Wildfire Watch · Fires, Wind and Risk on One Map

The base hotspot layer. Near real-time, one point per detection. Circle size is fire radiative power (FRP, in megawatts); colour is detection confidence.

Cross-checks Deepfire. A green ring means MTG saw the same fire; a dashed marker is a detection MTG made that Deepfire has not confirmed. Its 10-minute series is what makes the past steps of the timeline possible: they are estimates rebuilt from that series, not live readings.

Wind arrows, the fire-weather danger level (from wind, temperature and humidity), and the direction the forecast spread leans in.

A computed simulation, not a feed. The forecast outline is stretched along the wind; the dashed outline around it is the uncertainty. It is always labelled with the time of the model run.

Protected natural areas and population centres. They power the impact alerts and the nearest-town facts. Never real-time.

Left out on purpose. The public dataset is training imagery, not a live feed. It would add value only with a live camera network, which this page does not assume.