The accuracy of numerical weather prediction (NWP) depends on the accuracy of the inputted initial conditions (among other factors). Ensemble data assimilation (EnsDA) corrects those initial conditions through exploiting observations and probabilistic forecasts via Bayes’ rule. In principle, any observation can be assimilated as long as statistical associations connect the observable quantity with NWP quantities. EnsDA thus extracts corrections for multiple NWP grid boxes and variables from a single observation. Given its flexible force-multiplying corrective capabilities, EnsDA is a crucial component of NWP systems. As such, advancements in EnsDA have the potential to improve NWP.
In this colloquium, I will first provide a gentle introduction (no equations!) on EnsDA before discussing several efforts to advance EnsDA. The first effort is an interdisciplinary exploration into assimilating an astrophysical observation source (ground level muon fluxes) to improve NWP. The potential of using this novel observation source is demonstrated using a case study of the devastating record-breaking Severe Tropical Cyclone Freddy (2023). The second effort is to improve the assimilation of all-sky geostationary satellite infrared radiances through understanding and mitigating damaging side-effects resulting from inappropriate statistical assumptions. Finally, I will discuss an ongoing effort to improve the ensemble statistics used in EnsDA by exploiting user knowledge to cost-effectively generate large ensembles of NWP model states.