Wildfire risk and burn severity mapping for aerial fibre and power line routes
Aerial fibre and OPGW routes cross fire-prone terrain that standard engineering surveys ignore. Combining Sentinel-2 burn severity indices, VIIRS active fire detections and fuel-load time-series turns route corridors into scored risk profiles before, during and after a fire event.
Sensors
- Sentinel-2 MSI (ESA): 10 m resolution in visible bands, 20 m in SWIR bands 8A, 11 and 12 used for NBR and dNBR burn severity. Five-day revisit at the equator, improving to two to three days at mid-latitudes with both satellites. Ineffective through active smoke or thick cloud.
- VIIRS (Suomi-NPP and NOAA-20, NASA/NOAA): 375 m active fire product (VNP14IMG) with twice-daily global overpass. Minimum detectable fire size roughly 0.1 ha under clear sky. Near-real-time data latency typically under three hours from observation to public archive. Thermal bands at 3.75 µm and 11 µm.
- Sentinel-1 SAR C-band (ESA): Interferometric coherence change detection at 5 m to 20 m resolution, cloud-independent. Coherence loss between pre- and post-fire image pairs correlates with canopy structural damage. Six-day repeat at mid-latitudes; twelve-day globally. Does not directly measure thermal damage but provides the only reliable burn-extent signal during cloud-obscured events.
- MODIS Terra and Aqua (NASA): 500 m burned area product (MCD64A1) with monthly compositing and a global archive from 2000. Coarser than Sentinel-2 but provides two decades of fire history for probabilistic route scoring. Four daily overpasses combined across both satellites.
Why aerial cable sheaths have a thermal problem that engineers underestimate
Standard polyethylene outer sheaths on aerial fibre cables begin to deform at sustained temperatures above roughly 70 to 90 °C. OPGW (optical ground wire) aluminium cladding softens at higher temperatures, but the optical fibre inside is typically rated to 85 °C continuous. A crown fire moving at two to five metres per second through dry eucalyptus or pine canopy can produce radiant heat fluxes of 50 to 150 kW/m² at mid-canopy height. That is not a marginal exceedance. It is structural failure territory, often within seconds of flame contact.
The engineering implication is that the relevant question for route planning is not whether a fire will occur somewhere in a region, but whether fuel conditions along a specific corridor are sufficient to sustain the flame lengths and intensities that breach cable thermal thresholds. Canopy fuel moisture content below roughly 80 to 100 per cent live fuel moisture is the threshold most fire behaviour models associate with crown fire transition. Satellite-derived NDVI and SWIR reflectance time-series track the seasonal drying trajectory that drives that transition, weeks before any ignition.
What a floating roof gives away: reading NBR and dNBR along a route
The Normalised Burn Ratio (NBR) uses Sentinel-2 bands 8A (near-infrared, 865 nm) and 12 (SWIR, 2190 nm). Healthy vegetation has high NIR reflectance and low SWIR reflectance; fire-affected ground reverses that relationship. The differenced NBR (dNBR), computed between a pre-fire and post-fire image pair, produces a continuous severity index. Published USGS thresholds classify dNBR values above 0.66 as high severity, 0.27 to 0.44 as moderate-high, and below 0.1 as unburned or low severity. Those thresholds were developed for North American conifer forests and require local calibration for other biomes, which is an honest limit worth stating upfront.
For a route assessment, the useful product is not a single dNBR raster but a corridor profile: the distribution of severity classes within a defined buffer (typically 100 m to 500 m) around each span segment. Spans crossing high-severity patches in the MODIS historical archive and also sitting in high-fuel-load zones identified from recent NDVI are the ones that warrant detailed structural inspection or re-routing consideration. The combination of historical frequency and current fuel state is more informative than either alone.
The cloud problem during active events, and what SAR coherence actually tells you
Active wildfires generate dense smoke columns that render Sentinel-2 optical imagery unusable precisely when situational awareness matters most. VIIRS thermal anomaly products continue to detect fire fronts through moderate smoke at 375 m resolution, but they do not measure burn extent or severity on the ground below the canopy.
Sentinel-1 C-band SAR penetrates smoke and most cloud. When a fire removes canopy vegetation, the backscatter signal from the ground surface changes, and interferometric coherence between a pre-fire and a post-fire image pair drops sharply. Coherence values approaching zero indicate complete canopy loss; moderate coherence loss suggests partial scorch. The limitation is temporal: Sentinel-1 revisit is six days at mid-latitudes, so coherence change maps are available days after the fire front passes, not in real time. They are best used for post-event damage triage of affected spans rather than active fire tracking. Combining VIIRS thermal detections for real-time front position with Sentinel-1 coherence for post-event structural damage gives a more complete picture than either sensor alone.
Scoring a route before the fire season: the fuel-load time-series method
NDVI time-series from Sentinel-2 or MODIS, computed monthly over three to five years, reveal the seasonal biomass accumulation cycle for each corridor segment. Segments with persistently high NDVI through the dry season, or with anomalously rapid green-up following previous fire years, carry elevated fine-fuel loads. The relationship between NDVI and fuel load is not linear and varies by vegetation type, so the output is a relative risk rank across segments rather than a calibrated fuel-mass figure in tonnes per hectare.
The practical output for a network operator is a route-segment risk score updated at the start of each fire season, flagging spans where current-year fuel state exceeds the historical 80th percentile. That score can inform pre-season inspection scheduling, vegetation management contracts and insurance submissions. It does not predict ignition, which depends on human activity and lightning patterns that satellite data cannot resolve at span level. Satellize applies this method operationally on open Sentinel and VIIRS archives; the same analytical pipeline that supports the Tonga crop-estimation programme underpins the vegetation phenology processing here.
Post-fire: what the severity map tells a network operations team
After a fire event, a dNBR map at 20 m resolution can identify which specific spans crossed high-severity burn patches. That is not the same as knowing a cable has failed. Thermal damage to a sheath may not immediately affect signal transmission; optical fibre can survive moderate heating if the sheath retains its geometry. The severity map prioritises inspection order. Spans over high-severity patches get physical inspection first; spans over low-severity or unburned ground can wait.
A secondary post-fire risk is often underappreciated: burned slopes lose root cohesion rapidly, and debris flows or shallow landslides in the first wet season after fire can displace poles and anchor points even where the fire itself caused no direct cable damage. Sentinel-1 coherence monitoring through the post-fire wet season can detect ground movement at affected slopes, providing an early warning that complements the initial burn severity assessment. This is not InSAR-grade deformation measurement; it is a change-detection flag that warrants ground follow-up.
Honest limits and the numbers that matter
Sentinel-2 cannot image through cloud or active smoke. In tropical or subtropical fire seasons where convective cloud is persistent, optical burn mapping may be delayed by one to three weeks. Sentinel-1 coherence fills part of that gap but requires a pre-event image from the same orbital geometry, so archive availability matters. VIIRS active fire detections have a minimum detectable fire size of roughly 0.1 ha under clear conditions; small ignitions in early stages may be missed.
NBR thresholds published for one biome do not transfer cleanly to another. A dNBR of 0.5 in a California mixed-conifer forest and a dNBR of 0.5 in a West African savanna represent different physical conditions. Any route-scoring workflow that applies published thresholds without local validation is producing a first-order estimate, not a calibrated risk figure. That is still useful for ranking spans relative to each other, which is the primary application here, but it should not be presented as an absolute severity measurement.
Typical figures
| Burn severity spatial resolution | 20 m (Sentinel-2 SWIR bands for NBR/dNBR); 500 m (MODIS MCD64A1 historical archive) |
| Active fire detection resolution | 375 m (VIIRS VNP14IMG); 1 km (MODIS MOD14/MYD14) |
| SAR coherence resolution | 5 m to 20 m (Sentinel-1 IW mode, depending on processing multilook) |
| Optical revisit (route corridor) | 2 to 5 days cloud-free (Sentinel-2A+B combined at mid-latitudes); cloud cover may extend effective revisit to weeks in wet seasons |
| VIIRS near-real-time latency | Typically under 3 hours from overpass to public archive (NASA FIRMS) |
| Sentinel-1 revisit | 6 days at mid-latitudes; 12 days globally (single satellite) |
| Historical fire archive depth | MODIS burned area from 2000; Sentinel-2 from 2015; Sentinel-1 from 2014 |
| Minimum detectable active fire (VIIRS) | Approximately 0.1 ha under clear-sky conditions |
| Spectral bands used | Sentinel-2 bands 8A (865 nm), 11 (1610 nm), 12 (2190 nm) for NBR; VIIRS bands M13 (3.75 µm) and M15 (11 µm) for thermal detection |
| Deliverable formats | GeoTIFF severity rasters, GeoJSON route-segment risk scores, CSV span inspection priority lists, PDF seasonal risk reports |
Analytics Satellize can run
| Pre-season fuel-load risk score per route span | NDVI time-series anomaly detection (Sentinel-2 or MODIS) against 3 to 5 year baseline; percentile ranking of current-year biomass accumulation | GeoJSON layer with per-span risk percentile, updated at start of fire season |
| Active fire proximity alert | VIIRS VNP14IMG thermal anomaly buffered against route corridor geometry; alert triggered when active detection falls within configurable distance threshold | Near-real-time alert feed (email or API) with fire front coordinates and affected span IDs |
| Post-fire burn severity corridor profile | dNBR computed from Sentinel-2 pre/post image pair; severity class distribution extracted within 100 m to 500 m route buffer using published USGS dNBR thresholds with local calibration note | GeoTIFF severity raster plus CSV span-level severity summary for inspection prioritisation |
| Cloud-independent burn extent map | Sentinel-1 C-band interferometric coherence change between pre- and post-fire image pairs; coherence loss threshold applied to delineate canopy removal zones | GeoTIFF coherence-change layer delivered within one Sentinel-1 repeat cycle (6 to 12 days) of fire event |
| Post-fire ground instability flag | Sentinel-1 coherence monitoring on burned slopes through first post-fire wet season; change detection flags potential debris-flow or shallow-slide initiation zones near pole anchor points | Seasonal GeoJSON alert layer with flagged pole locations requiring physical inspection |
| Multi-decade fire frequency map for route corridor | MODIS MCD64A1 burned area product aggregated 2000 to present; fire return interval and cumulative burn frequency computed per 500 m grid cell intersecting route buffer | GeoTIFF fire frequency raster and PDF route risk summary for insurance or regulatory submission |
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.