FireWatch Europe

How FireWatch Europe Works

Every hotspot on our map starts as a raw satellite measurement, not a confirmed wildfire. This page explains, in plain language, exactly how we turn that raw data into what you see — what we combine, what we infer, and where each signal's limits are. Nothing here should replace official alerts or local authorities.

The raw data: three satellites, not one

We combine near-real-time detections from three NASA/NOAA VIIRS instruments — Suomi-NPP, NOAA-20, and NOAA-21 — instead of relying on a single satellite. Each one flies in a sun-synchronous polar orbit and crosses a given European location only a handful of times per day, at roughly the same local time each pass. Combining three satellites shortens the gaps between opportunities to spot a fire, but gaps of several hours — especially overnight — remain normal, not a sign that anything is wrong.

When two or three satellites detect essentially the same fire within ~375 metres (the size of one VIIRS pixel) and 10 minutes of each other, we merge those readings into a single hotspot rather than showing duplicates. The reading with the highest Fire Radiative Power (FRP) becomes the one you see, and the popup notes when a hotspot was "confirmed by" more than one satellite.

Persistence: detected once, or repeatedly?

Separately from the multi-satellite merge above, we track whether a hotspot keeps reappearing at roughly the same location (within ~1.5 km) across the full 48-hour window. A single, isolated reading is a much weaker signal than one that keeps showing up on pass after pass — the "Detections" count and "active since" text in each popup reflect this.

"Possibly fading" — an inference, not a confirmation

A hotspot that was previously detected repeatedly but hasn't been seen again in over 24 hours gets a dimmed marker and a "possibly fading" note. We chose 24 hours deliberately: because the satellites pass at roughly the same local time each day, we measured that a gap of just 8 hours would already mislabel the vast majority of perfectly normal, still-active fires as "fading" — purely from ordinary day-to-day timing, not because anything had changed. Even at 24 hours, this remains a guess: the satellite may simply have been blocked by cloud cover or the fire's own smoke. We never state or imply that a fire is out.

Trend per hotspot

Where enough history exists (at least a few of the last 30 days), the popup shows a small line chart of how many detections a given ~1 km location has had per day. This uses the same location-history data described below, just visualised over time rather than as a single flag.

Possible industrial source: three signals, one soft flag

Satellites detect heat, not fire specifically — refineries, gas flares, power plants, and industrial sites can trigger the exact same kind of reading as a wildfire. We flag hotspots that are likely industrial, but — unlike some other wildfire maps — we never remove them from the map or the list. The flag is informational only; you can optionally hide flagged hotspots with a filter, but the underlying data is never deleted or hidden by default.

We combine three signals to decide this:

  1. FIRMS Static Thermal Anomalies — investigated, not used. NASA FIRMS publishes a reference layer of known static heat sources, but we confirmed it is only available inside the FIRMS Fire Map viewer itself (Advanced Mode) — not through the publicly documented, key-authenticated WMS/WFS map services or the CSV data feed that this site is built on. We looked; it simply isn't exposed that way, so this signal is not part of our classification.
  2. OpenStreetMap: is there a known industrial site nearby? We check whether a hotspot falls on or within about 500 metres of a mapped industrial area, landfill, power plant, or industrial works in OpenStreetMap. We check the actual shape of these areas, not just their centre point — a large industrial complex can easily be hundreds of metres across, and the wrong hotspot could otherwise be missed entirely near its edges.
  3. Our own history: does this exact spot keep lighting up? We keep a rolling 30-day record of which ~1 km grid cells have produced a detection, and on how many separate days. A location seen on more than 15 distinct days in the last month is flagged as recurring. We deliberately count distinct days, not raw detections: with three satellites making multiple passes a day, a genuine, ongoing multi-day wildfire could otherwise rack up more than 15 individual detections within its first few days — which would wrongly flag a real, active fire as industrial. A fixed source like a flare or plant, by contrast, tends to show up on most days of the month.

Under very large, continent-wide fire activity, the OpenStreetMap check can be skipped for a cycle if too many separate regions would need checking at once — the map itself is never delayed or blocked by this secondary check.

"New since your last visit"

The list highlights hotspots that weren't present the last time you opened this site, using a small note kept in your browser's local storage — nothing is sent to a server, no account or notification permission is required, and nothing is shared with us or anyone else. This comparison is made once per page load, so hotspots that appear later during a long browsing session won't retroactively get flagged until you reload.

What we deliberately don't do

We don't send push notifications about individual fires. Satellite detections can lag reality by hours, and a location-based alert risks implying a certainty — "there is a fire near you right now" — that this kind of data cannot honestly promise. We'd rather be clear about that limitation than build a feature that could mislead someone in an actual emergency.

Sources