PATTERNS ACROSS TIME & PLACE
A longer view of fire.
Explore archived clusters through their observations, persistence and geography. Choose a place and compare years — every chart is calculated here, from the archive index alone.
What the index can (and cannot) tell us ↓Loading archive…
ARCH/H3TI/ARCHIVE.h3ti.zst · no individual event files are downloaded01
Clusters through the years
Click a bar to select that year; Shift-click to add or remove it.
02
Hotspot observations by year
Whole clusters are assigned to their first or last observation year.
03
How are cluster sizes distributed?
Log–log probability density. Equal-ratio bins: [1, 2), [2, 4), …
04
How common are large clusters?
Fraction with at least N observations. Compare tails across years.
05
The seasonal rhythm
Cluster counts by month of the selected first/last observation basis.
06
From first observation to last
Elapsed observation span; this does not imply continuous burning.
07
A latitudinal profile
Clusters per 10° band, using representative location. Not area-normalized.
08
Where are the clusters?
Top 15 locations in this selection. Click a bar to filter the country.
READING THESE CHARTS
Observations, not a census of fire.
What counts as size?
Size is the number of hotspot observations in an indexed cluster, not burned area or unique fire detections. Repeated satellite passes and overlapping sensors can observe the same fire. A cluster is an algorithmic grouping, not necessarily one physical fire.
Comparing years
The index stores first and last times, not each observation’s time. All observations in a cluster are therefore attributed to its selected first/last year, even when it crosses a year boundary. Annual bars retain all available years for context; the highlighted years drive the other charts. Dates use UTC.
Distributions and large clusters
The density chart divides each bin’s count by the cohort’s cluster total and bin width, allowing comparison across sample sizes. The exceedance curve is the fraction of clusters with at least the indicated size. Both axes are logarithmic; empty bins are omitted. Curves with many distinct sizes are sampled for display. Use the legend to show or hide years.
Approximate geography
Location is the center of the cluster’s first H3 cell, not its full footprint. Countries use coarse Natural Earth 1:110m boundaries; small islands, coasts and borders can be unmatched or approximate. Latitude counts are not adjusted for land area.
Coverage matters
Only ARCHIVE is included; active LIVE clusters are excluded. The newest years may be incomplete. Archive thresholds, sensor coverage, ingestion history and clustering choices affect all counts. Trends and straight segments on log–log plots alone establish neither stationarity nor a power law.
Made with
NASA FIRMS and IPMA / LSA SAF fire data. Developments, algorithms, data structures, deployment and design: Jean-Baptiste Filippi — CNRS / University of Corsica — Firecaster.org. Link with FET and FRP analysis: Ronan Paugam. Charts: Apache ECharts. Geography: Natural Earth (public domain).