two nutrients only slightly unbalanced with respect to the requirements
of the test species. Addition of nitrogen under this condition may
have provided a slight increase in response but at a level unidentifiable
as statistically significant; the relative abundance of nitrogen and
phosphorus then shifted so that phosphorus became limiting with the
result that when it was added with nitrogen present the response was
clearly identifiable. This pattern is classified here as representing
potential nitrogen limitation. In the remaining cases (40/166) some of
the results are confusing and difficult to interpret; in about half the
presence of negative effects provided evidence of occasional limitation
through toxicity or growth suppression by nutrients or interacting
factors. There was a single clear case of potential phosphorus limita-
tion.
The two criteria of maximum specific growth rate and maximum standing
crop were used in each assay to estimate the potential nutrient limita-
tion. Under the 0.01 level of statistical significance used here,
conclusions would be identical as to limiting nutrients in about half the
assays and would disagree somewhat in the rest, although some of the dis-
agreement may be seen as one or the other assay being more sensitive in
the particular instance. Reviewing Tables 52, 53, 54, and 55 it may be
seen that consistency of results seems to have been more a matter of the
particular stream rather than whether standing crop or growth was used
as the criterion, with more variable and erratic results in coastal
plain stream P-10 under both criteria. This apparent difference may be
In this paragraph, the term "significant" refers to statistical signifv
cance with p <. 0.01.
172
-------
a matter of stream heterogeneity. Thus these empirical results do not
confirm the theoretical argument that standing crop should be a more
satisfactory criterion than growth rate. It is interesting, however,
that the range of increase of growth rate seems to have fallen generally
within the predicted two-fold limitation.
In comparing results of samples taken from different points on the
same stream during a single day there is a surprising amount of apparent
disagreement as to potential nutrient limitation. We suspect that the
apparent differences are reliable (although they were not tested statis-
tically) because the method seemed to provide reasonably good precision
within any one assay, that is, it was usually able to detect statistically
significant differences with the modest level of replication used. The
differences within streams on the same day, which were more apparent
in stream P-10, may represent unexpected heterogeneity in the water of a
single stream.
In summary of the algal assay method as used to appraise potential
nutrient limitation, we found it provided apparently reliable information
for a single assay while results were somewhat inconsistent for assays
of water taken from different parts of the same stream on the same day.
It is possible that the precision of the method has allowed the detection
of unsuspected heterogeneity in flowing streams, and if this is true,
then the apparent potential nutrient limitation, as determined from a
single water sample, may not be a reliable indicator for an entire stream,
even a small one. Further, the nature of this variation in small streams
in space and time must be studied in greater detail. But whatever the
variation, the concept of potential limitation by nutrients as determined
for algae in such small streams, seems to apply to a hypothetical condi-
tion far downstream and under the unrealistic assumption that the nature
of the waters will not be changed in moving there.
173
-------
SECTION 7
BIOCHEMICAL OXYGEN DEMAND STUDIES
Samples for biochemical oxygen demand (BOD) tests were collected during
the period from November 11, 1974 through July 21, 1975. The samples were
collected from the F3 and P10 sites. These sites were F3K, F3M, F3A, F3W,
F3X, P10A, P10B, P10K, P10W, and P10Y. The verification numbers were respect-
ively 20, 22, 23, 24, 25, 1, 2, 3, 4, and 6. For each sample the following
information was determined:
1. Five day BOD
2. Ultimate BOD
3. Oxygen uptake rate
This data was determined for the samples at three temperatures, 10 °C,
20 °C and 30 °C, and for duplicate sets of samples, one set containing an in-
hibitor to prevent nitrification and a second set without inhibitor.
The data was entered in the main data set using the following data set
name arrangement. BOD data names are seven characters long. The.first three
are BOD. The.fourth character will be a "5" if the value is a five-day read-
ing or an "L" if the value is the ultimate calculated value. The fifth char-
acter will be an "I" if the sample contained a nitrification inhibitor or a
"U" if no inhibitor was added. The sixth and seventh characters indicate the
temperature at which the sample was run, 10 °C, 20 °C, or 30 °C. Thus,
"BODL120" is the ultimate calculated value for an inhibited BOD sample run at
20 °C.
The oxygen uptake rate data set names consist of five characters. The
first two, "KE" indicate that the rate was calculated using log base e. The
third and fourth indicate temperature and the fifth indicates if an inhibitor
was used. Thus "KEIOU" is the data set name for oxygen uptake rates deter-
mined for uninhibited samples at 10 °C.
TESTING PROCEDURES
The BOD was run using the membrane electrode method. A YSI (Yellow
Springs Instruments) membrane electrode was chosen due to ease of operation
and availability. To insure accuracy and reliability, the electrode tech-
nique was tested against the standard Winkler-azide modification, a wet chemi-
cal test for determining dissolved oxygen, outlined in Standard Methods' for
the Examination of Water and Wastewater (APHA, 1971). Electrodes, calibrated
daily by the Winkler-azide modification, never varied by more than two-tenths
of one part per million of dissolved oxygen from the chemical method (Figure
16.
174
-------
O1
o
o
oo
O WINKLER AZIDE MODIFICATION
DYSI MEMBRANE ELECTRODE
D
TIME (DAYS)
Figure 16. Biochemical oxygen demand comparison at various times for the winkler
azide modification vs. YSI membrane electrode method.
-------
The samples were run at three different temperatures and checked for dis-
solved oxygen concentration approximately every other day by the electrode
method. In order to reduce the relative errors, duplicates were set up at
each temperature: 10.°C, 20 °C, and 30 °C. Testing ten or more times in a
twenty-day period provided a complete sequence of data.
The samples were collected and delivered to the Sanitary Engineering Lab-
oratories at North Carolina State University a day or more after the actual
sampling. Preservation was accomplished by "icing" the samples. Samples were
run both with and without nitrification suppressor. The effects of several
nitrification suppressors were evaluated. The inhibitor marketed by the Hach
Company, No. 2533, proved to be the most effective suppressor of nitrogenous
oxygen demand. The Hach inhibitor was also tested for carbonaceous demand
and inhibition of cabonaceous demand. To evaluate the Hach inhibitor, a
solution containing a small BOD and no ammonia was prepared by distillation,
ion-exchange, and carbon absorption. Three BOD test sets were prepared using
this solution. The first set contained only the solution, the second had the
normal dose of Hach inhibitor added to the solution, and the third set con-
tained an inorganic ion-exchange media, permutit, to insure no nitrification.
The result showed no measurable differences in oxygen uptake between the three
sets; thus, verifying the Hach inhibitor as solely a nitrification suppressor.
Analysis and Results
The discussion in this section will be confined to the values observed at
20 °C. As expected, the values found at 10 °C were lower than those found at
20 °C and those at 30 °C was higher. The same patterns were found at each
temperature.
The ultimate BOD values found during this study were generally low, with
the highest being 14.0 mg/1 for samples without nitrification inhibitors and
13.2 mg/1 with the inhibitors added. The minimum values were 0.65 mg/1 and
0.55 mg/1 respectively. The lower values are questionable, however, as the
test requires that at least two mg/1 of oxygen be used and these low tests
did not use anywhere near this amount. The uninhibited values were, as ex-
pected, in general larger than the inhibited values. Very few BOD values
were observed in excess of 10 mg/1.
The oxygen uptake rates ranged from less than 0.01 per day to 0.20 per
day. Differences between the rates found for the uninhibited and inhibited
samples could not be discerned immediately.
Trends could not be readily observed due to the short-time period over
which the data was collected. However, it appears that the BOD values are
greater during the summer than during the winter. The overall low values of
BOD and oxygen uptake rates indicate the material causing the BOD was rela-
tively stable before entering the streams. This was probably due to the fact
that a large majority of the samples were collected during low flow periods
and thus any materials entering the streams and contributing to BOD were rela-
tively stale and thus stable.
176
-------
An attempt was made to use the QUAL II model for these stream sections
which were sampled for BOD and oxygen uptake rates. However, due to the low
oxygen uptake rates and short-stream sections sampled the model could not be
used effectively.
The data were then analyzed using the general linear models (GLM) procedure
of the SAS system. The first series of tests examined whether there was a
linear regression of BOD on chemical oxygen demand (COD) and of BOD on total
organic carbon (TOC). Both the uninhibited ultimate BOD and the ultimate in-
hibited BOD were used.
This analysis indicated a statistically significant (P = 0.025 or less,
see Table 57) linear regression on both COD and TOC for both inhibited and
uninhibited BOD. However, the low values of R-square and the high coefficients
of variation for all tests suggest that the correlation between these vari-
ables is of little practical significance.
The BOD data was also analyzed to see if there were consistent differences
between sampling sites and between visits using the SAS procedure GLM for a
two-way analysis of variance. The BOD again was the dependent variable with
site (verficiation) and visit used as classes. The model appeared to have a
reasonable fit with all values of R-square greater than 0.4; however, the
values found for the coefficients of variation were also large (all greater
than 45 percent), indicating a scattering of the data (Table 58).
An examination of the F-test data for the model indicates that differ-
ences in BOD values were detected with all models having F-values with prob-
abilities of finding larger values of F less than 0.015. The analysis in-
dicated statistically significant differences in BOD among visits (across
all sites), but no significant differences among sites (across all visits).
177
-------
TABLE 57. RELATIONSHIP OF BOD TO COD AND BOD TO TOG
00
Variable
Dependent
BODL120
BODLU20
BODL120
BODLU20
Dependent
Variable
BODL120
BODLU20
BOD5120
BOD5U20
BODG120
BODGU20
Independent
CODC
CODC
TOCC
TOCC
TABLE
F PR>F
2.05 0.0142
2.42 0.0029
2.05 0.0009
3.19 0.0001
3.79 0.0001
2.63 0.0012
F value
5.25
12.13
14.10
20.63
a greater F value
0.0251
0.0008
0.0003
0.0001
Coefficient
R-square of variation
0.072 84.16
0.1425 72.83
0.1531 75.25
0.1972 66.13
58. ANALYSIS OF BOD VARIATION
R-square
0.473
0.487
0.412
0.524
0.6228
0.509
Coefficient
of variation Verif
(%) F
70.71 1.12
62.01 1.44
147.5 0.72
114.14 0.74
52.92 1.22
45.662 1.08
Source
PR>F
0.3611
0.1780
0.8196
0.7983
0.2988
0.3893
Visit
F PR>F
1.93 0.0466
2.76 0.0038
2.41 0.0012
4.16 0.0001
4.05 0.0001
3.38 0.0006
-------
REFERENCES
Anon. A Million Random Digits With 100,000 Normal Deviates. The Rand Corp.,
Santa Monica. The Free Press, Glencoe, Illinois, 1955.
Anon. Algal Assay Procedure; Bottle Test. National Eutrophication Research
Program, EPA, Corvallis, Oregon, 1971. 82 pp.
APHA. Standard Methods for the Analyses of Water and Wastewater. 13 ed
American Public Health Association, Inc., New York, 1971.
Balducci, C. Phosphate Uptake Kinetics in the Chowan River, North Carolina.
M.S. Thesis, North Carolina State University, Botany Dept., Raleiqh,
North Carolina, 1977.
Barr, A. J., J. H. Goodnight, J. P. Sail, and J. T. Helwig. A User's Guide
to SAS 76. SAS Inst. Inc., Raleigh, North Carolina, 1976. 329 pp.
Bond, S., G. Cook, and D. Howells. Summary Report the Chowan River Project.
North Carolina Water Resources Research Institute, Raleigh, North
Carolina, pp. 1-36.
Buchanan, T. J., and W. 0. Somers. Discharge Measurements at Gaging Sta-
tions. Applications of Hydraulic Book 3, Department of Interior, U.S.
Geological Survey, Washington, D.C., 1969.
Busby,.M. W. Annual Runoff in the Conterminous United States. Hydrologic
Investigations Atlas HA-212, Department of the Interior, U.S. Geological
Survey, Washington, D.C., 1966.
Chow, V. T. Statistical and Probability Analysis of Hydrologic Data. Part
I, Frequency Analysis. In: Handbook of Applied Hydrology, V. T. Chow,
ed. McGraw Hill, New York, 1964. pp. 8-1 - 8-42.
Cochran, W. G. Sampling Techniques. 3 ed., John Wiley and Sons, Inc., New
York, New York, 1977. 428 pp.
Corp of Engineers. Chowan River Basin, Virginia and North Carolina (Phase I
Feasibility Report). Department of Army, Ft. Norfolk, 803 Front St.,
Norfolk, Virginia, January 1975.
Environmental Protection Agency. Methods for Chemical Analysis of Water and
Waste. Technology Transfer Publication, Environmental Monitoring Support
Laboratory, Cincinnati, Ohio, 1974.
179
-------
Eppley, R., J. Rogers, and J. McCarthy. Half-Saturation Constants for Up-
take of Nitrate and Ammonium by Marine Phytoplankton. Limnol. Oceanogr.,
14:912-920, 1969.
Fitzgerald, G. P. Bioassay Analysis of Nutrient Availability. In: Nutrients
in Natural Waters, H. E. Allen and J. R. Kramer, eds. Wiley-Interscience,
New York, 1972. pp. 147-171.
Forsythe, T. D. Identifying and Controlling Some Sources Of Error For a
Static Algal Bioassay With Applications. M.S. Thesis, Michigan State
University, East Lansing, Michigan, 1973.
Funs, G., S. Demmerle, E. Canelli, and M. Chen. Characterization of Phosphorus
Limited Algae (With Reflections on the Limiting Nutrient Concept). In:
Nutrients and Eutrophication, G. Likens, ed. Am. Soc. Limnol. Oceanogr.
Spec. Symp. I, 1972. pp. 113-132.
Gambell, A. W., and D. W. Fisher. Chemical Composition of Rainfall Eastern
North Carolina and Southeastern Virginia. Geological Survey Water Supply
Paper 1535-K, 1966.
Gambrell, R. P., J. W. Gillian, and S. B. Weed. The Fate of Fertilizer
Nutrients as Related to Water Quality in the North Carolina Coastal Plain.
UNC-WRRI Report 93, Water Resources Research Institute, Raleigh, North
Carolina, 1974. 151 pp.
Haan, C. T. Statistical Methods in Hydrology. The Iowa State University
Press, Ames, Iowa, 1977.
Hald, A. Statistical Theory With Engineering Applications. Wiley and Sons,
New York, 1952. 783 pp.
Harrison, W. G., and J. E. Hobbie. Nitrogen Budget of a North Carolina
Estuary. Report No. 86, N. C. Water Resources Research Institute,
Raleigh, North Carolina, 1974. 72 pp.
Hayne, D. W., A. M. Witherspoon, and T. R. Fisher. Algal Diversity and
Transport in Small Streams. In: Proceedings of Watershed- Research in
Eastern North America - A Workshop to Compare Results, D. L. Correll,
ed. Chesapeake Bay Center for Environmental Studies, Smithsonian Insti-
tution, Edgewater, Maryland, 11:593-610, 1977.
Hobbie, J. E., B. J. Copeland, and W. G. Harrison. Nutrients in the Pamlico
River Estuary, North Carolina, 1969-1971. Report No. 76, N. C. Water
Resources Research Institute, Raleigh, North Carolina, 1972.
Humenik, F. J., and M. R. Overcash. Discussion of Paper Entitled "Evaluation
of Methods for the Analysis of Physical, Chemical, and Biochemical
Properties of Poultry Wastewater" by Prakasam et al. Standardizing
Properties and Analytical Methods Related to Animal Waste Research,
Willrich, Miner, and Overcash, eds. ASAE Special Publication SP-0275,
1975.
130
-------
Hynes, H. B. The Ecology of Running Waters. University of Toronto, Ontario,
Canada, 1970.
Kuenzler, E. J., P. J. Mulholland, L. A. Ruley, and R. P. Sniffen. Water
Quality in North Carolina Coastal Plain Streams and Effects of Channel-
ization. UNC-WRRI Report 126, Water Resources Research Institute, Raleigh,
North Carolina, 1977.
Lloyd, M., and R. J. Ghelardi. A Table for Calculating the "Equitability"
Component of Species Diversity. J. Anim. Ecol., 33:217-225, 1964.
Lund, J. W. G. The Ecology of the Freshwater Phytoplankton. Biol. Rev.,
49:231-293, 1965.
Maclssac, J. J., and R. C. Dugdale. The Kinetics of Nitrate and Ammonia Up-
take by Natural Populations of Marine Phytoplankton. Deep Sea Res.,
16:45-57.
Maloney, T. E., W. E. Miller, and T. Shiroyama. Algal Responses to Nutrient
Additions in Natural Waters. I. Laboratory Assays. In: Nutrients .and
Eutrophication: The Limiting-Nutrient Controversy, G. E. Likens, ed.
ASLO, Inc., Lawrence, Kansas, 1972. 328 pp.
Nichols, K. H. Nutrient Phytoplankton Relationships in the Holland March.
Ecol. Monogr, Ontario, Canada, 46:179-199, 1976.
Nichols, K. H. and H. R. MacCrimmon. Nutrient Loading to Cook Bay of Lake
Simcoe From the Holland River Watershed. Inter. Rev. Ges. Hydrobiol.,
60:159-193, 1975.
Omernik, J. M. Nonpoint Source-Stream Nutrient Level Relationships, A
Nationwide Study. EPA-600/3-77-105, U.S. Environmental Protection Agency,
1977.
Overcash, M. R., A. Hashimoto, D. L. Reddell, and D. L. Day. Chemical Anal-
ysis ASAE Paper Mo. 74-4546. Standardizing Properties and Analytical
Methods Related to Animal Waste Research, Willrich, Miner, and Overcash,
eds. ASAE Special Publication SP-0275, 1975.
Pielou, E, C. Ecological Diversity. Wiley and Sons, New York, 1975. pp.
5-18.
Poole, R. W. An Introduction to Quantitative Ecology. McGraw-Hill, New
York, 1974. pp. 387-397.
Redfield, A. C., B. H. Ketchum, and F. A. Richards. The Influence of Organisms
on the Composition of Seawater. M. A. Hill, Interscience, New York, ed.
The Sea, 2:27-79, 1963.
181
-------
Riggs, H. C. Some Statistical Tools in Hydrology. In: Techniques of Water-
Resource Investigations of the U.S.G.S., USGPO, Washington, D.C., 1967.
Rubin, A. R., J. M. Stewart, and B. L. Carlile. Some Important Considerations
for the Land Application of Wastewater. UNC-WRRI Publication, Water
Resources Research Institute, Raleigh, North Carolina, 1978. 28 pp.
Saucier, W. J., A. H. Weber, and C. K. Bayne. Precipitation Variability
Over North Carolina. UNC-WRRI Report 84, Water Resources Research Insti-
tute, Raleigh, North Carolina, 1973.
Snedecor, G. W., and W. G. Cochran. Statistical Methods. 6 ed., Iowa State
University Press, Ames, Iowa, 1967. 593 pp.
Stanley, D. W., and J. E. Hobbie. Nitrogen Recycling in the Chowan River.
UNC-WRRI-121, Water Resources Research Institute, Raleigh, North Carolina,
1976. 60 pp.
Swamson, C. E., and R. W. Bachman. A Model of Algal Exports In Some Iowa
Streams. Ecology, 57:1076-1080, 1976.
U.S. Department of Commerce. National Oceanic and Atmospheric Administration,
National Weather Service. Forecast. In: The News and Observer, Raleigh,
North Carolina, 1971-76.
Vollenweider, R. A. A Manual on Methods for Measuring Primary Production in
Aquatic Environments. IBP Handbook No. 12, 2 ed., Blackwell Scientific
Pub., London, 1974.
Walsh, J. J. Relative Importance of Habitat Variables in Predicting the
Distribution of Phytoplankton at the Ecotone of the Antarctic Upwelling
Ecosystem. Ecol. Monogr., 41:291-309, 1971.
Webb, C. E. Biological Field and Laboratory Methods for Measuring the Quality
of Surface Waters and Effluents. EPA-670/4-73-001 (Section on Macro-
i nvertegrates), 1973.
Weiss, C. M., and R. W. Helms. The Interlaboratory Precision Test. An Eight
Laboratory Evaluation of the Provisional Algal Assay Procedure Bottle
Test. Proj. 16010DQT, EPA Water Quality Office, 1971. 70 pp.
Whitton, B. A. River Ecology: Studies in Ecology. University of California,
Berkeley, V. 2, 1975. 725 pp.
Winberg, G. G. Symbols, Units, and Conversion Factors in Studies of Freshwater
Productivity. Publ. IBP Central Office, London, 1971.
Wiser, E. W. Hydrologic Information Storage and Retrieval System. Technical
Bulletin 215, N. C. Agricultural Experiment Station, N. C. State Univer-
sity, Raleigh, North Carolina, 1975.
182
-------
Yates, F. Sampling Methods for Censuses and Surveys. 3 ed., Griffin, London,
1960. 440 pp.
Yentsch, C. S. Critical Mixing Depth: A Problem in the Measurement of Respir-
ation. In: Respiration of Marine Organisms, J. 0. Cech, D. W. Bridges,
and D. B. Morton, eds. TRIGOM, South Portland, Maine, 1975.
183
-------
PUBLICATIONS ASSOCIATED WITH PROJECT RESULTS
Humenik, F. J., M. R. Overcash, F. Koehler, L. Bliven, J. W. Gilliam, and
W. S. Galler. Nature and Impact of Stream Inputs on a Watershed Basis.
ASAE Paper 76-2564, Amer. Soc. of Agr. Engrs., Chicago, Illinois,
December 1976.
Bliven, L., F. Koehler, F. J. Humenik, and M. R. Overcash. Sampling Methods
to Measure Nonpoint Source Impact on Water Quality. ASAE Paper 77-4047,
Amer. Soc. of Agr. Engrs., Raleigh, North Carolina, June 1977.
Humenik, F. J., W. S. Galler, J. W. Gilliam, D. W. Hayne, D. H. Howells, M.
R. Overcash, and A. M. Witherspoon. An Overview of the Chowan River
Rural Runoff Study. In: Watershed Research in Eastern North America,
Cheasapeake Bay Center for Environmental Studies, Smithsonian Institution,
Edgewater, Maryland. 1:51-66. 1977.
Hayne, D. W., A. M. Witherspoon, and T. R. Fisher. Algal Diversity and Trans-
port in Small Streams. In: Watershed Research in Eastern North America,
Cheasapeake Bay Center for Environmental Studies, Smithsonian Institution,
Edgewater, Maryland. 2:593-610. 1977.
Hayne, D. W. Probability Sampling of Small Streams: Problems and Results.
In: Watershed Research in Eastern North America, Cheasapeake Bay Center
for Environmental Studies, Smithsonian Institution, Edgewater, Maryland.
2:809-830. 1977.
Humenik, F. J., F. A. Koehler, L. F. Bliven, J. W. Gilliam, and M. R. Overcash.
Sampling Methodologies for Assessing Rural Runoff. In: Watershed Re-
search in Eastern North America, Cheasapeake Bay Center for Environmental
Studies, Smithsonian Institution, Edgewater, Maryland. 2:791-808. 1977.
Overcash, M. R., L. F. Bliven, F. A. Koehler, J. W. Gilliam, and F. J. Humenik.
Nutrient Yield Assessment by Different Sampling Strategies. In: Water-
shed Research in Eastern North America, Cheasapeake Bay Center for Envir-
onmental Studies, Smithsonian Institution, Edgewater, Maryland. 1:365-
383. 1977.
Humenik, F. J., F. A. Koehler, F. L. Bliven, and M. R. Overcash. Nonpoint In-
puts From Agricultural Areas in a Southeastern Watershed. In: Pro-
ceedings of the 32nd Annual Meeting, Soil Conservation Soc. of Amer.,
Richmond, Virginia, 1977. pp. 243-250.
184
-------
Humenik, F. J.t M. R. Overcash, F. Koehler, L. Bliven, J. W. Gilliani, and
W. S. Caller. Nature and Impact of Rural Stream Inputs on Water Quality.
Transactions Amer. Soc. Agr. Engrs., 21(4):676-681, 1978.
Koehler, F. A., F. J. Humenik, E. P. Harris, and J. C. Barker. Simple.Sampler
Activation and Recording System. J. Environmental Engineering Division,
Amer. Soc. Civil Engrs., Vol. 104, No. EE5, 1978. pp. 1032-1035.
Humenik, F. J., L. Bliven, F. Koehler, and M. R. Overcash. An Analysis of
Rural Nonpoint Source Water Quality. ASAE Paper 78-2509, Amer. Soc. of
Agr. Engrs., Chicago, Illinois, December 1978.
Bliven, L., F. J. Humenik, F. Koehler, and M. R. Overcash. Monitoring Area-
wide Rural Water Quality. J. Environmental Engineering Division, Amer.
Soc. Civil Engrs., Vol. 105, No. EE1, 1979. pp. 101-112.
Humenik, F. J., M. R. Overcash, L. F. Horney, L. F. Bliven, and F. A. Koehler.
Areawide Assessment of Rural Stream Water Quality: A Statistical Sam-
pling Methodology. Environmental Protection Engineering, Wroclaw Tech-
nical University, Wroclaw, Poland, Vol. 5, No. 1, 1979.
Bliven, L. F., F. A. Koehler, L. F. Horney, M. R. Overcash, and F. J. Humenik.
Areawide Assessment of Rural Stream Water Quality: Data Analysis Tech-
niques. Environmental Protection Engineering, Wroclaw Technical Univer-
sity, Wroclaw, Poland, Vol. 5, No. 2, 1979.
Humenik, F. J., L. F. Bliven, M. R. Overcash, and F. Koehler. Rural Nonpoint
Source Water Quality in a Southeastern Wastershed. In Press, J.
Water Pollution Control Federation.
Bliven, L. F., F. J. Humenik, F. A. Koehler, and M. R. Overcash. Automated
Sampling Analysis of Rural Nonpoint Source Water Quality. Submitted
for Publication, Transactions, ASAE.
Koehler, F. A., F. J. Humenik, and L. F. Bliven. Discussion: Nutrient Budget
Analysis for Rend Lake in Illinois. Proceedings Paper 14616. Submitted
to J. Environmental Engineering Division, ASCE.
135
-------
APPENDIX: PHYSICO-CHEMICAL PARAMETER VALUES FOR ALL 30 STATIONS SAMPLED
All values are in mg/L except TEMP - °C,
COND - ymhos/cm, pH - pH units, and FLOW - mVsecond
Standard
Variable Samples Mean Minimum Maximum Deviation
Station 1
DO 37 7.96 1.6 14.7 3.41
TEMP 40 14.1 2 26 7.1
COND 38 70.2 20 152 35.6
pH 28 5.55 4.6 6.9 0.62
FLOW 50 0.021 0 0.70 0.098
COD 23 45.3 20 119 21.7
TOC 23 16.7 3 57 10.8
TP 35 0.216 0.05 0.79 0.174
NHa-N 7 0.129 <0.01 0.86 0.323
N03-N 35 0.260 <0.01 1.55 0.300
TKN 35 1.638 <0.04 7.35 1.458
Cl 35 11.86 2.8 39.0 9.24
SS 12 54.7 8 237 65.4
Station 2
DO 38 9.29 4.3 16.7 2.93
TEMP 40 13.3 2 24 6.2
COND 38 58.9 20 110 19.5
PH 28 5.65 4.6 6.7 0.57
FLOW 50 0.0090 0 0.12 0.021
COD 25 31.9 12 103 10.7
TOC 26 12.8 2 46 8.6
TP 38 0.199 0.06 0.75 0.177
NH3-N 8 0.016 <0.01 0.07 0.024
N03-N 39 0.364 <0.01 1.07 0.281
TKN 38 1.380 0.38 3.27 0.730
Cl 39 10.66 2.2 21.2 4.73
SS 14 22.8 4 61 18.6
Station 3
DO 35 9.52 3.1 18.0 3.10
TEMP 37 13.4 2 24 6.5
COND 35 64.6 25 154 29.5
pH 28 5.62 4.4 6.7 0.63
FLOW 51 0.014 0 0.314 0.046
186
-------
APPENDIX (continued)
Standard
Variable Samples Mean Minimum Maximum Deviation
COD 24 35.0 12 111 20.4
TOC 25 13.8 2 53 10.2
TP 34 0.184 0.04 0.86 0.170
NHS-N 8 0.196 <0.01 1.15 0.391
NO.-N 34 0.398 0.01 1.36 0.373
TKN 34 "1.694 <0.04 8.54 1.509
Cl 34 12.65 2.2 58.6 10.09
SS 13 23.4 6 50 13.6
Station 4
DO 43 8.59 0.5 17.2 3.42
TEMP 45 14.2 2 24 6.6
COND 43 66.9 25 200 31.3
pH 34 5.99 5.1 6.8 0.43
FLOW 50 0.018 0 0.240 0.040
COD 25 33.7 8 99 20.2
TOC 25 12.6 3 46 8.7
TP 39 0.217 0.01 1.43 0.244
NHj-N 8 0.176 <0.01 0.86 0.299
NO.-N 40 0.335 <0.01 1.03 0.249
TKN 40 1.496 0.24 3.92 0.859
Cl 40 10.66 1.5 19.9 5.25
SS 14 56.0 7 404 104
Station 6
DO 49 8.85 5.2 14.4 1.86
TEMP 51 14.9 2 25 5.7
COND 49 77.0 32 410 61.2
DH 34 5.99 5.1 6.8 0.43
FLOW 52 0.121 0.0080 2.47 0.350
COD 32 20.5 <4 68 17.5
inr n 9.2 2 31 0.8
|p 52 0.129 0.03 0.85 0.137
NHa-N 13 0.177 <0.01 1.15 0.390
N03-N 52 0.142 <0.0 J.44 0.220
TKN 52 0.996 <0.04 5.25 0.884
Cl 52 8.70 .5 19.2 2.66
SS 19 8.4 1 25 f't
Station 29
COD 23 26.8 8 68 13.4
TOC 24 13.0 5 34 6.7
107
-------
APPENDIX (continued)
Standard
Variable Samples Mean Minimum Maximum Deviation
TP 39 0.144 0.05 0.38 0.076
NH3-N 12 0.097 <0.01 0.57 0.161
N03-N 41 0.083 <0.01 0.59 0.124
TKN 41 1.316 0.28 3.23 0.796
Cl 41 8.75 3.0 25.6 3.73
SS 18 12.5 2 31 7.5
Station 7
DO 41 9.30 4.5 12.6 2.34
TEMP 44 14.1 4 24 6.4
COND 51 77.0 32 410 61.2
pH 29 6.08 4.6 7.5 0.65
FLOW 51 0.024 0 0.264 0.052
COD 29 28.2 8 73 14.6
TOO 29 12.8 1 45 8.6
TP 43 0.229 0.04 1.40 0.259
NH3-N 12 0.197 <0.01 1.68 0.474
NO.-N 45 0.564 <0.01 2.21 0.519
TKN 45 1.288 <0.04 3.90 0.916
Cl 45 10.94 3.2 22.2 4.22
SS 18 21.5 1 111 25.2
DO 46
TEMP 49
COND 48
pH 29
FLOW 51
COD 30
TOC 30
TP 48
NH3-N 13
NOS-N 49
TKN 49
Cl 49
SS 20
Station 8
8.35
14.7
180
6.24
0.011
34.1
16.4
0.480
0.164
1.017
1.789
19.01
26.2
1.5
4
34
4.3
0
12
1
0.05
<0.01
0.07
<0.04
3.7
4
12.4
26
1180
9.0
0.144
110
80
2.82
0.61
2.65
6.38
72.5
182
2.69
6.7
233
0.80
0.025
23.6
15.4
0.636
0.190
0.619
1.259
15.78
39.6
Station 9
DO 46 8.53 4.2 12.8 2.47
TEMP 49 14.5 4 24 6.6
COND 48 98.1 25 320 61.8
188
-------
APPENDIX (continued)
Standard
Variable Samples Mean Minimum Maximum Deviation
pH 30 6.15 4.8 6.8 0.49
FLOW 52 0.209 0.000082 4.90 0.749
COD 31 27.9 6 64 14.3
TOC 31 11.7 2 25 6.5
TP 52 0.180 0.04 0.73 0.146
NH3-N 14 0.132 <0.01 1.07 0.275
N03-N 52 1.062 0.05 2.54 0.654
TKN 52 1.366 <0.04 4.13 0.801
Cl 52 11.14 3.5 20.6 3.35
SS 20 27.4 6 75 17.1
Station 10
DO 46 7.15 1.7 13.5 3.01
TEMP 51 15.4 3 26 6.9
COND 50 86.1 30 190 38.3
pH 32 6.13 4.8 7.2 0.56
FLOW 52 0.886 0.040 12.42 1.935
COD 31 27.9 4 91 17.9
TOC 31 31.0 <1 39 9.0
TP 52 0.380 0.03 1-64 0.319
NH3-N 14 0.192 <0.01 0.87 0.277
N03-N 52 0.535 0.03 1.05 0.255
TKN 52 1.389 <0.04 3.36 0.736
Cl 52 8.95 3.5 14.9 2.48
SS 20 9.6 1 28 8.7
Station 11
DO 16 7.32 3.7 12.1 3.17
TEMP 19 10.3 5 23 5.3
COND 18 88.8 24 380 94.2^
12 0:&5 a' 0)061 0)0087
1°0 Is5'? ! " '"
T 7 2)951 0.06 33.22 8.085
h S'iSs 8:8! 8:§? 8
« " i™ °-8n4 ll-f li
Cl 17 14.89 2.0 «.0
189
-------
APPENDIX (continued)
Standard
Variable Samples Mean Minimum Maximum Deviation
Station 12
DO 44 6.45 0.2 12.8 3.43
TEMP 50 13.5 3 25 6.8
COND 49 67.1 22 200 38.2
pH 28 5.91 5.1 7.6 1.27
FLOW 52 0.180 0 3.31 0.602
COD 29 41.2 19 71 13.8
TOC 28 19.1 8 36 7.5
TP 47 0.218 0.05 2.29 0.386
NH3-N 12 0.219 <0.01 1.30 0.409
NO.-N 47 0.048 <0.01 0.63 0.099
TKN 47 1.351 <0.04 3.36 0.778
Cl 47 6.81 1.5 11.1 2.28
SS 19 12.6 1 29 8.0
Station 13
DO 47 7.60 1.9 12.3 2.77
TEMP 51 14.0 3 25 6.8
COND 49 59.9 25 170 24.4
pH 32 5.86 5.0 6.9 0.52
FLOW 52 0.998 0 18.54 3.368
COD 30 34.3 8 68 13.3
TOC 29 13.2 1 33 6.4
TP 47 0.129 0.05 0.33 0.071
NHj-N 12 0.140 <0.01 • 1.03 0.289
N03-N 47 0.464 <0.01 1.46 0.341
TKN 47 1.343 <0.04 4.77 0.894
Cl 47 9.37 3.9 30.5 3.80
SS 16 15.4 2 167 40.4
Station 14
DO 42 8.43 3.7 14.2 2.35
TEMP 47 13.4 2 22 5.9
COND 46 58.0 30 190 32.4
pH 25 5.77 4.5 6.9 0.52
FLOW 52 0.0319 0 0.28 0.0465
COD 27 25.9 8 44 11.2
TOC 27 11.4 3 23 5.6
TP 48 0.080 0.02 0.44 0.069
190
-------
APPENDIX (continued)
Standard
Variable Samples Mean Minimum Maximum Deviation
NH3-N 11 0.045 <0.01 0.18 0.067
N03-N 48 1.882 0.12 8.96 1.944
TKN 48 1.053 <0.04 2.80 0,517
Cl 48 8.45 2.8 12.1 1.90
SS 18 6.8 1 27 7.1
Station 15
DO 45 8.44 4.4 14.2 2.55
TEMP 49 13.8 0.5 23 6.3
COND 47 51.6 30 110 18.1
pH 31 6.05 4.8 7.4 0.56
FLOW 52 0.120 0 2.01 0.301
COD 30 18.4 4 37 8.6
TOC 30 7.8 1 20 4.7
TP 48 0.113 0.02 0.60 0.127
NH3-N 12 0.027 <0.01 0.09 0.029
N03-N 48 0.363 <0.01 1.13 0.269
TKN 48 1.043 <0.04 3.80 0.791
Cl 48 7.71 2.7 51.4 6.83
SS 18 4.7 1 28 6.1
Station 16
DO 45 6.96 1.4 12.2 3.21
TEMP 49 13.8 1 26 7.1
COND 48 55.0 25 150 25.6
pH 31 6.02 4.9 7.2 0.54
FLOW 52 0.249 0 1.96 0.377
COD 30 30.9 10 55 11.7
TOC 31 12.9 3 28 6.1
TP 48 0.122 0.04 0.60 0.092
NH3-N 12 0.089 <0.01 0.57 0.167
NOs-N 48 0.084 <0.01 0.56 0.120
TKN 48 1-212 0.28 339 0.706
Cl 48 6.29 2.4 8.8 1.29
SS 18 9.8 1 33 9-5
Station 17
DO 40 10.37 4.3 16.6 2.49
TEMP 41 12.3 3 23 5.6
COND 39 41.9 10 180 27.5
191
-------
APPENDIX (continued)
Variable
Samples
Mean
Minimum
Maximum
Standard
Deviation
PH
FLOW
COD
TOC
TP
NH3-N
N03-N
TKN
Cl
ss
DO
TEMP
COND
PH
FLOW
COD
TOC
TP
NH3-N
N09-N
TKN
Cl
SS
DO
TEMP
COND
PH
FLOW
COD
TOC
TP
NH.-N
N03-N
TKN
Cl
SS
27
50
26
26
40
11
42
42
42
16
45
46
44
28
50
27
27
44
10
44
44
44
17
48
49
47
30
52
30
30
49
12
49
49
49
20
6.17
0.0874
26.1
11.7
0.100
0.045
0.019
1.101
3.81
4.7
Station
9.46
12.8
48.1
6.10
0.0821
23.0
10.7
0.082
0.059
0.022
1.046
3.97
4.7
5.3
0
4
2
0.02
<0.01
<0.01
<0.04
1.8
1
18
2.3
2
10
5.2
0
6
4
0.02
<0.01
<0.01
<0.04
2.3
1
Station 19
9.18
13.3
51.5
6.14
0.219
27.8
10.4
0.109
0.052
0.024
1.078
4.17
1.7
2
10
5.0
0
8
4
0.02
<0.01
<0.01
<0.04
1.5
6.2
7.1
1.76
67
30
0.65
0.29
0.13
2.80
9.4
14
17.4
22
200
7.1
1.71
46
24
0.74
0.29
0.15
2.96
10.3
17
12.8
22
200
7.1
3.13
87
30
0.97
0.29
0.15
2.80
9.7
17
14
7
0.44
0.256
3
1
0.130
0.084
0.028
0.666
1.38
3.7
2.97
5.7
30.2
0.48
0.248
11
5
1
1
0.107
0.088
0.028
0.719
1.57
4.8
2.92
5.7
29.9
0.53
0.560
15.3
5.6
0.159
0.082
0.033
0.673
1.403
5.0
192
-------
APPENDIX (continued)
Variable Samples Mean Minimum Maximum Deviation
Station 20
00 47 9.49 4.6 13.2 2.14
TEMP 49 13.7 2 24 57
COND 47 49.5 20 330 54is
PH 30 6.10 5.1 7.0 0.53
FLOW 49 0.0819 0 0.99 0.165
COD 32 22.5 6 52 98
TOC 32 9.3 2 21 4.8
TP 51 0.119 0.02 0.74 0.114
NH3-N 13 0.064 <0.01 0.37 0.097
N03-N 51 0.062 <0.01 0.24 0.048
TKN 51 1.060 <0.04 3.42 0.706
Cl 51 3.68 2.2 19.6 2.42
SS 22 11.3 <1 56 14.7
Station 22
DO 48 9.75 5.1 14.0 2.12
TEMP 50 13.5 0.5 23 5.7
COND 48 45.9 20 150 23.8
pH 30 6.33 5.2 7.1 0.46
FLOW 49 0.201 0 1.65 0.377
COD 32 17.9 2 50 9.7
TOC 31 8.5 2 23 4.8
TP 49 0.105 0.02 0.76 0.116
NH3-N 13 0.018 <0.01 0.06 0.020
N03-N 50 0.054 <0.01 0.26 0.049
TKN 49 0.999 <0.04 6.38 0.985
Cl 50 3.45 1.5 6.8 0.98
SS 22 9.6 1 66 15.8
Station 23
DO 48 9.57 5.3 13.8 2.17
TEMP 50 13.6 0.5 24 5.8
COND 48 42.3 20 90 15.7
pH 30 6.33 5.5 6.9 0.39
FLOW 50 0.199 0 1.60 0.365
COD 32 21.2 4 73 13.9
TOC 32 8.9 1 22 5.4
TP 50 0.118 0.03 0.50 0.110
NH3-N 13 0.025 <0.01 0.08 0.026
N03-N 50 0.055 <0.01 0.28 0.054
193
-------
APPENDIX (continued)
Standard
Variable Samples Mean Minimum Maximum Deviation
TKN 49 1.034 <0.04 3.03 0.664
Cl 50 3.90 2.2 12.1 2.13
SS 22 10.1 1 83 18.8
Station 24
DO 48 9.46 4.3 13.4 2.27
TEMP 50 13.7 0.5 24 5.9
COND 48 43.2 20 97 17.1
pH 30 6.41 5.5 8.2 0.52
FLOW 49 0.380 0 3.93 0.744
COD 32 20.0 4 48 8.8
TOC 32 8.6 3 21 4.5
TP 51 0.103 0.03 0.64 0.101
NH*-N 13 0.019 <0.01 0.06 0.022
N03-N 51 0.058 <0.01 0.21 0.044
TKN 51 1.065 <0.04 3.35 0.764
Cl 51 4.10 2.0 20.3 2.76
SS 22 10.0 <1 76 17.3
Station 25
DO 50 9.53 4.0 13.9 2.35
TEMP 51 13.5 1.5 23 5.9
COND 49 43.1 20 110 16.4
pH 30 6.30 5.2 7.0 0.56
FLOW 52 0.580 0 6.33 1.213
COD 33 23.4 4 76 13.9
TOC 33 8.5 1 23 4.6
TP 52 0.094 0.03 0.46 0.076
NH,-N 13 0.027 <0.01 0.08 0.029
N03-N 52 0.058 <0.01 0.24 0.052
TKN 52 1.153 0.23 3.57 0.781
Cl 52 4.95 1.5 55.3 8.39
SS 22 10.7 1 88 21.8
Station 28
COD 23 20.7 4 50 11.0
TOC 23 9.0 1 22 5.7
TP 41 0.079 0.03 0.14 0.026
NH3-N 12 0.022 <0.01 0.06 0.022
N03-N 42 0.044 <0.01 0.15 0.037
194
-------
APPENDIX (continued)
Standard
Variable Samples Mean Minimum Maximum Deviation
TKN 42 1.094 <0.04 4.23 0.741
Cl 42 3.48 1.6 7.3 0.96
SS 21 9.0 1 79 16.9
Station 26
DO 50 9.38 4.1 14.1 2.58
TEMP 52 13.4 3 24 6.2
COND 50 42.7 20 88 14.0
pH 31 6.26 5.1 7.2 0.48
FLOW 52 0.247 0 4.86 0.706
COD 32 21.5 8 40 9.0
TOC 32 9.5 1 21 4.7
TP 51 0.114 0.02 0.78 0.151
NH3-N 14 0.029 <0.01 0.08 0.027
NOs-N 51 0.048 <0.01 0.20 0.038
TKN 51 1.152 <0.04 3.84 0.939
Cl 51 4.61 2.0 21.1 3.39
SS 22 11.2 <1 94 20.1
Station 27
DO 49 9.21 2.2 13.4 2.89
TEMP 51 13.2 3 22 6.0
COND 49 50.6 25 108 20.2
PH 32 6.30 5.4 7.0 0.47
PLOW 52 0.195 0.000008 1.36 0.313
COD 32 18.1 4 39 8.9
mr 19 68 1 22 4.b
TP 52 0.150 0.02 1.15 0.248
NH3-N 14 0.023 <0.01 0.09 0.028
N03-N 52 0.038 <0.01 0.12 0.029
TKN 52 1.120 <0.04 4.19 0.787
Cl 52 6.07 2.0 32.0 4.41
SS 22 6.1 1 20 5'4
Station 30
DO 35 9.21 5.5 13.6 238
TEMP 37 15.7 1 26 6.3
COND 38 68.0 30 220 _ 31.^
n T7
pH 17
FLOW 38
5 6 7.6 0.50
3-° .
. . ?7
0.09 0 1-77
195
-------
APPENDIX (continued)
Variable
Samples
Mean
Minimum
Maximum
Standard
Deviation
COD
TOC
TP
NH3-N
N03-N
TKN
Cl
ss
DO
TEMP
COND
PH
FLOW
COD
TOC
TP
NH3-N
N03-N
TKN
Cl
SS
DO
TEMP
COND
PH
FLOW
COD
TOC
TP
NH3-N
N03-N
TKN
Cl
SS
19
19
38
14
38
38
38
17
35
38
38
19
38
19
19
38
14
38
38
38
18
35
38
38
19
38
19
19
38
14
38
38
38
18
20.7
8.3
0.108
0.035
0.084
1.127
5.08
5.1
Station
9.09
14.5
42.2
6.26
0.169
18.7
6.6
0.102
0.021
0.166
0.963
3.51
11.0
Station
8.92
14.6
43.3
6.17
0.278
13.9
6.1
0.102
0.013
0.082
0.919
4.00
4.3
8
2
0.04
<0.01
<0.01
0.15
2.6
<1
31
6.2
1
20
5.1
0.00066
4
1
0.03
<0.01
<0.01
<0.04
2.1
2
32
6.1
1
19
4.8
0.015
4
1
0.02
<0.01
<0.01
<0.04
2.3
1
71
28
0.34
0.17
0.87
3.63
19.2
12
13.1
23
110
7.0
2.75
75
26
0.35
0.07
0.43
2.93
6.9
30
13.0
24
100
6.7
3.16
55
31
0.50
0.04
0.28
3.73
29.3
13
14.5
5.6
0.072
0.055
0.151
0.924
2.67
3.4
2.06
6.2
17.3
0.50
0.441
14.9
5.4
0.065
0.022
0.035
0.622
0.969
9.2
2.08
6.5
17.1
0.535
0.588
10.7
6.5
0.104
0.015
0.073
0.730
4.40
3.3
196
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1, REPORT NO.
EPA-600/3-80-035
2.
3. RECIPIENT'S ACCESSIOt*NO.
4. TITLE AND SUBTITLE
Probability Sampling to Measure Pollution from Rural
Land Runoff
5. REPORT DATE
February 1980 issuing date
6. PERFORMING ORGANIZATION CODE
'_AUTHOR(S)
F.J. Humenik, D.W. Hayne, M.R. Overcash, J.W. Gilliam,
A.M. Witherspoon, U.S. Galler, and D.H. Howells
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
North Carolina State University
Raleigh, North Carolina 27650
1O. PROGRAM ELEMENT NO.
A1MH1E
11. CONTRACT/GRANT NO.
R803328
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Research Laboratory—Athens GA
Office of Research and Development
U.S. Environmental Protection Agency
Athens, Georgia 30605
13. TYPE OF REPORT AND PERIOD COVERED
Final. 7/74-12/78
14. SPONSORING AGENCY CODE
EPA/600/01
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The feasibility of probability sampling in describing quality of rural water
not affected by point sources is examined. The study site was a portion of the
Chowan River Basin in Virginia and North Carolina. Flow was measured along with dis-
solved oxygen, temperature, conductivity, and pH. All samples were analyzed for ni-
trate plus nitrite nitrogen, total Kjeldahl nitrogen, total phosphate and chloride.
Flow was highly variable in time and space; concentration was less so. For estimating
basin-wide mean values with modest budgets, grab sampling was more flexible and provi-
ded better precision of estimation than did automated sampling under the same budget.
With high-budget studies (several hundred thousand 1975 dollars) this relationship
may be reversed. This comparison is stated on the restrictive basis of estimating
mean values; automated sampling may have other advantages with other kinds of studies.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS
c. COS AT I Field/Group
Agriculture
Sampli ng
Statistical analysis
Water pollution
14A
68D
13. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Repot
SECURITY CLASS ft
UNCLASSIFIED
215
?S (This page)
22. PRICE
EPA Form 2220-1 (9-73)
197
» U.S. GOVERNMENT WHITING OFFICE: 1MO-657-U6/5593
------- |