Design of Trend Monitoring Networks

The design of trend monitoring networks requires specification of both sampling frequencies and station locations. The extended Kalman filter, an algorithm that incorporates the imprecise knowledge of stream quality dynamics, is used to predict the variance of prediction of eight water quality constituents (DO, PO₄-P, NO₃-N, NO₂-N, NH₃-N, fecal coliform, and temperature) as a function of distance downstream from a station. By setting a criterion of maximizing average trend detection power over a river basin, optimal combinations of sampling frequency and number of stations are determined for a fixed constraint on the total basin-wide sampling effort. The results show that, in general, station placement is far less critical than the number of stations used. For most small- to moderate-sized basins and typical constraint levels faced by managing agencies the optimum sampling program will have very few stations (often only one) with relatively high sampling frequencies at each station.