specify the desired number of samples per stratum, or at least to understand the balance between sample size and power. Although this was not explicitly addressed in the presentations, it is presumably the whole reason for stratifying in the first place. It allows optimizing the sampling strategy by focusing on the most vulnerable systems. Notice that shifting samples to vulnerable strata does not mean that we are ignoring the CWSs where the concentrations are not high. Nor would stratification necessarily bias the distribution upward. We still get the entire distribution, with all concentrations high and low. We're just making sure that the estimate of its right tail is especially good.
In some of the Agency's discussion, it was inferred that small CWSs were more vulnerable than larger CWSs. Studies done by SAP members suggested that this was not accurate to generalize to the size of the water system. Smaller watersheds tend to be more vulnerable to higher concentrations of pesticides than larger watersheds, but this does not translate to system size (often measured by the population served or the volume of water produced).
4.) EFED is proposing to use a geographic information systems (GIS) watershed characterization tool for identifying CWSs in the domains and strata.
Does the SAP agree with the usefulness of this tool for these purposes?
We are aware of limitations in the accuracy of the data used in the site selection tool.
Is the panel aware of better ways of handling the accuracy and precision limitations of this data?
Is the Panel aware of alternative data sources that may not have these limitations?
The development of a technically sound approach for estimating human exposure to pesticides from surface-derived drinking water depends critically upon the development and use of the proposed GIS watershed characterization tool. The approach appears to be building on extensive national data sets and therefore appears to maximize the use of readily-available information. One SAP member noted that the initial step for this effort should be a characterization of the surface water sources for each CWS using the GIS tool. Thus, an objective and defensible approach to determining how limited resources should be allocated needs to be developed. It was emphasized that this population characterization should be viewed as a first step in the development of a sound monitoring strategy. The FIFRA SAP agrees with the Agency that there are limitations about use of these data.
The availability of new national GIS layers such as the National Hydrologic Data (NHD) and the National Elevational Data (NED) makes this characterization more accurate (and therefore more useful) than ever before. Automatic delineation of watersheds is now computationally feasible, although this may be subject to large error in areas of low relief. A concern is the quality of other types of information and the appropriate scale at which it was collected (e.g. the pesticide use data). This is especially the case as it pertains to small watersheds both to the availability of information at that scale (e.g., 8-digit hydrologic units may not be appropriate in many cases) as well as the accuracy of these data if they are available. There is not