A systematic review and multivariate classification of sediment transport models: integrating PRISMA and factor analysis of mixed data

Le résumé fourni par la source

Erosion and sediment transport are still major challenges to the sustainability of agriculture, water quality, and aquatic ecosystem functioning, especially with changing climatic conditions. Understanding sediment transport processes is critical for researchers, policymakers, and environmental managers interested in formulating effective watershed-management and erosion-control solutions. This work presents a systematic review of sediment-related erosion, delivery, and transport models following the PRISMA 2020 protocol (Preferred Reporting Items for Systematic Reviews and Meta-Analyses). A formal PRISMA-based screening set of 1344 records was assessed against predefined eligibility criteria, resulting in 138 peer-reviewed studies and 73 distinct sediment-related models, including soil-loss, sediment-yield, sediment-delivery, and sediment-transport models. To enable structured comparison across environmental contexts, we applied Factor Analysis of Mixed Data (FAMD), a statistical method suited to integrate qualitative descriptors such as spatial scale, temporal resolution, and data requirement with binary process-output indicators coded as presence/absence variables, including sediment yield, runoff, peak flow, soil loss, and erosion–deposition outputs. This enabled classification of models based on their operational context and dominant process representation. The analysis identified major patterns in model structure and applicability, grouping models according to their spatial and temporal resolution as well as data requirements. An additional heatmap summarizes the suitability of each model across spatial and temporal scales, providing a scale-explicit guide for model selection. To increase decision-making value, we benchmarked representative models using performance metrics Nash–Sutcliffe Efficiency (NSE) and coefficient of determination (R 2 ) derived from validation studies. Benchmarking values, compiled from heterogeneous published sediment-validation studies, varied widely both within and between clusters; because each value is a single, context-dependent validation statistic, they are reported descriptively rather than tested for statistical significance and are not interpreted as a ranking of model accuracy. This integrated framework offers a reproducible, evidence-informed approach for selecting sediment transport models tailored to environmental and operational demands, highlighting the importance of considering spatial extent, temporal resolution, and data availability in model selection. Overall, the five clusters form a clear structure–application gradient, ranging from simple, empirical plot- and field-scale models with low data requirements to process-rich, spatially distributed, data-intensive catchment models. Coupled with the scale-explicit suitability heatmap — which lets users read off candidate models directly from a target spatial and temporal scale — this yields a reproducible framework that helps shortlist structurally plausible model classes from a user's target scale and data availability, subject to local validation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.