Providing Fuzzy Neural Models for Soil Temperature Using Meteorological Data in two Different Climates

The purpose of this study is to provide a model for determining the soil temperature at depths of 5, 10, 20, 30, 50 and 100 cm using a fuzzy-neural network method in two cities of Arak and Rasht. Independent variables included dry air temperature, humid air temperature, humidity, dew point, vapor pressure, pressure, day and hour of sampling, precipitation, radiant energy and ground air temperature. After determining the test and training data for each soil depth, in each city, the type of inference engine and its parameters were determined separately and the engine was constructed. Comparison of model error with similar attempts and regression method showed a decrease in modeling error. The error rate of the model at 5, 10, 20, 30, 50 and 100 cm depths in Rasht was 36/1, 19/1, 22/1, 17/1, 60/1 and 33/1 degrees C Grad and in Arak, respectively, were 0.002, 1.90, 1.61, 1.53, 1.44, and 1.61 ° C; respectively. The results showed that the models of soil surface temperature for 5 to 30 cm in depth in Rasht were more accurate than Arak city. The justification for this is that the climate of Rasht is wet, as a result of the higher velocity of temperature transfer in wet soil, weather parameters are faster and more effective at soil temperature, but at a depth of 50 and 100 cm in both cities with differences They have little or no precision. This indicates that the depths of the soil are less affected by the surface of the earth, and the type of climate has not been more effective in soil simulation at the depths.