Using the Ensemble Kalman Filter to update a fast surrogate model for flow forecasting

Nadia Schou Vorndran Lund, Maurizio Mazzoleni, Henrik Madsen, Ole Mark, Peter Steen Mikkelsen, Dimitri Solomatine, Morten Borup

    Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

    170 Downloads (Pure)

    Abstract

    Many cities face issues with rain induced flooding and combined sewer overflows, which can be addressed by using hydrodynamic models. These models are often simplified in a real-time setting to make them faster, and their performance can be improved by using data assimilation. In this study we use the Ensemble Kalman Filter to update a simplified model of a small area of Copenhagen, Denmark. The model is
    evaluated using perfect rain data for one summer month in 2016, and flow forecasts are used to quantify the performance of the update. We found that the 1-60 minutes forecast can be improved by updating the model. Having a small noise on the rain gives slightly worse results on a short forecast horizon and slightly better forecasts on a longer horizon. The forecast performance is also dependent on which model parts are updated.
    Original languageEnglish
    Title of host publicationProceedings of the 11th International Conference on Urban Drainage Modelling
    EditorsG. Mannina
    Place of PublicationPalermo, Italy
    Publication date2018
    Pages710-714
    Publication statusPublished - 2018
    Event11th International Conference on Urban Drainage Modelling - Palermo, Italy
    Duration: 23 Sept 201826 Sept 2018

    Conference

    Conference11th International Conference on Urban Drainage Modelling
    Country/TerritoryItaly
    CityPalermo
    Period23/09/201826/09/2018

    Keywords

    • Data assimilation
    • Surrogate model
    • Flow forecasting

    Fingerprint

    Dive into the research topics of 'Using the Ensemble Kalman Filter to update a fast surrogate model for flow forecasting'. Together they form a unique fingerprint.

    Cite this