Research

The aim of this project is to assess the impact of summer low flows on the remobilization of pollutants from river sediments. The analyses are carried out for eastern Austria, where agricultural input and the predicted risk of climate warming on low flow and water temperature are particularly high. The innovative combination of data-based models with laboratory experiments and water quality monitoring allows an overall assessment of the sediment-related risk of quality impairment along the water network. The added value of the derived information is presented for three selected catchments and discussed with stakeholders with regard to water management relevance. From this, recommendations for future climate scenarios will be developed.

Spatiotemporal models are a common approach to a wide range of environmental problems. In this context it can be distinguished between a single spatiotemporal model, temporal functions correlated in space or spatial functions correlated in time. Applications of time series extended to a spatial scale are widely used in modelling air pollution, whereas spatial methods extended to a temporal scale can be found in, e.g., soil moisture modelling or interpolation of meteorological variables. However, the application of spatiotemporal models to streamflow is rare. This can be explained by the nested characteristics of streamflow catchments, the tree-wise structure of river networks and spatial and temporal variability of co-variables, such as physiographic catchment characteristics or meteorological variables. Considering these conditions, it is essential to develop and improve spatiotemporal models for streamflow that take the specific spatial and temporal variability into account. Therefore, the project’s aim is to adapt spatiotemporal methods for streamflow and to extend temporal and spatial models to a space-time framework. The main hypothesis of the doctoral proposal is the following: Spatiotemporal models for streamflow can be developed that yield better predictions than (i) point-wise methods in time or space, (ii) and spatiotemporal methods that do not consider the river network topology. To assess this science question in detail, the project is split up into four work packages and each is completed by the submission of a paper to a peer-reviewed journal.

In a one-years-extension of our previous project “Robust Risk Estimation”, we focus on multivariate aspects and dynamics of extreme events so far not yet covered in this detail. In all our reference applications, i.e., in financial risks of a bank, public health (hospital length of stay and costs), and hydrology (river discharge), there are important questions where these aspects cannot be ignored but rather have to be accounted for. With thin empirical evidence available to this end, misspecification becomes a central issue in the corresponding applications. As a remedy to some extent, we propose to enhance our robust procedures applying robust likelihood techniques to adjust our procedures in a way that they can adapt to minor to moderate model misspecifications. This continues our successful work on theoretical foundation, development and application of robust procedures for risk management of complex systems in the presence of extreme events. In particular this extends the applicability of our software infrastructure in R developed in the current project with its powerful set of diagnostic tools to assess influence and outlyingness of data.