Exploring soil organic carbon dynamics based on climatic change in the Central Black Sea Region through machine...

Recent studies have been conducted to mitigate the effects of global climate change by re-sequestering carbon released into the atmosphere into the soil and increasing soil organic carbon (SOC) levels. The present study analyses the organic carbon content of soil samples collected in different periods (2005 (I. Period) and 2022 (II. Period)) within the Central Black Sea climate zone. Climate parameters Mean Annual Precipitation (MAP) and Mean Annual Temperature (MAT) and climate indices (Emberger Pluviothermic Ratio (QE), UNEP-UNCCD Aridity Index (AI), De Martonne Drought Index (IDM), Fournier Climate Aggressiveness Index (MFI), and Bagnouls-Gaussen Drought Index (BGI)) were used in the analysis. Projections for the III. period (2034) were estimated using the Time Series SARIMA model. Additionally, this study aimed to evaluate the predictability of SOC content in the future period (III) using various machine learning algorithms, including Multivariate Linear Regression (MLR), Support Vector Machine Regression (SVR), Artificial Neural Networks (ANN), Random Forest Regression (RF), and XGBoost Regression. According to the SARIMA time series model, the RMSE values obtained for the prediction of mean precipitation and temperature ranged between 1.377-1.817 mm and 23.595-58.039 °C, respectively. Among the machine learning models, the lowest error rate in predicting the SOC content for 2034 was achieved with the RF algorithm (RMSE: 0.39, MAE: 0.28, R²: 0.61), and IDM was identified as the most influential parameter in the prediction. The predicted SOC values for 2034 were found to be lower levels (1.46-1.95%) compared to previous periods. The SARIMA model projected an increase of approximately 1-1.5 °C in future temperature values, while precipitation was expected to decrease. The aridity index indicated that regions previously not at risk of desertification during periods I and II were projected to become more susceptible to desertification by 2034.