What happened the past two days (especially Saturday) is a good example of a growing problem and a dirty little secret in the meteorological world. We often know more about upcoming weather patterns than we share with you in a simplistic “high-low-no rain/rain” forecast.
The slide above is from the Dr. Cliff Mass presentation to our local AMS chapter just 2 weeks ago. I hadn’t really given it much thought until he made it the cornerstone of a push towards using more “uncertainty language” in a forecast.
1. Regional models were clearly showing the possibility for afternoon showers or thunderstorms somewhere in the western valleys of northern Oregon or SW Washington both Friday and Saturday afternoons. And the Portland/Vancouver Metro Area seemed to be a good place for that to occur.
Here are the 12z RPM and WRF-GFS models from Friday morning, showing later Friday afternoon/evening…looks wet to me on both, although the placement was off, especially on the WRF-GFS (2nd image). We actually did okay with Friday here at FOX-12, mentioning the chance for afternoon rain or thunder.
Now on Saturday morning, our 12z RPM showed almost exactly what occurred just 14 hours later; although it was a few hours earlier than reality in the end:
The WRF-GFS wasn’t quite as good, but it was hinting at some sort of action coming off the Coast Range (not correct):
2. Yet all the forecasts (including mine) downplayed the threat enough Saturday that the thunderstorms were probably a surprise for most of the public. We didn’t make it abundantly clear that an evening thunderstorm in the city was quite possible. Yes, a relatively small chance, but the chance was there. On Friday night at 8/10/11pm I showed our RPM and mentioned that I thought thunderstorms would probably stay on the far east side of the metro area, close to the Cascades. I thought there would be a slightly shift to more westerly upper-level flow later in the day keeping the developing storms out there. Clearly location was wrong on that forecast. But when I saw the skies darken to my northwest and heard a few rumbles, it definitely wasn’t a surprise. It was more like “I wish I would have included a chance for the entire metro area.”
The amount of weather information we receive nowadays is incredible, but breaking that down into a forecast that goes just a few minutes on TV (and for a whole state) is a challenge. Maybe more important, forcing that info into a pretty little cloud/sun/rain graphic is even tricker. We are sure able to bring uncertainty information verbally into a weathercast and through text on a website. We need to work harder getting that information out.
Now Dr. Mass also suggested we not show a 7 Day forecast like this:
His point being that we can’t REALLY nail a forecast within a degree or two (regularly) more than a couple days out, so we should be giving temperature ranges instead of specific numbers. But this ship sailed a long time ago…television 7 day forecasts have specific numbers and that won’t change.
These 7 Day forecasts also don’t give the public any idea about our confidence in certain patterns either. For example, models are showing some variance (right now) on Thursday and Friday. Some are pushing the rain in a bit quicker on Thursday, but others waiting until late Friday. So I’m quite confident on the Monday-Wednesday forecast, and pretty confident on the general pattern change to wet and cool next weekend. But in between is a period with significant uncertainty. You sure don’t see it on the graphic; and there are plenty of numbers, text, and logos on the graphic already. So where would you show lower confidence on this graphic? There is no easy way, for now it has to be verbal on television or webcasts.
Weather forecasts are getting better and better, but we have work to do when conveying that increased forecast accuracy don’t we?
Chief Meteorologist Mark Nelsen