This is a question that has two dimensions, each with a range of different situations. There is no one-size-fits-all answer to adequately address this problem. Ideally, the final choice of sampling frequency should be determined on a individual basis based on the chemical of interest, the hydrologic system of interest and data quality determination.
The first dimension is that of the chemicals. There are two sets of considerations embedded in this. One is based on pesticide fate properties and this has been discussed in the Agency’s background document. The other is the timing and number of applications. For some chemicals on some crops, there is usually only one application. For other chemicals there may be up to six applications a year. For urban areas, there may essentially be a continual application in a given watershed throughout the active growing season. These two factors together (persistence and application) will govern how long of a duration the pesticide source will exist in a given basin. For many agricultural chemicals, particularly in the mid and northern latitudes, the pesticide source will be strongest for a single four-month period (April to July). For other examples, such as pesticides in other environments (urban pesticides and the southern latitudes and the west coast) the pesticide source may be active for two sets of four month periods. Each of these generalized situations ideally might be sampled in a different manner to adequately capture the annual mean concentration.
Given the high cost of analysis, the number of samples that are analyzed should be limited. The chemographs presented in the various documents clearly support the concept of variable sampling intensities for different seasons. A priori knowledge on the fate properties (especially half-life, times of chemical application, hydrology, etc.) should be used to optimize sampling intensity.
The second dimension is that of the hydrologic system. Streams and rivers need to be considered separately from reservoirs and lakes. Reservoirs and lakes are probably the easiest situation to sample. The pesticide concentrations will change relatively slowly - as a function of pesticide inputs, pesticide persistence, turnover time, and hydraulic residence times. For long-lived pesticides, the sample frequency may be relatively low and still result in an adequate representation of the annual mean. The ratio of the chemical half-life and hydraulic residence time may provide guidance on the minimum sampling frequency required. Residence time reflects a reservoir-oriented perspective; a much larger problem is short term variability in flowing water systems. Streams and rivers, on the other hand, are much more dynamic than lakes and reservoirs. Rivers and streams form a continuum from drainage ditches to continental rivers. The duration of a specific pesticide peak concentration will vary throughout this continuum. Sampling frequencies will need to be greater in flowing water systems (except the continental rivers), perhaps by a factor of three to four or even more, to obtain comparable confidence intervals around the annual mean concentration.
It might be possible to obtain more information for a given amount of sampling by using daily flow data (at stations where it is available) as an auxilliary variable. This approach is