Abstract
The soil temperature plays a strong role in the critical processes and also has a critical role in surface energy balancing and saving. On the other hand, soil temperature parameters, such as depth of frost penetration, are essential indicators of climate in agriculture, construction and installation of drainage and water piping networks. In this paper, the H-step ahead temperature prediction of soil at the difference depth (5, 10, 20, 30, 50 and 100 centimeters) is modeled. Data of average relative humidity, minimum daily temperatures, average daily temperature, solar radiation and historical data of the previous soil temperature were used for predicting the soil temperature. A compression study among three prediction algorithms, including ordinary least square, multi-layer perceptron and local linear models, are carried out to obtain the best prediction algorithm. On the other hand, to find a reliable and efficient prediction of h-step ahead of soil temperature in various depths, three different prediction models, which is used several input data, are designed. Proposed models are evaluated by several confirmed cases. By evaluating the accuracy criteria and comparing prediction models considering these indices, the best prediction model is introduced. Finally, numerous diagrams and tables are presented to measure the fitted model.