Offline-Coupling of the Lagrangian Particle Dispersion Model FLEXPART to ICON
- 1Empa Swiss Federal Laboratories for Materials Science and Technology, Air Pollution/Environmental Technology, Dübendorf, Switzerland (stephan.henne@empa.ch)
- 2Federal Office for Meteorology and Climatology MeteoSwiss, Zurich Airport, Switzerland
Over the last years, the numerical weather prediction (NWP) and climate model ICON has become the operational forecasting tool for several national weather services and research groups. Operational analysis or re-analysis fields from NWP models are often used as input for offline-coupled atmospheric chemistry and transport models. Among the latter, Lagrangian Particle Dispersion Models (LPDMs) allow computationally efficient simulations, especially for point sources such as hazardous releases and for receptor-oriented studies such as determining the sensitivity of a concentration observation to upstream surface fluxes (i.e., estimating concentration footprints). One frequently used LPDM is the FLEXible PARTticle (FLEXPART) model, which is available for inputs from different global and regional NWPs (e.g., ECMWF-IFS, WRF, COSMO). Although these versions differ in the applied horizontal and vertical coordinate systems, they have in common that they interpolate gridded NWP output from a rectangular grid to particle positions. In contrast, ICON solves its state variables on a triangular grid. To make best use of ICON output, a direct interpolation from its native grid to particle positions is required. However, compared to a rectangular grid, where interpolation can be done in a straightforward fashion applying bi-linear or bi-cubic interpolation, interpolation from a triangular grid requires additional considerations concerning the choice of interpolation stencil and weight calculations.
Starting from FLEXPART for COSMO, which shares the same vertical grid system with ICON, we revised and generalized how FLEXPART interpolates from grid input to particle positions. Four different direct interpolation methods were implemented: next neighbor (containing triangle), inverse distance weight, barycentric interpolation, and radial basis function interpolation. The resulting FLEXPART version is runs efficiently with outputs from both COSMO and ICON. Next to the direct implementation, we also evaluated an indirect coupling in which ICON output is first interpolated onto a COSMO-like, staggered grid and then used as input for FLEXPART-COSMO.
Both direct and indirect FLEXPART-ICON approaches were thoroughly evaluated by comparison of individual plume simulations resulting from point sources. As a reference simulation, the same point sources were simulated with the Aerosols and Reactive Trace gases (ART) extension of the ICON model. We discuss differences in the performance between the direct and indirect approach and between the interpolation methods. Computational costs for the different approaches are evaluated and trade-offs between model performance and computational efficiency are discussed.
How to cite: Henne, S., Kaufmann, P., Emmenegger, L., and Brunner, D.: Offline-Coupling of the Lagrangian Particle Dispersion Model FLEXPART to ICON, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-14799, https://doi.org/10.5194/egusphere-egu24-14799, 2024.