Catchment-scale variability and driving factors of fine sediment deposition: insights from a coupled experimental and machine-learning-based modeling study

Abstract

PurposeFine sediment deposition is an important component of the catchment sediment budget and affects river morphology, biology, and contaminant transfer. However, the driving factors of fine sediment deposition remain poorly understood at the catchment scale, limiting our ability to model this process.MethodsFine sediment deposition and river reach characteristics were collected over the entire river network of three medium-sized (200–2200 km2) temperate catchments, corresponding to 11,302 river reaches. This unique database was analyzed and used to develop and evaluate a random forest model. The model was used to predict sediment deposition and analyze its driving factors.ResultsFine sediment deposition displayed a high spatial variability and a weak but significant relationship with the Strahler order and river reach width (Pearson coefficient r = −0.4 and 0.4, respectively), indicating the likely nonlinear influence of river reach characteristics. The random forest model predicted fine sediment deposition intensity with an accuracy of 81%, depending on the availability of training data. Bed substrate granularity, flow condition, reach depth and width, and the proportion of cropland and forest were the six most influential variables on fine sediment deposition intensity, suggesting the importance of both hillslope and within-river channel processes in controlling fine sediment deposition.ConclusionThis study presented and analyzed a unique dataset. It also demonstrated the potential of random forest approaches to predict fine sediment deposition at the catchment scale. The proposed approach is complementary to measurements and process-based models. It may be useful for improving the understanding of sediment connectivity in catchments, the design of future measurement campaigns, and help prioritize areas to implement mitigation strategies.