Fatemeh Nooshin Nokhandan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Fatemeh Nooshin Nokhandan has a total of 3 co-authors. Here are the top co-authors by number of shared publications:
- Erzsébet Horváth
Department of Physical Geography, Eötvös Loránd University, Pázmány Péter sétány 1/c, H-1117 Budapest, Hungary
1 shared publication - Kaveh Ghahraman
Department of Physical Geography, Eötvös Loránd University, Budapest, Hungary
1 shared publication - Ágnes Novothny
Department of Physical Geography, Eötvös Loránd University, Pázmány Péter sétány 1/c, H-1117 Budapest, Hungary
1 shared publication
Latest Publications
- Open Access
- 6 December 2025
- Published by MDPI AG in Remote Sensing
- Vol. 17 (24) , 3950
Advancing Soil Erosion Mapping in Active Agricultural Lands Using Machine Learning and SHAP Analysis
Soil erosion is a significant land degradation process in Hungary, especially in agricultural regions. This study assesses soil erosion susceptibility in a loess-covered, intensively cultivated area near Úri and Mende (central Hungary) using Random Forest and Light Gradient Boosting Machine (LightGBM) models. A balanced erosion inventory (500 erosion-affected and 500 non-erosion points) and thirteen geo-environmental factors were used to generate erosion susceptibility maps. Permutation importance and Shapley Additive Explanations (SHAP) identified slope, land use/land cover (LULC), and NDVI as the most influential predictors. The susceptibility maps indicate that 43% (Random Forest) and 46% (LightGBM) of the study area fall within the High and Very High susceptibility classes, with croplands being the most vulnerable. Random Forest achieved AUROC = 0.90, Overall Accuracy = 0.81, RMSE = 0.38, MAE = 0.14, and Kappa = 0.70 for the test dataset; LightGBM achieved AUROC = 0.91, Overall Accuracy = 0.82, RMSE = 0.39, MAE = 0.16, and Kappa = 0.67 for the test dataset. The results identified erosion-prone areas and confirm the reliability of the models. They also highlight the key driving factors as critical determinants of erosion susceptibility. The findings provide a solid foundation for designing targeted soil conservation measures and supporting sustainable land management strategies in central Hungary.
- 6 December 2025
- Published by MDPI AG in Remote Sensing
- Vol. 17 (24) , 3950
Open AccessAdvancing Soil Erosion Mapping in Active Agricultural Lands Using Machine Learning and SHAP Analysis
Soil erosion is a significant land degradation process in Hungary, especially in agricultural regions. This study assesses soil erosion susceptibility in a loess-covered, intensively cultivated area near Úri and Mende (central Hungary) using Random Forest and Light Gradient Boosting Machine (LightGBM) models. A balanced erosion inventory (500 erosion-affected and 500 non-erosion points) and thirteen geo-environmental factors were used to generate erosion susceptibility maps. Permutation importance and Shapley Additive Explanations (SHAP) identified slope, land use/land cover (LULC), and NDVI as the most influential predictors. The susceptibility maps indicate that 43% (Random Forest) and 46% (LightGBM) of the study area fall within the High and Very High susceptibility classes, with croplands being the most vulnerable. Random Forest achieved AUROC = 0.90, Overall Accuracy = 0.81, RMSE = 0.38, MAE = 0.14, and Kappa = 0.70 for the test dataset; LightGBM achieved AUROC = 0.91, Overall Accuracy = 0.82, RMSE = 0.39, MAE = 0.16, and Kappa = 0.67 for the test dataset. The results identified erosion-prone areas and confirm the reliability of the models. They also highlight the key driving factors as critical determinants of erosion susceptibility. The findings provide a solid foundation for designing targeted soil conservation measures and supporting sustainable land management strategies in central Hungary.