The inspiration for this blog came from a Facebook message from Alan Gurley: "Matt, I am not being critical, but why is this front (the one that will bring storms for the Egg Bowl) so hard to predict? Everybody has been changing their mind.
Good question, and valid point. The answer is 3-part.
1) Scrutiny. If I make a forecast for a random Wednesday in July, and that forecast changes a little, does anybody notice? Maybe if you have a big event planned for that day. If not, it might get lost in the shuffle. Since the Egg Bowl is a big deal for nearly every sports fan in the state, it is a forecast that is under intense scrutiny. Just imagine how many times you'd catch forecast changes if every drop/degree mattered every day (I've revealed the dirty little secret no meteorologist wants to admit).
2) Model Accuracy. This may sound like "passing the buck," but it cuts to the core of the issue. Forecast models are rarely accurate. When looking more than 4-5 days out, they are usually very bad. This makes it border-line remarkable that we've been SO CLOSE on the timing of storms (afternoon/evening) for a 3-4 hour event. Believe it or not, if I can nail a cold front and thunderstorm timing from 1 week out within 6 hours, I am super excited. That just doesn't happen often, no matter how good you are. Forecast models have increasingly poor resolution the farther in the future that they project. A week out, many forecast models only produce outputs for every 12 hour period, meaning anything in between 6pm and 6am is all lumped together. It's up to the meteorologist to decipher timing/type/severity within that timeframe. Also, the same "blob" of precipitation is valid from Tupelo to Natchez, so the resolution is like going from an HDTV to an animated flip-book.
3) Model Overload. Believe it or not, it's possible for a forecaster/meteorologist to look at too much data. There are only about 3 forecast models that go beyond 5-6 days out. Within that 4-5 day timeframe there are 10's of models, all depicting drastically different things. This can lead to an overload of conflicting information that can be hard to sort through. The number-1 rule for a forecaster to remember is that the model is ALWAYS WRONG. It's the first thing I teach an intern or a new forecaster. The model might be just a little wrong, or a lot wrong. When you get to looking at 10-20 outputs of wrong data, you can get overloaded. When this system even a week out, each of the models had a drastically different solution. It's just in the last 24 hours that the solutions have all converged on a single idea. Is more better? It can certainly be more frustrating.
All of that applies to forecasts of situations that are reasonable to grasp: severe storms, snow, wind, rain.... the normal stuff. That's a preface to say, when things get really messy, as could be the case the next few days, where we'll have a "cut-off low," all bets are off.
And what's my prediction for this low pressure system that could actually move backward?
Forecasters will get it somewhat wrong.
