ESTIMATION OF SOIL TEMPERATURE FROM AIR TEMPERATURE USING REGRESSION MODELS, ARTIFICIAL NEURAL NETWORK AND ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM (CASE STUDY: KERMANSHAH REGION)
N Parsafar, Safar Marofi · 2011
In order to develop a simple and rational relationship between air temperature and soil temperature at different depths and to compare to Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) a study was conducted during 1992-2005. Air temperatures and soil temperature at depths of 5, 10, 20, 30, 50 and 100 centimeters were measured at Kermanshah station. To determine the optimum functional relationships between parameters the statistical criteria of correlation coefficient, RMSE and MAE were used. Based on correlation coefficient and error parameters, results showed that the regression methods of the third and second degree equations, linear, power and logarithmic had the best estimations, respectively. Also, results showed that the best and worst estimations between air temperature and soil temperature were observed, at 5 and 100 cm soil depths respectively. Results of this study produced a second degree equation and a linear equation (noting their simpticities of application in comparison with the third degree equation) for each soil depth. Based on the correlation coefficients and errors if can be said that obtained this equation is usable for soil depth of 100 cm with poor precision, but in the case of other depths has to high accuracy. The comparisons of regressions, ANN and ANFIS results indicated that ANN estimated more accurately soil temperature.