Neural network predictions of pollutant emissions from open burning of crop residues: Application to air quality forecasts in southern China

Open burning of crop residues is a strong seasonal source of air pollutants in many parts of China, but the large day-to-day variability of the associated emissions pose a great challenge for air quality forecasts. Here we developed back-propagation neural network (BPNN) ensembles to forecast the daily fire pixel counts in Southern China during the month of January. The BPNN ensembles were trained using daily assimilated surface meteorological data (including air temperature, relative humidity, pressure, and winds) and daily fire pixel observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) for the month of January during the years 2003-2012. We showed that the BPNN ensembles successfully forecasted the day-to-day variability and the interannual variability of fire pixel counts over Southern China in January of the years 2013-2015, with correlation coefficients of 0.6-0.8 against the MODIS observations. We used the forecasted daily fire pixel counts in January 2014 and January 2015 to scale the climatological January biomass burning emissions from the Fire Inventory from NCAR (FINN) and applied the resulting forecasted daily biomass burning emissions to drive the WRF-Chem regional air quality model. The use of BPNN-ensemble-forecasted daily biomass burning pollutant emissions led to significant improvements in the daily forecasts of PM2.5 concentrations in Southern China for January 2014, with the mean bias of the simulated surface PM2.5 concentrations reduced from -9.1% to -1.2%. We repeated the sensitivity simulations for January 2015 and also found a modest improvement when using the forecasted daily biomass burning pollutant emissions (mean bias of the simulated surface PM2.5 concentrations reduced from -5.8% to -2%). Our approach can be applied to other source regions of biomass burning emissions to improve the accuracy of daily air quality forecasts.