Short-term forecasting of soil temperature using artificial neural network

Abstract

Soil temperature is one of the most important meteorological parameters that plays a critical role in land surface hydrological processes. In the current study, artificial neural network (ANN) models were developed and tested for 1 day ahead soil temperature forecasting at 5, 10, 20, 30, 50 and 100 cm depths. Antecedent soil temperatures plus concurrent and antecedent air temperatures were used as inputs for the ANN models. Soil and air temperature data were collected from two Iranian weather stations located in humid and arid regions for the period 2004-2005. The models' accuracies were evaluated using the Nash-Sutcliffe co-efficient of efficiency, the correlation co-efficient, the root mean square error and the mean bias error between the observed and forecasted soil temperature values. The Nash-Sutcliffe co-efficient of efficiency values >0.94 and correlation co-efficient >0.96 for all the ANN models show that the models can be applied successfully to provide accurate and reliable short-term soil temperature forecasts.