Abstract Full Text:PDF Recommended articles Mail to Author How to Cite

¹ Batı Akdeniz Tarımsal Araştırma Enstitüsü, Antalya
² Ondokuz Mayıs Üniversitesi Ziraat Fakültesi Tarımsal Yapılar ve Sulama Bölümü, Samsun DOI : 10.16882/derim.2019.539794 Viewed : 2140 - Downloaded : 992 The presence of water in the soil, movement, evaporation and air capacity, decomposition events, microbiological activity, root respiration and vegetative activity are all under the influence of soil temperature. In this study, it is aimed to estimate the soil temperature values using Artificial Neural Networks (ANN) method. Monthly average soil temperature values and other meteorological data in different soil layers (5, 10, 20, 50 and 100 cm) of the Central Black Sea region provinces were obtained from the General Directorate of Meteorology for the years 1971-2015. A three-layer feed-forward ANN structure was created and the Levenberg-Marquardt (LM) algorithm was applied for ANN learning. Monthly meteorological data education data, monthly data test data from 1991- 2000, monthly data between 2001-2015 were used as validation data. Based on climate data and soil layer, 10 different ANN models were created. For the results obtained at different depths in all stations, the coefficient of determination (R²) is between 0.85-0.99, the standard deviation of the estimation error (RMSE) is between 0.24- 3.74 and the mean absolute error (MAE) is between 0.01-2.33. As a result of the study, it was observed that ANN models yielded successful results in the monthly soil temperature calculations of Middle Black Sea Provinces. Keywords : Soil temperature; Artificial neural network; Model