Estudio de las relaciones lluvia escorrentía mediante redes neuronales

Abstract

Abstract: This research analyzes the rainfall-runoff relationship of the Rircay River basin, Ecuador, using artificial neural networks of the multilayer perceptron type (ANN-MLP) as an alternative to traditional hydrological models. Daily precipitation and streamflow records from 2004 to 2014 were used, integrating hydrometeorological information, time lags, antecedent soil moisture conditions, and simplified physical variables of the basin. Through the spatial delimitation of the study area, land use and land cover were characterized, followed by the application of the SCS hydrological classification method and the calculation of the CN curve number. The model was trained under a supervised scheme and evaluated using performance metrics. Among the results obtained were values ​​of MAE = 5.61 m³/s, RMSE = 8.78 m³/s and R2 = 0.786, demonstrating a behavior capable of reproducing the temporal variability of flow that the data represent.