Fuel-Driven Wildfire Risk Mapping Over CONUS to Guide Targeted Resources Allocation

Dynamic factors such as precipitation, fuel accumulation, and live fuel moisture content all play a role in determining the risk of any given wildland area to fire potential, while evolving on differing timescales. Fuel compositions within a landscape may evolve slowly over decades or dramatically change within weeks depending on the fuel category (i.e., trees versus grasses). Likewise, LFMC can vary substantially within differing vegetation types at the same site due to plant physiological adaptions to water stress. Despite the dynamic nature of fuels, most operational wildland fire management systems are dependent on in-situ sampling or CONUS-scale maps updated at a nominal 5-year cadence, which fails to capture both the broad-scale and high temporal resolution needed for CONUS-scale fire surveillance. Recent large-scale fires have demonstrated how quickly fuel conditions can change and manifest into explosive fire conditions, emphasizing the even greater need for up-to-date fuels information. To address this operational need, we propose the development of three ML algorithms using existing remote sensing assets and ground-based measurements. The proposed project will address the pre-fire life-cycle stage by combining multiple remote sensing and ground-based measurements with machine learning (ML) to provide rapid updates (daily to monthly) at fine spatial scale (~30-meter) for three key fuels intelligence gaps in the operational landscape: (1) fuel category (i.e., grasses, mixed forests, etc.), (2) canopy structure, and (3) live fuel moisture content (LFMC). These fuels intelligence products will be produced at CONUS-scale and will serve as input to a daily extreme wildfire risk analysis based on ensemble 3D fire behavior simulations, meteorological conditions and the climatology of risk at local scale. Together, the new fuels maps and wildfire risk index will provide a comprehensive view of how fuel conditions are changing and the subsequent implications for enhanced risk, enabling better coordination amongst wildland land and fire management at regional and national scale. All three proposed ML products will leverage various NASA, USGS and ESA remote sensing assets in conjunction with existing ground measurements to generate broad-coverage maps. The fuel category algorithm will ingest monthly accumulated observations from ESA's Sentinel-1A C-band synthetic aperture radar (SAR) and USGS Landsat multispectral imagery to estimate fuel category. To estimate canopy structure, we will first post-process ICESat-2 measurements to produce a higher spatial-resolution and more accurate canopy height (CH) and fractional canopy coverage (CC) product. We will then train a ML algorithm to estimate on a per-pixel basis CC and CH using monthly accumulated SAR data from both Sentinel-1A (C-band) and NISAR (L-band) and Landsat imagery. Finally, we will use a time series of SAR (Sentinel-1A and NISAR), Landsat, VIIRS and SMAP measurements over the previous 90-days to estimate daily LFMC on a per pixel basis. By using both active and passive microwave imaging sensors as well as optical imagery, we will directly sense both the dry and wet mass of vegetation, and therefore enable more accurate estimation of LFMC. To translate dynamic fuel status maps into daily assessment of fire risk, we will perform ensembles of QUIC-fire, a new 3D fire behavior model, for a variety of fuel, weather and topography input conditions. 3D QUIC-fire simulations will then be postprocessed so that incident fuel and weather conditions can be mapped to an indicator of extreme fire behavior, and thus, the ML-generated fuel conditions maps can be quickly translated to fire risk at CONUS-scale. This project intends to enhance the science of mapping fuel conditions using remote sensing and ML, while also specifically addressing the needs of the operational community, with all proposed methods potentially extensible to a global, operational fire surveillance system.

Enhances the capabilities for existing science instruments for monitoring pre-fire, active-fire, and post-fire situations, reduces the power and mass of these instruments, and enables unprecedented observations in support of wildfire science through distributed observing systems and the information technologies needed for their support.

This is a mature technology (TRL 7+) — the realistic path in is usually NASA's Technology Transfer Program: licensing an existing NASA patent, or a Space Act Agreement to use NASA facilities/expertise directly. NASA also runs a startup licensing program with no upfront fee for companies formed to commercialize a specific NASA technology.

None of these are guaranteed paths for this specific project — TechPort itself doesn't have an "apply" button. Reaching out to the contact(s) above with a specific question is usually the fastest way to find out what's actually open.