The assessment of two smoke modelling systems for forecasting local air quality impacts from smouldering peat fires

On 17 March 2018 major grassland fires in south-west Victoria, Australia ignited several peat bogs. The peat fires smouldered for 40 days, generating substantial amounts of smoke. The surrounding communities were exposed to significant concentrations of fine particles (PM2.5), resulting in the need for active interventions (such as the relocation of schools) in order to protect vulnerable communities. The peat fires provided a unique opportunity to review the capabilities of two air pollution forecasting models to assess the impact from the peat fires on nearby communities: the Air Quality Forecasting System (AQFx) and the Accident Reporting and Guiding Operational System (ARGOS). Both systems are used in Victoria to inform emergency management response strategies and community warnings. A key configuration change was made to the smoke emissions module in AQFx from simulating emissions from a planned burn to a sub-surface peat fire. Emissions were derived by using heat maps generated from aerial imagery data and estimates of fuel load determined by the peat bulk density and depth of the peat. The results indicated that AQFx successfully captured most smoke plume events during the simulation period, despite some errors in timing and magnitude. Accurate forecasting was most challenged by calm conditions, and meso-scale meteorological transition events. The ARGOS model performed better at capturing smoke plume dispersion during a meso-scale meteorological transition event due to meteorological forecasts being updated every 6 hours compared to a 24-hour update in AQFx. However, the source geometry in ARGOS meant that emission rates were concentrated within a small number of release points resulting in relatively narrow plumes with a likely overprediction of higher end PM2.5 concentrations. The AQFx system was better suited to area emissions. The ARGOS and AQFx models showed different strengths in providing timely information to emergency response agencies to better manage smoke impacts from smouldering peat fires on communities. The ARGOS model can be set up quickly with preliminary emission estimates that can be adjusted as more accurate and updated information becomes available. Compared to the ARGOS model, AQFx is better suited for area emissions. Implementation of a rapid update cycle in AQFx would further improve forecasts especially during meteorological transition events.