Assimilation, Blending and Ensembling

 

This component of the project focuses on enhancing the robustness of AQFx smoke forecasts through the following approaches:

1. Blended Forecast Initialization: A new forecast will be initialised by blending the previous day’s AQFx forecast with a global assimilated chemical weather forecast. The improved smoke observation network will be used to spatially blend each data set, providing an optimal initial estimate of smoke concentrations at the start of each AQFx forecast.
2. Optimisation of Smoke Emissions Using Ensemble Kalman Filtering: This approach systematically explores parameter combinations that influence smoke emissions. AQFx is run in a fast tracer mode to simulate smoke transport and forecast accuracy is evaluated using both ground-based and satellite observations. The optimised model parameters then persist for the duration of the AQFx forecast.
3. Integration of Near-Real-Time Observations: Near-real time observational data will be incorporated in the system to support forecast blending. These enhanced forecasts are made available to state control centres and public-facing smoke information apps.
4. Ensemble Modeling:  Multiple models or simulations are combined to improve predictive performance and quantify forecast uncertainty.