Next generation of global water risk modelling

Observations show that climate change has increased the frequency and severity of extreme weather events around the globe. Droughts are becoming more severe and flooding now affects many regions of the world with increasing frequency. This intensification in a changing climate on top of the natural variability is one of the big challenges currently facing humanity. Measures needed for coping with these issues include an intelligent use of hydrological models. A new field is the development and application of such models at the global scale, which is enabled by the vast increases in available data and computing power. Amongst other things, these models are being used in climate change impact assessments, forecasting of floods and seasonal water availability, and decision-making for dam projects. With applications in insurance, adaption and emergency response, global hydrological models thus are a very valuable tool.

Physically based models offer an alternative tool to other types of hydrological models with some attractive advantages. They provide a closer representation of physical processes occurring in catchments. This means that they are theoretically more robust for simulations under non-stationary conditions, such as climate and land use change. Simpler, conceptual models are often favoured for global modelling, as they are easier and faster to set up and run. They also typically perform well after calibration. However, truly global modelling means simulating ungauged catchments, where direct calibration often is not possible. This presents a challenge to conceptual models, as it is often difficult to specify all of the required parameters.

Physically based models have large data and computing power requirements which has historically acted as a barrier to their use. However, the number and quality of continental and global scale datasets available for configuring and driving a physically based hydrological model has grown immensely in recent years. This is particularly so for rainfall. Available data sources include gauge-based, satellite, reanalysis and blended products. There is also an increasingly good collection of soil and land use data, often drawing on satellite products, but direct information on 3D geology remains harder to obtain for all countries. For model evaluation, traditional measurements such as river flow gauging is increasingly complemented by remote sensing. For example, patterns of soil moisture variation can now be investigated at unprecedented resolution. Using a physically based model to aid comparison of input and evaluation provides a way to explore the information content of different data products.

This research will therefore develop a global physically based hydrological modelling system to improve existing operational hydrological forecasts. The aims of this project are to:

Create a global, physically based hydrological modelling system by adapting SHETRAN for deployment on a cloud computing platform such as Azure.

Improve automated validation in areas with limited or no flow gauge data through use of alternative data sources.

Assess the information content of different global and national hydrological datasets and their suitability for global scale analysis.

Assess the value of physically based modelling for operational hydrological forecasting by collaborating with the European Centre for Medium-Range Weather Forecasts (ECMWF), comparing the performance of their Global Flood Awareness System (GloFAS) with the physically based modelling system set up in this research.