Comparison of accuracy, variability and relative bias for two Central Analytical Laboratories for the National Atmospheric Deposition Program
Gregory Wetherbee1, Camille Danielson2, Martin Shafer3, Sybil Anderson4, Nina Gartman5, Chris Worley6 and Amy Mager7
Implementation of new data-collection methods and systems, including changes in laboratories for the National Atmospheric Deposition Program (NADP) requires great care to prevent shifts in the analytical data, which could then be interpreted as real changes in the chemical composition of the atmosphere. In 2018, the NADP moved centralized support operations, including the Central Analytical Laboratory (CAL), from the University of Illinois (UI) to the University of Wisconsin (UW). This change in laboratory could potentially introduce biases into the chemical wet-deposition data for the National Trends Network (NTN – major ions, ammonia, bromide, and pH), the Atmospheric Integrated Research Monitoring Network (AIRMoN, same analytes as NTN plus orthophosphate), and Ammonia Monitoring Network (AMoN, ambient ammonia in air). The former CAL at UI (I-CAL) and the new laboratory at UW (W-CAL) collaborated with the U.S. Geological Survey’s Precipitation Chemistry Quality Assurance Project (PCQA) to implement comparison studies to identify potential shifts in analytical data.
The PCQA supplied 30 natural matrix samples to I-CAL, W-CAL, and four additional volunteer laboratories to evaluate laboratory accuracy and precision. These samples were amended with bromide and phosphate to ensure detection. PCQA also used these solutions to created low-volume (2 – 69 mL) samples. Monthly PCQA interlaboratory-comparison program data were also evaluated to assess W-CAL performance. Additionally, 75 NTN samples were shared between I-CAL and W-CAL for assessment of relative bias. Results for the PCQA monthly samples, the additional 30 natural matrix spike samples, and the low-volume samples were compared to most probable (median) values (MPVs) for each test solution. Results from I-CAL and W-CAL for the 75 shared NTN samples were compared directly. Metrics for acceptance criteria based on current I-CAL quality-control protocols were prescribed in a Readiness Verification Plan (RVP) before the comparison studies began.
Results for monthly PCQA samples indicated acceptable precision and bias for both I-CAL and W-CAL per RVP metrics. Analyses of the shared 75 NTN samples indicate a slight positive bias for W-CAL data relative to I-CAL data for calcium, sodium, and potassium and a slight negative bias for ammonium, chloride, nitrate, specific conductance, and hydrogen ion. Median differences were small and of the same magnitude of analytical detection limits. No bias was observed for W-CAL data for magnesium, bromide, sulfate, and phosphate in the 75 NTN samples. Similar, biases in W-CAL results were observed for the 30 natural matrix spike samples.
Analysis of low-volume samples by W-CAL showed median relative percent biases ranging from: -13% to -1.1% for ammonium, -45% to -0.3% for nitrate, and -2.9% to 2.8% for sulfate. The largest negative relative percent biases were for the smallest sample volumes (2 – 6 mL) and for analytes at concentrations near analytical detection; primarily potassium, bromide, and orthophosphate.
1U.S. Geological Survey, wetherbe@usgs.gov
2Wisconsin State Laboratory of Hygiene, Camille.Danielson@slh.wisc.edu
3Wisconsin State Laboratory of Hygiene, mmshafer@wisc.edu
4University of Illinois, IL State Water Survey, sybilma@illinois.edu
5University of Illinois, IL State Water Survey, ngartman@illinois.edu
6Wisconsin State Laboratory of Hygiene, Chris.Worley@slh.wisc.edu
7Wisconsin State Laboratory of Hygiene, amy.mager@slh.wisc.edu