Situation Report: Bogota Cerros Orientales, Wildfire Risk Profile

Situation Report: Fire Risk in the Bogotá Cerros Orientales

Capybara Center · Early Warning System (EWS) · Alejandro Feged-Rivadeneira, Tecnológico de Monterrey · 22 July 2026

Bottom line. Two conditions decide fire in the Cerros Orientales, and they must be read together. Where a fire can occur depends on fuel, terrain and moisture: under the current super–El Niño dry season, 45 percent of the roughly 80,000 hectare fuel body is receptive. How large a fire grows depends on how uniform the vegetation is: homogeneous fuel carries fire, a diverse mosaic breaks it. Where both conditions coincide — receptive now and structurally primed to grow large — the Cerros hold about 19,000 hectares, from which thirteen ranked patches concentrate the risk into treatable sites. High páramo is a separate case: it ignites often but burns small, so it belongs in monitoring and rapid response rather than being treated as either a large-fire target or as fire-free. Because these fires are human-ignited and give no seasonal moisture warning, the effective lever is spatial: treat the fuel where the two conditions meet.

~80,000 ha

connected fuel body in the Cerros Orientales

45% receptive

of the fuel body passes all necessary conditions under the current dry season

19,000 ha

act-now: receptive and structurally primed to grow large — the priority surface

40–340 m

the band out from the city edge where fuel uniformity most governs fire size; smoother fuel means larger fire, a genuine property of the vegetation

970 ha

high páramo that is receptive: frequent ignition, small fires, slow recovery — a monitoring class

13 sites

ranked intervention patches, each with local firebreaks at its fuel-connectivity necks

Two conditions, read together

Fire prevention needs two maps that are usually collapsed into one. The first is receptivity: given fuel, terrain and moisture, can a fire take hold at all? This is a necessary-conditions question, answered by the weakest of seven necessary conditions at each pixel — a location is only as receptive as its most limiting factor. Four conditions are static (fuel presence, fuel load, elevation, slope) and three are dynamic (precipitation anomaly, surface-moisture drought, and the combined humidity modifier). Under the current dry season the dynamic terms lift the receptive fraction to 45 percent of the connected fuel body, concentrated on the lower, drier, sun-exposed flanks. Elevation lowers receptivity in the high country but does not zero it: the fire record shows the high páramo does ignite, so it stays in the receptive class rather than being excluded outright.

The second map is size potential: once a fire starts, will it stay small or run? This is a question about fuel continuity, and it is invisible to the coarse products that answer the first. Both maps now exist at compatible resolution, and crossing them is what isolates the real target.

How large a fire grows: vegetation uniformity

The size a fire reaches is governed by how uniform the vegetation it burns is. Using satellite vegetation structure at 10 metre resolution, each location is described by a high-dimensional signature of what the vegetation is, and local heterogeneity is how varied those signatures are within a neighbourhood. Larger fires burn less varied vegetation. The mechanism is direct and visible in the fuel itself: uniform fuel carries fire, a diverse mosaic breaks it.

The 10 metre vegetation-structure field around real fires. The largest fires burned nearly uniform vegetation (solid blue interiors); small fires sit in varied mosaic (red mottling). Colour: blue = homogeneous fuel, red = diverse mosaic.

This is a real property of the vegetation, not a by-product of nearness to the city. At the built edge, a neighbourhood that mixes houses, roads and trees can read as varied even where the vegetation is uniform, and fires there stay small because they are caught and suppressed quickly; measured over vegetation alone, that city-edge distortion falls away and the relationship holds. The effect is moderate and spatially specific. It is strongest in a band running roughly 40 to 340 metres out from the built edge, far enough that a fire can spread but where the vegetation genuinely varies, and it fades both at the immediate fringe, where every fire is small regardless of fuel, and deep in the interior. The size-potential layer in this report is built on the vegetation-only measure, with the immediate urban fringe left blank because uniformity cannot be read reliably there.

Fuel uniformity is a genuine driver of fire size, distinct from the city-edge effect. Left: fire size rises with distance from the city. Middle: the uniformity reading is distorted only at the built fringe (grey), and measuring over vegetation alone (green) leaves the interior unchanged. Right: within equal-distance bands, more uniform fuel still carries a higher large-fire rate.

The priority surface: where both conditions meet

Overlaying the two maps separates the fuel body into operational classes. A location that is receptive and homogeneous is the priority: fire can start there now and would grow large. A homogeneous but not-yet-receptive location is a watch zone. A receptive but diverse location will burn small. High páramo is treated as its own class — receptive, but held out of the large-fire target because its fires stay small (see below). The cross yields about 19,000 hectares in the act-now class, a further 17,500 on watch, and roughly 970 hectares of receptive páramo to monitor. This is the surface the city should target, and it exists only because the two conditions were kept separate and then combined, rather than blended into a single index.

Left: receptivity — where fire can occur. Middle: size potential — how large it grows (homogeneous fuel dark). Right: the crossing. Dark red is the act-now surface, receptive and primed to grow large; light blue is watch; the darker blue marks receptive high páramo, a frequent-ignition but small-fire monitoring class.

Thirteen ranked intervention sites, placed where fires actually start

Where a fire would grow is only half the question; in an ignition-limited, human-dominated regime, where a fire starts matters just as much. Every one of the recorded fires sits on the western urban flank, a median of 150 metres from the built edge, so the intervention sites are ranked on the interface: ignition likelihood, from the density of historic fires, combined with size-potential and receptivity. The result is thirteen sites strung along the western wildland-urban interface, a median of about 440 metres from the built edge, just beyond the immediate fringe, in the middle band where an ignition is both likely and sitting on fuel whose structure genuinely governs how far it spreads. These are where fuel breaks, thinning, early-detection cameras and rapid-response staging return the most, because they sit close to the ignition source and on fuel that can carry a fire, rather than on distant fuel a fire would first have to reach.

The thirteen intervention sites (dark red, numbered) over the historic-fire density. Fires concentrate along the western wildland-urban interface; the sites sit on that interface, just beyond the immediate built edge, where an ignition is likely and the fuel can carry it.

Ranked sites (exportable). The full set of thirteen ignition-aware sites — rank, coordinates, primed area, elevation, receptivity, size-potential, ignition-density, distance to the built edge and land cover — is provided as an exportable GeoJSON (cerros_intervention_sites.geojson, WGS84) for direct use in QGIS, ArcGIS or web maps, and is also carried, with the land-cover and heterogeneity layers and the historic fires, in the Google Earth file (cerros_key_layers.kmz). The sites concentrate along the western city fringe. A note on coverage: the fire record is the Bomberos de Bogotá dataset, which covers the city's jurisdiction on the western flank; the eastern La Calera side is a separate municipality and largely outside that record, so its low fire count is partly a coverage gap, not proof it does not burn.

Each of the thirteen sites at close range, on strongly exaggerated relief over the vegetation-uniformity field (blue = homogeneous fuel, red = diverse mosaic; grey = built or non-vegetation, where uniformity is not read). Green segments mark local firebreaks at fuel-connectivity necks; the label under each panel is the land cover where the district survey maps it. The Google Earth file carries these at full zoom for field use.

What the homogeneous fuel is, and where páramo fits

The dangerous fuel has a name. Read against the district's own land-cover map (Cobertura de la Tierra 2018, IDECA and the Jardín Botánico, mapped at 1:5000), the most uniform vegetation in the reserve is Andean shrubland, then native secondary forest, then plantation; bare soil and roads are the most varied of all. The risk of a large fire is therefore homogeneous connected canopy wherever it occurs — shrubland, native forest and plantation alike — not one land-cover type or planted species. Where the survey assigns a class, 99.7 percent of the priority hectares fall on real combustible woody fuel and effectively none on bare ground, water or built surfaces.

Left: vegetation uniformity by land-cover class; green bars are combustible woody fuel. Shrubland and native forest are the most uniform fuels; bare soil and roads the least. Middle: priority sites over the mapped reserve, falling on real woody fuel. Right: of priority sites checked against land-cover, 99.7 percent is combustible woody fuel and effectively none is non-fuel.

Páramo needs its own reading, because it behaves differently from the woody fuels. It is uniform vegetation, so a purely structural view would flag it as large-fire ground; but the fire record contradicts that. Normalised by area, páramo has the second-highest ignition frequency of any cover in the reserve, about 5.9 fires per thousand hectares, just below plantation and above both shrubland and native forest — it plainly burns, and often. What it does not do is burn large: páramo fires run a median of about 2 hectares with a large-fire rate near 14 percent, because the wet, cold, low grass fuel does not sustain spread the way continuous canopy does. Native secondary forest is the opposite — it ignites least often but produces by far the biggest fires. The practical consequence is that páramo should be managed as a frequently-ignited, small-fire, slow-recovering cover: kept under monitoring and rapid-response, since even small burns set back páramo for decades, but not carrying the fuel-break and thinning effort that the large-fire act-now blocks demand.

Fire frequency (left, normalised by the area of each cover) and fire size (right) by real vegetation cover. Páramo ignites frequently but stays small; native forest ignites rarely but burns large. This is why páramo is treated as a monitoring class rather than a large-fire target or as fire-free.

Ground record and the absence of a seasonal warning

The Bomberos incident record grounds the satellite view. Ground and satellite agree the season runs December to March, but the satellite misses roughly three quarters of real fire-weeks — 96 percent of fires fall below the coarse-satellite ~21 hectare detection floor — which is why a fuel-and-conditions layer, not detection alone, is the right basis for prevention. Unlike moisture-limited regions where soil-moisture forecasts fire weeks ahead, Bogotá shows no such lead. This is an ignition-limited, human-dominated regime: the intervention is structural and rapid-response, not seasonal soil-moisture forecasting.

Ground record versus satellite. A: season agrees, fires peak in the dry-season wetness trough. B: no soil-moisture lead. C: 96 percent of fires fall below the coarse-satellite detection floor. D: fire events on shaded relief.

The absence of a predictive lead does not make moisture irrelevant — it sets today's receptivity even if it cannot forecast next month's. So it is worth tracking where the current season sits against history. On the ERA5-Land soil-moisture record, 2026 has already fallen to about the driest tenth of the 2015–2023 range for the date as the Cerros drop into the mid-year (July to September) dry trough, with the 2024 super–El Niño year showing how much deeper that trough can still go. This is a current-conditions monitor, not an early-warning signal: it says how primed the fuel is now, not when a fire will come, and it is consistent with the elevated receptive fraction.

Bogotá Cerros moisture monitor. A: ERA5-Land soil moisture (layer 1) for 2026 against the 2015–2023 P10–P90 envelope, the normal, and the 2024 El Niño extreme, dropping toward the early-September dry trough. B: MODIS NDWI surface wetness, noisier but with the same downward turn. C: for the last weeks, 2026 soil moisture sits in the driest tenth of the record for the date.

Recommendations

Target prevention on the act-now surface, not on receptivity or fuel uniformity alone: the thirteen ranked sites are where both the conditions to catch and the structure to spread are present, and the local firebreaks show where within each block to cut. Treat the watch surface as the pre-season work list, since it converts to act-now as the dry season deepens. Handle the high páramo separately: keep it under camera monitoring and rapid-response staging so its frequent small ignitions are caught early, given how slowly páramo recovers, rather than committing fuel-break and thinning resources there. Because the large-fire property is homogeneous connected canopy rather than a single cover type, the long-run lever is to break that continuity: interrupting the largest homogeneous blocks with diverse mosaic lowers the size-potential term directly and permanently, shrinking the priority surface itself. And because there is no seasonal moisture lead to lean on, situational awareness should rest on this fuel-and-conditions crossing together with the ground record and finer active-fire detection.

Notes and limits. The receptivity layer is built from the seven necessary conditions and calibrated so its connected fuel body (about 80,000 ha) and receptive fraction (45 percent) match the operational Capybara run; elevation lowers but no longer zeroes receptivity in the high country, so páramo remains in the receptive class. The size relationship is correlational but holds after controlling for slope, elevation, distance to infrastructure and retamo. The land-cover and fire-frequency figures cover the eastern reserve strip the survey mapped (about 43,500 ha); roughly three quarters of that is left unclassified in the source and is treated as potential fuel, not as non-fuel, and the páramo fire counts are small (single tens), so they are indicative rather than a full census. Satellite vegetation-structure data begin in 2017, so the size test is strongest for recent fires. All results are reproducible from the archived scripts and per-fire tables.

FAIR Digital Object — Capybara Center EWS
title: Fire risk in the Bogota Cerros Orientales
creator: Alejandro Feged-Rivadeneira, Tecnologico de Monterrey
date: 2026-07-22
sources: Capybara v6.0 receptivity (necessary conditions, Liebig minimum); AlphaEarth Satellite Embedding v1 (10 m vegetation structure); IDECA/JBB Cobertura de la Tierra 2018 (1:5000 land-cover); UAECOB Bomberos forest-fire record 2009-2025; ESA WorldCover; SRTM; MODIS/VIIRS active fire
methods: seven-condition receptivity (min-rule) calibrated to the operational fuel body and receptive fraction, with softened elevation term; vegetation-structure heterogeneity measured over vegetation only (built pixels excluded), with distance-to-built quintile stratification confirming the fuel-uniformity effect on fire size is independent of proximity to the city (p<0.001); multivariate large-fire size model controlling slope, elevation, distance-to-built and retamo; land-cover validation; paramo reclassified as frequent-ignition small-fire monitor class from the fire record; receptivity x size-potential crossing; patch-ranked intervention sites (exported as GeoJSON) with topological firebreaks
data-sha256: a20ec6cbdd7d114cdd7d078d2f279a621cd568a2605ba4bf259222b48fd48f50
license: CC-BY 4.0

Capybara Center situation report. Prepared for the Early Warning System project. Institutional affiliation: Tecnológico de Monterrey. Reproducible scripts and per-fire tables archived with the project.