Enhanced hillslope storage is utilised in "natural" flood management in order to retain overland storm run-off and to reduce connectivity between fast surface flow pathways and the channel. Examples include excavated ponds, deepened or bunded accumulation areas, and gullies and ephemeral channels blocked with wooden barriers or debris dams.

The performance of large, distributed networks of such measures is poorly understood. Extensive schemes can potentially retain large quantities of run-off, but there are indications that much of their effectiveness can be attributed to desynchronisation of sub-catchment flood waves. Inappropriately sited measures may therefore increase, rather than mitigate, flood risk. Fully distributed hydrodynamic models have been applied in limited studies but introduce significant computational complexity. The longer run times of such models also restrict their use for uncertainty estimation or evaluation of the many potential configurations and storm sequences that may influence the timings and magnitudes of flood waves.

Here a simplified overland flow-routing module and semi-distributed representation of enhanced hillslope storage is developed. It is applied to the headwaters of a large rural catchment in Cumbria, UK, where the use of an extensive network of storage features is proposed as a flood mitigation strategy. The models were run within a Monte Carlo framework against data for a 2-month period of extreme flood events that caused significant damage in areas downstream. Acceptable realisations and likelihood weightings were identified using the GLUE uncertainty estimation framework. Behavioural realisations were rerun against the catchment model modified with the addition of the hillslope storage. Three different drainage rate parameters were applied across the network of hillslope storage.

The study demonstrates that schemes comprising widely distributed hillslope storage can be modelled effectively within such a reduced complexity framework. It shows the importance of drainage rates from storage features while operating through a sequence of events. We discuss limitations in the simplified representation of overland flow-routing and representation and storage, and how this could be improved using experimental evidence. We suggest ways in which features could be grouped more strategically and thus improve the performance of such schemes.