11 Jul 2022
11 Jul 2022

Improving trajectory calculations by FLEXPART 10.4+ using deep learning inspired single image superresolution

Rüdiger Brecht1, Lucie Bakels2, Alex Bihlo3, and Andreas Stohl2 Rüdiger Brecht et al.
  • 1Department of Mathematics, University of Bremen
  • 2Department of Meteorology and Geophysics, University of Vienna, Josef-Holaubek-Platz 2
  • 3Department of Mathematics and Statistics, Memorial University of Newfoundland

Abstract. Lagrangian trajectory or particle dispersion models as well as semi-Lagrangian advection schemes require meteorological data such as wind, temperature and geopotential at the exact spatio-temporal locations of the particles that move independently from a regular grid. Traditionally, this high-resolution data has been obtained by interpolating the meteorological parameters from the gridded data of a meteorological model or reanalysis, e.g. using linear interpolation in space and time. However, interpolation errors are a large source of error for these models. Reducing them requires meteorological input fields with high space and time resolution, which may not always be available and can cause severe data storage and transfer problems. Here, we interpret this problem as a single image superresolution task. That is, we interpret meteorological fields available at their native resolution as low-resolution images and train deep neural networks to up-scale them to higher resolution, thereby providing more accurate data for Lagrangian models. We train various versions of the state-of-the-art Enhanced Deep Residual Networks for Superresolution (EDSR) on low-resolution ERA5 reanalysis data with the goal to up-scale these data to arbitrary spatial resolution. We show that the resulting up-scaled wind fields have root-mean-squared errors half the size of the winds obtained with linear spatial interpolation at acceptable computational inference costs. In a test setup using the Lagrangian particle dispersion model FLEXPART and reduced-resolution wind fields, we demonstrate that absolute horizontal transport deviations of calculated trajectories from "ground-truth'' trajectories calculated with undegraded 0.5° x 0.5° winds are reduced by at least 49.5 % (21.8 %) after 48 hours relative to trajectories using linear interpolation of the wind data when training on 2° x 2° to 1° x 1° (4° x 4° to 2° x 2°) resolution data.

Rüdiger Brecht et al.

Status: final response (author comments only)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • CEC1: 'Comment on egusphere-2022-441', Juan Antonio Añel, 23 Aug 2022
    • AC1: 'Reply on CEC1', Rüdiger Brecht, 01 Sep 2022
    • AC2: 'Reply on CEC1', Rüdiger Brecht, 13 Sep 2022
      • CEC2: 'Reply on AC2', Juan Antonio Añel, 13 Sep 2022
        • AC3: 'Reply on CEC2', Rüdiger Brecht, 11 Oct 2022
          • CEC3: 'Reply on AC3', Juan Antonio Añel, 11 Oct 2022
            • CC1: 'Reply on CEC3', Andreas Stohl, 11 Oct 2022
              • CEC4: 'Reply on CC1', Juan Antonio Añel, 25 Oct 2022
                • AC4: 'Reply on CEC4', Rüdiger Brecht, 24 Nov 2022
                • AC5: 'Reply on CEC4', Rüdiger Brecht, 24 Nov 2022
                • AC6: 'Reply on CEC4', Rüdiger Brecht, 24 Nov 2022
  • RC1: 'Comment on egusphere-2022-441', Anonymous Referee #1, 12 Sep 2022
  • RC2: 'Comment on egusphere-2022-441', Anonymous Referee #2, 14 Nov 2022

Rüdiger Brecht et al.

Rüdiger Brecht et al.


Total article views: 769 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
583 152 34 769 6 4
  • HTML: 583
  • PDF: 152
  • XML: 34
  • Total: 769
  • BibTeX: 6
  • EndNote: 4
Views and downloads (calculated since 11 Jul 2022)
Cumulative views and downloads (calculated since 11 Jul 2022)

Viewed (geographical distribution)

Total article views: 715 (including HTML, PDF, and XML) Thereof 715 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 29 Nov 2022
Short summary
We use neural network based single image super resolution to improve upscaling of meteorological wind fields to be used for particle dispersion models. This deep learning based methodology improves the standard linear interpolation typically used in particle dispersion models. The improvement of wind fields leads to substantial improvement of the computed trajectories of the particles.