I’m still struck by how poor the day-ahead smoke forecast is for my area. Via Airnow.gov. Separately, I remain surprised at how quickly the actual AQI can change.
So this is, in effect, an extension of Post #1803. I still have no clue why there is such a high variance.
With this last batch of bad air, in the past few days, I felt that I got less than a day’s warning. At one point, the forecast was for a next-day AQI of about 65, while the actual AQI the next day went above 100.
This is not a knock on the NWS, or NOAA. This is an observation on the level of accuracy for the AQI — and maybe the inherent variability of the AQI. Otherwise, NWS forecasting for the DC area seems excellent.
Sometimes you don’t know what you don’t know.
In this case, I don’t even know what question to ask.
By and large, what I’m talking about is small-area forecasts of, the ground-level impact of, the 2023 Canadian forest fire smoke plumes, in the Washington DC area.
Why are those numbers so all-over-the-place?
Is it as mundane as lack of data points? I noted a couple of oddities. The U.S. Airnow map doesn’t have a whole lot of data points (circles) on it. In addition, different websites providing a current AQI reading for Vienna, VA seem to rely on different AQI monitoring stations. Just how thin on the ground are the Federally-referenced AQI monitoring stations? Are they thin enough on the ground to limit the sort of particulate-flow forecasting that they can do?
Is it something as exotic as the butterfly effect? The forest fires are, in some sense, point sources. They are 500 miles from here, and up. Maybe a little error in measuring that parcel of air leads to big errors in just a few days.
Is it something inherent in the complexity of atmospheric mixing? Maybe the forecasters know where it will be (latitude and longitude), they just aren’t sure how much of it’ll be passing by at ground level.
Is it possible that our weather forecasting models do not, themselves, track individual parcels of air? They divide the atmosphere into a grid, then use physical laws to calculate changes over time in the weather, at the points on that grid. Do they implicitly track many individual parcels of air, as they do that?
Maybe it was always this hard to track wind speed and direction, I just never noticed?
Whatever it is — whether that’s real volatility, or whether that’s due to failure-to-predict — things change on a dime. Compared to what I expect.
At root, my problem is that I assumed the AQI readings would be as forecast-able as the rest of the weather. That’s not so. At least, not around here. I should plan accordingly.
