APIFLAME v2.0 trace gas and aerosol emissions from biomass burning: application to Portugal during the summer of 2016 and evaluation against satellite observations of CO (IASI) and AOD (MODIS)

Abstract

<jats:p>Abstract. Biomass burning emissions are a major source of trace gases and aerosols. Wildfires being highly variable in time and space, calculating emissions requires a numerical tool able to estimate fluxes at the kilometer scale and with an hourly time-step. Here, the APIFLAME model version 2.0 is presented. It is structured to be modular in terms of input databases and processing methods. The main evolution compared to the version v1.0 is the possibility to merge burned area and fire radiative power (FRP) satellite observations to modulate the temporal variations of fire emissions and to integrate small fires that may not be detected in the burned area product. Accounting for possible missed detection due to small fires results in an increase ranging from ∼ 5 % in Africa and Australia to ∼ 30 % in North America, on average over the 2013–2017 time period based on the Moderate-Resolution Imaging Spectroradiometer (MODIS) collection 6 fire products. An illustration for the case of south-western Europe during the summer of 2016, marked by large wildfires in Portugal, is presented. Emissions calculated using different possible configurations of APIFLAME show a dispersion of 75% on average over the domain during the largest wildfires (8–14/08/2016), which can be considered as an estimate of uncertainty on emissions (excluding the uncertainty on emission factors). Corresponding enhancements of aerosols and carbon monoxide (CO) simulated with the regional chemistry transport model CHIMERE are consistent with observations (good timing for the beginning and end of the events, &amp;amp;pm; 1 day for the timing of the peak values) but tend to be overestimated compared to observations at surface stations. On the contrary, vertically integrated concentrations tend to be underestimated compared to satellite observations of total column CO by the Infrared Atmospheric Sounding Interferometer (IASI) instrument and aerosol optical depth (AOD) by MODIS, which allow regional scale evaluation. This underestimate is lower close to the fire region (5 % to 40 % for AOD depending on the configuration, and 8–18 % for total CO) but rapidly increases downwind. For all comparisons, better agreement is achieved when emissions are injected higher into the free troposphere using a vertical profile as estimated from observations of aerosol plume height by the MISR satellite instrument (injection up to 4 km). The overestimate compared to surface sites and underestimate compared to satellite observations point to uncertainties not only on emissions (total mass and daily variability) but also on their injection profile and on the modelling of the transport of these dense plumes. </jats:p>

Machine Learning-Based Improvement of Aerosol Optical Depth from CHIMERE Simulations using MODIS Satellite Observations

“…The accuracy of the CTM simulations depend on the precision of the inputs e.g., emissions of atmospheric constituents, wind, vertical velocity fields and also on the assumed aerosol properties (e.g., microphysical and optical properties). Therefore, simulations of the aerosol spatial distribution are prone to bias compared to observations [7,8]. The biases are linked to uncertainties in the physical parameterizations of the model, input data, and numerical approximations [9].…”