Valoración multiescalar de un modelo de regresión para el pronóstico de sequías en la microcuenca del río Chulco usando predictores climáticos y descomposición Wavelet

Abstract

Drought is one of the most important hydrological events because of the magnitude of the scope; which has environmental, social and economic impact. Thus, drought forecast for water resources management is essential. Machángara River sub-basin provides different ecosystem services, excelling water supply for different sectors. Therefore, a predictive model in the microbasin of Chulco River was carried out using climatic indices and meteorological variables. First, three types of droughts: meteorological, agricultural, and hydrological were characterized through the Standardized Precipitation Index (SPI), Effective Reconnaissance Drought Index (eRDI) and the Standardized Flow Index (SSI), respectively. Also, in order to analyze time series at the component level, Wavelet Discrete Transform (DTW) was applied to decompose the indices. Then, the performance of the model was evaluated by applying K-Fold and Hold-out validation. The results indicate that (a) climatic indices, meteorological variables, and drought indices correlate with the different types of droughts. However, its influence depends on the response time to the precipitation deficit and the hydrological cycle. (b) The most favorable validation process is the random type (K-Fold). (c) For the SPI 1 and SPI 3 index, the best predictors correspond to local variables, while climatic indices are used at 6 and 12-month scales. (d) The eRDI and SSI indices are best predicted using the same indices. It indicates dependence between drought types