GRAZING IMPACT ON LAND DEGRADATION IN THE CURVATURE SUBCARPATHIANS (ROMANIA) – PRELIMINARY RESULTS
Gabriel MINEA1,2*, Nicu CIOBOTARU1,2, Gianina NECULAU1,2, Georgiana TUDOR1,2, Sevastel MIRCEA1,3
1Research Institute of the University of Bucharest, University of Bucharest, 90 Panduri Street, 7 Sector 5, 050107, Bucharest, Romania
2National Institute of Hydrology and Water Management, 97 E Bucureşti - Ploieşti Road, Sector 1, 013686, Bucharest, Romania
3University of Agronomic Sciences and Veterinary Medicine of Bucharest, 59 Marasti Blvd, Sector 1, 011464, Bucharest, Romania
*Corresponding author: Gabriel Minea (gabriel.minea@unibuc.ro)

Abstract
In various environments, unsustainable grazing is a pressure factor on land and water resources quality. We deal with grazing impact on land degradation based on machine learning algorithms (e.g., random forest) using 18 explanatory variables (e.g., grazing density, tree-cover density, slope, grassland probability, etc.) in the Curvature Subcarpathians (Romania). The data for model training, validation, and testing consisted of over 4300 sampling points taken by Google Earth imagery, within the study area (6792 sq km), on binary form, erosion or not erosion. The model has been implemented in R environment, with randomforest package, which offered good results in terms of accuracy (92%) and ROC curve, indicating a class A performance model on testing data. Here, we report preliminary results and found that the main explanatory variables to land degradation are tree cover density, slope factor, land use land cover type, profile curvature, and aspect of the hills while grazing density and grassland probability playing a lesser role in the land degradation process. Also, we discovered that the Curvature Subcarpathians geomorphologic region is highly exposed to the erosion process, affecting the hillslopes, grazing fields, and agricultural land as well. The less exposed areas to erosion are the forested areas and the large river valleys (e.g., Buzău river alluvial fan). Overall, based on our approach we concluded that land degradation in the Curvature Subcarpathians region is unlikely affected by grazing.
This work was supported by a grant of the Romanian Ministry of Education and Research, CNCS - UEFISCDI, project number PN-III-P1-1.1-TE-2019-1180, within PNCDI III.

Keywords: grazing, land use, machine learning, Curvature Subcarpathians