Investigating the impact of fires on atmospheric composition using reactive gas and aerosol assimilation

The European Centre for Medium-range Weather Forecasts’ (ECMWF) Composition - Integrated Forecast System (C-IFS) provides global operational forecasts and re-analyses of atmospheric composition at high spatial resolution (~80 km). In this thesis the C-IFS system, with Global Fire Assimilation System (GFAS) emissions, is used to assess the impact of fires on tropospheric ozone and aerosol concentrations. The skill of the model and the impact of data assimilation is evaluated using independent aircraft, ground and satellite observations. In particular, observations from the 2012 South American Biomass Burning Analysis (SAMBBA) field campaign is used as a case-study. This work represents the first evaluation of trace gas observations from SAMBBA. C-IFS simulations show that 4% of the total tropospheric column is produced by fire emissions, with significantly larger contributions closer to fire sources. Compared to both SAMBBA aircraft measurements and Ozone Monitoring Instrument (OMI) satellite data (0-6 km column) O3 concentrations are underestimated in the tropics. Evaluation of precursor emissions show that this is due to a combination of NOx flux underestimation in GFAS and the lack of a prescribed injection height, partly acting through a subsequent limited influence of PAN. Carbon monoxide, a tracer of biomass burning transport, is well captured by C-IFS implying that GFAS detects the majority of fires although the occurrence of small ones appears underestimated. The assimilation of satellite reactive gas measurements from OMI, Measurements Of Pollution In The Troposphere (MOPITT) and the Microwave Limb Sounder (MLS) generally improved the C-IFS comparison against SAMBBA and other independent data. The O3 tropospheric representation was improved by MLS assimilation but a significant negative bias against OMI (0-6 km column) remained. The chemical data assimilation impacts non-assimilated species such OH with increases up to 8% in the tropics. However, compared to observations this degraded the OH Northern/Southern hemisphere ratio and the impact of the OH change on other species was minimal. Fires in C-IFS are shown to dominate the global carbonaceous aerosol budget during the tropical fire season. Aerosol concentrations are slightly underestimated by C-IFS compared to SAMBBA. The relative contribution of black carbon (BC) and organic matter (OM) is similar to that observed in the eastern savannah but the contribution of BC is overestimated in the western deforestation region. The use of the suggested ×3.4 GFAS scaling factor leads to a large positive bias in the model aerosol compared to SAMBBA. The assimilation of aerosol optical depth (AOD) generally decreases concentrations near fire-sources, due to this scaling, but still increases concentrations in remote regions. This suggests an inaccurate model loss of aerosol is a cause of the bias. The analysis/model bias used to characterise a more realistic local scaling factor which is shown to be different for BC and OM, and for AOD metrics compared to aerosol mass