A dominant processes framework of hydrological modelling across scales

The standard procedure for representing different hydrological settings in hydrological modelling is to adjust the model parameters but leave the model structure unchanged. There are indications that the suitable choice of the model structure is just as important, as the hydrological processes may differ vastly, depending on the setting. The main hypothesis investigated in this project is that a single model structure cannot capture the wide variety of hydrological processes. Specific model structures are needed at the right scales, dependent on the dominant controls. The overall approach consists of using the dominant hydrological processes to identify a suitable structure of hydrological models. The processes are identified by combining the downward approach of classifying hydrological processes according to their controls, and an upward approach of hydrological simulation. The dominant processes approach to modelling will involve the use of a classification scheme based on the idea of the "mapping" between landscape structure and model structure. The framework is based on a hierarchy of spatial scales where the largest-scale processes are considered to be the main controls and the effects become more subtle as one goes down in spatial scale. The process controls so identified are indexed by hydrological signatures such as the seasonality of flow which are used as a basis for identifying a suitable model structure. The model structures will hence differ in terms of the components they consist of. For example, they may or may not represent processes such as rapid macropore flow and shallow groundwater response, and may involve different soil moisture-runoff relationships. The method will be developed and tested for three regions with strong hydrological gradients. Two of the regions lie in Peru where annual precipitation ranges from a few millimetres per year in the desert areas of the coast to more than 4000 mm/yr in the Andes mountains. The third region is Austria where annual precipitation ranges from 400 mm/yr in the Pannonian lowlands of the East to 3000 mm/yr at the northern fringe of the Alps. Also, landforms and the soil/geological settings differ vastly. It is anticipated that the gradients in these regions will allow to more clearly identify the strengths and weaknesses of the method than would be possible in regions with spatially more uniform controls. The project breaks new scientific ground in at least three areas: A hierarchy of dominant process controls will be identified based on the idea of co-evolution of soils, vegetation and landscapes. Methods from comparative hydrology will be used to learn from differences in different parts of the landscape and map them into the modelling space. Finally, a new modelling framework will be developed where the model structure depends on the dominant process controls. The new model framework will be an important contribution to hydrology and useful for better understanding the water balance under current and future climate regimes, to analyse the effects of changes in other controls (e.g. land use) and understand the interplay of hydrological and ecological processes at a range of scales.

Runoff predictions are needed for many purposes including hydrological design, flood warning and water resources management. Predictions are based on hydrological models that are usually tailored to the local conditions by adjusting the model parameters but leaving the model structure unchanged. The main hypothesis investigated in this project was that a single model structure cannot capture the wide variety of hydrological processes. Specific model structures are needed, dependent on the dominant controls. The overall approach consisted of using dominant hydrological processes to identify a suitable structure of hydrological models. The dominant processes were inferred by classifying hydrological response through both large scale climatic controls (e.g. precipitation) and smaller scale subsurface controls (e.g. geology, soils). Catchments in Peru with contrasting precipitation regimes were compared with catchments in Austria and other countries. The results suggest that, in the dry Peruvian lowlands of the North, the strength of the runoff seasonality is smaller than that of precipitation due to a relatively short rainy period from January to March and catchment storage. In the Austrian transect, the strength of the runoff seasonality is greater than that of precipitation due to the influence of snowmelt in April to June. At the event scale (e.g. for floods), climate is very important through storm type and antecedent soil moisture, and geology is very important through soil characteristics. Four study catchments were selected along a climate gradient in Austria. Hydrologic response units (HRU) were defined to discriminate areas dominated by surface runoff, interflow and deep groundwater flow. Depending on these processes and the seasonal runoff regime, different model structures were selected for each of these four catchments. Again based on these processes, a priori parameter sets, in particular response times, were specified. Model tests investigated whether the right model structure gave better runoff simulations than the structure from one of the other catchments. The overall findings confirm the project hypothesis that the model structure matters for the runoff predictions. Choosing a tailor-made model structure based on dominant processes is of substantial value if one cannot calibrate the model to observed runoff data. This implies that the framework developed in this project is most relevant for ungauged basins. The new model framework will be useful for better understanding the water balance under current and future climate regimes, to analyse the effects of changes in other controls (e.g. land use) and understand the interplay of hydrological and ecological processes.