Don't miss our weekly PhD newsletter | Sign up now Don't miss our weekly PhD newsletter | Sign up now

  Riverine flood forecasting using hybrid model-machine learning approaches


   School of Ocean and Earth Sciences

This project is no longer listed on FindAPhD.com and may not be available.

Click here to search FindAPhD.com for PhD studentship opportunities
  Prof J Sheffield  No more applications being accepted  Competition Funded PhD Project (Students Worldwide)

About the Project

Flood risk in the UK and globally has been increasing due to the impacts of climate change, causing rivers to overflow and coastal areas to experience storm surges. This has resulted in a higher frequency and severity of floods, endangering lives, damaging infrastructure, and causing significant economic losses. Flood risk is expected to continue to increase with climate change, with the most vulnerable of society expected to bear the largest impacts. Responding to this challenge requires a range of interventions to reduce risks, including early warning via flood forecasting. Flood forecasts are traditionally generated using physical hydrological models driven by weather forecasts, however, these are generally quite uncertain especially in complex environments, difficult to update and expensive to run multiple simulations to represent the uncertainties. The use of data-driven AI and machine learning methods could potentially transform forecasting by providing enhanced accuracy, reliability and speed compared to traditional methods, and with potential to combine with physical models in hybrid approaches (Slater et al., 2023). The literature on AI-based hydrological forecasting has generally been limited to specific flood case studies, although is now accelerating with the availability of large-sample datasets (e.g. Nevo et al., 2022). Its potential as a broader, more robust and more applicable approach for forecasting floods, needs to be explored, and especially for complex events including compound with potentially multiple drivers.  

The overall aim of this project is to develop and test a framework for data-driven hydrological prediction, focused on flooding, by integrating large and diverse datasets with state-of-the-art machine learning methods for hydrological prediction, and test this for a set of case studies in the UK and Europe. 

1. Review existing AI approaches including hybrid methods, focusing on emerging developments and application of AI to general forecasting problems, and comparison with traditional approaches in terms of predictive skill, data requirements and utility. The review is required to establish a benchmark against which to evaluate the proposed framework. 

2. Develop a data-driven forecasting framework. Based on an existing hydrological prediction framework using dynamical models, the project will develop data-driven ML methods, testing different approaches such as RF, SVM, ANN, and RNN including convolutional LSTM networks that include long-term dependencies. 

3. Test in a set of case studies, including identification of relevant forecast metrics and datasets with partners. Experiments will be carried out for a range of predictive metrics across scales and for a range of configurations, to understand how predictability varies over space and time scales and with different drivers, and as a function of data availability/quality, and how climate change impacts can be represented. 

The following are essential and will be provided by the supervisory team if not by the FLOOD-CDT training: 

  • Large dataset processing 
  • Machine learning methods 
  • Hydrological forecasting approaches and analysis methods

HOW TO APPLY

Apply online: Search for a Postgraduate Programme of Study (soton.ac.uk). Select programme type (Research), 2024/25, Faculty Environmental and Life Sciences, select Full-time or Part-time, next page select “PhD Ocean & Earth Science (FLOOD CDT). In Section 2 of the application form you should insert the name of the supervisor.

PhD FLOOD CDT – full-time (programme length of 48 months) – code 9215

PhD FLOOD CDT – part-time (programme length of 84 months) – code 9216


Computer Science (8) Environmental Sciences (13) Geography (17) Mathematics (25)

Funding Notes

The CDT will provide at least 56 fully funded PhD studentships over 4 cohorts, with first entry of 16 doctoral researchers staring in October 2024. The studentship will cover UK course fees and an enhanced tax-free stipend of year for 3.5 years along with a budget for research, travel, and placement activities. Details of the studentship amount can be found on the NERC web-site: https://www.ukri.org/apply-for-funding/studentships-and-doctoral-training/get-a-studentship-to-fund-your-doctorate/#contents-list

References

Slater, L. J., Arnal, L., Boucher, M.-A., Chang, A. Y.-Y., Moulds, S., Murphy, C., Nearing, G., Shalev, G., Shen, C., Speight, L., Villarini, G., Wilby, R. L., Wood, A., and Zappa, M.: Hybrid forecasting: blending climate predictions with AI models, Hydrol. Earth Syst. Sci., 27, 1865–1889, https://doi.org/10.5194/hess-27-1865-2023, 2023.
Nevo, S., Morin, E., Gerzi Rosenthal, A., Metzger, A., Barshai, C., Weitzner, D., Voloshin, D., Kratzert, F., Elidan, G., Dror, G., Begelman, G., Nearing, G., Shalev, G., Noga, H., Shavitt, I., Yuklea, L., Royz, M., Giladi, N., Peled Levi, N., Reich, O., Gilon, O., Maor, R., Timnat, S., Shechter, T., Anisimov, V., Gigi, Y., Levin, Y., Moshe, Z., Ben-Haim, Z., Hassidim, A., and Matias, Y.: Flood forecasting with machine learning models in an operational framework, Hydrol. Earth Syst. Sci., 26, 4013–4032, https://doi.org/10.5194/hess-26-4013-2022, 2022.

How good is research at University of Southampton in Earth Systems and Environmental Sciences?


Research output data provided by the Research Excellence Framework (REF)

Click here to see the results for all UK universities