Efficient use of a genetic algorithm for long-term groundwater monitoring design

This paper summarizes a methodology for designing long-term monitoring plans using groundwater fate-and-transport simulation, global mass estimation, and a genetic algorithm. Kriging and inverse distance weighting are the plume interpolation methods used to attain global mass estimates. Kriging provides the most accurate global mass estimates but has the drawback of having an increased computational complexity relative to inverse distance weighting. A hybrid method demonstrates how initial solutions found using inverse distance weighting can be refined using kriging to substantially reduce computational effort. Theoretical relationships available in the evolutionary computation literature were used to set the control parameters for the genetic algorithm, substantially reducing the number of trial runs required to ensure optimal or near optimal solutions. Results from the test case show that sampling costs could be reduced by as much as 60 percent without significant loss in accuracy of the global mass estimates.