Statistical modelling of floods is among others used for estimating design floods (e.g. the 200-year flood) required for dam safety, area planning, and infrastructure design. Robust and reliable models are needed. Statistical modelling faces several challenges including data quality, sample size, and prediction in ungauged locations.
The Norwegian Water Resources and Energy Directorate (NVE) is responsible for the national recommendation for design flood estimation and providing data and tools for such estimates (e.g. nevina.nve.no). Results from master projects can form the basis for improving guidelines or provide recommendations for best practices.
Example projects:
POT analysis of floods - Peak over threshold (POT) flood data will be extracted using two approaches: a fixed threshold for all years and a threshold depending on the season. Floods will be classified according to the flood-generating processes using a precipitation-runoff model. The data will be used to analyse flood trends (frequency, magnitude, flood generation processes) and extreme value models (both standard models and mixture distributions) will be fitted to the data to estimate design floods. New models will be evaluated and compared with existing benchmark models.
Metastatistical Extreme Value (MEV) analysis of floods – A key challenge with standard approaches for estimating floods is that the available data set might be too small. This can lead to either underestimation or overestimation of the magnitude of rare floods, or to considerable uncertainty. The meta-statistical approach uses both ordinary high flows and flood peaks to overcome this challenge (Mushtaq et al., 2022). Moreover, a characterization of the statistical properties of ordinary high flows allows for identifying cases for which unexpectedly large floods may occur, although they have not been observed yet. Possible thesis topics in this field include:
- Investigation of temporal or spatial variations in the statistical properties of ordinary high flows, with the aim of identifying locations where extreme floods can be expected, or changes that indicate the possible occurrence of extreme floods;
- Combining knowledge of runoff generation processes with a metastatistical approach to flood estimation (Mushtaq et al., 2023). Ordinary discharge peaks will be categorized into types based on the runoff generation process which triggered them (e.g., snowmelt, rain), analogously to what done for the POT approach. Suitable statistical distributions will be selected to represent different types of ordinary discharge peaks. The metastatistical approach will be then applied to obtain flood frequency and magnitude emerging from the combination of different types of ordinary discharge peaks.
The data will be used for analysing trends on floods (frequency, magnitude, flood-generating processes). New models will be evaluated and compared to existing benchmark models.
- Mushtaq, S., A. Miniussi, R. Merz, and S. Basso (2022), Reliable estimation of high floods: a method to select the most suitable ordinary distribution in the Metastatistical Extreme Value framework, Advances in Water Resources, 104127. https://doi.org/10.1016/j.advwatres.2022.104127
- Mushtaq, S., A. Miniussi, R. Merz, L. Tarasova, F. Marra, S. Basso (2023), Prediction of extraordinarily high floods emerging from heterogeneous flow generation processes, Geophysical Research Letters, 50, e2023GL105429. https://doi.org/10.1029/2023GL105429
Risky rivers - inferring the peril of extreme river floods from normal discharge dynamics. Recent work (Basso et al., 2023) revealed the possibility to predict the occurrence of extreme river floods by means of simple metrics of discharge dynamics. Ongoing changes of these metrics in Norway indicate a rising peril of extreme floods. In this project, we will study changes in space and time of the predictors of extreme floods and investigate whether they align with large floods occurred in the past. The goal is to understand whether such simple metrics can be used as early warning indicators of the possible occurrence of extreme floods in rivers.
- Basso, S., R. Merz, L. Tarasova, and A. Miniussi (2023), Extreme flooding controlled by stream network organization and flow regime, Nature Geoscience, 16, 339–343. https://doi.org/10.1038/s41561-023-01155-w
Analysis of widespread floods - Analyzing and improving of the co-occurence of floods at multiple locations is critical for handling flood risk since it allows us to assess how large areas can be affected. We suggest to build on methods developed by Brunner et al. (2020) and Jiang et al. (2020). In Brunner et al. (2020) a methodology for creating an event occurrence matrix across locations is suggested, and it is possible to estimate this matrix for floods of different return periods. Jiang et et al. (2020) propose an analysis similar to principal component analysis (PCA) for extremes that allows us to look at patterns describing the spatial co-occurrence of extremes.
For the analysis both observed flood data and simulated flood data from a hydrological model can be used to analyse widespread floods along a linear infrastructure or within specific areas and assess how frequently a flood of specific return periods might occur depending on the length of the infrastructure or size of a region.
REFERENCES
- Brunner, M L, S Papalexiou, M P Clark and E Gilleland (2020): How probable is widespread flooding in the United States? Water Resources Research, 56: e2020WR028096.
- Jiang, Y, D Cooley and M F Wehner (2020): Principal component analysis for extremes and application to U.S. precipitation. Journal of Climate, 33: 6441-6451. hb..-------
