Multi-model comparisons in river flow forecasting

Rationale and Objective

In climate research, weather prediction, and air quality studies, combining forecasts from ensembles of numerical models with statistical techniques has been shown to produce more reliable and accurate predictions. Ensemble methods enhance reliability by leveraging the diversity among models, which can reduce forecast errors and improve accuracy.

Despite this, in river flow forecasting, there is a common practice of relying solely on the output of a single hydrological or a combination of hydrological and hydraulic models. This approach is particularly problematic during extreme, unprecedented events where different models may show diverging responses.

Within JCAR, the goal is to develop common technology that combines various hydrological and hydraulic models to provide flood and drought forecasts. This involves comparing forecasts using different models in various regional river basins.

Approach

The project will involve developing a system to compare forecasts generated by different hydrological and hydraulic models. This system will evaluate the forecasts qualitatively against single-model forecasts and assess how multi-model forecasts might influence decision-making.

To achieve this, the Deltares-FEWS technology will be utilized to enable parallel simulations of the models. The models to be tested will include those currently used in different countries.

The student will begin by reviewing existing literature and will receive training in the use of the FEWS software. The project will explore the comparison of mature modeling software alongside tools with lower Technology Readiness Levels (TRLs).

To succeed in this project, the student should be proficient in Python or another scripting language like Julia or R, and should have completed courses in hydrology, hydrological modeling, or water resources management.

Deliverables

• A thesis addressing the research questions
• Scripts and datasets used to generate the results
• Documentation enabling others to reproduce the results

Requirements

• Successfully completed courses in hydrology, hydrological modeling, or water resources management
• A reasonable familiarity with Python or another programming language such as Julia or R