Soil erosion implications on future food security are gaining global attention because in many areas worldwide there is an imbalance between soil loss and its subsequent deposition. Soil erosion is a complex phenomenon af fected by many factors such as climate, topography and land cover (in particular forest resources, natural vegetation and agriculture) while directly influencing water sediment transport, the quality of water resources and water storage loss. A modeling architecture, based on the Revised Universal Soil Loss Equation, is proposed and applied to evaluate and validate at regional scale potential and actual soil water erosion, enabling it to be linked to other involved natural resources. The methodology benefits from the array programming paradigm with semantic constraints (lightweight array behavioural contracts provided by the Mastrave library) to concisely implement models as composition of interoperable modules and to process heterogeneous data.