A parallelized multi-objective particle swarm optimization model to design soil sampling network

Abstract

Optimization of soil sampling network is a complex optimization problem, which must reconcile a series of conflicts such as survey budget, sampling efficiency and sampling barriers, etc.. High computational cost of this problem motivated the applications of parallel computation algorithms. Our study proposes a parallelized multi-objective particle swarm optimization model (PMOPSO), which combines minimum mean kriging variance and minimum survey budget as the objectives. The model was applied to optimize soil sampling network of Hengshan County in loess hilly area in China. The performance of the PMOPSO model was compared to that of sequential MOPSO. The results indicate that the PMOPSO model can improve the computational efficiency and fitness values of the objectives significantly at the expense of the convergence rate.