Soil Temperature Prediction for Konya, Türkiye: Machine Learning Approaches

Authors : Kazım Kumaş, Ali Özhan Akyüz

Pages : 76-88

View : 66 | Download : 82

Publication Date : 2024-12-31

Article Type : Research Paper

Abstract :Soil temperature is a critical parameter for agriculture meteorology applications. Although highly accurate, direct measurement may not be practical over large areas. The measurement process can also be costly and time-consuming. On the other hand, variables such as surface and soil properties that affect soil temperature can make it difficult to predict with physical models. Machine learning methods can overcome various limitations and predict targeted variables using complex non-linear relationships in the data distribution. For this purpose, it is used in many fields. Machine learning approaches are sensitive to input data and require many training data. This paper studied 5, 10, 20, and 50 cm soil temperature values of Konya province between 1960 and 2ied using machine learning algorithms (k-nearest neighbors, adaptive boosting, gradient boosting, light gradient boosting machine (LGBM)). The models were trained using data from 1960 to 2017, and the years 2019, 2020, and 2021 were predicted. In line with the successful results achieved, these models were used to predict the years 2022, 2023, 2024, and 2025.
Keywords : : Konya, Machine Learning, Prediction, Soil Temperature