This paper investigates the potential of data mining techniques to predict daily soil temperatures at 5-100 cm depths for agricultural purposes. Climatic and soil temperature data from Isfahan province located in central Iran with a semi-arid climate was used for the modeling process. A subtractive clustering approach was used to identify the structure of the Adaptive Neuro-Fuzzy Inference System (ANFIS), and the result of the proposed approach was compared with artificial neural networks (ANNs) and an M5 tree model. Result suggests an improved performance using the ANFIS approach in predicting soil temperatures at various soil depths except at 100 cm. The performance of the ANNs and M5 tree models were found to be similar. However, the M5 tree model provides a simple linear relation to predicting the soil temperature for the data ranges used in this study. Error analyses of the predicted values at various depths show that the estimation error tends to increase with the depth.
Modeling soil temperature based on Gaussian process regression in a semi-arid-climate, case study Ghardaia, Algeria
