HOUSTON – Almost any time of year our long range models, especially the American model, will forecast something a week or more out that raises alarms but is also suspect. For instance, a nor’easter, a super strong cold front, or a tropical system can show up on the model and that is exactly what happened last Friday. The morning run of the American model came up with this beauty one week out--a healthy tropical storm moving into Louisiana:
No other global model showed this -- not the European, German or Canadian. Also, the upper level wind forecast over the Gulf of Mexico is very unfavorable for tropical development right now (tropical systems like calm upper level conditions and the winds right now are strong!).
The lack of support from other models and the surrounding conditions are two big clues that this was just a modelcane: the model showing a hurricane that’s not likely. I first heard the term “modelcane” from Bill Read, former National Hurricane Center director. The American model is especially guilty of this and the Louisiana tropical system was completely gone the next morning although the model did forecast another gulf low 10-11 days out:
And this morning’s latest run? A lot of rain for us the next few days, but no tropical Gulf systems at all.
So What’s Going On?
The problem is what’s called “convective feedback” which is another way to say the American model gets too excited about storm development. You know that a spark can cause a fire (huge wildfires in California have started from the tiniest flames!). The American Model tends to see a “spark” or a cluster of showers over warm, tropical waters -- usually the Caribbean which is plenty warm right now. The model assumes that showers over warm water will continue to form and the more they form the more likely the pressures will fall and the more organized the storms will become. And the more organized they become, the more likely they’ll start to spin around a common center due to the natural spin of the Earth and, thus, a tropical low shows up in the forecast. This problem has been around for 20 years and I’m not sure why the upgrades to the American model cannot seem to work this one out, but I suspect it’s all about the initial conditions and initial data input. And that’s why we have to give the model a few runs to see if we’re looking at a modelcane or a serious threat. There isn’t a ton of data over ocean waters like there is land mass. The more data you have, the better your forecast.
The European model can also be wrong in its forecast (before you start jumping to conclusions), but seems to have a better data initialization then any other model out there. This is why it’s better to forecast by looking at all the models along with common sense synoptics (like water temperature, upper level winds, other systems that might hamper development, moisture content of the surrounding air, for example).
Speaking of the European, hang on to your umbrellas this week and watch for flooding. Here’s the numerical output through the week showing 5-7″ of rain across the area:
A wet week for sure and a good time to say “turn around, don’t drown!”