Predicción de la precipitación a partir de variables meteorológicas utilizando modelos de regresión funcional.

Abstract

In this investigation we use Functional Linear Regression Models where the response variable belongs to the scalar field, while the explanatory covariates have a functional structure in the integrable square space. Previously, exploratory analysis techniques and outlier detection, within the framework of Functional Data Analysis, are discussed. Then, three different types of functional linear regression with scalar response involving a single explanatory covariate are applied, and the adaptation of a fourth model involving two explanatory covariates at the same time is carried out. The proposed methodology is used with the purpose of choosing the best adjustment to predict rainfall in different corn producing areas in Ecuador, by means of functional explanatory covariates such as temperature and wind speed. These results will be very useful to calculate the value of the effective precipitation, fraction of water used by the crop, from which the value of the real evapotranspiration of the plant is obtained and, more importantly, the irrigation requirements are determined at each stage of corn production to understand the behavior of the productivity of this crop through its adequate physiological development.