Improving the modeling of hydroecological indices

There is an increasing awareness about the importance of ensuring that there is enough water in quantity and quality for future generations. It is important, however, not taking only human needs into account, but to consider also ecological and ecosystem requirements. Ecohydrological indices play here an important role, since they describe streamflow characteristics that are thought to be biologically relevant. While these indices can be estimated from measured discharge time series, it is necessary to resort to modelled discharge time series when dealing with ungauged basins or when the goal of the analysis is to assess the impact of different actions or changes (e.g. dam construction, climate change, change of land cover) on the ecological potential of a stream. Since the changes have not taken place yet, there are no measured time series and we must estimate the values of the indicators based on modelled results. To successfully support the evaluation of streamflow indices it is necessary to know how well models are able to replicate the ecohydrological indices and what we can do to improve their capabilities for this task. It is further important to study how we can obtain reliable model results when there are no data for the situation that needs to be modelled (as it happens when modelling potential or future changes, and when modelling ungauged sites). This project deals with some aspects related to both the above mentioned challenges, and so it aims at increasing our ability to predict the effects of changes on the ecological characteristics of the streams. This is achieved using tools for model structural diagnosis, calibration and uncertainty analysis, which can help expand our knowledge about how models work and assist us to more reliably interpret the results we get from them in both gauged and ungauged basins. The objectives of this study are: 1) developing and testing a methodology for assessing which hydrological model structures are able to reproduce observed combinations of ecohydrological indices, 2) improving the reliability of estimates of ecohydrological indices in ungauged basins by evaluating methods developed for engineering hydrology statistics and 3) providing an improved assessment of the uncertainties of estimated ecohydrological indices by making a better use of information on model errors.

Ecohydrological indicators describe biologically relevant characteristics of streams, for example, the monthly magnitude of river discharge or the number of high flow pulses. These indicators are important for defining the management objectives of streams and for assessing the impact of natural and anthropogenic changes on their ecological properties. The aim of the project was to develop new approaches for modelling these indicators and to improve, in this way, the estimates of these indicators in gauged and ungauged catchments and under changed conditions. The main results and findings of this project are: - The development and testing of a framework for evaluating the abilities of different models for modelling several indicators in a large number of catchments. - The difficulties of hydrological models for reproducing changes in ecohydrological indicators are in many cases the result of model deficiencies and can therefore not be addressed by improving model calibration approaches. - Century-long meteorological time-series derived from reanalysis datasets achieve a good model performance for some catchments in the NW, but are still not good enough for modelling large parts of the country. It could be further observed that model performance tended to decrease as the time to the calibration period increases when the model was forced with the interpolated datasets, but that this was not evident when using the reanalysis datasets. - Ungauged catchments influenced by snow can be best modelled with physically based models or with temperature based approaches which include a radiative component. It was further evident that the snowmelt models were able to achieve a high correlation between the modelled and observed discharge in ungauged basins, but that there are still large uncertainties in the amount of snowmelt, reflected in relatively large biases. Finally, the results showed that the parameter estimates could be significantly improved when using satellite derived data.