Forest fire regimes are changing faster than the tools used to predict them. In the U.S., the impact of drought, a century of accumulated fuel, and an expanding wildland-urban interface have produced fire behavior that operational models, built on homogeneous representations of fuel, struggle to predict. Advanced fire models can capture that behavior, but need input data that describes fuels in three dimensions, and at a meter scale, across forest and urban landscapes. Today, those data barely exist.
In a recent technology challenge, CTrees scientists introduced a new approach to model 3D fuels. The submission to the 3D Surface Fuels and Vegetation Modeling Prize Challenge (opens in a new tab), a competition run by SERDP/ESTCP with the Naval Air Warfare Center Training Systems Division and the Central Florida Tech Grove, won third place. The entry team was led by CTrees’ CEO and chief scientist Sassan Saatchi and included CTrees scientists and experts Zhihua Liu, Bryan Shaddy, Chris Mihiar, and Aijing Li.
The delivered arrays, validation, and interactive 3D models are now public at 3dfuels.ctrees.org (opens in a new tab).
Modeling ladder fuels that accelerate fire spread
Ladder fuels, the shrubs and small trees between the forest floor and the canopy, are one of the decisive factors in whether a fire reaches the crowns. But ladder fuels are difficult to measure: Remote sensing products struggle to detect shrubs beneath the canopy, and ground measurement, while accurate, does not scale across landscapes.
CTrees’ 3D technique uses lidar to model fuels in three layers: canopy, ladder, and ground. The model combines all three strata in a single, one-meter-scale voxel (a volumetric, 3D pixel) that includes information on bulk density, moisture, live fraction, and surface area to volume. The voxels are represented in a 3D array that shows how fuels are distributed in the landscape, and the entire package is delivered in a format that next-gen fire models can read directly.
Since CTrees’ model is built on public data (USGS 3DEP airborne lidar), it can scale to multiple ecosystems. CTrees’ entry includes examples from mixed conifer forest near South Lake Tahoe, chamise chaparral in the San Jacinto Mountains, and a loblolly pine and sweetgum plantation in the Georgia Piedmont.
Why 3D fuel modeling matters for wildfire resilience
Mapping 3D fuels across the continental U.S. is an important step toward wildfire resilience. Fire is an inherent part of many western ecosystems, and Indigenous peoples have long used cultural burning to steward them. A century of fire suppression has led to fuel buildup, and bringing good fire back to the landscape is now a priority. Mapping 3D fuels at scale helps land managers plan and design prescribed burns where they are needed most, in a cost-effective way.
Observed 3D fuel structure is the foundation of a broader CTrees roadmap for wildfire resilience technology that includes fuel moisture observation and fire behavior forecasting. An AI fire model is only as good as the fuel state it is conditioned on, which is why the measured layer matters most.
CTrees thanks SERDP/ESTCP, NAWCTSD and the Central Florida Tech Grove, and the subject matter experts who reviewed the entries. No single approach gets the field to operational 3D fuels, and the challenge was designed to bring that community together. We look forward to working with the other finalists, and with military and civilian land managers, particularly on the field measurements that would close gaps for validation.
Interested in partnering with CTrees on technology for wildfire resilience? Please reach out via our contact form.
Technical note: Science for 3D modeling of forest fire fuels
CTrees’ 3D fuel modeling grows out of two decades of work by our scientists on measuring vegetation structure. The methods behind each step have been published, tested against field measurements, and used in diverse ecosystems, from California’s mixed conifer forests, to Yellowstone’s lodgepole pine forests, and the Brazilian Cerrado. The Adaptive Mean Shift 3D (AMS3D) model has been independently benchmarked against competing approaches. In its competition entry, CTrees carried the science down to the meter scale, into three dimensions, and into the near-ground layer that fire actually moves through.
References:
Aubry-Kientz, M., Dutrieux, R., Ferraz, A., Saatchi, S., Hamraz, H., Williams, J., Coomes, D., Piboule, A., & Vincent, G. (2019). A Comparative Assessment of the Performance of Individual Tree Crowns Delineation Algorithms from ALS Data in Tropical Forests. Remote Sensing, 11(9), 1086. https://doi.org/10.3390/rs11091086 (opens in a new tab)
Ferraz, A., Saatchi, S., Mallet, C., Jacquemoud, S., Gonçalves, G., Silva, C. A., Soares, P., Tomé, M., & Pereira, L. (2016). Airborne Lidar Estimation of Aboveground Forest Biomass in the Absence of Field Inventory. Remote Sensing, 8(8), 653. https://doi.org/10.3390/rs8080653 (opens in a new tab)
Ferraz, A., Saatchi, S., Mallet, C., Meyer, V. (2016). Lidar Detection of Individual Tree Size in Tropical Forests. Remote Sensing of Environment, Vol. 183, 318-333. https://doi.org/10.1016/j.rse.2016.05.028 (opens in a new tab)
Ferraz, A., Saatchi, S., Longo, M., and Clark, D. B. (2020). Tropical Tree Size–Frequency Distributions from Airborne Lidar. Ecological Applications 30(7):e02154. https://doi.org/10.1002/eap.2154 (opens in a new tab)
García, M., Saatchi, S., Casas, A., Koltunov, A., Ustin, S. L., Ramirez, C., & Balzter, H. (2017). Extrapolating Forest Canopy Fuel Properties in the California Rim Fire by Combining Airborne LiDAR and Landsat OLI Data. Remote Sensing, 9(4), 394. https://doi.org/10.3390/rs9040394 (opens in a new tab)
