Comparison of Mamdani and Sugeno Fuzzy Models for Soil Temperature Estimation

Abstract

Soil temperature is an important variable that directly impacts the growth of plants. In this study, the feasibility of predicting soil temperature at different depths using Multiple Linear Regression (MLR), Mamdani fuzzy inference system, and Artificial Neuro-Fuzzy Inference System (ANFIS) was assessed. The soil temperatures below the ground were measured at depths of 10, 20, 50, and 100 cm. The study was conducted in Kastamonu province of Turkiye. Meteorological data was obtained from the State Meteorological Service. Air temperature and soil depth data were used as input in order to predict the soil temperature as output. For the training set, 2 out of 7 available stations were selected and the other 5 stations were used as the testing data. In order to assess the negative/positive effect of the selected training sets mentioned here, another scenario that used %70 of the total data for training and the remaining %30 for testing was developed. In order to test model performance, mean absolute error, root-mean square error, coefficient of determination, and Nash-Sutcliffe efficiency were employed. Taylor diagrams and violin plots were used in order to compare the types of models. The results showed that ANFIS was the most successful model, followed by MLR and Mamdani. It was concluded that soil temperature at different depths could be estimated using air temperature and depth data.