The U.S. Geological Survey (USGS) used seven distinct programs to provide external quality-assurance monitoring for the National Atmospheric Deposition Program / National Trends Network (NTN) and Mercury Deposition Network (MDN) during 2005-06. Overall variability of NTN data was estimated using a co-located-sampler program. Variability and bias of NADP chemical analysis data were estimated using two separate interlaboratory-comparison programs, one for each network. Bias in MDN sample analysis data for total mercury (Hg) was evaluated using a blind-audit program. A separate blind-audit study was conducted in 2006 for the NTN chemical analysis data to evaluate laboratory analytical detection limits. The sensitivity of NADP measurements was evaluated using a field-audit program for NTN and a system blank program for MDN. Overall measurement sensitivity and laboratory analysis sensitivity are evaluated by comparison to data-quality objectives (DQOs) for each network. Field-audit results for 2005-06 indicate that DQOs for NTN overall measurement sensitivity were met during 2005-06. Network maximum contamination levels (NMCLs) have been increasing in NTN samples since 2005 for all constituents except magnesium, sulfate, and hydrogen ion, but increases in 3-year moving average NMCLs between 2002 and 2006 were less than the 10 percent. NMCLs determined from the field-audit data are compared to analytical detection limits to evaluate sample analysis sensitivity. Although the DQO for NTN sample analysis sensitivity was not attained for magnesium, ammonium, chloride, nitrate, and sulfate in selected 3-year time periods prior to 2005, the DQO was met for all constituents during 2005-06. Interlaboratory-comparison program results indicate variability and bias in NTN data are low relative to data from other participating laboratories. Central Analytical Laboratory (CAL) sulfate and hydrogen-ion data were slightly positively biased during 2005, and CAL potassium and hydrogen-ion data were slightly positively biased during 2006. CAL analyses of deionized-water samples indicated possible low-level potassium contamination in NTN samples during 2005-06. CAL data were within statistical control during at least 90 percent of 2005-06 with few exceptions. CAL analyses of synthetic wet-deposition solutions were within +10 percent of most probable values (MPVs) except for selected sodium and potassium analyses. A blind-audit study, in which the CAL analyzed quality-control samples disguised as real samples, confirmed CAL’s reported detection limits. In fact, CAL detection limits for nitrate and sulfate might actually be lower than reported by CAL. ... Three pairs of co-located samplers were used to estimate overall variability of NTN wet-deposition measurements in terms of median absolute error (MAE). MAEs were less than 15 percent for nitrate and sulfate concentrations, specific conductance, and collector catch for water year 2005 (WY05), whereas MAEs for these analytes were less than 8 percent for WY06. Consistent with co-located sampler data collected during previous water years, MAEs for cations were higher than for anions. Median absolute difference (MAD) values for the co-located samplers indicated that NADP DQOs for uncertainty were met for most analytes during WY2005-06. For the MDN system-blank program, the median system-sample minus bottle-sample Hg concentration difference was 0.027 nanogram per liter (ng/L), which is 21 percent of the HAL method detection limit (MDL). The calculated NMCL for the combined 2005 and 2006 system-blank differences is approximately 0.42 ng/L, which is an order of magnitude higher than the 2004 NMCL of approximately 0.04 ng/L. This implies that contamination levels in MDN samples might have increased during 2005-06 as compared to 2004. In response, USGS submitted followup diagnostic samples to either confirm or reject this conclusion. The diagnostic results failed to confirm an increase in MDN sample contamination. The NMCL (0.42 ng/L) is less than the 10th percentile of all 2004-06 MDN data, and thus the draft DQO for Decision Rule 1 for sensitivity has been met. The NMCL is more than 2 times the MDL (0.13 ng/L) reported by the HAL for 2006. Therefore, the sensitivity of the HAL analytical measurements is acceptable per the draft DQO Decision Rule 2. DQO decision rules are described in the report. For the MDN interlaboratory comparison program, HAL reported data with the lowest variability among the participating laboratories during 2005 but had the second highest variability during 2006. The median difference between HAL-reported concentrations and MPVs was small (-0.25 ng/L) during 2005, and no significant (α=0.05) bias was detected. During 2006, however, the HAL data were negatively biased. Control charts were used to show that HAL reported four values (less than 4 percent of all results) outside statistical control. The 2005 median Hg concentration for HAL blanks (0.94 ng/L) is approximately 11 percent of the median result of 8.80 ng/L for all valid 2005 MDN samples associated with measurable wet deposition. The 2006 median Hg concentration for HAL blanks (0.28 ng/L) is approximately 3 percent of the median result of 9.50 ng/L for all valid 2006 MDN samples associated with measurable wet deposition. The MDN blind-audit program officially began during 2006 to evaluate accuracy of HAL Hg analyses. The median total Hg percent recovery for 13 blind-audit samples was 97 percent with an f-pseudosigma (nonparametric standard deviation analogue) of 11 percent. Sample volume, which affects how much Hg mass is available for analysis, did not correlate with Hg percent recovery. Overall, the external quality-assurance program results indicate that HAL analytical performance meets MDN DQOs. Overall MDN measurement sensitivity also meets DQOs and is adequate to distinguish between measurement noise and environmental signals.
|Published - Jan 2009
|ISWS Data Case Study 2009-01