Wildfire data · working analysis
In North America and Australia, forest burn is climbing steeply — up 58% and 48% decade-on-decade — while the headline "total area burned" figure barely moves or falls outright. Europe is the exception. The gap between those two numbers is the story.
If you ask "are fires getting worse?" and answer with total hectares burned, you get a reassuring answer in most of the world. But that total is dominated by grass, shrub and savanna, which burn on a completely different logic — and on a completely different scale — than forest does. Separate the two and the reassurance mostly evaporates. Not everywhere, though: Europe is a genuine counter-example, and it's worth understanding why before building a story on this.
The estimate is the change in average annual forest burn between 2002–11 and 2015–24. The line shows how much that estimate moves when each decade is repeatedly resampled. It makes the signal, and the uncertainty created by extreme seasons, visible together.
Forest burned area, later decade versus earlier decade
Point = change in the two decade averages. Thin line = 95% bootstrap interval from the annual observations. This is descriptive uncertainty, not a climate-attribution test.
Canada is the clearest increase in this record. Europe is a real counter-signal. Australia and Greece rise, but their intervals show how strongly the result depends on a small number of extreme seasons.
Read forest and total burned area against each other. Australia's total fell 21% while its forest burn rose 48%. North America's total rose 14%, while forest burn rose 58%. Europe is the exception: both measures fell, while forest's share of the burn was broadly unchanged (6.7% → 7.1%).
Both lines are indexed so that each region's own 2002–2011 average = 100. That puts total and forest on one scale, so the gap between the lines is the whole point.
Total burned area vs. forest burned area, indexed to each region's 2002–2011 average
Index = 100 means "same as that region's first-decade average". The solid line is forest; the dashed line is all land cover combined.
The thick horizontal bars are each decade's average — the gap between the blue bar and the 100 line is the decade shift, underneath the year-to-year noise. Y-axis is clipped at 400: Australia's 2019 forest index reaches 733 (marked ▲).
The indexed view shows direction. This shows magnitude — and it is dominated by a handful of catastrophic years.
Annual forest area burned, by country
Million hectares of forest land cover burned per year, 2002–2024. Portugal is included at true scale — which is itself the point about how the Iberian story differs in magnitude from the boreal one.
2023 in Canada alone accounts for 30% of the country's entire 23-year forest burn. Australia's 2019–20 Black Summer accounts for 51% of its top-three total.
Here is the composition of what actually burns. Forest is a thin slice of the total — so the total tracks savanna and grassland, not forest.
What burns: annual burned area by land cover type
Each region's average annual burned area, 2015–2024, split by land cover. All three bars share one scale, so lengths are directly comparable.
Forest is 15.4% of North America's burned area, 7.1% of Europe's, and just 1.7% of Australia's — where shrub and grassland alone are 87%, and grass, shrub and savanna together 97%. A wet-then-dry cycle in the Australian rangelands moves the national total by more than any forest fire can, which is why the headline number tells you almost nothing about forests.
This record cannot show a smooth annual trend. It can show whether recent seasons more often clear each country's own high historical threshold, and how much total burn is concentrated in a few catastrophic years.
How often did forest burn exceed the earlier decade's high threshold?
An extreme season is one above the 90th percentile of that country's 2002–11 forest-burn record. Each dot is one season, so this is a frequency count, not an area total.
The threshold is deliberately country-specific. It asks whether high seasons became more frequent within each country, not whether Portugal and Canada have the same-sized fires.
Share of 23 years of forest burn concentrated in the three worst years
Of all forest area burned 2002–2024, the percentage that came from just three fire seasons.
This is the severity half of the story. In Greece, three summers produced three-quarters of two decades of forest fire.
Everything above stops at 2024 because that's where the MODIS land-cover dataset ends. But 2025 is now a complete season, and it broke the "Europe is the exception" pattern in Iberia hard enough that it's worth checking before anyone repeats that finding. Retrieved 17 July 2026, directly from EFFIS and CWFIS — a different data source from the chart above, on purpose (see the note below the chart).
Annual burned area, Iberia and Greece: EFFIS fire perimeters, 2006–2026
Hectares burned in fires of roughly 30 ha or larger, as mapped by EFFIS. 2026 (rightmost, hatched) is a partial season — it will keep growing through the autumn.
Spain's 2025 (393,079 ha) is its highest year in this 2006–2026 record — ahead of 2022. Portugal's 2025 (278,917 ha) is its second-highest, behind only 2017. Greece, by contrast, is having a quiet year so far. The 9–10 July Almería fire (Los Gallardos, ~3,200–3,800 ha, 13 dead — the third-deadliest wildfire in Spanish history) sits inside Spain's still-climbing 2026 bar.
Canada, year-to-date area burned vs. historical averages
Cumulative area burned by 17 July each year — so every bar is a fair, same-date comparison, not a full-season total.
2026 is already 41% above the 20-year same-date average and 15% above the 10-year average. It's still below the 5-year average only because the 5-year window contains 2023, the worst season on record — a reminder that one extreme year can distort a short average more than it clarifies anything.
France, quietly, already had a bigger July than all of 2025. By 17 July, France's 2026 season (41,650 ha) had already surpassed its entire 2025 total (36,951 ha) — with the rest of the summer still ahead. Not dramatic yet, but worth watching alongside Spain and Portugal.
Methodology note: EFFIS maps individual fire perimeters ≥ ~30 ha from a mix of satellite and agency data — a different method from the MODIS land-cover pixel classification used in sections 1–4 above. The two series use different units, different fire-size cutoffs, and don't include a land-cover split, so they are never plotted on the same axis and shouldn't be quoted against each other. 2026 totals are running counts as of the retrieval date and will rise before the season ends.
Three angles that the data actually supports, in rough order of strength:
"Wildfires are declining globally" is a real, citable fact — global burned area is down 18% decade-on-decade — and it gets used in bad faith fairly often, because it is true of the total and says nothing about the thing people actually mean. Africa's savanna converting to farmland drives most of the global decline. Meanwhile boreal forest fire is escalating hard. Both facts are true at once, which makes this a story about a measurement rather than a debunk of anyone.
Be careful how far you push it, though: this dataset does not show a uniform global forest-fire surge. It shows a strong boreal signal (Canada +326%), a clear Australian and US signal (+48%, +59%), and a European forest trend that is actually negative (−28%). Greece is the one European country here with a sharp forest increase (+69%); Spain is flat (+3%). A piece that claims "forests everywhere are burning more" would be overreaching.
Canada's forest burn is up 326% decade-on-decade, and one year — 2023 — carries 30% of the entire 23-year record. It is the clearest single-country escalation signal in this dataset, and it is recent enough that people remember the smoke reaching New York.
Portugal's total burned area is down 19% decade-on-decade, and its forest burn is up only 17% — a modest number next to Canada's. But 59% of Portugal's 23-year forest burn came from three years (2003, 2005, 2017). The Portuguese story isn't "more fire", it's "fire that, when it goes, goes catastrophically" — which points at fuel accumulation, rural abandonment and extreme fire weather rather than at more ignitions. That's a different, harder, better story than the annual-total chart would suggest, and it needs ICNF data to tell properly.