Modeling of Soil Temperature Using Meteorological Factors, Multivariable Regression and Artificial Neural Networks (Case Study: Bandar- Abbas Synoptic Station)

Soil temperature is an important and influential parameter on plant growth that is not measured regularly and continuously at all weather stations. Thus, soil temperature data is lacking. Soil temperature differs in depth and is affected by ambient. The aim of this study is modeling soil temperature in the different depths (5, 40, 20, 30, 50 and 100 cm) and clay- sandy texture using meteorological factors, multivariate regression and artificial neural network in the Bandar Abbas synoptic station during 1993-2017. Result showed that the air temperature, pan evaporation and dew point have the highest correlation coefficient with soil temperature. The mean absolute error (MAE) is 1.09-1.88°C (from 10 to 100 cm of soil depth) in the multivariate regression model while it is 1.17-1.85 °C in the artificial neural networks. Thus, multivariate regression model is proposed due to the simplicity and the lack of significant difference with artificial neural network model. This model can be used in similar regions to predict soil temperature in different depth.