Improved physical process representation of river networks in Global Flood Models

Lead Research Organisation: University of Leeds
Department Name: Civil Engineering


Quantifying flood hazard is an essential component of resilience planning, emergency response, and insurance. Efforts have recently intensified to estimate flood hazard for the whole planet with global flood models (GFMs). A key area of improvement for these GFMs is in their representation of global river networks. All current GFMs use a simplified drainage network derived automatically from analysis of the topography. This simplified network is easy to use, but errors result in mismatches with real rivers. The emergence of new, global, high resolution Digital Elevation Models (DEMs) and remote sensing data sets can help address these issues, but questions remain about how to operationalize the incorporation of this new data in the GFMs.

This PhD will aim to answer two key questions: How to identify, collect, and incorporate the relevant new data and, does this result in model output improvements? After reviewing current river datasets, the researcher will then develop methods to identify and collect this data, and incorporate it into the current GFM river networks. Due to the global nature of the datasets, this will require automated and consistent methods, while capturing the most relevant hydraulic process details and integrating these into one dataset. Methods will then be tested under different rivers and climates to ensure global applicability and quantify the improvements to output. The researcher will have direct access to a Fathom Global Ltd.'s state-of-the-art GFM framework, and work alongside their experienced developers during a 4 months secondment.

The work carried out in this PhD has the potential to improve the performance of GFMs and as a result increase their applicability. The researcher will publish all findings in open scientific publications. This multidisciplinary project includes; remote sensing, computer methods, statistics,machine learning, river hydraulics, geomorphology and flood management.


10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
NE/R008949/1 30/09/2017 29/09/2021
1964274 Studentship NE/R008949/1 30/09/2017 29/09/2020 Mark Viktor Bernhofen
Description The first key finding of this study was that Global Flood Models perform fairly well when compared against historical flood events. In the regions studied, the models that better represent the physics of fluid flow tend to perform better. This study also highlighted the importance of performing a collective validation when comparing multiple models. Collectively validating the models enables one to comparatively judge the performance of the models against a reference 'real' event.
The second key finding of this study relates to the river representation in Global Flood Models. The models differ in terms of the smallest river for which they model flooding. This study showed that global flood exposure estimates can differ by up to a factor of two depending on what 'minimum river size' was chosen.
Exploitation Route Users of Global Flood Models can make use of both the comparative validation that was carried out and the research into river size representation. Both these pieces of work could help to inform users about what the most appropriate flood model to use is.

The validation work mentioned previously can also be expanded on. That initial validation looked at only two flood events in two countries. Expanding the validation to include more flood events globally would be an important direction to take this work. The remainder of this research project is going to be focussed on that.
Sectors Environment