Quantification of Transport Errors in regional CO2 inversions using a physics-based ensemble of WRF-Chem simulations

Atmospheric inversions can be used to assess biosphere-atmosphere CO2 surface exchanges, but variability among inverse flux estimates at regional scales remains significant. Atmospheric transport model errors are presumed to be one of the main contributors to this variability, but have not been quantified thoroughly. Our study aims to evaluate and quantify the transport errors in the Weather Research and Forecasting (WRF) mesoscale model, recently used to produce inverse flux estimates at the regional scale over the NACP Mid-Continental Intensive (MCI) domain. We evaluate transport errors with an ensemble of WRF simulations using different physical parameterizations (e.g., atmospheric boundary layer (ABL) schemes, land surface models (LSMs), and cumulus parameterizations (CP)). Modeled meteorological variables and atmospheric CO2 mixing ratios are compared to observations (e.g., radiosondes, wind profilers, AmeriFlux sites, and CO2 mixing ratio towers) available in the MCI region for summer of 2008. Comparisons to date include simulations using two different land surface models (Noah and Rapid Update Cycle (RUC)), three different ABL schemes (YSU, MYJ and MYNN) and two different cumulus parameterizations (Kain-Fritsch and Grell-3D). We examine using the ensemble as a proxy for the observed model-data mismatch. Then we present a study of the sensitivity of atmospheric conditions to the choice of physical parameterization, to identify the parameterization driving the model-to-model variability in atmospheric CO2 concentrations at the mesoscale over the MCI domain. For example, we show that, whereas the ABL depth is highly influenced by the choice of ABL scheme and LSM, the mean horizontal wind speed is mainly influenced by the LSM only. Finally, we evaluate the variability in space and time of transport errors and their impact in atmospheric CO2 concentrations. Future work will be to describe transport errors in the MCI regional atmospheric inversion based on the transport uncertainties estimated by this ensemble, for the year 2008.