Optimal design of additional sampling pattern for drinking-water quality control

A crucial step in complementary studies of the water quality determination is to design the additional sampling pattern or, in other words, to determine the number and location of the additional samples. Since location and number of samples highly affect decisions’ uncertainties, and because sampling process is quite costly and time-consuming, optimization of the sampling pattern will enhance the efficiency and productivity. Solution of such an optimization problem requires defining an objective function and constraints. There exist many previous studies regarding locating additional samples in other environmental problems wherein the objective function is defined as the minimization of the kriging variance (based on the problem nature), but the point is that kriging variance is not sensitive to local variability. Since manner and extent of small-scale variations are both important and necessary in water quality studies, it is required to resolve this shortcoming of the traditional objective function. Solution is to make use of the combined variance consisting of kriging and local variances. In this study, the applicability and efficiency of the minimization of combined variance as the objective function of the additional sampling was adopted and proved for a salt marsh (east of Iran) on the basis of a simulated annealing-based algorithm. It was shown, practically, that the locational distribution of additional sampling points is quite logical and more compatible with experts’ proposed methods using this objective function (compared to the traditional one).