Modeling of daily soil temperature using synoptic data and neural network

Soil moisture, as the soil hydrologic parameters, can be affected by soil temperature and controls various hydrological processes. Given the importance of this issue, in this study, the efficiency of artificial neural network was studied to simulate soil temperature at 5- 100 cm depth. Recorded meteorological parameters in the Isfahan synoptic station were used to simulate the soil temperature at different depths. The structure of the neural network was formed with an input layer, a hidden layer and an output layer and network training was done by Levenberg–Marquardt algorithm. Also test and error was done to determine a number of suitable neurons in hidden layer. The results showed that error in both neural network and ANFIS model increases with depth increase that can be due to the weak correlation between soil temperature changes in the lower layers and climatic parameters.