(Introduction):

Deforestation and unsustainable management of agricultural and livestock production systems in mountainous areas have caused land degradation and a progressive reduction in the provision of ecosystem services.

(Objective):

This paper elaborates erosion susceptibility mapping applied at the local scale in the Claro river subbasin in the Fila Cruces mountain range, in Southern Pacific Costa Rica.

(Methodology):

The spatial analysis was conducted using a geographic information system (GIS) environment employing two techniques, logistic regression and artificial neural networks, as well as remote sensing tools. Five conditional factors were finally evaluated for the models: land use, geomorphology, slope gradient, distance to streams, and the Normalized Difference Vegetation Index (NDVI). The erosion susceptibility maps were validated through the receiver operating characteristics (ROC) function.

(Results):

The artificial neural network model showed higher predictive power than the logistic regression method based on the calculated value of the Area under the Curve (AUC). The factors with the greatest explanatory power varied depending on the model used.

(Conclusions):

The erosion susceptibility maps showed a high ecological alteration in terms of the probability of occurrence of erosion processes, especially in the upper subbasin, lands mostly occupied by cattle ranching, and a steeply sloped morphology.