Modeling fuel loads and predicting fire behavior: Novel applications of remote sensing technology

Researchers are using site-specific data, multi-scale remotely sensed data relating to fuels (e.g. shrubs, litter, duff), weather data, and information gathered from literature sources and external databases to model fuel dynamics and inform prescribed fire management. 

The overall goal of this research project is to develop a novel, user-friendly approach to mapping fuel structures, predicting fuel combustion, and modeling smoke plumes and fire emissions. Outputs of the research will be at multiple scales - from individual trees to ecological subunits and management units - to allow for the seamless transition across scales relevant to land use managers charged with predicting smoke plumes and forecasting emissions from their prescribed fire operations. The research project began in 2020 and is planned to continue through Fiscal Year 2025. Expected data products include multi-scale measurements of fuels, fuel consumption, heat release, and smoke plumes that are needed by fire and smoke modelers to improve the next generation of physics-based models of fire behavior and emissions.

Traditional models of forest fuels generally don’t account for the spatial heterogeneity inherent to forests, including gaps in vegetation, landform changes, and tree groupings. This forest heterogeneity results in variation in fuel loads and fire behavior that is often missed. In this project, researchers are using high resolution remote sensing data – light detection and ranging (lidar) data in particular - to characterize tree aboveground biomass across entire forest landscapes. High spatial resolution of the remote sensing data helps better represent the spatial heterogeneity on the landscape and therefore better predict fuel loads of different fuel components and how they change over time.

Plant litter under the forest canopy is an important, though often missed, factor in predicting fire behavior. This research project uses novel approaches to better characterize heterogeneous surfaces fuel sources, such as litter.

The research team is integrating remotely sensed data with object-based analysis, machine learning techniques, and ecological-based concepts to develop different predictive models of fuel loads. Taking an “object-based” approach means that data and images are being analyzed at a scale that is ecologically relevant and linked to real world objects such as a tree or a forest stand. This approach allows manager to predict and interpret the results of prescribed fire at multiple relevant scales and compare treatment options.

Spatially explicit characterization of both canopy and surface fuels is a novel approach. In forest ecosystems, the surface fuels are the major contributors to fire spread and consumption, especially on sites under prescribed fire management. Unfortunately, surface fuels are extremely difficult to estimate, especially under canopy cover, due to the low sensitivity of overhead remote sensing systems. In this project, researchers are bringing in ancillary information from external databases, such as fire history, and using models to infer the “missing” information, such as surface fuel loads. 

For example, the researchers are mapping litter loads at high spatial resolution by characterizing first tree-level foliage biomass that is then the main contributor to annual litterfall; subsequent litter accumulation is then modeled, relying on ecological-based concepts. Additional fuel components comprising surface fuel beds that the researchers are working to better estimate are the wood debris originating from trees and partially to fully decomposed litter. These are all critical to develop fire behavior models and predict fire consumption and emissions during prescribed fire.

Research Objectives:

1. Estimate fuel loads in ways that more accurately account for the spatial heterogeneity of longleaf pine forests and at scales that are relevant to managers, such as trees, ecological subunits, and management units.
2. Develop fire behavior models to predict fire combustion/consumption and resulting emissions at multiple spatial scales and under variable fuel moisture and fire weather scenarios.
3. Determine the spatial resolution of fuel and fire data that accurately forecasts smoke plume development and emissions at multiple operational scales.