Appendixes and Data Quality Assurance


Evaluation of Quality-Assurance Data


Quality-assurance (QA) water samples consisted of field and laboratory blanks, field and laboratory (churn split) replicates, matrix spike samples, and standard reference samples. Appendix A lists the various types of QA samples for each type of data; appendix B provides the QA results for dissolved organic carbon (DOC) and optics; appendix C shows QA data for disinfection by-products (DBPs) in finished water; appendix D shows QA data for DBP formation potential (DBPFP) measurements; appendix E shows QA results for nutrients.


Organic carbon and optical properties—QA samples for carbon and optical properties included four laboratory replicates (churn split) and two field replicates (appendix B). While the average relative percent differences (RPDs) for replicate samples were within 5–6 percent for total fluorescence, DOC, ultraviolet absorbance (UVA), fluorescence index (FI), and four of five carbon component loadings, higher average RPDs were found for component C5 (11 percent), spectral slopes (8–11 percent), humic index (HIX) and peaks B, T, and N (20–36 percent). The one finished-water replicate had some relatively high RPDs (appendix B), but in many cases, these represented small actual differences in concentrations that were at or near the detection limit. 


More variability in C5 values produced higher standard deviations compared with other components, possibly because the protein-like organic matter is highly variable amongst sites and seasonal trends (Yamashita and Tanoue, 2003). Reproducibility and precision of measurements of protein-like peaks in natural waters are reduced at shorter wavelength, because of interference in background fluorescence (Yamashita and Tanoue, 2003). Instability of the protein peaks can be due to changes in the complex structures of the protein. The fluorescing component of proteins found in organic matter resides in the residuals of the protein folds. Tryptophan and tyrosine display high anisotropies that are usually sensitive to protein conformation and the extent of motion during the excited state. This leads to highly variable natural lifetimes of proteins and makes it difficult to have accuracy in the reproducibility of the fluorescence of these particular fluorophores. 


DBP—QA samples for DBPs included “in-house” samples prepared by the laboratory and one blank, three replicates, and one standard reference sample submitted blindly during the project. Each set of 10 samples analyzed by the laboratory included 1 blank sample, pre- and post-CCV (continuing calibration verification) samples, and 1 matrix‑spiked sample (regular sample spiked with 0.01 mg/L). 


The 1 blank sample submitted blindly contained no detections, and 12 in-house laboratory blanks for chloroform and bromodichloromethane also resulted in no detections. The standard reference sample for DBPs indicated a high percent relative difference for a few compounds, notably bromoform and chloroform (appendix C). The percent recovery for chloroform was just 87 percent, and although this may represent a low bias, this represented only a small difference between expected and reported concentrations of 0.01 mg/L (appendix C). In-house matrix spike samples (0.01 mg/L) for these two compounds resulted in average percent recoveries of 98 ± 13 percent and 104 ± 9 percent, respectively, whereas recovery ranges were 74–125 percent and 78–117 percent. The RPDs among replicate spike samples were about 3 percent for both compounds (Adriana Gonzalez-Gray, Alexin Laboratories, written commun., 2012). Field replicate values showed more variation in trihalomethanes (THMs) (notably chloroform) compared with haloacetic acids (HAAs), although absolute differences between replicates was again small, 0.001–0.004 mg/L (appendix C).


DBP Formation Potentials—One blank sample contained a low-level concentration of chloroform at the detection limit of 0.002 mg/L (appendix B). Given the much higher concentrations present in environmental samples, this low-level detection does not affect the results or interpretations. Field DBPFP replicates showed variations of 5–10 percent for THMs and 5–20 percent for HAAs; the highest relative differences were for trichloroacetic acid (TCAA [appendix D]). Samples not meeting the upper acceptable range of 30 percent generally had low DPB concentrations (less than 0.010 mg/L), thus the absolute differences were small. Differences between field duplicates were largest for chloroform differences between field duplicates were largest for chloroform, which is more volatile compared with the other DBPs, and thus most prone to loss due to the occasional air bubble. The occurrence of bubbles in some DBPFP samples may have resulted in a low bias for the highly volatile compounds such as chloroform, but as the standard reference results presented below show, this does not appear to be a widespread issue.


The generally accepted range for THM and HAA recoveries is between 70 and 130 percent. The RPD between expected and reported results for the seven DHBA samples tested was 0.2–16.2 percent (average 7.6 percent), well within this range (appendix D). Errors associated with these analyses include the preparation of 3,5-dihydroxy-benzoic acid (DHBA) standard solution from a solid powder, steps involved with chlorine dosing and quenching, potential loss of the volatile THMs due to bubble formation or air leaks, and variations in the determination of the THM compounds by the laboratory. In addition, because of high concentrations of chlorinated DBPs in the formation potential samples, analyses usually required dilution. Although this additional step could introduce errors, no QA or quality control (QC) issues were apparent in replicate or standard reference samples. Nevertheless, some loss of the more volatile compounds, particularly chloroform, could have occurred during sample transfer for the high-concentration samples requiring this dilution step and may cause a low bias in reported concentrations.


Nutrients—The nutrient data had considerable QA issues, including numerous reported total phosphorus (TP) concentrations that were less than soluble reactive phosphorus (SRP) concentrations and unacceptably high variation (high RPDs) in laboratory replicate “churn-split” samples for SRP, TP, ammonium (NH₄), silica (Si), total particulate carbon (TPC), and total particulate nitrogen (TPN) (appendix E). Nitrate concentrations appeared more reliable; laboratory split replicates were within 5 percent of one another. Given the concerns about the quality of the nutrient data, these data were utilized sparingly during analysis, and nutrients were not good explanatory variables during the study. 


The average RPDs for TPC and TPN for three laboratory replicate split samples also were relatively high—25 and 34 percent, respectively (appendix E). Given that TOC concentrations were calculated by summing the TPC and DOC concentrations, the potential errors in TPC values could affect calculations of the percentage of particulate carbon compared to dissolved, which is duly noted.


The appendix data files are included in an Excel^© Workbook and are available for download at http://pubs.usgs.gov/sir/2013/5001. This workbook consists of the following worksheets.


Appendix A. Number and Type of Quality-Assurance Samples.


Appendix B. Quality-Assurance Data for Dissolved Organic Carbon and Selected Optical Properties.


Appendix C. Quality-Assurance Data for Disinfection By-Products in Finished Water.


Appendix D. Quality-Assurance Data for Disinfection By-Product Formation Potentials.


Appendix E. Quality-Assurance Data For Nutrients and Total Particulate Carbon.


Appendix F. Spearman Rank Correlations for Select Groups of Samples and Sites.


Appendix G. Discrete Data Used in the Analysis, Including Watershed, Finished-Water, and Jar-Test Samples.


First posted February 11, 2013