Application of Monte Carlo sampling and Latin Hypercube sampling methods in pumping schedule design during establishing surrogate model

Abstract

For creating surrogate model of the groundwater numerical simulation model of Jinquan Industrial Park in Inner Mongolia, the application of Monte Carlo sampling method and Latin Hypercube Sampling method in pumping test design is studied. Firstly make sure the pumping load of each pumping wells obeys uniform distribution, then generate Monte Carlo samples and Latin Hypercube samples according to their own methods. Analyses these two results comparing with each other, it suggests that in small sample size, Monte Carlo sampling method has a low sampling efficiency, low coverage of the sampling value to population and needs a large amount of calculation, while Latin Hypercube Sampling method relatively improves the sampling efficiency, coverage of the sampling value to population, and reduce the workload. Latin Hypercube Sampling method has better practicability in this job.