Stochastic search algorithms for optimal design of monitoring networks

The design of efficient monitoring networks is critical for a better understanding of environmental, ecological, and epidemiological processes. In this paper, we develop for the optimal design of monitoring networks a new hybrid genetic algorithm (HGA) which combines the standard genetic algorithm (GA) with a local search (LS) operator. We compare the performance of our HGA with two other stochastic search algorithms, a simulated annealing (SA) algorithm and a standard GA. Specifically, we consider the reduction of pre-existing large-scale monitoring networks, when the optimality criterion is the maximization of the entropy of the included stations. The algorithms were tested on a set of simulated datasets of different sizes, as well as on a real application involving the downsize of a large-scale environmental monitoring network. In each of the cases considered the HGA outperformed the other two algorithms. Copyright (C) 2009 John Wiley & Sons, Ltd.