Hydrological regionalisation - Towards a coherent framework

Regionalisation methods are needed in hydrology to estimate hydrological fluxes, state variables and model parameters at locations without measurements. The conceptual basis of many of the regionalisation methods currently in use is disconnected. Also, some of the methods have not fully realised their potential to represent hydrological processes at the regional scale. There is hence a need for more coherence among methods and a need for enhancing the process basis of individual methods. The aim of the project is to assess the relative merits of existing regionalisation methods in an Austrian context; to develop regionalisation methods that are driven by process considerations rather than by the availability of the data and / or model concepts; and to reconcile various regionalisation methods with different conceptual bases for similar problems. Three work packages are planned on statistical methods, deterministic methods and combined statistical-deterministic methods. In work package 1, existing statistical methods of regionalisation that are based on auxiliary data will be compared and extended in terms of the process indicators that are used to represent event based and flow duration based hydrological variables. A geostatistical regionalisation method will be developed that accounts for the spatial organisation of the landscape into nested catchments. These two types of methods will be compared and combined. In work package 2, parameters of event based deterministic catchment models will be regionally analysed in a first step. In a second step, parameters of a continuous deterministic catchment model will be regionalised assisted by satellite snow data. These two types of methods will, again, be compared and made consistent. In work package 3 on combined statistical-deterministic methods, probabilistic estimates from deterministic models will be obtained in a regionalisation mode. The predictive uncertainty of the methods examined and developed in this project will by analysed by comparative sensitivity analyses and explicit error analyses through jack-knifing. A range of digital data sets covering all of Austria will be used in this project all of which are already available to the applicant and can be used in the proposed project. The project breaks new scientific ground both by increasing the understanding of individual processes and methods at the regional scale and by synthesising methods that so far have not been related to each other. The complementary approach is important as it is expected to exploit the strengths of individual methods and build on their synergies. Examining more than one approach to the same problem allows inter-comparisons between the methods which enhances the credibility of each of the methods used. Credibility of the analysis methods translates directly into the credibility of the process understanding obtained and the predictive methods developed and/or examined in this project.

Regionalisation methods are needed in hydrology to better understand how the water flows in the landscape even if no measurements are available. This involves curiosity driven research in the earth sciences and has implications for better understanding the earth system. This also involves applied research as is needed for design of hydraulic structures such as dams and bridges and for managing the water resources. This is a particularly timely issue in the context of climate change. In the past, the methods to estimate how water flows in the landscape and in the rivers have been somewhat disconnected. The purpose of the project was to advance the process based methods for Austrian conditions and to better link the various methods in order to more accurately estimate the water flow. The project has made a number of major advances. First, the project has analysed new data that help understand the regional patterns of runoff in the streams as well as their drivers. The novel thing was that the seasonality (i.e. the timing within the year) of rainfall, runoff and floods was analysed at a regional scale. This seasonality was, among other things, related to the weather types. The seasonality method allowed identification of the drivers of runoff changes and floods in the Alps and the Carpathians. The new data types were exploited by new statistical methods. The novel idea here was to account for how similar the water flow is along a river reach. Second, the new data types were used in hydrological models to be able to predict the water flow at locations where no runoff data are available. Satellite data on snow patterns and on soil moisture patterns were used for this purpose. Use of these data allowed for more accurate predictions of the water flows at a regional scale. Third, the statistical methods were combined with the models. This enabled the prediction of the water flows in rivers including their probability, i.e. some statement of how likely it is that a certain flow (or river level) is exceeded. In the final part of the project a synthesis of the previously developed approaches was performed. This was based on combining a multitude of data types of rainfall, runoff and floods in a statistical modelling framework. This is a paradigm shift in hydrology as, previously, individual data have been analysed in isolation. To highlight the focus on a synthesis of processes the term "flood frequency hydrology" was coined. This new framework is more elaborate than the traditional flood frequency statistics. It is more accurate as a wider range of data and processes are accounted for. All methods and concepts have been tested on extensive hydrological data sets in Austria and neighbouring countries. These tests showed that the new combined approach exploits the strengths of the individual component methods and builds on their synergies.