The Agency’s background document correctly calls for inclusion of CWS that could be influenced by urban areas to be included in the survey. Finally, prometon may be in a unique situation. It is widely used for non-agricultural purposes in both agricultural and urban environments. It has been widely observed in streams across the United States, but is not listed on any of the standard pesticide use surveys. All of these examples point to the limitations of using agricultural pesticide use alone as a determinant of the CWS populations, but also suggest that the resulting errors will not be too great. Of course, as part of the basic design a few reference (background) watersheds, with low or no pesticide use, could be included as a quality assurance measure.
10.) One option proposed in the design framework is a census of the facilities serving the largest cities. These facilities are believed to have little agriculture in their watersheds, but reflect the drinking water for a large percentage of the population.
Does the SAP believe sampling CWSs serving very large populations is useful even if
they are expected to be of low vulnerability?
The SAP agrees that sampling the large population CWSs is necessary. One of the main objectives of the study is to determine the human exposure to pesticides from surface waters. Therefore, sampling CWSs serving the very large populations should be carried out with an inclusion probability of 1.0. If sampling of large population CWSs is not implemented, it would be an assumption in this study that these CWSs have low levels of pesticides and impose an insignificant human exposure.
The question of whether to sample or not sample large population CWSs is an issue of science policy and must balance the risks associated with making decisions based on validated or un-validated assumptions. However, for CWSs where there is verification of little use and occurrence, sampling might be performed at a reduced frequency. This would also seem dependent upon the choice of primary sampling domain. If pesticide use is the primary domain, and none of the largest systems are in the domain for that pesticide, then they may not be necessary. Also, the large systems often do more complete sampling and monitoring. There may already be data for at least some compounds. It could at least guide a choice of frequency. Large systems might be willing to bear the burden of some limited sample collection and monitoring. Some balanced approach will be necessary, because large systems are complex and the monitoring could be very costly if not designed carefully.
11.) Modeling should allow the Agency to reduce the extent of future drinking water surveys, and help to better identify areas at higher risk.
Are there any changes to the design which will better allow the survey to support
model development and testing?
The Agency's proposal to develop a stratified random sample of CWSs with oversampling of