The impact of turbulence parameterization in high-resolution inverse modeling of synthetic greenhouse gases with the Lagrangian particle dispersion model FLEXPART-COSMO

Synthetic greenhouse gases contribute currently about 10% to anthropogenic radiative forcing, and their future impact depends on the replacement of compounds with long lifetimes by compounds with short lifetimes and negligible global warming potential (GWP). Furthermore, chlorine and bromine-containing synthetic gases are the drivers of stratospheric ozone destruction. Therefore, observing the atmospheric abundance of synthetic gases and quantifying their emission sources is critical for predicting their related future impacts and assuring successful regulation.

Regional-scale atmospheric inverse modeling can provide observation-based estimates of greenhouse gas emissions at a country and continental scale and, consequently, support the process of forecasting and regulation. Inverse modeling is based on three main components: Source sensitivities derived from atmospheric transport models, observations, and an inversion framework. Within the Swiss National Science Foundation project IHALOME (Innovation in Halocarbon Measurements and Emission Validation) we increase the spatial resolution of the Lagrangian particle dispersion model FLEXPART-COSMO from 7 km to 1 km in order to enhance localization of Swiss halocarbon emissions based on newly available observations from the Swiss Plateau at the Beromünster tall tower. The transport model is driven by the meteorological fields of the regional numerical weather prediction model (NWP) COSMO run at MeteoSwiss.

The higher-resolution model exhibits increased three-dimensional dispersion, and as a result, is unable to reproduce the variability seen in the observations and in the 7 km model at the tall tower site Beromünster for a well-studied validation tracer (methane). Because the TKE (Turbulent Kinetic Energy) values do not differ significantly between the two model versions, head-to-head comparisons of parameterized turbulence cannot fully explain the concentration discrepancies. Comparisons of wind fluctuations resolved on the grid-scale suggest that the dispersion differences may originate from a duplication of turbulent transport, on the one hand, covered by the high-resolution grid of the Eulerian model and, on the other hand diagnosed by FLEXPART's turbulence scheme. In an attempt to tune FLEXPART-COSMO’s turbulence scheme at high resolution, we scale FLEXPART's parameterized turbulence so it matches the TKE computed in COSMO. Test simulations with the scaled FLEXPART turbulence show remarkable improvements in the high-resolution model's ability to predict the observed tracer variability and concentration at the Beromünster tall tower. We further introduce new equations in FLEXPART's turbulence scheme for each component of the variations of the winds in order to mimic the TKE produced by the turbulence scheme of COSMO and hence resolve the part of the turbulence spectrum which is unresolved by the high-resolution model. Compared to the coarse resolution simulations, simulations with scaled turbulence result in a more realistic and pronounced diurnal cycle of the tracer and overall improved correlation with observations.

Concluding, the increasing resolution of NWP models may lead to the representation of the part of the turbulence spectrum by the models themselves. In these models, big eddies (most likely related to convection) are partly resolved and do not require additional parameterization. The turbulence schemes of the past, developed for coarse resolution models, should be revisited to include this effect.