A genetic algorithm's novel rainfall distribution method for optimized hydrological modeling at basin scales

Abstract

Rainfall has a dominant role in rainfall-runoff models, with the rendering of these models depending both on the accuracy of the data and on the way that rainfall is spatially allocated. The research proposes a methodological framework where a genetic algorithm (GA)-based method responsible for the spatial distribution of gauge observations at the basin scale is coupled with the HEC-HMS hydrological model to produce simulated discharges of high accuracy. The custom-developed GA is used to divide a 2D space, adhering to specific criteria, into polygonal geometries that represent gauge zones of influences, similar to the Thiessen polygon method concept. Consequently, a collection of vectorial polygonal areas, equivalent in number to the employed monitoring stations, is produced with the areal weights to be used for distributing the rainfall across the case study basin and subsequently to force the hydrological simulations. The generated gauge weights are validated for a different temporal precipitation event. The final outputs expressed through a series of statistical measures, clearly demonstrate the effectiveness of the specific methodology (e.g. R2 and Nash–Sutcliffe are larger than 0.83 and 0.73). The methodology can foster accurate hydrological simulations, especially in cases where there is a limited number of rainfall stations and corresponding observations.