Redes neuronales artificiales y funciones de transferencia aplicados a la estimación de humedad de suelo y temperatura del aire

Abstract

Statistical, mathematical and artificial neural networks techniques were used to estimate the behavior of the following variables: the air temperature (AT) and the soil moisture (SM). The AT plays an important role in the heat interchange between the land surface and the atmosphere, whereas the soil moisture helps to maintain the heat balance between the land surface and the atmosphere. These two parameters are especially important in the maintenance of the biosphere. Models for hourly estimation of the AT and hourly estimation of the SM were developed for 15 field stations in Puerto Rico. These models were generalized to regions that exhibit similar atmospheric characteristics to each station by using remote sensing information and competitive neural networks, obtaining hourly maps of AT and SM over Puerto Rico. The experimental and validation process show that the proposed methodologies are a powerful tool for the estimation of the mentioned variables and a valuable contribution in the study of the behavior of these variables.