Thesis Code: T12

Advancing Water Systems Management: Bias Correction and Synthetic Forecast Generation

Short description:

In water systems management, perfect forecasts hold great promise for improvement, especially over medium- to long-term periods (1). However, despite advancements in weather prediction and extreme event forecasting, inherent biases often limit the practical utility of real forecasts in operational contexts. This challenge becomes particularly evident in sub-seasonal lead times, typically around one month ahead, where forecast information could be most valuable for operational decision-making. Additionally, these forecasts are usually available for relatively short periods, spanning only a few decades (2).

To address these challenges, bias correction and synthetic forecast (3) generation have emerged as promising solutions. By leveraging state-of-the-art techniques in bias correction and utilizing historically observed data, bias-corrected synthetic forecasts can be accurately generated. These forecasts provide larger datasets for Machine Learning training and validation, enhancing their utility in water systems management.

The primary objective of this thesis is to develop bias-corrected synthetic forecasts for medium-term time ranges using cutting-edge methodologies. These forecasts will then undergo validation in one or more selected case studies (e.g. Lake Como, Italy; Folsom Lake, California; Lake Kariba, Zambia/Zimbabwe). The testing process will involve the application of advanced methods, including the utilization of a Direct Policy Search (DPS) approach like Evolutionary Multi-Objective DPS (EMODPS) and the implementation of Stochastic Model Predictive Control (SMPC) (4). Through this comprehensive approach, the student is expected to carry out the following activities:

  1. Literature review: reviewing the state of the art of forecast bias correction and synthetic forecast generation
  2. Design of traditional bias correction methodologies
  3. Generation of bias-corrected syntetic forecasts with state of the art techniques
  4. Application of these forecasts with EMODPS and SMPC optimization methods

References

  1. M. Giuliani, M. Zaniolo, P. Block, and A. Castelletti, ‘Data-driven control of water reservoirs using an emulator of the climate system’, IFAC-PapersOnLine, vol. 53, no. 2, pp. 16531–16536, Jan. 2020, doi: 10.1016/j.ifacol.2020.12.771.
  2. M. Giuliani, L. Crochemore, I. Pechlivanidis, and A. Castelletti, From skill to value: isolating the influence of end-user behaviour on seasonal forecast assessment. 2020. doi: 10.5194/hess-2019-659.
  3. M. A. Nayak, J. D. Herman, and S. Steinschneider, ‘Balancing Flood Risk and Water Supply in California: Policy Search Integrating Short-Term Forecast Ensembles With Conjunctive Use’, Water Resources Research, vol. 54, no. 10, pp. 7557–7576, 2018, doi: 10.1029/2018WR023177.
  4. A. Mesbah, “Stochastic Model Predictive Control: An Overview and Perspectives for Future Research,” in IEEE Control Systems Magazine, vol. 36, no. 6, pp. 30-44, Dec. 2016, doi: 10.1109/MCS.2016.2602087.

Relevant courses and knowledge:

Natural Resources Management

Number of students:

1 or 2

Requisites:

The student should be comfortable with data handling and programming skills (preferably Python, Matlab).