USGS WRD banner

Concentrations and annual fluxes for selected water-quality constituents from the USGS National Stream Quality Accounting Network (NASQAN), 1996-2000

By Valerie J. Kelly, Richard P. Hooper, Brent T. Aulenbach, and Mary Janet

Quality-Assurance Results

Download instructions || Field-Blanks || Replicates || Field-Matrix Spikes

An important component of the NASQAN Quality Assurance program consists of the collection and evaluation of quality control (QC) samples, including field blanks, replicates, and field-matrix spikes for pesticides. This section provides a brief discussion of QC results, with an emphasis on identifying the implications of these results for load calculations. Separate data tables are available for field blanks (nutrients, organic carbon, major ions and trace elements, pesticides) field-matrix spikes and parameter code definitions and units of measure; replicate data are provided in the ambient data tables.


Data from field blanks were evaluated according to the following objectives: (1) to determine the potential for contamination from the combined processes of sampling, processing, and laboratory analysis; and (2) to identify ambient data that were likely to be associated with contamination.

Results from the field blanks indicate that the potential for systematic contamination across the network was negligible for most constituents. Measured concentrations were less than the minimum reporting level (MRL) for more than 90 percent of all field-blank data. Where detections were observed, results were evaluated in the context of analytical precision and the effect on the associated ambient data. Contamination was considered to be significant where blank values both exceeded twice the MRL and represented a large component (i.e. at least 10 percent) of the corresponding ambient concentration. Only one percent of the field-blank samples contained concentrations that exceeded both of these criteria, primarily associated with four constituents: ammonia-nitrogen (pcode 00608), orthophosphate (00671), dissolved organic carbon (DOC) (00681), and zinc (01090) (Figure 3).

For ammonia-nitrogen and orthophosphate, 13 and 7 percent of field-blank samples, respectively, had detections that exceeded twice the MRL. More than half of these samples contained concentrations that exceeded 10 percent of the associated ambient data, although the relevant ambient concentrations were generally quite low (Figure 4). For both constituents, ambient concentrations associated with the higher potential for contamination were primarily less than 0.05 mg/L. In other words, the blank results for these constituents indicate that considerable analytical 'noise' occured for ambient data in these concentration ranges. These results indicate that appropriate caution should be utilized in interpreting ambient data for ammonia-nitrogen and orthophosphate at concentrations less than 0.05 mg/L. No editing of data was done for load calculations for these constituents, however, because larger error in the low end of the concentration range does not greatly impact estimates for loads.

Results for DOC and zinc were more problematic. While 11 and 14 percent, respectively, of the field-blank samples contained concentrations that exceeded twice the MRL for these constituents, every one of these samples contained concentrations that were greater than 10 percent of the associated ambient data (Figure 5). Furthermore, the potential for significant contamination was not confined to the low concentration range for ambient data. While it is difficult to define the potential for bias from contamination precisely for any individual sample from the field-blank data, these results imply that a small but real potential exists that any value within the entire range of ambient concentrations could potentially be affected by contamination. Consequently, all data for DOC and zinc should be interpreted with considerable caution. Additionally, because of the extreme uncertainty surrounding concentrations for DOC and zinc, estimated loads for these constituents were not calculated for this report.


Analysis of replicate results focused on evaluation of the significance of sampling variability as measured by the replicate samples, i.e., precision error resulting from the sampling, processing, and analytical process. This variability was compared with the total variability as measured by the full range of ambient data. This analysis showed that sampling error due to the measurement process was significantly less than temporal variability inherent in the ambient data (Figure 6). Standard deviations from replicate samples were generally smaller by about one order-of-magnitude than those for the ambient data. The distinction was slightly less pronounced for trace elements, presumably reflecting the increased analytical variability associated with values that are less than or close to the MRL. Suspended sediment showed the largest sampling error, although the general relationship between sampling and temporal variability remained the same. These results imply that error from the measurement process is a negligible component of the total error for load estimates, which is predominantly a factor of infrequent sampling of the range of ambient conditions.

Suspended sediment samples were collected at most stations for two separate analyses: (1) routine concentration, analyzed by District Sediment Laboratories (80154); and (2) sediment concentration associated with sediment chemistry, analyzed by the Georgia Sediment Chemistry Laboratory (50279). The two separate parameter codes reflect different procedures used for sediment separation in the analytical process. Large differences (greater then 30 percent relative to the mean) were observed in about 30 percent of the samples overall, especially pronounced during 1996 (Figure 7). The observed differences were determined to result primarily from temporal variability over the duration of sample collection, as they could not be attributed to differences in sampling and analytical procedures (unpublished data). These results are consistent with the relatively large sampling error that was observed in the routine replicate samples for suspended sediment, and indicate that significant temporal variability for sediment can exist over relatively short time frames.

Field-Matrix Spikes

Results from field-matrix spikes were evaluated to describe accuracy and precision for pesticides. Field-matrix spikes are samples that are fortified in the field with known concentrations of analytes. They are particularly useful for identifying the potential for bias due to matrix interference and degradation during shipping to the laboratory. Field spikes were collected at most stations at least once per year, generally during the period of pesticide run-off to the stream, when concentrations were expected to be relatively high. Results are reported as recoveries of spike added, in percent, and indicate the ranges of uncertainty associated with pesticide data.

The median recovery for all field-spike samples was 98 percent, with the 25th and 75th percentiles equal to 83 and 111 percent, respectively. Ninety percent of all data were between 37 and 151 percent. Examination of field-spike recoveries by individual pesticide compound shows that most show a similar pattern, that is, recovery generally between 80 and 110 percent (Figure 8). Six pesticides show a consistent low bias (median less than 80 percent recovery): desethyl atrazine (04040), DDE (34653), phorate (82664), terbufos (82675), disulfoton (82677), and permethrin (82687). In contrast, four pesticides show a potential for consistent high bias (median greater then 120 percent recovery): tebuthiuron (82670), carbofuran (82674), carbaryl (82680), and azinphos-methyl (82686).

To evaluate the potential impact for data interpretation of the wide range observed for field-spike recoveries, the results were classified according to the ambient concentrations. Samples were included only where the added spike concentration was 0.1 ug/L. A low concentration range was defined as less than twice the added spike, i.e. less than 0.2 ug/L (N=7,255); the high concentration range consisted of those samples with concentrations greater than 0.2 ug/L (N=95). The sum of concentrations for all pesticides in each unspiked sample provided a measure of the ambient pesticide matrix for the spike samples. Significant differences (p<.0001) were observed between these two groups for both the field-spike recoveries and the pesticide matrix (Figure 9). Recoveries associated with high concentration ranges were biased high (median = 126 percent), while those associated with low concentration ranges were not (median = 97 percent). Variability between the two groups was also significantly different (p<0.0001), with much larger variability observed for recoveries in the high concentration range. These results indicate that the potential for matrix interference may increase with pesticide concentration, and that higher concentrations are frequently biased high. As a consequence, caution should be utilized when interpreting pesticide data at higher concentrations.

Return to Results