Automatic interpretation of pumping tests data using metaheuristics

Abstract

Pumping tests data interpretation is of major importance in groundwater engineering. It is traditionally performed in a subjective manner by means of standard type curves. In this paper, an automatic interpretation of time-drawdown data has been proposed based on two algorithms, the real-coded genetic algorithm and differential evolution. The proposed approaches combine metaheuristic algorithms with an appropriate analytical drawdown solution depending upon the nature of the considered aquifer system, leaky or naturally fractured rock aquifers. The standard error of estimate (SEE) was used as a performance criterion to evaluate the discrepancies between predicted and observed drawdown data in different pumping time periods. Both of the proposed metaheuristic algorithms provide accurate aquifer parameters. For all analyzed pumping tests data, the differential evolution yielded the most accurate results with an improvement in SEE values ranged from 0.2 to 50% compared to previously published results, and exhibits speed and robustness. Furthermore, a new full range numerical evaluation of the Hantush well function based on a tanh-sinh quadrature scheme has been proposed. Our results indicate that our method is accurate, since the maximum relative error was found to be equal to 0.00036%, and practical, since it is free from special functions and could be easily incorporated within an optimization technique to analyze transient time-drawdown data.