greece_fire
2024-09-21 → 2026-09-20 · 730 days
Detections
70,981
Peak day
2,005
Median day
45
Flagged
19 (2.6%)
iberia_fire
2024-09-21 → 2026-09-20 · 730 days
Detections
196,071
Peak day
13,329
Median day
99
Flagged
18 (2.5%)
Seasonal profile — greece_fire
Mean detections per day by month. Orange bars are months that produced at least one flagged day; the curve is what makes a winter anomaly legible.
Seasonal profile — iberia_fire
Mean detections per day by month. Orange bars are months that produced at least one flagged day; the curve is what makes a winter anomaly legible.
Do the context features separate anything?
Flagged days versus the rest. A detector that only separates magnitude is not separating anything interesting — these columns are the test.
| Feature | Flagged | Unflagged | Ratio |
|---|---|---|---|
| Mean detections | 2,281.2 | 132.8 | 17.17× |
| Mean FRP per detection | 27.6 | 7.5 | 3.68× |
| Night-time share | 50% | 60% | 0.82× |
| Cell concentration | 5% | 17% | 0.29× |
| Satellites seeing it | 4 | 3.11 | 1.29× |
Flagged days carry roughly 3.7× the intensity per detection and concentrate less — large events spread across many hexagons rather than one. FRP pools MODIS and VIIRS, whose scales differ ~6× for the same fires, so intensity features carry an instrument-mix confound.
Cryosphere — observed vs normal
The gap between the two series is the finding. Antarctic has run below the 1981–2010 normal for the entire window.
Solid: observed. Dashed: 1981–2010 normal. The vertical gap is the deficit.
Solid: observed. Dashed: 1981–2010 normal. The vertical gap is the deficit.
Strongest departures in the record
Ranked by robust z-score. The top row is a winter event, which is the point: with a seasonal baseline, an out-of-season day can outrank a megafire.
| Region | Date | Detections | z | Night | Cell conc. | FRP/det | Severity |
|---|---|---|---|---|---|---|---|
| iberia_fire | 2026-02-25 | 1,184 | 22.5 | 31% | 3% | 6.9 | extreme |
| iberia_fire | 2025-08-15 | 13,329 | 18.7 | 38% | 1% | 59.3 | extreme |
| iberia_fire | 2026-07-03 | 2,238 | 18 | 65% | 4% | 42.4 | extreme |
| iberia_fire | 2026-02-26 | 998 | 17.9 | 18% | 2% | 8.4 | extreme |
| greece_fire | 2025-08-13 | 1,980 | 17.5 | 64% | 3% | 21 | extreme |
| greece_fire | 2025-08-12 | 2,005 | 17.2 | 58% | 3% | 44.4 | extreme |
| iberia_fire | 2025-08-17 | 9,429 | 14.2 | 60% | 1% | 42.4 | extreme |
| iberia_fire | 2026-02-27 | 774 | 13.8 | 58% | 4% | 4.9 | extreme |
| iberia_fire | 2025-08-16 | 8,669 | 13 | 48% | 1% | 63.8 | extreme |
| greece_fire | 2024-09-30 | 808 | 12.5 | 54% | 10% | 33.4 | extreme |
| greece_fire | 2025-07-26 | 1,502 | 11.7 | 42% | 4% | 34.4 | high |
| iberia_fire | 2026-02-24 | 622 | 10.9 | 25% | 3% | 9.3 | high |
| iberia_fire | 2025-02-20 | 630 | 10.8 | 11% | 3% | 7.9 | high |
| iberia_fire | 2025-08-19 | 6,144 | 10.4 | 69% | 2% | 26.1 | high |
| greece_fire | 2025-07-27 | 1,251 | 9.8 | 69% | 5% | 16.9 | high |