Error Assessment of Air Quality Forecasting through Chemical Data Assimilation over Southern and Eastern Africa: Characterizing Background and Observation Covariance Errors

Air pollution in Eastern and Southern Africa (E&SA) presents a severe public health concern, contributing to over 23,000 premature deaths annually and exacerbating respiratory ailments for millions. Despite this, air quality forecasting remains challenging due to sparse observational infrastructure. To address these challenges, we are developing an advanced air quality forecasting framework using the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) and its associated data assimilation system (WRF-DA). Our approach incorporates VIIRS aerosol optical depth (AOD) assimilation, with background error statistics derived through the community Gridpoint Statistical Interpolation (GSI). Meteorological inputs are sourced from the Global Forecast System (GFS), while monthly anthropogenic emissions from CAMS and real-time fire emissions from the Fire Inventory from NCAR (FINN) enhance our 48-hour forecasts at 15 km resolution. The forecasts are demonstrated for a case study in June 2022, capturing wildfires, dust storms, and local anthropogenic emissions. We used an updated AOD error estimate using AERONET stations and evaluate the forecast capabilities by comparing the base and assimilation runs against AirNow PM2.5 observations and AERONET observations. Additionally, we assess the impact of observation covariance and background error on the assimilated forecasts and provide insights into pollution source attribution. This work discusses the improvements for operational air quality forecast in data-sparse regions like E&SA.