9/17/26 Thursday 4 am
Models struggled in the evening runs (again). Overnight, the GFS corrected somewhat. I am still waiting for the latest European overnight run; the evening run is worthless.
I should explain this further. Most weather models have four run times per day. 12z, 18z, 00z, and 06z (if you are unfamiliar with Z time, that is 6 am, noon, 6 pm, and 12 am). That is for our time zone; you add or subtract if you are not in our time zone.
Historically, I have found the 12z and 00z runs to be more accurate. That wasn’t the case last night, when both models struggled. It is usually a timing issue due to model lag times.
By model lag time, I mean what happens in the real weather while the model generates its forecasts. So if the Euro starts processing data at midnight and there is a slight unexpected change in the atmosphere in Arizona 30 minutes later, the whole forecast for our area could end up completely wrong (bad data in, bad data out).
For the GFS (US bellwether model), the midnight run is released/available between 3:30 am and 5:15 am. The European model is higher resolution (longer compute times) and is released/available between 5:15 am and 6:15 am.
Both the European and GFS have AI versions with compute times less than half of the base models. I am not a fan of AI models (yet) because they use early AI technology that is very old (in the AI world). 30 minutes is a ridiculously long compute time for AI. If these were constantly upgraded, compute time would be instantaneous. I haven’t dug into the Google weather models yet, but I’ve heard good things.
I may revisit all of these topics from time to time, but it seemed appropriate to talk about them while waiting for the European.
5:45 am…
The Euro is out, and it initialized a little better than the last run. The precipitation forecasts from midnight to 6 am are just slightly higher than what has actually occurred.
Both models show light, steady rain, with intermittent, moderate to occasionally heavy rain continuing today and tonight. The models already start to diverge in 18 hours, with the GFS showing rain continuing for the next 36-48 hours. The Euro is similar, but it shows a break overnight; then it brings another wave through tomorrow morning.
I still don’t have a great handle on the weekend, but if we end up with as much rain as forecast, I think it is overoptimistic to expect immediate drying. At least for Saturday.
I will post again after the morning models generate.