Towards a Dynamic Multiscale Wildfire Digital Twin

Climate change is creating highly favorable regional environments globally, that are inducing wildland fires at explosive spread with historically greater scale and intensity. These wildfires are destroying wide swaths of homes and related resources, transporting toxic smoke damaging to human health and turning regional forests into Savannahs. Recent studies have indicated we are at a tipping point where the feedback effects of deforestation will accelerate global warming beyond current efforts to curb anthropomorphic CO2. As part of a new NASA formed Fire-Tech program for the development of real time fire observing instruments and wildfire models, we have been awarded the opportunity to develop a NASA Wildfire Digital Twin. Digital twins simulate complex processes that occur on multiple scales with interacting physical systems driven by real time observations. Wildfires are examples of such physical, multiscale, interactive processes that draw on the local topography, surface vegetation and regional atmospheric states to sustain fire spread and produce distant health impacts through the atmospheric propagation of the chemical and aerosol smoke particles. We will present the first case study results of the NASA Unified Weather Research Forecast (NUWRF) based Wildfire Digital Twin (NWDT) simulation employing an interactive fire spread module (SFIRE). The NWDT model, when fully developed, will need to run nested atmospheric grids over North America at 3km, the US at 1km and a sub-cloud resolving resolution of covering a large wildfire at 300m that combines real time data assimilation for the SFire spread module at 10-100 m resolution. SFire can be coupled to large eddy simulation runs to form deep integrated plumes allowing NWDT to track atmospheric smoke over the CONUS. We further are implementing a real-time dynamic fire data assimilation forecast cycle to conduct 'what if' digital twin scenarios. Initial results of a machine learning FourCastNet model emulation of the NWDT aimed for subsequent sub-km resolutions employing their Adaptive Fourier Neural Operators will also be presented. Finally, we will present recent regional hospital data collected on the increased health impacts traceable to toxic smoke.