Fire emissions are estimated, not fixed
Smoke models need estimates of where fires are burning, how much material they emit, and how high the plume rises. Satellite fire detections and fire radiative power help build those estimates, but clouds, dense smoke, canopy, and observation timing can obscure activity. A new detection or revised estimate can substantially change the next run.
Operational acreage and containment are not direct smoke-emission measurements. A contained fire can still produce smoke, and a large incident can have variable activity. WFIGS reports provide context but do not set the model concentration by themselves.
Small weather changes move plume edges
Wind speed and direction vary across height, terrain, and time. Atmospheric stability and mixing determine whether smoke stays elevated or reaches the surface. Rain and deposition remove particles. New weather observations can shift those processes enough to move a narrow plume across a city.
The effect is most visible near sharp boundaries. Two successive model runs can agree on the regional event but disagree on which side of a boundary contains one grid cell. Use neighboring cells and several time steps to understand that uncertainty.
Fresh guidance should revise the plan
Smoke Window checks for the newest complete extended run when the page loads. It keeps a result provisional until all 17 three-hour frames are verified, then computes the answer from the chronological sequence. A refresh later may install a different run and choose the frame nearest the current time.
Treat forecast changes as part of responsible use. Compare the new run with current AirNow data and official local information. Confidence should come from converging evidence and stable trends, not from an old screenshot or an exact number.
A new satellite look can change emissions
Fire detections are not continuous perfect observations. Satellite overpasses, clouds, canopy, dense smoke, and sensor geometry affect what is seen. When a previously obscured active area appears, the emission estimate can rise in the next model cycle. When activity wanes or a detection is reassessed, it can fall. These revisions occur before atmospheric transport is considered, so the starting plume can change even under similar predicted winds.
Plume-rise assumptions change the route
Smoke injected high into the atmosphere encounters different winds and may travel farther before reaching the surface. Smoke kept near the ground follows another path and interacts more with terrain and mixing. Models estimate plume rise from fire energy and atmospheric stability, both uncertain quantities. A change in injection height can shift the downstream surface corridor even if the incident location and total emitted mass look similar.
Weather observations update the atmosphere
Each HRRR cycle assimilates newer information about wind, temperature, moisture, clouds, and pressure. Small adjustments compound through the forecast, especially along fronts, coastlines, and mountain terrain. A plume boundary may move tens of miles while the larger regional event remains recognizable. Compare runs at the same valid time to separate a true forecast revision from the ordinary movement seen between different forecast hours.
Why hour 48 moves more than hour 3
Near-term guidance begins close to the latest observed atmospheric state. By hour 48, errors in winds, mixing, fire behavior, and removal processes have had more time to grow. A later frame may still be useful for directional planning, but precise bands and boundaries deserve less confidence. If an important event is two days away, use the long lead as an early signal and plan to refresh as shorter-lead runs become available.
Read revisions without losing trust
A forecast update is evidence that the system incorporated newer inputs, not an attempt to hide uncertainty. Smoke Window displays the model run time and checks for the newest complete 48-hour cycle on page load. Keep the old and new questions distinct: what did the earlier run predict from its inputs, and what does the newer run predict now? Decisions should use the newest complete guidance alongside current AirNow and official local information.
Bottom line: why smoke forecasts change
Forecast change is expected when observations, fire estimates, or meteorology change. Judge an update by whether it clearly states its run and valid time and whether the new regional pattern is physically plausible—not by whether it preserves yesterday’s answer. A transparent system should make revisions visible. Reloading Smoke Window obtains the newest complete cycle, while the timeline and map retain discrete evidence users can compare rather than hiding uncertainty behind a fixed promise.
A careful reading of why smoke forecasts change
Smoke forecasts change because both fires and the atmosphere change—and because every new model run begins with newer observations and revised estimates. Keep that immediate answer attached to the valid time, selected geography, model run, and product limits described above. Those details are part of the result, not fine print, and they should travel with any shared interpretation of this guide.