Estimating Soil Temperature Using Modern Methods of Data Analysis

Soil temperature as one of the important environmental parameters, is measured only in synoptic weather stations and three times per day, while the other meteorological parameters such as air temperature and relative humidity are measured most in meteorological stations except rain guage stations. In this research which based on three methods of multivariate linear regression, artificial neural network and k-nearest neighbor were evaluated to estimate the soil temperature at different depth based on air temperature and relative humidity using daily data from synoptic meteorology stations of Gorgan from 1996 to 2005. The results showed that with increasing soil depth, the effect of meteorological parameters and estimation accuracy will decrease.more accuracy of shallow depth soil temperature is due to the greater influence of climatic factors on soil temperature and less time delay of heat transfer from the surface to the deeper depths. By comparing the observed and predicted soil temperature values from the various models, it will be concluded that the best performance is obtained from the k-nearest neighbor algorithm which has higher R2 and less RMSE. Artificial neural network and regression models, are in second and third place to predict soil temperature.