Fire over the native dry forest of western Formosa
Satellite reconstruction of the wildfire event of 24 to 28 September 2026 in western Formosa: the magnitude of the fire, the land cover over which it occurred, and its territorial context. All measured with open data, reproducible pixel by pixel — and corroborated by three independent satellite systems.
GeoFuego Case · Copernicus — Territorial Intelligence · published 30/09/2026 · updated 04/10/2026 with the post-peak measurement (scene of 02/10) · Data: VIIRS/FIRMS (NASA), Sentinel-2 (ESA/Copernicus), ESA WorldCover 2021, Open-Meteo · Licence CC BY-NC 4.0
11,786
VIIRS heat spots in Formosa in September — the worst month of the year, 3.7× August
68%
of active hotspots across all of Argentina were in Formosa on 28/09 (83% of radiated energy)
44,359 ha
of native dry forest (monte) under the active-fire footprint (VIIRS) — 82% of the footprint
9,644 ha
of measured burned area (dNBR Sentinel-2, post-peak 02/10) in the central-eastern complexes — partial: the western cores still without a clean scene
1 · The event
Between 24 and 28 September, western Formosa recorded the highest fire activity of the year. Detections rose from normal dry-season levels to a pronounced peak within three days: 1,086 hotspots on 26/09 → 2,148 on 27/09 → 3,153 on 28/09, with a maximum radiated energy (FRP) of 68,209 MW. On 29/09 a cold front with rainfall (18 mm over the fire zone) halted activity: hotspots fell to 25 in a single day.
At the event's peak, more than two thirds of Argentina's active hotspots were recorded in the province of Formosa.
Left: daily escalation of hotspots in Formosa (bars) and their share of the national total (line) — reaching 68% on 28/09. Centre: land-cover composition under the fire footprint. Right: September (11,786 hotspots) dwarfs all other months of 2026.The same escalation, in the field: each orange point is a VIIRS active-fire detection (NASA FIRMS, 375 m) over native dry forest cover (green, ESA WorldCover). From 25 to 28 September detections rise from 378 to 3,153 in one day, concentrated over the dry Chaco of the central-western Formosa region. These are active-fire hotspots, not confirmed burned area (see methodology).
2 · The geographical scale
The fire was not a single blaze but a complex of 86 simultaneous active hotspots distributed across the dry Chaco of central-western Formosa. The 5 most intense complexes concentrated 72% of all the radiated energy. The epicentre of the event —confirmed by the official coverage— was Pozo de Maza (Ramón Lista department) and the settlements of La Madrugada, La Rinconada, Isla García, El Chañaral, and Los Cieneguitos.
Left: land cover (ESA WorldCover 2021) of western Formosa with the VIIRS fire footprint in red. The Bañado La Estrella stands out as the band of wetland (turquoise) crossing diagonally NW–SE; the agricultural frontier appears in magenta to the southwest. Right: fire falls mostly on native dry forest (monte) (red) rather than other cover types (yellow); circles mark the most intense complexes.
Cross-validation — three satellite systems agree
The most intense complex in our analysis (Bermejo nucleus, ~23.5°S 61.5°W, 48,222 MW of radiated energy) is independently flagged by Google's wildfire alert system, fed by its own satellite sources. This is not the reading of a single instrument: NASA VIIRS, ESA Sentinel-2, and Google all point to the same nucleus and agree on its location and severity.
Google Maps alert for the same event: "Serious fire — Bermejo, Formosa, Argentina. From satellite sources. Approximate burned area: 288 km²" over the La Rinconada / Vaca Perdida / Pozo de Maza area — exactly the complex our analysis identifies as the most intense in the province.
Third-party corroboration. Google, with its own satellite sources, independently flags the same epicentre (Bermejo) as a serious fire. The agreement is on the existence, location, and severity of the event and its main nucleus — not a comparison of figures: each system measures a different quantity (Google estimates burned area; our footprint marks VIIRS active fire), but all three sources describe the same phenomenon.
The epicentre, photographed during the fire
Sentinel-2 passed over the Bermejo nucleus on 25 September, with the fire in full escalation, and left a direct record: the same forest that five days earlier was intact, burning.
The fire front over the Bermejo River (10 km window, 23.46°S 61.59°W). Left: intact forest on 20/09, with the Bermejo River winding through. Right: five days later, a column of smoke rises from a combustion nucleus on the river bank and drifts kilometres southward over the settlements. True colour, same processing for both panels — the difference is the fire.The same moment, in shortwave infrared (Sentinel-2 B12/B8A/B04, false colour). The bright orange is thermal emission from active flames — the fire front seen through smoke, not a retouch. Around it, in dark reddish tones, what had already burned.
3 · Land cover under the fire: the native dry forest (monte)
Crossing the active-fire footprint with the reference land cover (ESA WorldCover, 10 m), the result is the central finding of this study. The hectares in the table represent the area of each cover type under the fire footprint, not confirmed burned area (see methodology):
Cover under the fire footprint
Ha under the footprint
% of footprint
% in the region
🌳 Native dry forest (monte)
44,359
82.4%
67.4%
🌾 Grassland
7,761
14.4%
24.0%
🌱 Cropland
633
1.2%
5.9%
💧 Herbaceous wetland
605
1.1%
1.3%
🏜️ Shrubland
323
0.6%
0.9%
🌊 Permanent water
169
0.3%
0.4%
Forest over-representation in the footprint. Native dry forest (monte) covers 67% of the region, yet concentrated 82% of the fire footprint: fire was recorded over forest at a proportion greater than its availability in the landscape. The active-fire footprint (VIIRS) is a proxy for the affected area, not identical to the final burned area, and the cover corresponds to ESA WorldCover 2021 — both caveats are detailed in the methodology section.
4 · The burn scar, up close
To verify that the satellite signature corresponds to actual burning and not noise, the dNBR (differenced normalised burn ratio, Sentinel-2 at 20 m) was computed over a 14×14 km window of the central nucleus, comparing a clean pre-fire image (15/09) with a post-fire image (27/09, 100% clear sky).
Left to right: true colour before (15/09), true colour after (27/09) where burn scars appear as dark/magenta patches, SWIR composite highlighting fresh burning, and the dNBR map quantifying severity. Burn scar detected in this single window: 924 ha verified.
Scope of that first measurement. The 27/09 image was prior to the 27–28/09 peak: the 924 ha were a validated proof of method, not the total. The post-peak measurement came with the 02/10 scene — that follows.
Update 04/10 — the post-peak measurement (scene of 02/10)
On 02/10 Sentinel-2 published the first clean pass after the peak (0% cloud cover over the central-eastern province). With it we closed the dNBR for the two central-eastern complexes:
Above: central-eastern complex (~24.4°S 59.8°W), 36×36 km window mosaicked over two tiles (pre 15/09 → post 02/10): 6,486 ha burned. Below: south-eastern complex (~24.8°S 59.5°W), 20×20 km (pre 17/09 → post 02/10): 3,158 ha. White gaps are residual clouds — declared, not filled.
Severity (USGS thresholds)
Central-east
South-east
Total
Low (0.10–0.27)
4,413 ha
1,837 ha
6,250 ha
Moderate (0.27–0.44)
901 ha
460 ha
1,361 ha
High (0.44–0.66)
978 ha
532 ha
1,511 ha
Severe (>0.66)
193 ha
329 ha
522 ha
Total burned area
6,486 ha
3,158 ha
9,644 ha
Crossing the dNBR burn mask (not the VIIRS footprint) with ESA WorldCover: 6,846 of the 9,644 ha burned were native dry forest (monte) (71%), of which 1,391 ha burned at high or severe intensity. The remainder was primarily grassland (~2,600 ha).
What is missing — declared, not estimated. The two most intense western cores (Bermejo / Pozo de Maza — the one flagged by Google — and the neighbouring western nucleus) are still without a usable post-fire scene: the 30/09 pass came out 100% cloudy and the next one arrives around 05/10. The 9,644 ha is the measured floor of the event, not the provincial total: if Google's satellite estimate for the Bermejo focus alone (288 km²) is in the right order of magnitude, the final total will be several times larger. This will be completed with the next clean pass.
5 · The forest before and after, in true colour
The figures give the scale; the optical images show the change on the ground. These are Sentinel-2 images at 10 metres resolution over three nuclei of native dry forest (monte) (confirmed as forest by WorldCover), comparing cover from 15/09 with that from 27/09.
Central-eastern Chaco forest (~24.5°S 59.8°W). Left: continuous, healthy forest (green) on 15/09. Centre: the same area on 27/09, with a brown-purple patch of burned vegetation. Right: the scar highlighted — 315 ha of burned native forest in an 8 km window.Central Chaco forest (~24.4°S 59.9°W). The patch of forest on the right side turns from green to dark brown between 15 and 27/09. Scar: 351 ha.North-west, Pozo del Tigre area (~24.8°S 60.0°W). Mosaic of native forest, wetland edge, and agricultural frontier. Scars (red) surround water bodies and advance over forest patches — ~1,020 ha in this window.
What these images show and what they do not. These three windows total ~1,700 ha visible as of 27/09 — only what had burned through the 26th, in three 8 km sections. They are the direct visual record of burning on native forest; the closed provincial figure will come with the full dNBR from the post-peak scene.
Update 04/10 — before/after with the post-peak scene
With the clean 02/10 pass, the before/after of the central-eastern complex was completed: the same frame, 17 days apart, with the fire in between.
Central-eastern complex (6 km window, 24.46°S 59.91°W). Left, 15/09: plots and Chaco forest green. Right, 02/10: the plots completely carbonised and the scar extending over the forest to the west — ~750 ha burned in this section alone, ~480 of them over forest.South-eastern complex (7 km window, 24.93°S 59.45°W). The same method applied to the second measured complex: pre from 17/09, post from 02/10 — part of the 3,158 ha quantified there by dNBR.
6 · The territory: land cover and protected areas
Bañado La Estrella — the wetland on the fire's edge
The fire footprint borders the Bañado La Estrella, the second-largest wetland in Argentina (~400,000 ha) and the third in South America. It is a Provincial Nature Reserve (Formosa Law 1471/2005), an Important Bird and Biodiversity Area, and one of the country's 7 Natural Wonders. It harbours more than 300 bird species — nearly half of those recorded in Argentina — along with yacarés, carpinchos, curiyú, aguará guazú, and giant anteater. It is the fishing source for the Wichí, Pilagá, and Qom communities.
Protected areas exposed
The Reserva Natural Formosa (Administración de Parques Nacionales (National Parks Administration), 9,005 ha, Bermejo department) is the most exposed national protected area: at the south-western edge, within range of the event, in the heart of the Dry Chaco. Parque Nacional Río Pilcomayo (RAMSAR site, 51,889 ha) was far away, to the north-east.
The affected ecosystem — Dry Chaco
The dominant cover under the fire footprint is the native forest of the Dry Chaco, one of the most extensive xeric forest masses in South America.
Quebracho coloradoQuebracho blancoPalo santoAlgarroboYaguaretéOso hormigueroTatú carretaPecarí quimilero — species of the Chaco native forest of the area, several declared Natural Monument.
7 · Population of the area
The affected western strip —Patiño, Ramón Lista, Matacos, and Bermejo departments— totals approximately 132,000 inhabitants (2022 Census), mostly dispersed rural population: criollo settlements and indigenous communities.
Department
County seat
Inhabitants (2022)
Patiño
Las Lomitas
81,819
Ramón Lista (epicentre: Pozo de Maza)
Gral. Mosconi
17,729
Matacos
Ing. Juárez
17,032
Bermejo
Laguna Yema
15,644
Western Formosa is a heartland of indigenous presence: Wichí (predominant in the burned zone), Pilagá, Nivaclé, and Qom peoples. Their livelihoods are historically tied to the native forest (hunting, gathering, honey, timber, chaguar fibre) and to the wetland (fishing).
Human toll of the event (official and journalistic sources). No fatalities reported. Around 30 people — adults and minors — received assistance for smoke inhalation. 3 families affected and at least 1 dwelling destroyed. The Pozo de Maza fire was declared contained on 29/09. The official satellite estimate for that focus was 96.55 km² (>9,000 ha) of hot zones.
8 · Methodology and data integrity
What we measured and how.
Fire activity: VIIRS detections (S-NPP, NOAA-20/21) from NASA's FIRMS system, ingested and aggregated by province in GeoFuego's own database. FRP = Fire Radiative Power, a proxy for intensity.
Fire footprint (~53,900 ha): 375 m grid (VIIRS pixel resolution) marked where there were ≥1 detections during the 24–28/09 peak. It is the active-fire footprint, a defensible proxy for the affected area — not identical to the final burned area.
Land cover: ESA WorldCover 2021 (10 m), resampled by majority class to the fire grid. Its date (2021) means that part of the 'forest' may have changed since then.
Burn scar (dNBR): Sentinel-2 at 20 m, NBR=(B8A−B12)/(B8A+B12), pre−post difference, USGS severity classes. Post-peak: scene of 02/10 (0% clouds over the central-east). Burned forest = dNBR mask > 0.10 × WorldCover. Accurate but limited to windows with a clear post-fire scene.
What we do NOT claim. The provincial total remains open: the western cores do not yet have a clear post-fire scene, and their hectares are neither estimated nor interpolated — they will be measured with the next clean pass (~05/10). The 9,644 ha measured is a floor, not a total. We always distinguish 'active-fire footprint' (VIIRS) from 'burned area' (dNBR) — they are not the same, and saying otherwise would inflate the figure.
A documented case by GeoFuego in collaboration with Copernicus — Territorial Intelligence. Open data, reproducible methodology; each satellite figure is derived from public data (NASA FIRMS, ESA Copernicus Sentinel-2, ESA WorldCover, Open-Meteo). Context: INDEC Censo 2022, Administración de Parques Nacionales (National Parks Administration), Ley 26.331 de Bosques Nativos (Native Forest Law). Licence CC BY-NC 4.0.
This study is a satellite analysis for dissemination and prevention — it is not an official alert and does not replace the Servicio Nacional de Manejo del Fuego (National Fire Management Service) or Defensa Civil (Civil Defence). In the event of an active fire, call 100 (Fire Brigade) or 911.
GEOFUEGO · satellite monitoring — a product of AgentNEO.