Skip to main content

Multi-model comparisons in river flow forecasting (Theme: technology development)

Are you interested in this MSc topic? Send an email with your CV to Jaap.Kwadijk@deltares.nlm.j.booij@utwente.nl, frederiek.spernaweiland@deltares.nl 

Participating institutes: Deltares, University of Twente

Areas under investigation: several international basins

Rationale and objective

In climate research, weather prediction as well as in air quality studies, combining forecasts produced by ensembles of the same and/or multiple numerical models with statistical techniques have been demonstrated more reliable and accurate than the forecasts of individual models. Ensemble methods have been proved to outperform single-model predictors due to the diversity among the models that they combine. It has been proven that the error of multi-model ensemble is smaller than that of the ensemble members and it tends to be reduced when combining various types of models. 

In view of this, it is remarkable that in river flow forecasting, the hydrological response of weather events and thus also decisions for issuing warning and organising response, the common practice is to depend on the output of one single hydrological or a single combination of a hydrological and a hydraulic computational models. It is known that different models show diverging responses particularly when applied outside the range they have been validated for. This is typically relevant under very extreme, unprecedented, events. And these are the events the results of these forecasting systems are the most relevant since a lot is at stake.  A similar practice is applied for the design of measures to mitigate the consequences of (future) extreme floods and droughts.  

Approach

Within JCAR we aim at developing common technology based on various hydrological and/or hydraulic models to provide flood and drought forecasts assess. In view of this, a group of MSc students will be assigned to develop a system able to compare forecasts using various hydrological and hydraulic models in different regional river basins. Using this system these forecasts will be evaluated on both their qualitative performance compared to single model based forecasts as well as to what extend the results of these multi-model forecasts would have led to different responses. The Deltares-FEWS technology will be used to enable parallel simulations of these models. The different hydrological and hydraulic models applied will be a.o.  those models actually used the different countries. 

The student will start reviewing the literature available on the subject and will be trained in the use of the FEWS software. We foresee the comparison of a combination of mature modelling software as well as tools with lower TRLs. To successfully execute these MSc assignments the students should be familiar with python or another modelling language such as Julia or R 

Deliverables

The student will deliver a thesis that answers the research questions, as well as the scripts and datasets that were used to get these answers, such that we can reproduce the results.

Requirements

  • Successfully completed courses on hydrology and/or hydrological modellling and/or water resources management
  • A reasonable acquaintance with Python or other script languages for programming such as Julia or R

Supervision 

Deltares will provide daily supervision. The MSc thesis project will be part of a larger effort on machine learning in hydrology. A monthly stipend will be available and based on the ECTS.