INTEGRATING REMOTE SENSING AND PREDICTIVE MODELING TO ASSESS SOIL ERODIBILITY IN BRAZIL

Soil plays a crucial role in providing essential ecosystem
services, yet it is a non-renewable resource threatened by
erosion. This study aims to leverage remote sensing and
machine learning techniques to accurately map soil
erodibility across Brazil, improving upon traditional methods
that often rely on low-resolution data. We calculated the K
factor, which represents soil erodibility, using highresolution
maps of soil attributes generated from a
comprehensive database. The findings indicate significant
variability in erodibility values, ranging from 0.01 to 0.12 Mg
mm MJ⁻¹ ha⁻¹, with 23% of soils classified as very high
erosion risk and 35% as high risk. Additionally, the analysis
of Bare Soil Frequency (BSF) reveals that perennial crops
exhibit lower soil exposure compared to annual crops, with
sugarcane displaying the highest rates of soil exposure. These
results underscore the urgent need for sustainable agricultural
practices to mitigate soil erosion and safeguard this vital
resource in high-risk areas.