Comparison of observed and DEM-driven field-to-river routing of flow from eroding fields in an arable lowland catchment.

• Flow/sediment routing strongly controlled by interactions with landscape elements. • Despite this, observed routes replicated using only DEM and steepest-descent model. • But other attributes of flow, controlled by landscape elements, not well simulated. • Therefore models need explicit representation of interaction with landscape elements. • Inadequate representation explains findings of previous model validation studies. Field-to-river flow of runoff and sediment in a lowland arable catchment in the south of England is explored from both field and modelling perspectives. Routes observed to be taken by flow and sediment on five study areas include many interactions between flow and 'landscape elements' (LEs), including those (field boundaries, paths, roads) of anthropogenic origin. We were able to satisfactorily replicate observed flow routes using a simple steepest-descent-with-overtopping model with a 5 m DEM. This was unexpected, considering the narrowness of linear LEs such as paths and tracks. However LE attributes showed considerable sensitivity: changing just one attribute of a single FE-flow interaction notably altered the route taken by simulated flow, while changing LE attributes notably affected synthetic hydrographs for flow reaching the river, suggesting similar impacts upon transported sediment reaching the river. Thus while simple steepest-descent and overtopping permits satisfactory replication of observed flow routes, it is likely that more explicit representation of LE-flow interactions is necessary in order to adequately capture the dynamics of field-to-river runoff and sediment transport, as must be done by catchment-scale erosion models. This will enable such models to better represent runoff speed and volume, and the flux and size distribution of transported sediment, with the aim of overcoming some broad limitations of such models as noted in earlier model validation studies. Finally, we consider the representation of some LE-flow interactions in several catchment-scale models, and discuss the ways in which such representation might be improved. [ABSTRACT FROM AUTHOR]