There is much policy interest in the possible linkages that might exist
between land use and downstream fluvial flood risk. On the one hand,
this position is sustained by observations from plot- and field-scale
studies that suggest land management does affect runoff. On the other,
upscaling these effects to show that land-management activities impact
upon flood risk at larger catchment scales has proved to be elusive.
This review considers the reasons for why this upscaling is problematic.
We argue that, rather than it reflecting methodological challenges
associated with the difficulties of modelling hydrological processes
over very large areas and during extreme runoff events, it reflects the
fact that any linkage between land management and flood risk cannot be
generalized and taken out of its specific spatial (catchment) and
temporal (flood event) context. We use Sayer's (1992) notion of a
`chaotic conception' to describe the belief that there is a simple and
general association between land management and downstream flood risk
rather than the impacts of land management being spatially and
temporally contingent in relation to the particular geographical
location, time period and scale being considered. Our argument has
important practical consequences because it implies that land-management
activities to reduce downstream flood risk will be different to
traditional flood-reduction interventions such as levees. The purpose of
demonstration projects then needs careful consideration such that
conclusions made for one project are not transferred uncritically to
other scales of analysis or geographical locations.