BUSLE: A probabilistic tool for forecasting sediment inflow to large reservoirs from soil loss

Occurrence of geodynamic processes like rainfall-runoff as external or seismic processes as internal, lead to frequent unmanageable problems globally. Heightened extreme rainfall events markedly boost soil erosion and hazards associated with soil erosion. The early forecasting of the main processes involved in soil erosion assists in the early identification of soil erosion risk and can reduce damage by implementing suitable actions (preventive, corrective and palliative). This research focuses on forecasting the average likelihood of soil erosion and posterior sediment inflow to large reservoirs through a Machine Learning (ML) method. This approach comprises a Bayesian Causal Reasoning (BCR) development for proposing an innovative stochastic computation of the Universal Soil Loss Equation (USLE) model that is called BUSLE (Bayesian USLE). An Object-Oriented Bayesian Networks (OOBNs) system has been developed, entirely based on automatic learning process approach and mathematical computations. This BUSLE tool comprises the following parameters: R (Rainfall Erosivity): K (Soil Erodibility): L and S correspond to Topography, Length and degree of slope respectively, C (Crop management) and P (Conservation practices). For this, annual records for all parameters were collected from different sources and repositories. The total soil loss (A) is computed as the multiplication of all the parameters for each sub-basin. BUSLE comprises a total of 10 Bayesian nets, one pear each sub-basin, and a final Master Network that collects the information for all sub-basins. This Master net implements the overall module of erosion and sediment inflow to large reservoirs prediction. This research is applied to the Rules reservoir catchment (Granada province, SE Spain). Results indicate that soil loss is up to 2.75 Mm ³ for the studied period, which is a very low fraction of the total reservoir silting/colmatation. BUSLE allows to assist in water policy-level decision-making, and researchers can additionally evaluate various scenarios alongside the BUSLE tool to improve the prediction accuracy of soil erosion and reservoir silting likelihood.