How does landscape organization determine complex flood runoff behavior? Lessons from geomorphology based mapping and modeling of dominant runoff processes in meso-scale alpine catchmens

Abstract:

In his 1986 paper 'Looking for Hydrologic Laws,' Dooge argued that "most problems in catchment hydrology fall in the category of complex systems with some degree of organization," and suggested meso-scale laws of flood hydrology could be derived from integrating hydrologic knowledge obtained in micro-scale and macro-scale studies. Mapping the spatial distribution of dominant runoff processes (DRP) is one way of describing catchment complexity and organization. Here, we describe how field experimentation and monitoring campaigns have guided development of a classification and hydrological modeling tool for mapping and simulating DRP in mountainous terrain, and evaluate how landscape organization explains flood behavior in meso-scale basins in the Swiss Alps.

Field data indicated that some landforms, although steep, store most rainfall long enough to strongly dampen flood runoff. Areas with such large storage capacity could contribute little during small events but react unexpectedly strongly during extreme events, thereby complicating prediction of extreme floods if they comprise large parts of the basin. The classification scheme of Scherrer and Naef (2003) was extended to allow mapping of dominant runoff processes and their damping strength in mountainous terrain. This mapping tool relates infiltration, storage, and drainage characteristics to hillslope formation processes determined from geomorphological maps, digital elevation models and aerial photography.
Maps of dominant runoff processes were obtained for three meso-scale catchments with strongly contrasting flood runoff behavior: one catchment with fast flood runoff response, one with damped flood behavior, and one with a possible step-change in runoff formation processes, i.e., much stronger reaction during the four largest events on record. The differences in catchment-scale flood behavior were well explained by the mapped spatial organization of dominant runoff processes.

Next, the dominant runoff process (DRP) maps were used as the basis for a spatially distributed rainfall-runoff model specifically designed to reflect our improved understanding of deep runoff formation processes. The model was subjected to three tests to evaluate its usefulness for prediction of extreme floods. First, the model parameterization was based on small scale observations and catchment scale runoff during small events, and then used for predicting extreme flood events (differential split-sample test). Second, the obtained parameters were used to predict extreme floods in the other catchments (proxy-basin differential split-sample test). Both tests gave satisfactorily results, given the uncertainty in precipitation inputs. The third test aimed at evaluating how our understanding of catchment organization improves prediction of flood behavior, by conducting the same differential split-sample test with lumped models calibrated to a small flood event in a Monte Carlo approach. This revealed that the distributed model is particularly useful in our catchment that has the largest difference between small and large floods, because the lumped models suffer from overfitting and show a much wider range of predictions.

The presented work indicates: (a) maps of dominant runoff processes may provide a useful description of catchment organization and complex flood behavior; (b) the developed framework allows meaningful transfer of small-scale hydrologic understanding to the catchment scale; and (c) uncertainties of estimating extreme floods may be reduced by incorporating understanding of hillslope-scale dominant runoff processes.

References:
Scherrer, S., and F. Naef (2003), A Decision Scheme to Indicate Dominant Flow Processes on Temperate Grassland, Hydrological Processes 17: 391--401, doi:10.1002/hyp.1131