Wildfire fuel-load mapping for property exposure assessment
Spaceborne lidar, SWIR reflectance and multispectral vegetation indices combine to estimate where combustible material accumulates at the wildland-urban interface, scoring properties by proximity to high-load corridors before underwriters or planners ever visit the site.
Sensors
- GEDI (Global Ecosystem Dynamics Investigation): Full-waveform lidar aboard the ISS. Footprints of approximately 25 m diameter, spaced 60 m along-track and 600 m across-track. Measures canopy top height, canopy cover fraction and vertical foliage profile, which are the primary inputs to canopy fuel-load estimation. Coverage is limited to latitudes between roughly 51.6° N and 51.6° S.
- Sentinel-2 MSI (ESA Copernicus): Ten-metre visible and near-infrared bands; 20 m SWIR bands at 1610 nm (Band 11) and 2190 nm (Band 12). SWIR reflectance is sensitive to leaf and canopy moisture content, providing a proxy for live fuel moisture. Five-day revisit at the equator with two satellites, enabling seasonal moisture tracking through the fire season.
- Landsat 8/9 OLI: 30 m multispectral bands including SWIR at 1570 nm and 2110 nm. Sixteen-day single-satellite revisit, extended to eight days with both Landsat 8 and 9 operating together. Useful for long-term fuel accumulation trends and for calibrating vegetation indices against a multi-decade archive stretching back to 1972 via earlier Landsat missions.
- Planet SuperDove: 3 m resolution, eight spectral bands including red-edge and near-infrared, daily revisit over most land areas. Resolves individual tree crowns and shrub patches within interface zones, enabling fine-scale fuel-type classification that coarser sensors cannot achieve. Commercial tasking; archive depth varies by region.
- MODIS / VIIRS (NASA/NOAA): 250 m to 375 m resolution, near-daily global coverage. Not used for per-property mapping, but the multi-year NDVI and EVI time series from these sensors underpin regional fuel-accumulation baselines and fire-history context that calibrate the higher-resolution analysis.
What the canopy is hiding, and why it matters to an underwriter
Surface and canopy fuel loads are the primary determinant of fire intensity and rate of spread. Crown fire behaviour in conifer stands, for instance, is strongly governed by canopy bulk density and the vertical continuity of fuel from the ground layer upward. An insurer pricing a property in the wildland-urban interface without a fuel-load estimate is essentially pricing fire risk from a postcode alone.
Satellite data cannot replace a ground-based fuel survey, and it does not try to. What it can do is characterise the spatial pattern of fuel accumulation across tens of thousands of hectares at a cost and speed that ground crews cannot match. The honest constraint is that satellite methods estimate fuel load indirectly, through proxies: canopy height, vegetation density, spectral moisture content and fuel-type class. Uncertainty grows in dense multi-storey canopies where the sensor cannot see the surface layer, and in areas with persistent cloud cover that interrupts the SWIR moisture time series.
Reading moisture from reflected light
Water absorbs strongly at 1400 nm and 1900 nm, with secondary absorption features at the wavelengths Sentinel-2 Band 11 (1610 nm) and Band 12 (2190 nm) are designed to capture. Live fuel moisture content (LFMC) correlates with the ratio of SWIR to near-infrared reflectance: drier vegetation reflects more strongly in SWIR relative to NIR. Published studies using Sentinel-2 SWIR bands have demonstrated LFMC retrieval errors in the range of 20 to 30 percentage points under open canopy conditions, which is sufficient to distinguish critically dry fuel states from adequately moist ones.
The practical limitation is canopy closure. Once a conifer canopy exceeds roughly 70 to 80 percent cover, SWIR reflectance increasingly represents the upper canopy rather than the understorey shrubs and dead surface fuel that actually carry a ground fire. In those zones, the SWIR moisture index must be interpreted alongside the canopy structure data from GEDI rather than used alone.
What a floating roof gives away: canopy height from spaceborne lidar
GEDI's full-waveform lidar decomposes the return pulse from a 25 m footprint into contributions from ground, understorey and canopy layers. From this, it derives canopy top height, relative height metrics (RH50, RH75, RH98) and plant area index. These are the inputs the fire-behaviour modelling community uses to estimate canopy bulk density, a key parameter in the Scott and Reinhardt crown fire initiation model and its successors.
The spatial sampling of GEDI is sparse compared with airborne lidar. The 600 m cross-track spacing means that individual footprints must be used statistically, combined with Sentinel-2 and Landsat spectral predictors in a wall-to-wall regression model, to produce continuous canopy height maps at 10 to 30 m resolution. This is a well-established method in the forestry remote-sensing literature. The result is an estimate, not a measurement, and the confidence interval widens in structurally complex mixed-species stands.
For properties at the urban fringe, the canopy height model also identifies ladder-fuel conditions: situations where low shrubs connect vertically to tall trees, creating pathways for fire to climb from the surface into the crown. These are the configurations that produce the rapid, erratic fire spread that has driven catastrophic losses in recent interface fires.
Scoring properties: from fuel maps to exposure indices
A fuel-load map on its own is not an underwriting product. The translation step requires combining fuel-load and moisture estimates with topographic slope and aspect (which control fire-spread rate and direction), prevailing wind climatology, and the spatial relationship between each insured property and the surrounding fuel matrix.
The standard approach draws on published fire-spread models such as Rothermel's surface fire spread equations, which relate rate of spread to fuel load, moisture, slope and wind. Satellite-derived inputs feed these equations to generate fire-spread corridor estimates: the probable paths a fire would follow under design-wind conditions. Each property is then scored by its position relative to those corridors and by the fuel load within a defined radius, typically 100 m to 500 m depending on the terrain and the insurer's risk appetite.
Satellize applies this pipeline to open-constellation data for interface zones, producing per-parcel exposure scores that can be ingested directly into underwriting platforms. The methodology is structurally similar to the approach used in the Tonga crop-estimation programme, where spectral indices are translated into field-level economic outputs rather than fire-risk scores.
Honest limits: what the satellite cannot resolve
Several factors reduce confidence in satellite-derived fuel assessments. First, dead surface fuel, including fallen branches, dry grass and duff, is largely invisible to optical sensors once it is beneath a closed canopy. It is the most important carrier of ground fire and the hardest to map remotely. Second, fine-scale heterogeneity within a 10 m pixel, such as a cleared defensible space immediately around a house, is averaged away unless Planet SuperDove imagery is used at 3 m resolution.
Third, GEDI's ISS orbit means that any given location is sampled only occasionally, and the sparse footprint grid introduces spatial interpolation error that can reach several metres in canopy height estimates. Fourth, cloud cover in humid interface regions, parts of the Pacific Northwest or southern Chile for instance, can create gaps of weeks in the Sentinel-2 SWIR time series during the fire season itself. SAR backscatter from Sentinel-1 can partially fill those gaps for vegetation-density estimates, but it does not provide a direct moisture proxy.
These are not reasons to avoid satellite fuel mapping. They are reasons to be specific about what the product is: a scalable, repeatable, cost-effective first screen that identifies which properties warrant closer attention, not a substitute for ground-truthed fuel surveys in the highest-risk parcels.
Typical figures
| Canopy height spatial resolution | GEDI footprints ~25 m diameter; wall-to-wall model output at 10–30 m using spectral predictors |
| Fuel-moisture index resolution | 20 m (Sentinel-2 Band 11/12); 30 m (Landsat 8/9 OLI SWIR) |
| Fine-scale fuel-type classification | 3 m (Planet SuperDove, commercial licence required) |
| Revisit for moisture time series | 5 days (Sentinel-2 dual satellite); 8 days (Landsat 8+9 combined) |
| GEDI latitude coverage | 51.6° N to 51.6° S; polar regions not covered |
| SWIR moisture retrieval accuracy | Live fuel moisture content error ~20–30 percentage points under open canopy; degrades under closed canopy >70–80% cover |
| Spectral bands used | NIR, SWIR-1 (1610 nm), SWIR-2 (2190 nm), Red-Edge, NDVI, EVI; GEDI RH50/RH75/RH98 waveform metrics |
| Archive depth | Sentinel-2 from 2015; Landsat from 1972; GEDI from April 2019; Planet varies by region |
| Delivery format | GeoTIFF fuel-load and moisture rasters; GeoJSON per-parcel exposure scores; PDF summary report |
| Latency from image acquisition to scored output | Typically 3–7 days for a new area of interest; faster for pre-processed regions |
Analytics Satellize can run
| Wall-to-wall canopy fuel-load map | GEDI RH metrics and plant area index combined with Sentinel-2 NDVI and SWIR bands in a random-forest regression model to predict canopy bulk density and canopy base height at 20–30 m resolution | GeoTIFF raster layer, updated seasonally or on request |
| Live fuel moisture content (LFMC) index | Sentinel-2 SWIR-to-NIR ratio (equivalent to the Normalised Difference Water Index) calibrated against published LFMC retrieval curves; time-series stack for seasonal trend | Monthly raster stack with anomaly flags for critically dry conditions; GIS layer |
| Fuel-type classification map | Supervised classification of Planet SuperDove eight-band imagery using training samples from published fuel-type libraries (e.g. FBFM40 classes); validated against Landsat spectral signatures | GeoTIFF fuel-type raster at 3 m resolution for defined area of interest |
| Fire-spread corridor model | Rothermel surface fire spread equations parameterised with satellite-derived fuel load, slope from a digital elevation model, and regional wind climatology; outputs probable spread paths under 90th-percentile wind conditions | GeoJSON corridor polygons with rate-of-spread estimates; PDF map |
| Per-parcel wildfire exposure score | Spatial join of cadastral parcel boundaries with fuel-load raster, LFMC anomaly layer and fire-spread corridors; score computed as weighted combination of proximity, fuel density and moisture deficit | CSV or GeoJSON parcel-level score table, ingestible into underwriting platforms |
| Fuel-accumulation change detection | Landsat multi-year NDVI and SWIR time series analysed for progressive fuel build-up in previously disturbed areas (post-fire regrowth, abandoned agricultural land); change magnitude quantified against baseline | Annual change-detection report with flagged high-accumulation polygons |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.