To main content

Local Drift Forecasting with Simplified Ocean Models and Multi-Level Data Assimilation

Abstract

Search-and-rescue operations at sea are supported by fast predictions of drift trajectories that are classically based on operational ocean models. This thesis promotes a complementary approach for drift forecasting based on computationally efficient methodologies utilising simplified models and ensemble-based data assimilation. Simplified ocean models aim to capture only the most relevant dynamics for short time horizons and they are hence more computationally efficient than complex operational models. Herein, the rotational shallowwater equations and a massively parallel simulation framework are exploited for the simplified modelling. Given the inherent inaccessibility of the true dynamics of the ocean, both presently and in the future, large ensembles of simplified models can be run to account for this spatio-temporal uncertainty in local forecasts. Such ensemble-based representations enable the incorporation of observations of ocean currents by data assimilation techniques as new measurements are available. Consequently, the uncertainty in the prediction is typically reduced. In this work, methodologies behind such an on-demand system for local short-term drift trajectory prediction are considered. The investigations include different modelling and assimilation techniques suitable for search-and-rescue scenarios. The first part of this thesis synthesises the background and the description of the general concepts, whereas the second part consists of the scientific papers. The contributions in this thesis reach from the discussion of numerical solvers for shallow-water simulations, over mathematical modelling for simplified ocean dynamics, to tailored data assimilation methods for sparse in-situ observations and settle with the advancement of computational efficient data assimilation, building on the foundations of multilevel Monte Carlo methods and simulations on different resolutions.
Read the publication

Category

Doctoral thesis

Language

English

Author(s)

Affiliation

  • SINTEF Digital / Mathematics and Cybernetics
  • Norwegian University of Science and Technology
  • OsloMet - Oslo Metropolitan University

Year

2024

Publisher

Norges teknisk-naturvitenskapelige universitet

Issue

2024:241

ISBN

9788232680733

View this publication at Norwegian Research Information Repository