random differences might be that samples from A experience wider temperature fluctuations resulting in more variable degradation histories. If the measurement errors are different because of the random components have different sizes, then there would be an impact on the optimal allocation of sampling effort. Precision can be measured as the reciprocal of the standard error.
7.) Annual mean pesticide concentrations occurring at any CWS vary from year to year. A multi-year study would help to quantify year-to-year variability but is more costly.
Does the SAP have any suggestions for assessing annual variability given the financial
constraints of the survey?
Would drawing out the survey over three years (with the same number of samples per
CWS) improve it?
It is clear that both spatial and temporal variability are important in characterizing pesticide residues in drinking water. It is clear from the Agency’s background document that a significant amount of variability in pesticide concentrations in surface waters will be related to year-to-year changes in the dynamics of climate fluctuations, usage trends, and pest pressure. The answers to the questions asked have more to do with the relative importance (magnitude) of temporal variability than spatial variability.
The SAP is in favor of incorporating some degree of temporal sampling in the study. A multi-year survey effort is required to meet either of the goals of the survey--exposure estimates and model construction. One suggestion from the Agency was to draw the survey out over three years. There are a number of ways of doing this. One approach is to sacrifice spatial coverage by reducing the number of CWSs surveyed to a third of that proposed, with each CWS sampled for a longer period of time. Another approach is to simply sample only one-third of the total CWS in any one year with a non-overlapping set in any one year, though this approach could loose information on year-to-year changes for individual systems.
With no apparent historical data on year-to-year trends in pesticide levels for a wide range of CWSs, there appears to be little that can be done except to extend the study for several years to collect data to estimate year-to-year variability in annual average pesticide levels and upper percentiles of pesticide concentration distributions in CWSs. The number of extra years to collect data will be a function of the cost and the amount of year-to-year variability that is observed in CWS pesticide levels. The issue of whether the particular year of a one year study is representative or not could be evaluated by comparison with longer-term data series such as NAWQA or the Heidelberg College data from Ohio. Long-term precipitation and flow information would also offer a useful perspective, though this approach could use information in year-to-year changes for individual systems.
One issue that should be addressed is to not only monitor CWSs in future years that exhibited high pesticide concentrations but a random sample across all CWSs in the baseline study since CWSs with high concentrations may be either in high years or just be at elevated levels in