The Prediction through Artificial Intelligence Approach and Geographic Information Systems (GIS) of the Soil Temperature in Kastamonu Province, Turkey

Abstract

Abstract In the present work, the average soil temperature of Kastamonu province was predicted by artificial neural networks approach employing data gained from five various meteorological measurement districts located in provincial borders. Twenty-two years of (2000–2021) monthly average atmosphere temperature data achieved from soil depths (5, 10, 20, 50, and 100 cm) have been utilized for artificial intelligence structure. It has been compared monthly average soil temperature for Cide, Devrekani, İnebolu, Kastamonu City Center, and Tosya stations. Measured and estimated soil temperature values have been exceedingly related to the Correlation Coefficient values (R2), Mean Absolute Error (MAE), Mean Square Error (MSE), and Average Relative Deviation (ARD). As a result, the estimated soil temperature findings were in the acceptable range with the measured data with average R2 values of 0.9851, 0.9456, 0.9712, 0.9691, and 0.9586 for Cide, Devrekani, İnebolu, Kastamonu CC, and Tosya, the respectively. MAE of 0.6808°C to 0.6848°C, ARD of 0.010–10.674% and MSE of 0.144 and 4.109 at all measurement districts where insignificant error tendency is very clear.