Development of an innovative approach for calculating soil erosion by applying machine learning algorithms

Soil erosion is a critical issue to human development and the need of innovative approaches for improved prediction and monitoring is crucial, since the consolidated available empirical methods pose several limitations in soil erosion studies. This thesis investigated the potential application of advanced technologies in soil erosion modeling through the integration of machine learning algorithms, advanced remote sensing and high-resolution data, and cloud computing infrastructures. Indeed, machine learning proved effective results in capturing complex, non-linear relationships between erosion factors, hyperspectral remote sensing enabled detailed mapping of erosion prone areas by detecting key soil properties, while cloud computing platforms like Google Earth Engine allowed large-scale, multi-temporal geospatial analyses. Four case studies were used to demonstrate the potential application of new methods through a progression from simpler, small-scale methods to more complex, innovative approaches, each varying in application size, resource demands, and intrinsic complexity addressing the lack of field validation with a comparison with similar studies. The outcomes underscored the potential of integrating ML, advanced and high-resolution data, and cloud computing to enhance soil research, proposing a new approach to calculate and monitor soil erosion considering the advantages of a combined use of those technologies.