Prediction of Drought with Artificial Neural Networks and Genetic Algorithms Using Precipitation by Remote Perception

Abstract

The aim of the study is to determine the performance of the regional agricultural drought prediction by the model of ANN (Artificial Neural Networks) type NARX, using the SPI (Standardized Precipitation Index), SPEI (Precipitation Index Standardized Evapotranspiration), VCI (Vegetation Condition Index) and GCI (Global Climate Indexes). There have been determined 10 homogeneous regions through RAF (regional frequency analysis) and L-moments, defining the most arid region and the index representing their respective time scale (SPEI 6 months) which responds to the growth and development of vegetation in the basin correlation Pearson equal to 0.58. Monthly rainfall and temperatures correspond to PISCO data prepared by SENAMHI-Peru, with space resolution of 0.05 degrees. For prediction, they have determined two groups, the first to build the model with 80% of the registration and validation of the model and the hypothesis with the remaining 20%. The results have been satisfactory prediction accepting the null hypothesis.