Survey of India states + LGD/Bharatmaps districts (DataMeet / ramSeraph mirror)fetched 9 Nov, 08:47 IST · fetch 44144
0.164
Short-wave infrared 1 darkened (dB11)
0.194
Short-wave infrared 2 darkened (dB12)
0.101
Burned-area index after (BAIS2)
0.93
Brightness after
0.32
Greenness before, after (NDVI)
0.21, 0.12
Burn ratio change (dNBR), NBR after
0.17, −0.19
The detection rule
Sentinel-2 L2A, cropland only: NIR and SWIR darkened since the previous clear pass (dB8A>0.07, dB11>0.07, dB12>0.03) and the field now looks charred (BAIS2>=0.78, brightness<=0.34); passes at most 6 days apart; clusters of at least 0.52 ha whose mean post-burn NBR is below 0.0 (drops flooded fields and cloud shadows). Candidates: tillage can look similar.
When a burn counts as unseen
unseen = no FIRMS detection within max(1 km, pixel half-diagonal + 200 m) of the scar between the clear pass before it and 3 h after the pass that shows it
Known look-alikes: cloud and smoke-plume shadows in partly cloudy passes and flooded fields with mean post NBR just below 0 still pass, often with high confidence; tillage of wet soil can mimic char; a hazy pre pass can exaggerate the darkening.
Weather between the passes
Open-Meteo best-match model, 48 of 48 hours
Afternoon wind, mean5 km/h
Afternoon mixing height, mean561 m
Rain in the window0.0 mm
Wind at 10 m, km/hMixing height, m
Arrows point where the wind blew. A low mixing height keeps smoke near the ground.
Sentinel-2 scenesS2C_T43RFQ_20261109T053257_L2A and S2A_T43RFQ_20261111T053939_L2A
Candidate unseen burns as a whole: audit precision 87% (95% CI 68%–96%, n = 23), checked by eye.
How often this rule is right: when people checked a random sample of candidate unseen burns by eye, the audit precision was 87% (95% CI 68%–96%, n = 23).
