Quantifying fine‐grained sediment sources in the River Axe catchment, southwest England: application of a Monte Carlo numerical modelling framework incorporating local and genetic algorithm optimisation

Abstract

Abstract Increasing recognition of the deleterious environmental effects of excessive fine sediment delivery to watercourses means that reliable sediment source assessment represents a fundamental component of catchment planning targeting the protection of freshwater resources and their ecological integrity. Sediment tracing or fingerprinting approaches have been increasingly used to provide catchment scale sediment source information, but there is a need to continue refining existing procedures especially with respect to uncertainty analysis during mass balance modelling. Consequently, an updated Monte Carlo numerical modelling framework was devised and tested, incorporating both conventional and robust statistics coupled with random and Latin Hypercube Sampling (LHS) together with local and genetic algorithm (GA) optimisation. A sediment sourcing study undertaken in the River Axe catchment, southwest England, suggested that the use of robust statistics and LHS with GA optimisation generated the best performance with respect to predicting measured bed sediment geochemistry in six out of eight model applications. On this basis, the catchment‐wide average median sediment source contributions were predicted to be 38 ± 1% (pasture topsoils), 3 ± 1% (cultivated topsoils), 37 ± 1% (damaged road verges) and 22 ± 1% (channel banks/subsurface sources). Using modelling frameworks which provide users with flexibility to compare local and global optimisation during uncertainty analysis is recommended for future sediment tracing studies. Copyright © 2011 John Wiley & Sons, Ltd.