Charles G. Crawford

Estimation of suspended-sediment rating curves and mean suspended-sediment loads

Suspended-sediment loads are often estimated from an empirical relation between suspended-sediment load (L) and streamflow (S). This relation is usually defined as a power function, L = aSb, and is referred to as a suspended-sediment rating curve. This function can be formulated as either a linear or non-linear model to find the solution of the rating-curve parameters (a and b). Formulation of the power function as a linear model requires a logarithmic transformation to linearize the function and a subsequent correction for transformation bias. Rating-curve parameter estimates for both the bias-corrected, transformed-linear or non-linear models can be obtained by the method of least squares. Each model has distinct advantages and disadvantages. A unique solution of the parameters of the transformed-linear model may be obtained algebraically. These parameter estimates have some optimal properties when certain attainable conditions are met. However, the parameter estimates must be corrected for transformation bias when obtained this way. Parameter estimates obtained for the non-linear model do not require a correction for transformation bias. However, these estimates must be obtained by iterative methods which do not always converge to a solution. In addition, the residual errors of the non-linear model typically are not identically distributed throughout the range of streamflow values. This problem adversely affects the precision of the parameter estimates. Weighted non-linear least squares can be used to improve the parameter estimates for the non-linear model, but the weights must be approximated and their appropriate form may be difficult to determine. A simulation study was done to evaluate: (1) the accuracy and precision of parameter estimates for the bias-corrected, transformed-linear and non-linear models obtained by the method of least square; (2) the accuracy of mean suspended-sediment loads calculated by the flow-duration, rating-curve method using model parameters obtained by the alternative methods. Parameter estimates obtained by least squares for the bias-corrected, transformed-linear model were considerably more precise than those obtained for the non-linear or weighted non-linear model. The accuracy of parameter estimates obtained for the bias-corrected, transformed-linear and weighted non-linear model was similar and was much greater than the accuracy obtained by non-linear least squares. The improved parameter estimates obtained by the bias-corrected, transformed-linear or weighted non-linear model yield estimates of mean suspended-sediment load calculated by the flow-duration, rating-curve method that are more accurate and precise than those obtained for the non-linear model.

Mississippi River Nitrate Loads from High Frequency Sensor Measurements and Regression-Based Load Estimation

Accurately quantifying nitrate (NO3-) loading from the Mississippi River is important for predicting summer hypoxia in the Gulf of Mexico and targeting nutrient reduction within the basin. Loads have historically been modeled with regression-based techniques, but recent advances with high frequency NO3- sensors allowed us to evaluate model performance relative to measured loads in the lower Mississippi River. Patterns in NO3- concentrations and loads were observed at daily to annual time steps, with considerable variability in concentration-discharge relationships over the two year study. Differences were particularly accentuated during the 2012 drought and 2013 flood, which resulted in anomalously high NO3- concentrations consistent with a large flush of stored NO3- from soil. The comparison between measured loads and modeled loads (LOADEST, Composite Method, WRTDS) showed underestimates of only 3.5% across the entire study period, but much larger differences at shorter time steps. Absolute differences in loads were typically greatest in the spring and early summer critical to Gulf hypoxia formation, with the largest differences (underestimates) for all models during the flood period of 2013. In additional to improving the accuracy and precision of monthly loads, high frequency NO3- measurements offer additional benefits not available with regression-based or other load estimation techniques.

Factors Affecting Stream Nutrient Loads: A Synthesis of Regional SPARROW Model Results for the Continental United States

Abstract We compared the results of 12 recently calibrated regional SPARROW (SPAtially Referenced Regressions On Watershed attributes) models covering most of the continental United States to evaluate the consistency and regional differences in factors affecting stream nutrient loads. The models – 6 for total nitrogen and 6 for total phosphorus – all provide similar levels of prediction accuracy, but those for major river basins in the eastern half of the country were somewhat more accurate. The models simulate long-term mean annual stream nutrient loads as a function of a wide range of known sources and climatic (precipitation, temperature), landscape (e.g., soils, geology), and aquatic factors affecting nutrient fate and transport. The results confirm the dominant effects of urban and agricultural sources on stream nutrient loads nationally and regionally, but reveal considerable spatial variability in the specific types of sources that control water quality. These include regional differences in the relative importance of different types of urban (municipal and industrial point vs. diffuse urban runoff) and agriculture (crop cultivation vs. animal waste) sources, as well as the effects of atmospheric deposition, mining, and background (e.g., soil phosphorus) sources on stream nutrients. Overall, we found that the SPARROW model results provide a consistent set of information for identifying the major sources and environmental factors affecting nutrient fate and transport in United States watersheds at regional and subregional scales.

Long-term trends in alkalinity in large rivers of the conterminous US in relation to acidification, agriculture, and hydrologic modification

Alkalinity increases in large rivers of the conterminous US are well known, but less is understood about the processes leading to these trends as compared with headwater systems more intensively examined in conjunction with acid deposition studies. Nevertheless, large rivers are important conduits of inorganic carbon and other solutes to coastal areas and may have substantial influence on coastal calcium carbonate saturation dynamics. We examined long-term (mid-20th to early 21st century) trends in alkalinity and other weathering products in 23 rivers of the conterminous US. We used a rigorous flow-weighting technique which allowed greater focus on solute trends occurring independently of changes in flow. Increasing alkalinity concentrations and yield were widespread, occurring at 14 and 13 stations, respectively. Analysis of trends in other weathering products suggested that the causes of alkalinity trends were diverse, but at many stations alkalinity increases coincided with decreasing nitrate+sulfate and decreasing cation:alkalinity ratios, which is consistent with recovery from acidification. A positive correlation between the Sen-Thiel slopes of alkalinity increases and agricultural lime usage indicated that agricultural lime contributed to increasing solute concentration in some areas. However, several stations including the Altamaha, Upper Mississippi, and San Joaquin Rivers exhibited solute trends, such as increasing cation:alkalinity ratios and increasing nitrate+sulfate, more consistent with increasing acidity, emphasizing that multiple processes affect alkalinity trends in large rivers. This study was unique in its examination of alkalinity trends in large rivers covering a wide range of climate and land use types, but more detailed analyses will help to better elucidate temporal changes to river solutes and especially the effects they may have on coastal calcium carbonate saturation state.

Regional and temporal differences in nitrate trends discerned from long-term water quality monitoring data

Riverine nitrate (NO3) is a well-documented driver of eutrophication and hypoxia in coastal areas. The development of the elevated river NO3 concentration is linked to anthropogenic inputs from municipal, agricultural, and atmospheric sources. The intensity of these sources has varied regionally, through time, and in response to multiple causes such as economic drivers and policy responses. This study uses long-term water quality, land use, and other ancillary data to further describe the evolution of river NO3 concentrations at 22 monitoring stations in the United States (U.S.). The stations were selected for long-term data availability and to represent a range of climate and land-use conditions. We examined NO3 at the monitoring stations, using a flow-weighting scheme meant to account for interannual flow variability allowing greater focus on river chemical conditions. River NO3 concentration increased strongly during 1945-1980 at most of the stations and have remained elevated, but stopped increasing during 1981-2008. NO3 increased to a greater extent at monitoring stations in the Midwest U.S. and less so at those in the Eastern and Western U.S. We discuss 20th Century agricultural development in the U.S. and demonstrate that regional differences in NO3 concentration patterns were strongly related to an agricultural index developed using principal components analysis. This unique century-scale dataset adds to our understanding of long-term NO3 patterns in the U.S.

Watershed Regressions for Pesticides (WARP) Models for Predicting Stream Concentrations of Multiple Pesticides

Watershed Regressions for Pesticides for multiple pesticides (WARP-MP) are statistical models developed to predict concentration statistics for a wide range of pesticides in unmonitored streams. The WARP-MP models use the national atrazine WARP models in conjunction with an adjustment factor for each additional pesticide. The WARP-MP models perform best for pesticides with application timing and methods similar to those used with atrazine. For other pesticides, WARP-MP models tend to overpredict concentration statistics for the model development sites. For WARP and WARP-MP, the less-than-ideal sampling frequency for the model development sites leads to underestimation of the shorter-duration concentration; hence, the WARP models tend to underpredict 4- and 21-d maximum moving-average concentrations, with median errors ranging from 9 to 38% As a result of this sampling bias, pesticides that performed well with the model development sites are expected to have predictions that are biased low for these shorter-duration concentration statistics. The overprediction by WARP-MP apparent for some of the pesticides is variably offset by underestimation of the model development concentration statistics. Of the 112 pesticides used in the WARP-MP application to stream segments nationwide, 25 were predicted to have concentration statistics with a 50% or greater probability of exceeding one or more aquatic life benchmarks in one or more stream segments. Geographically, many of the modeled streams in the Corn Belt Region were predicted to have one or more pesticides that exceeded an aquatic life benchmark during 2009, indicating the potential vulnerability of streams in this region.

Comparison of gas chromatography/mass spectrometry and immunoassay techniques on concentrations of atrazine in storm runoff

Gas chromatography/mass spectrometry (GC/MS) and enzyme-linked immunosorbent assay (ELISA) techniques were used to measure concentrations of dissolved atrazine in 149 surface-water samples. Samples were collected during May 1992–September 1993 near the mouth of the White River (Indiana) and in two small tributaries of the river. GC/MS was performed on a Hewlett-Packard 5971 AUse of brand names is for identification purposes only and does not constitute endorsement by the U.S. Geological Survey, the Uniroyal Chemical Company, or Wichita State University. with electron impact ionization and selected ion monitoring of filtered water samples extracted by C-18 solid phase extraction; ELISA was performed with a magnetic-particle-based assay with photometric analysis. ELISA results compared reasonably well to GC/MS measurements at concentrations below the Maximum Contaminant Level for drinking water set by the U.S. Environmental Protection Agency (3.0 μg/L), but a systematic negative bias was observed at higher concentrations. When higher concentration samples were diluted into the linear range of calibration, the relation improved. A slight positive bias was seen in all of the ELISA data compared to the GC/MS results, and the bias could be partially explained by correcting the ELISA data for cross reactivity with other triazine herbicides. The highest concentrations of atrazine were found during the first major runoff event after the atrazine was applied. Concentrations decreased throughout the rest of the sampling period even though large runoff events occurred during this time, indicating that most atrazine loading to surface waters in the study area occurs within a few weeks after application.

Regression models for explaining and predicting concentrations of organochlorine pesticides in fish from streams in the United States

Empirical regression models were developed for estimating concentrations of dieldrin, total chlordane, and total DDT in whole fish from U.S. streams. Models were based on pesticide concentrations measured in whole fish at 648 stream sites nationwide (1992-2001) as part of the U.S. Geological Surveys National Water Quality Assessment Program. Explanatory variables included fish lipid content, estimates (or surrogates) representing historical agricultural and urban sources, watershed characteristics, and geographic location. Models were developed using Tobit regression methods appropriate for data with censoring. Typically, the models explain approximately 50 to 70% of the variability in pesticide concentrations measured in whole fish. The models were used to predict pesticide concentrations in whole fish for streams nationwide using the U.S. Environmental Protection Agencys River Reach File 1 and to estimate the probability that whole-fish concentrations exceed benchmarks for protection of fish-eating wildlife. Predicted concentrations were highest for dieldrin in the Corn Belt, Texas, and scattered urban areas; for total chlordane in the Corn Belt, Texas, the Southeast, and urbanized Northeast; and for total DDT in the Southeast, Texas, California, and urban areas nationwide. The probability of exceeding wildlife benchmarks for dieldrin and chlordane was predicted to be low for most U.S. streams. The probability of exceeding wildlife benchmarks for total DDT is higher but varies depending on the fish taxon and on the benchmark used. Because the models in the present study are based on fish data collected during the 1990s and organochlorine pesticide residues in the environment continue to decline decades after their uses were discontinued, these models may overestimate present-day pesticide concentrations in fish.

Century-scale perspective on water quality in selected river basins of the conterminous United States

..........................................................................................................................................................

Datasets for the Report Entitled: "A Method for Addressing Differences in Concentrations of Fipronil and Three Degradates Obtained by Two Different Laboratory Methods"

This report provides data input and computation results for a method developed by Crawford and Martin (2017) to address differences in concentrations of fipronil and three degradates obtained by two different laboratory methods. Data are arranged in 9 tables that include water-quality site information, laboratory recovery data, laboratory analyses results and measured water-sample concentrations analyzed by the two laboratory methods, and estimated concentrations from the older method removing the effects of method differences.