Prepared in cooperation with the
Arkansas Soil and Water Conservation Commission
Water-Resources Investigations Report 02-4187
Water-Resources Investigations Report 02-4187
U.S. DEPARTMENT OF THE INTERIOR
Water-Quality and Discharge Data 4
Water-Quality, Biological, and Habitat Assessment 10
Benthic Macroinvertebrate Communities and Physical Habitat 23
Figure 1. Map showing location of study area 3
2. Photograph showing cotton field storm monitoring station at Coon Bayou Tributary near Tillar, Arkansas 5
3. Photograph showing forested storm monitoring station at Boeuf River Diversion Canal near
Hudspeth, Arkansas 6
4. Hydrograph and hyetograph showing discharge and rainfall record of storm at the cotton field site 07369654 Coon Bayou Tributary near Tillar, Arkansas, November 1-2, 1996 6
5. Boxplot showing ranges and distributions of suspended solids and nutrient concentrations at
25 ambient sites sampled during a period of relatively low streamflow in November and
December 1994 11
6. Boxplots showing ranges and distributions of dissolved chloride, suspended solids, and nutrient
event-mean concentrations at cotton field and forested sites 17
7. Graph showing measured and estimated total nitrogen storm loads 18
8. Graph showing sampled concentrations with percent volume of catfish pond drained 20
9. Graph showing relation of biological condition score to habitat score 30
10. Graph showing relation of biological condition score to dissolved oxygen concentration 33
11. Graph showing relation of biological condition score to suspended solids concentration 33
Table 1. Water-quality and rapid bioassessment sampling sites 4
2. Water quality of samples from the 25 ambient sites 12
4. Water quality of composited storm runoff samples from the cotton field site 15
5. Water quality of composited storm runoff samples from the forest site 16
6. Mean and median event-mean concentrations by sites 18
7. Regression coefficients and error statistics for percent load regression equations at the
forested site 19
8. Regression coefficients and error statistics for load regression equations at the cotton
9. Regression coefficients and error statistics for load regression equations at the forested
10. Estimated annual yields for 1996 calendar year 20
11. Water quality of grab samples collected during draining of the catfish pond 21
12. Selected constituent loads from catfish pond drainage, April-May 1995 22
13. Abundances, tolerance values, and feeding groups of macroinvertebrate taxa collected from 25
sites in the Boeuf River Basin and a reference site 24
14. Bioassessment metrics for samples collected from sites in the Boeuf River Basin and a
reference site 28
15. Comparison of bioassessment metrics for samples collected from sites in the Boeuf River
Basin to a reference site 29
16. Comparison of biological condition scores for samples collected from sites in the Boeuf River
Basin to a reference site 31
17. Relative percent difference values for metrics and scores associated with duplicate samples 32
18. Physical habitat score assessment of the Boeuf River and its tributaries 32
19. Correlations between water quality and biological condition scores 33
Water-quality and biological samples were collected at several sites in the Boeuf River Basin between November 1994 and December 1996. Water-quality and benthic macroinvertebrate community samples were collected and habitat was measured once at 25 ambient monitoring sites during periods of seasonal low flow. Water-quality storm-runoff samples were collected during 11 storm events at two sites (one draining a cotton field and one draining a forested area). Water-quality samples were collected at one site during the draining of a catfish pond.
Water-quality samples from the 25 ambient sites indicate that streams in the Boeuf River Basin typically are turbid and nutrient enriched in late fall during periods of relatively low flow. Most suspended solids concentrations ranged from about 50 to 200 milligrams per liter (mg/L), most total nitrogen concentrations ranged from about 1.1 to 1.8 mg/L, and most total phosphorus concentrations ranged from about 0.25 to 0.40
mg/L.
Suspended solids, total nitrogen, total ammonia plus organic nitrogen, total phosphorus, and dissolved orthophosphorus concentrations from samples collected during storm events were typically higher at the cotton field site than at the forested site. Estimated annual yields of suspended solids, nitrogen, and phosphorus were substantially higher from the cotton field than from the forested area. Dissolved chloride concentrations typically were higher at the forested site than from the cotton field site. Typically, the suspended solids and nutrient concentrations from the 25 ambient sites were lower than concentrations in runoff from the cotton field but higher than concentrations in runoff from the forest area. Concentrations of sulfate, chloride, suspended solids, and some nutrients in samples from the catfish pond generally were greater than concentrations in samples from other sites. Total phosphorus, orthophosphorus, and fecal coliform bacteria concentrations from the catfish pond generally were lower than concentrations in samples from other sites.
Biological condition scores calculated using macroinvertebrate samples and U.S. Environmental Protection Agency Rapid Bioassessment Protocol II indicated that most of the 25 ambient sites would be in the "moderately impaired" category. However, substantial uncertainty exists in this rating because bioassessment data were compared with data from a reference site outside of the Boeuf River Basin sampled using different methods. Several metrics indicated that communities at most of the ambient sites are composed of more tolerant macroinvertebrates than the community at the reference site.
Habitat assessments (using Rapid Bioassessment Protocol II) indicated the reference site outside the Boeuf River Basin had better habitat than the ambient sites. Physical habitat scores for the 25 ambient sites indicated that most ambient sites had poor bottom substrate cover, embeddedness values, and flow and had poor to fair habitat related to most other factors. Most habitat factors at the reference site were considered good to excellent.
Part of the variation in biological condition scores was explained by physical habitat scores and concentrations of suspended solids and dissolved oxygen. However, a considerable amount of variability in biological condition scores is not explained by these factors.
The Boeuf River Basin has undergone major land changes during the last century. Deforestation of bottomland hardwoods, increased agricultural land-use, and channelization of natural stream geomorphology for flood control and irrigation are the main changes that have occurred within the basin. Data collected by the Arkansas Department of Environmental Quality (formerly Arkansas Department of Pollution Control and Ecology) in the Boeuf River Basin indicate that aquatic life is impacted by runoff of silt and nutrients from agricultural activities (Arkansas Department of Pollution Control and Ecology, 1996; 1998). Discharge from aquaculture reservoirs within the Boeuf River Basin also may impact the water quality of the receiving streams.
In 1994, the U.S. Geological Survey (USGS), in cooperation with the Arkansas Soil and Water Conservation Commission, began a study to assess the water quality in selected drainages within the Boeuf River Basin. The objectives of this study were to sample concentrations of dissolved chloride, suspended solids and nutrients from three different land-use drainages (cotton field, catfish pond, and forested); estimate individual storm loads, annual loads, and yields with each of the three drainages; and collect and compile baseline data on benthic macroinvertebrate communities and water quality for 25 sites in the basin.
The purpose of this report is to describe the results of a water-quality and biological investigation of selected drainages in the Boeuf River Basin of southeastern Arkansas. Water quality is assessed using water-quality data collected during three types of sampling efforts--a synoptic sampling of 25 sites during relatively low flow conditions during November through December 1994, stormwater runoff sampling of a forested area and a cotton field during January 1995 through December 1996, and sampling of a catfish pond discharge in April and May of 1995. Biological benthic macroinvertebrate and habitat information for the 25 synoptic sites also was used in the assessment. Thus, four aspects of water quality in the Boeuf River Basin were sampled--late-fall to winter low flow, stormwater runoff from areas of negligible and more intensive agricultural land practices, periodic discharges from catfish ponds, and biological communities integrating the effects of water quality and habitat conditions.
Samples were collected between November 1994 and December 1996. Water-quality data were collected one time per site during November and December 1994 at 25 sites distributed throughout much of the basin. Discharge and water-quality data were collected downstream from a cotton field and downstream from a forested area during 11 storms during January 1995 through December 1996. Data also were collected during the draining of a catfish pond in April and May 1995. Benthic macroinvertebrate and physical habitat data were collected at the 25 sites sampled in November and December 1994 as part of a bioassessment (Plafkin and others, 1989).
The Boeuf River Basin (fig. 1) is located in the Mississippi Alluvial Plain physiographic province in southeastern Arkansas. The drainage area of the Boeuf River at the Arkansas-Louisiana State line is 755 square miles (Yanchosek and Hines, 1979). The basin drains from north to south through a network of channels, canals and ditches.
The land use in this area is predominately agriculture and aquaculture. In Desha and Chicot Counties (which include most of the basin), the primary crops grown in 1995 were soybeans (26 percent of the land in the two counties), cotton (16 percent), and rice (10 percent) (Arkansas Agricultural Statistics Service, 2001a). Catfish and minnows are produced by commercial aquaculture operations in the area. Area of catfish ponds in Desha and Chicot Counties in 1999 was 11,600 acres; this is about 35 percent of the total catfish pond acreage in Arkansas (Arkansas Agricultural Statistics Service, 2001b).
The authors express appreciation to Mr. Jimmy Appleberry and the Dermott Hunting Club for allowing the installation and operation of water data collection equipment on their properties, and to Mr. David Yocum for allowing access to his property to perform data collection.
Water-quality and streamflow data were collected at 25 synoptic sites, 2 stormwater runoff sites, and 1 catfish pond drainage site (fig. 1, table 1). Biological (macroinvertebrate) and physical habitat data also were collected at the 25 ambient sites. The 25 ambient sites were selected primarily based on biological sampling criteria to be described later.
Water-Quality and Discharge Data
Water-quality samples at the 25 ambient sites were collected in conjunction with the macroinvertebrate bioassessment. Sample collection, processing, and preservation methods followed guidelines outlined by Shelton (1994). Prior to water sampling, the stream was divided into equal-width-increments (EWI). This EWI procedure resulted in 10 sampling points across the cross section. Water was collected at each sampling point with a Teflon/polypropylene depth integrating sampler. Samples were collected in November and December 1994 during a period of relatively low flow.
Water-quality analyses included specific conductance, pH, temperature, dissolved oxygen, turbidity, suspended solids, sulfate, chloride, fluoride, and nutrients. Specific conductance, pH, temperature, and dissolved oxygen were determined on-site using field meters. Analyses of the remaining constituents were conducted at the USGS Water Quality Laboratory in Ocala, Florida.
1 |
|||
2 |
|||
3 |
|||
4 |
|||
5 |
|||
6 |
|||
7 |
Diversion Canal Boeuf River beside Highway 293 near Hudspeth, Ark. |
||
8 |
|||
9 |
|||
10 |
|||
11 |
|||
12 |
|||
13 |
|||
14 |
|||
15 |
|||
16 |
|||
17 |
|||
18 |
|||
19 |
|||
20 |
|||
21 |
|||
22 |
|||
23 |
|||
24 |
|||
25 |
|||
26 |
|||
27 |
Two sites were selected to monitor storm runoff water quality. One site (site 26) was downstream from a homogenous agricultural land use (cotton) and the other site (site 8) was downstream from a forested area. The agricultural site selected was located in west-central Desha County. The monitoring station for the agricultural site was established on a ditch that drains approximately 76 acres of a cotton field and is a tributary to Coon Bayou. The forested site was located in north-central Chicot County. The monitoring station for the forested site was established on a ditch that drains approximately 1,230 acres and drains into the Boeuf River Diversion Canal. It is possible that the size of the basin that drains to the location of the monitoring station could change with extreme magnitudes of storm runoff events. The ditch intersects other ditches at three places within the basin and during events with extreme runoff water may be diverted into or out of the ditch.
Streamflow-gaging stations (Buchanan and Somers, 1974) were installed at both storm runoff sites (figs. 2 and 3). A sharp-crested rectangular weir was installed at the cotton-field site and a 24-in. diameter culvert was installed at a dirt road crossing downstream of the forested site. Stage-discharge ratings (Kennedy, 1984) were developed using indirect measurement computations for weirs (Hulsing, 1984) and indirect measurement computations for culverts (Bodhaine, 1982). The ratings were verified using current-meter streamflow measurements (Buchanan and Somers, 1984). Stage data were collected at 0.01-foot increments at both sites using float and stilling-well combinations (Buchanan and Somers, 1974). Staff gages were installed and read during storm events to verify stage data. Continuous stream-stage data were measured and recorded on an electronic data logger in 15-minute increments. Using the stage data and the stage-discharge ratings, continuous discharge values were computed (Kennedy, 1989) and recorded in 15-minute increments. Continuous rainfall data were collected at both sites using tipping-bucket rain gages. Rainfall data were collected at 0.01-inch increments and recorded at 15-minute increments.
2. Cotton field storm monitoring station at Coon Bayou Tributary near Tillar, Arkansas.
3. Forested storm monitoring station at Boeuf River Diversion Canal near Hudspeth, Arkansas.
Automatic samplers using peristaltic pumps powered by 12-volt batteries were installed at each site and used to collect storm-water samples. The sample water was pumped into four 1-gallon glass bottles contained within the automatic sampler. A data logger was used to control the automatic sampler. The data logger was programmed to read the stage, compute a discharge, and compute the volume of water that passed the monitoring station during 15-minute increments. Each time a designated volume of water passed the monitoring station the data logger would send a voltage pulse to the sampler causing it to activate and take a sample. With this setup, the sampler would take a flow-weighted composite sample throughout the storm hydrograph. The hydrograph and hyetograph in figure 4 show a typical sampled storm with rainfall intensity, instantaneous discharge, and sample activations plotted against time.
At the cotton field site, the entire storm-runoff hydrographs were sampled with the automatic sampler for every storm event. Because of the length of the storm runoff events at the forested site, the entire hydrograph was sampled only once. During the other sampled events, samples were taken at least until the peak of the hydrograph had passed.
Water-quality analyses were performed on the flow-weighted composite samples from both storm runoff sites between January 1995 and December 1996 for specific conductance, pH, turbidity, suspended solids, sulfate, chloride, fluoride, nutrients, and fecal coliform and fecal streptococcus bacteria. Bacteria concentrations were determined at the USGS laboratory in Little Rock, Arkansas. Specific conductance and pH were determined on-site for the composite sample. Other constituents were analyzed at the USGS Water Quality Laboratory in Ocala, Florida.
The catfish pond site was in central Chicot County. The pond is used to raise fingerling catfish. It is a levee type pond that drains into the Main Ditch Canal of the Boeuf River when the drainpipe valve is open. Samples were collected from the pond effluent as the pond was being drained between April 17 and May 7, 1995. The volume of water in the pond before draining was approximately 10.8 acre-feet. Effluent samples were collected using a polyethylene churn splitter at the drainpipe. Two samples were collected on April 17 after the drainpipe valve was opened and one sample was collected on April 18. The catfish-pond manager closed the valve on April 19 through April 25 to allow for time to seine the fish out of the pond. After the valve was reopened, samples were collected daily from April 26 through April 28. The valve was closed again from April 29 through May 3. After the valve was reopened, daily samples were collected from May 4 through May 7 when the pond completed draining. Ten samples were collected as the pond drained.
The pond water-surface level was flagged on the side of the levee before the drain valve was initially opened and each time a sample was collected. A total station surveying instrument was used to collect data to calculate the pond areas and volumes that corresponded with the water-surface level initially and at the time each sample was collected.
During an initial reconnaissance conducted in early November 1994, 40 sites within the study area were evaluated based on absence or presence of stream discharge and recent anthropogenic disturbances. Twenty-five sites were selected for bioassessment using the U.S. Environmental Protection Agency (USEPA) Rapid Bioassessment Protocol II (Plafkin and others, 1989). The selected sites had discharges of at least 0.5 cubic feet per second (ft 3 /s) and no evidence of recent bank, substrate, or channelization disturbances. Homogeneous dispersal of sites throughout the study area was limited by a lack of sites meeting these criteria in some areas.
During November and December 1994, macroinvertebrate communities were qualitatively sampled using a D-frame kick net with a mesh size of 425 microns. A reach length of 25 to 50 meters (82 to 164 feet) typically was sampled for 60 minutes. Reaches were sampled for less than 60 minutes (sample collection time) when the stream reach had a very homogeneous habitat and all available habitat within the reach could be adequately sampled in less than 60 minutes. In these situations, sample collection time was reduced to a minimum of 30 minutes. Samples were collected by kick netting, dipping, and hand picking from bottom substrates, vegetation, and rip rap.
Samples were sorted and organisms were enumerated at the site. The original sample was dispensed into a 5-gallon container and mixed until contents were judged homogeneous. Then a small aliquot was transferred into a white picking pan measuring approximately 15 × 20 inches. Organisms in the subsample pan were removed, labeled, and preserved in a container of 10 percent formalin. If the first aliquot contained at least 100 invertebrates the subsample was complete. However, if a minimum of 100 invertebrates was not found in the first aliquot, the sample again was mixed and a second aliquot was processed. Subsample aliquots were processed until at least 100 organisms were removed or the sample was completely processed. At two sites (sites 12 and 16), the sample in the 5-gallon container was thoroughly mixed and then split into two subsamples. Comparison of the results of these duplicate samples provided some measure of the variability of the processing steps following sample collection.
Stream habitat was assessed using methods described in Plafkin and others (1989). Habitat parameters (bottom substrate or available cover, substrate embeddedness, flow, channel alteration, bottom scouring and deposition, run-to-bend ratio, bank stability, bank vegetative stability, and streamside cover) were given a rating score according to qualitative parameter descriptions. All parameter scores were then summed to calculate a total score.
At the cotton field site, forest site, and catfish pond, water-quality and discharge data were used to compute constituent loads associated with storms (cotton field and forest sites) or the draining of the pond. Boxplots and the Wilcoxon rank sum test (Helsel and Hirsch, 1992) were used to summarize and compare the water-quality data. USEPA Rapid Bioassessment Protocol II data-analysis methods (Plafkin and others, 1989) were used for an assessment of biological conditions at the 25 ambient sites.
Data from the flow-weighted composite water-quality samples provided the mean concentrations of constituents for the sampled event. At the cotton field site, the entire storms were sampled; therefore, the sample event-mean concentrations were the same as the storm event-mean concentrations.
Loads for the cotton field site were computed using the flow-weighted composite samples for dissolved chloride, suspended solids, total nitrogen, total ammonia plus organic nitrogen, total phosphorus, and dissolved orthophosphorus. The loads were computed using the sampled event-mean concentrations (EMCs) and the runoff volumes (RVs). The RVs were computed for each event by multiplying the event-mean discharges by the length of time each event occurred. A load for a given sampling event (i) is computed by the equation
LOAD i is constituent load (in pounds) for event i;
EMC i is event-mean concentration of constituent (in milligrams per liter) for event i;
RV i is runoff volume (in cubic feet) for event i, and;
6.245 x 10 -5 is the conversion from milligrams per liter to pounds per cubic foot.
Because the EMCs sampled at the cotton field site represent the entire storms sampled, the computed loads at this site are considered storm loads.
Constituent loads computed for the forested site are for the sampled event and not the complete storm event except for the one complete storm event that was sampled at the forested site. The sampled event was subdivided and used to develop relations between percent storm load for a constituent and percent runoff volume, which then were used for estimating storm loads. The relations developed were second-order polynomial regression equations in the form of:
PERCLOAD i is percent of storm load computed for storm i;
b 0 ,b 1 ,b 2 are regression coefficients; and
PERCRV i is percent of storm runoff-volume sampled for storm i.
Storm loads for the forested site for each constituent were computed by dividing the sampled event loads by the estimated PERCLOAD for each storm.
The standard error of estimates (SE) and the coefficient of determination (R 2 ) were computed for each equation. The SE is a measure of the error about the regression. A smaller SE indicates a more precise prediction. The R 2 is the proportion of the variation in the response variable explained by the explanatory variables. A greater R 2 indicates a better fit.
The EMCs for the forested site were computed using the estimated storm-load and the storm-runoff volumes. A storm EMC for a given storm (i) is computed by the equation
EMC i is event-mean concentration of constituent (in milligrams per liter) for storm i;
LOAD i is pollutant load (in pounds) for storm i;
RV i is runoff volume (in cubic feet) for storm i, and;
1.601 x 10 4 is the conversion from pounds per cubic foot to milligrams per liter.
Mean EMCs of all the storm samples combined, were computed for dissolved chloride, suspended solids, total nitrogen, total ammonia plus organic nitrogen, total phosphorus, and dissolved orthophosphorus at both sites. Three methods were used to compute the mean EMCs: volume-weighted, logarithmic-transformed, and arithmetic. The volume-weighted mean EMC was computed using the equation
VWMEMC is volume-weighted mean EMC of a
constituent (in milligrams per liter) for
a site.
The logarithmic-transformed mean EMCs were computed by transforming the EMCs to base-10 logarithms, summing the values, dividing by the number of storms, and retransforming the value. The arithmetic mean was computed by summing EMCs and dividing by the number of storms.
Regression equations were developed to estimate storm loads for unsampled storms for suspended solids, total nitrogen, total ammonia plus organic nitrogen, total phosphorus, and dissolved orthophosphorus at both sites. Size of the data set limited the number of explanatory variables that could be used in the regression analysis. Explanatory variables used were RV and a seasonal factor for the cotton field site, and RV for the forested site. The seasonal factor was used to explain the agricultural condition of the cotton field. Regression equations were not developed for estimating storm loads for dissolved chloride because there is not a good correlation between the dissolved chloride storm loads and the RVs.
For the cotton field and forested sites the regression equations, using a logarithmic transformation (log base-10) of the response and explanatory variables are in the following form:
SEASON is 0 for storms occurring March 1 through July 15, and 1 for storms occurring July 16 through February 28.
The retransformation of a log-transformed regression model provides a consistent estimator of median response but systematically underestimates the mean response (Miller, 1984). Therefore, a bias-correction factor (BCF) needs to be included in the retransformed regression equation if an unbiased estimate of the mean is to be obtained. A BCF was computed for each equation by using a smearing estimate that is a nonparametric method based on the average residuals in original units (Duan, 1983). After applying the BCF to equation 6, the form of the equation becomes
To estimate annual loads at both sites for the 1996 calendar year, the appropriate storm load equations were applied to all of the unsampled runoff-producing storms during the year. Base-flow samples collected at the forested site were used to estimate the base-flow loads that occurred during the year at that site. Annual loads were estimated by summing the sampled loads, the estimated unsampled loads, and at the forested site the base flow loads that occurred during the year.
The constituent loads released during the draining of the catfish pond were determined by multiplying the concentrations by the change in volume of water that occurred between samples. The total loads from the drainage were determined by summing the individual sampled loads.
Selected water-quality data were graphically summarized and compared using boxplots. The Wilcoxon rank sum test (a non-parametric test comparing ranked data) was used to compare selected water-quality data from storms at the cotton field and forest sites.
The USEPA Rapid Bioassessment Protocol II data-analysis methods (Plafkin and others, 1989) require comparisons to a reference site. No suitable bioassessment reference sites representative of relatively undisturbed conditions were found in the Boeuf River Basin. All sites evaluated were affected by land clearing, channelization, and bank disturbances. Boat Gunwale Slash, a least-disturbed reference stream for the Mississippi Alluvial Plain (Delta) ecoregion (Bennett and others, 1987), was used in this assessment as the reference stream for rapid bioassessment comparisons.
Benthic macroinvertebrate data for Boat Gunwale Slash were obtained directly from Bennett and others (1987). The habitat parameter and total scores for the reference site were calculated based upon interpretation of the physical description presented in Bennett and others (1987).
Organisms from the Boeuf River Basin samples were identified to the family level using dichotomous keys (Merritt and Cummins, 1984; Pennak, 1989), enumerated, categorized by tolerance value and functional feeding group (Plafkin and others, 1989; Merritt and Cummins, 1984; and Lenat, 1993), and entered (along with data from the reference site) into a rapid bioassessment protocol metric calculation spreadsheet template provided by the USEPA (Howell, 1989). The spreadsheet subsequently was modified by the USGS to add some families that were present in the study area samples but not included in the provided spreadsheet. Taxa richness, family biotic index, ratio of scraper to filtering collector abundance, ratio of EPT (Ephemeroptera, Plecoptera, and Trichoptera) to Chironomidae abundances, percent contribution of dominant family, EPT index, ratio of shredder to total abundance, and the community loss index are the seven metrics used in the benthic invertebrate Rapid Bioassessment Protocol II (Plafkin and others, 1989). The spreadsheet provided by the USEPA calculated these seven metrics except that, for this study:
(1) ratio of scraper to filterer plus scraper abundance replaced scraper to filtering collector abundance, and
(2) ratio of EPT (Ephemeroptera, Plecoptera, and Trichoptera) to Chironomidae plus EPT abundance replaced ratio of EPT (Ephemeroptera, Plecoptera, and Trichoptera) to Chironomidae abundance.
A metric value "normalized" to the reference site metric value then was calculated for most metrics at each of the Boeuf River Basin sites. The normalized metrics were calculated as a ratio of the metric value at the reference site and the metric value at the Boeuf River Basin site; except the biological condition score for the percent contribution of dominant family was expressed as the actual percent contribution and the community loss index was not compared to the reference station, because a comparison to the reference station is incorporated into the index. Each metric value obtained was given a metric score of 0, 3, or 6, based on criteria given in Plafkin and others (1989). For each site, a biological condition score was calculated by summing the metric scores. Biological condition scores were compared to the biological condition score for the reference site. A biological condition category then was assigned to each bioassessment site based on the comparison of biological condition scores at the bioassessment site to the reference site score and criteria in Plafkin and others (1989).
Relations between benthic macroinvertebrate communities, physical habitat, and water quality were examined using the Spearman's correlation test (Helsel and Hirsch, 1992). Biological community scores were tested for correlation with physical habitat scores and with water-quality values. The strongest correlations also were examined using x-y plots.
Metrics and scores for duplicate samples from sites 12 and 16 were compared using relative percent difference. Relative percent difference was calculated using the formula
where RPD is relative percent difference (percent), and a and b are the values
associated with the duplicate samples from a site.
WATER-QUALITY, BIOLOGICAL, AND HABITAT ASSESSMENT
Concentrations, loads, and other water-quality associated results for the ambient, cotton field, forest, and catfish pond sites are described in this section. Benthic macroinvertebrate and habitat results also are discussed.
Water-quality samples from the 25 ambient sites (which were at locations where recent bank, substrate, or channelization disturbances were not evident) indicate that streams in the Boeuf River Basin typically are turbid and nutrient enriched during the late fall during periods of relatively low flow (fig. 5, table 2). Most suspended solids concentrations (residue at 105 degrees Celsius) ranged from about 50 to 200 mg/L. Most total nitrogen concentrations ranged from about 1.1 to 1.8 mg/L. Much of the nitrogen was ammonia plus organic nitrogen, which typically ranged from about 0.8 to 1.5 mg/L. Total phosphorus and dissolved orthophosphorus typically ranged from about 0.25 to 0.40 mg/L and 0.10 to 0.25 mg/L, respectively.
Data from previous investigations suggest that concentrations of suspended solids and some nutrients in the Boeuf River Basin are higher than in much of the Mississippi Alluvial Plain in Arkansas. Median concentrations of total suspended solids, total phosphorus, and dissolved phosphorus from samples collected during all seasons at several sites in the Mississippi Alluvial Plain (Petersen, 1988; Petersen, 1992) usually were lower than 50 mg/L, 0.25 mg/L, and 0.10 mg/L, respectively. Median concentrations of total ammonia plus organic nitrogen and total nitrogen for the sites in the Boeuf River Basin were similar to medians reported by Petersen (1988, 1992). Total suspended solids and nutrient concentrations for the sites in the Boeuf River Basin generally were similar to concentrations for sites sampled in previous USGS investigations (Bryant and others, 1978; Lamb, 1979; Petersen, 1981). However, in two basins, suspended solids (Flat Bayou) or total phosphorus (L'Anguille) concentrations were substantially lower than concentrations from the Boeuf River Basin. Data collected as part of the U.S. Geological Survey's National Water-Quality Assessment (NAWQA) Program (Coupe, 2002) in the Mississippi Embayment study unit (an area primarily containing the Mississippi Alluvial Plain of Arkansas, Louisiana, Mississippi, and nearby states) indicate that the November-December sample concentrations of total nitrogen, total phosphorus, and dissolved orthophosphorus from the Boeuf River Basin are slightly higher than November-December sample concentrations from several other streams in the Mississippi Alluvial Plain.
Values listed as water-quality standards or guidelines by the Arkansas Pollution Control and Ecology Commission (1998) occasionally were exceeded (or, in the case of dissolved oxygen were less than the standard) in samples from the ambient sites (table 2). However, conditions listed in the standards sometimes provided exceptions or specific sampling criteria that must be met to legally apply the standard. For example, the numeric turbidity standard only applies to turbidities resulting from "waste discharges or instream activity", the total phosphorus guideline does not apply in waters "highly laden with natural silts...which reduce the penetration of sunlight needed for plant photosynthesis...", and the chloride and sulfate standards are based on multiple samples collected over 30 to 360 days. Therefore, the following comparisons to the standards and guidelines are for general comparison and do not necessarily imply violation. The standard for turbidity in channel-altered Delta (Mississippi Alluvial Plain) streams (75 nephelometric turbidity units) was exceeded at about half of the sites. The total phosphorus guideline (0.1 mg/L) was exceeded at all but one site. Chloride (160 mg/L) and sulfate (30 mg/L) standards for the Boeuf River Basin were exceeded at one and two sites, respectively. The primary season dissolved oxygen standard (5 mg/L) was not met at two sites.
Typically the suspended solids and nutrient concentrations from the ambient sites (sampled during a period of relatively low streamflow) were lower than concentrations in runoff from the cotton field but higher than concentrations in runoff from the forest area. These differences indicate that suspended solids and nutrient concentrations in the Boeuf River Basin are affected by streamflow and land use.
5. Ranges and distributions of suspended solids and nutrient concentrations at 25 ambient sites sampled during a period of relatively low streamflow in November and December 1994.
is 
.