BlindSpot looks between the passes

Fields are lit after the fire satellites pass over, so their fire counts miss most of the burning.

72%of the burnt area Sentinel-2 has mapped since 2 Oct had no fire alert: 3,85,016 ha in 2,30,478 burn scars, the candidate unseen burns.

Audited precision 87% (95% CI 68%–96%, n = 23). Some candidates are tilled or flooded fields, so read the share as an upper bound.

Now 19:13 IST. The hand is in the blind span. No fire satellite over Sangrur now. Next: Terra at 20:49 IST.

Gold: when a polar fire satellite (VIIRS, MODIS) reported a fire in Punjab or Haryana in the last 3 days. The bite: 16:30 to 20:20 IST, 3 h 50 min, no fire alert from any of them.

Red: fire activity by hour from GK-2A, a geostationary satellite that scans every ten minutes but is too coarse to count fires (an activity index), mean of the last 3 days. 79% of it fell outside the gold, 66% inside the bite.

Sentinel-2 photograph of fields near Raipur Peerbakhshwala, Kapurthala district, Punjab, on 7 Nov 2026, with the burn scars mapped so far outlined.

By 7 Nov: 323 burn scars touching this frame (1,415 ha). 45 had a fire alert (white); 278 had none, candidate unseen burns (gold).

Near Raipur Peerbakhshwala, Kapurthala district, Punjab. Sentinel-2, true colour. Outlines: burn scars mapped since 8 Oct.

Farmers raking paddy stubble in a hazy field in Punjab. — Raking paddy stubble before a burn, Punjab. Photograph: CIAT.

79% of the fire activity GK-2A saw in the last 3 days fell outside the hours in which a polar satellite reported a fire.

GK-2A is a geostationary satellite that scans Punjab every ten minutes; the activity it sees peaks at 17:00–18:00 IST. Its figure is an activity index, not a count of fires. The polar satellites’ count is NASA FIRMS: 173 fire detections in the 24 hours to 19:13 IST in Punjab, Haryana, Delhi and Chandigarh, the newest seen at 14:17 IST (FIRMS adds each pass an hour or more after it).

Today’s fire-satellite passes over Sangrur, 13 Nov, IST, the list the clock’s live line counts down; paler rows are still to come
In view	Satellite	Sensor	Minutes
01:18–01:21	NOAA-20	VIIRS	3
01:59–02:07	NOAA-21	VIIRS	8
02:36–02:44	Suomi NPP	VIIRS	7
02:57–03:03	NOAA-20	VIIRS	6
04:52–04:57	Aqua	MODIS	5
09:52–09:56	Terra	MODIS	4
12:36–12:38	NOAA-20	VIIRS	3
13:16–13:24	NOAA-21	VIIRS	7
13:53–14:01	Suomi NPP	VIIRS	7
14:14–14:20	NOAA-20	VIIRS	6
15:50–15:55	Aqua	MODIS	6
20:49–20:55	Terra	MODIS	6

The plains of Punjab and Haryana from space on 13 Nov 2026, VIIRS on NOAA-20, true colour. — The Punjab plains on 13 Nov, VIIRS on NOAA-20, true colour. NASA GIBS.

The scars keep the record

A burnt field stays black on Sentinel-2 for days, whatever hour it was lit. We match every scar against every fire alert.

The slips below are the largest of the 2,30,478 candidate unseen burns in 46 districts, mapped up to the pass of 12 Nov, one per village. Each opens its case file.

Tell me about burn scar case 379724 near Sikanderpur, Jalandhar district: did any fire satellite see this field burn, and where did its smoke go?

Burn scar case 379724 near Sikanderpur in Jalandhar district was a candidate unseen burn not detected by any fire satellite, and its specific historical smoke trajectory is not tracked.
Zero fire satellites detected this burn, classifying it as a candidate unseen burn,.

7 of 7 figures checked against the data. BlindSpot’s investigator, Gemini 3.5 Flash-Lite, answered recently.

Harvested, not yet burnt

A tractor with a seeder sowing wheat into rice stubble in Punjab, seen through leaves. — Sown without burning: a seeder in rice stubble, Punjab. Photograph: CIAT.

A relative ranking from 0 to 100, not a probability: it orders today's fields by how likely they are to burn next. On this season's burns, the part of that ordering the backtest could test (neighbourhood and history; timing and rain cannot be reconstructed for past passes) is modest (AUC 0.640 over 387 pass pairs, 95% CI 0.595-0.693; 0.5 is chance). Timing is untested; today's full score follows the tested ordering at Spearman 0.67, and 47% of its top 1% are also in the tested top 1%.

From the ranking’s method, as the API states it.

Tested on the burns that followed: AUC 0.64 over 387 pairs of passes (95% CI 0.60–0.69; 0.5 would be chance), 0.62 on the latest pair alone.

Villages in the ranking’s order, first at the top, ranked 13 Nov, 17:23 IST from 2,61,123 harvested fields.
Rank	District	Village	Cut, not burnt	Fields
1	Barnala	near Karamgarh	475 ha	191
2	Bathinda	near Gumti Kalan	474 ha	271
3	Bhiwani	near Siwana	535 ha	83
4	Barnala	near Harbanspura	417 ha	158
5	Bathinda	near Selbrah	453 ha	216
6	Barnala	Mehta	423 ha	197
7	Ludhiana	Chakkar	370 ha	185
8	Ludhiana	Lakha	377 ha	168
9	Sangrur	near Chamar Tibbi	371 ha	192
10	Sangrur	near Namol	347 ha	196
11	Jind	Kalwan	347 ha	94
12	Barnala	near Khudi Khurd	340 ha	207

A seeder can sow wheat into standing stubble. We rank the fields that are cut but not burnt, so the machines reach them before the match does.

One day over Punjab

Of the 5,813 fire detections of the last 7 days, 98% were made on daytime passes.

The afternoon pass
2 fire satellites cross Punjab between 12:28 and 14:36 IST and count what is burning.
Punjab at the afternoon pass, 2 Nov. VIIRS on Suomi NPP, true colour with its fire detections in red, NASA GIBS.
The evening peak
GK-2A’s ten-minute scans put the peak of fire activity at 17:00–18:00 IST. In the last 3 days no polar satellite reported a fire in Punjab or Haryana from 16:30 to 20:20 IST, the span that held 66% of that activity.
Rice straw burning, Punjab. Photograph: CIAT.
The night drift
Smoke rides the evening wind towards Delhi. Our particle model follows it hour by hour on two weather models and shows where they disagree.
The plains from Punjab to Delhi, 13 Nov. VIIRS on NOAA-20, true colour, NASA GIBS.
Delhi under a lid
At 18:30 IST the mixing layer over Delhi was 28 m deep: a lid that keeps what arrives near the ground.
Smog at New Delhi railway station, a winter morning. Photograph: Sumita Roy Dutta.
The night pass
Between 01:14 and 03:18 IST the satellites pass again and see embers, not the fields already burnt.
Punjab at night, 12 Nov. VIIRS Day/Night Band on NOAA-20, NASA GIBS.
Days later
Sentinel-2 photographs the fields again. The scar is there, alert or no alert.
Case 2405504, 12 Nov, short-wave infrared. Sentinel-2.
The case file
Each candidate unseen burn becomes a record: where, how big, between which two passes.
Case 2316788, 12 Nov, true colour. Sentinel-2.
Help, not blame
Fields cut but not burnt rise to the top of the list, so a seeder can reach them first.
A zero-till seeder, Punjab. Photograph: FatherofGod.

Where it comes from

Fires and haze over Punjab and Haryana on 13 Nov 2026, MODIS on Terra, true colour, with VIIRS fire detections in red. — Punjab and Haryana, 13 Nov: MODIS on Terra, true colour, with VIIRS fire detections in red. NASA GIBS.

Fires: NASA FIRMS active fires CREAMS, ICAR-IARI: official residue-burning events KMA GK-2A AMI L1B full disk, via NOAA Open Data on AWS
Orbits: CelesTrak GP orbital elements
Pictures of the ground: Microsoft Planetary Computer STAC: Sentinel-2 L2A Element 84 Earth Search STAC Sentinel-2 L2A pixels Planetary Computer data/tiler API ESA WorldCover Sentinel-2 L2A COGs on AWS
Air, measured: OpenAQ v3 data.gov.in CPCB real-time AQI
Air, models: Copernicus CAMS via Open-Meteo air quality IITM Decision Support System for Delhi
Weather: Open-Meteo forecast
Places and boundaries: Survey of India states + LGD/Bharatmaps districts OpenStreetMap places via the Overpass API Open-Meteo geocoding Local Government Directory names and codes Census 2011 village points GeoNames cities15000 Wikidata Query Service
The investigator: Google Gemini API

21 of 22 sources are on schedule. Running late: IITM Decision Support System for Delhi.

Delhi’s PM2.5 now: 180 µg/m³, the median of 38 stations.

Air crossed the burn belt: traced back hour by hour, the air reaching Delhi spent 72% of its last day over Punjab and Haryana’s burning belt (every model run carried the air over the burn belt and past fires as they burned). That is where the air has been, not a measure of how much of its PM2.5 is smoke; local sources weigh too (see the method).

Ask why

Ask whether today’s smoke came from Punjab. The investigator checks the wind, the fires and the scars on the live map, and says what the data cannot settle.

Open mission control

BlindSpot looks between the passes

When the satellites look, and when the fields burn

The scars keep the record

Harvested, not yet burnt

One day over Punjab

The afternoon pass

The evening peak

The night drift

Delhi under a lid

The night pass

Days later

The case file

Help, not blame

Where it comes from

Ask why

One day over Punjab

The afternoon pass

The evening peak

The night drift

Delhi under a lid

The night pass

Days later

The case file

Help, not blame