Factors controlling sediment yield at the catchment scale in NW Mediterranean geoecosystems

Purpose This study aimed to (1) increase understanding of the relation between sediment yield and environmental variables at the catchment scale; (2) test and validate existing and newly developed regression equations for prediction of sediment yield; and (3) identify how better predictions may be obtained. Materials and methods A correlation and regression analysis was performed between sediment yield and over 40 environmental variables for 61 Spanish catchments. Variables were selected based on availability and expected relation with diverse soil erosion and sediment transport processes. For comparison, the Area Relief Temperature (ART) sediment delivery model was applied to the same catchments. Sediment yield estimates obtained from reservoir surveys were used for model calibration and validation. Results and discussion Catchment area, catchment perimeter, stream length, relief ratio, Modified Fournier Index, the RUSLE's R factor, and catchments percentage with poor vegetation cover showed highest correlations with sediment yield. Stepwise linear regression revealed that variables representing topography, climate, vegetation, lithology, and soil characteristics are required for the best prediction equation. Although calibration results were relatively good, validation showed that the models were unstable and not suitable for extrapolation to other catchments. Reasons for this unstable model performance include (1) lack of detail and quality of the data sources; (2) large variation in catchment characteristics; (3) insufficient representation of all relevant erosion and sediment transport processes; and (4) the presence of nonlinear relations between sediment yield and environmental variables. The nonlinear ART model performed relatively well but systematically overpredicted sediment yield. A model reflecting human impacts, including dams and conservation measures, is expected to provide better results. This, however, requires significantly more input data. Conclusions Although important insight is obtained into the relation between sediment yield and environmental factors, prediction of sediment yield at the catchment scale requires alternative approaches. More detailed information is required on land cover (change), and the effect of soil conservation measures. Validation of regression equations is a necessity, and better predictions are obtained by nonlinear models.