A single global burned-area line can hide what is being counted. The source separates burned area across land-cover types, and those series should remain distinct when read. We compare 2002–2006 with 2021–2025 for the source-provided World entity, using a five-year mean for each category. The result is neither a total of all land nor a count of individual fires.
The later mean is lower in each selected category
Each selected land-cover series has a lower mean in 2021–2025 than in 2002–2006. Forest is one concrete example: mean annual burned area was 32.5 million hectares in the earlier window and 27.9 million in the later one. The other categories also declined in the reviewed extract, but the comparison does not establish why.
Separate series, separate comparisons
| Land-cover type | Earlier-window mean | Later-window mean | Difference |
|---|---|---|---|
| Forest | 32.5 million hectares | 27.9 million hectares | -4.7 million hectares |
| Savannas | 193.2 million hectares | 154.8 million hectares | -38.4 million hectares |
| Shrublands and grasslands | 183.4 million hectares | 144.4 million hectares | -39.0 million hectares |
| Croplands | 40.4 million hectares | 25.3 million hectares | -15.1 million hectares |
The rows answer a narrow question: how does each source-defined category compare with itself across the selected windows? They are not components that can safely be summed into a global total. Nor does hectares burned say how many fires occurred, how severe they were, how much carbon they emitted, or what damage people experienced.
How to read the period means
Each value is the arithmetic mean of annual source observations within the 2002–2006 or 2021–2025 window for the OWID_WRL World entity and one land-cover series. The displayed difference is the later mean minus the earlier mean. Our World in Data presents the Global Wildfire Information System data; the table makes a separate calculation for each category and does not calculate a global category total.
Why compare multi-year windows
Burned area can move sharply from one season to the next. Comparing multi-year means reduces the influence of any single unusually high or low annual observation, but it does not turn the result into a trend model. Every annual observation in a window receives equal weight. The comparison does not adjust for changes in climate, vegetation, land management, population, suppression capacity, or satellite performance. It is a descriptive summary of two selected parts of the source series and should be read alongside the full chart rather than as a forecast.
Land-cover categories describe the type of surface associated with the detected burned area. They do not classify the ignition source, the legal status of the event, or whether a burn was planned. A lower estimate in one category can coexist with severe fire seasons in particular places, and a worldwide mean can conceal large regional differences. The source chart is therefore useful for comparing its global categories over time, but not for evaluating a local hazard, estimating insured losses, or judging the ecological consequences of a specific event.
A source-version caveat for the newest observations
The reviewed Our World in Data extract identifies the Global Wildfire Information System as its source and contains observations through the end of the later window used here. The separate GWIS public technical page still describes an earlier coverage endpoint. Because those public pages are not synchronized, the newest observations should be understood as values in the reviewed OWID-distributed, GWIS-labelled extract—not as a release that Cite Supply independently matched to a current GWIS publication page. This distinction matters for provenance even though it does not change the arithmetic shown in the table.
Detection and interpretation limits
The source notes that satellite detection can miss burned area and fire occurrence, particularly in historical observations. Changes in observed area can reflect measurement and detection limits as well as real conditions. This comparison does not establish a universal decline in wildfire risk or harm and gives no explanation for changes in the estimates.
Source data are attributed to the Global Wildfire Information System and presented through Our World in Data. The figures and period means are a customised presentation; consult the linked licence and source chart for the data terms, attribution details, and context.