Friday afternoon

I’m not much of a fan of extended outlooks and/or long range forecasting.  Nevertheless, I do check a model that portends to go out weeks and months.  That model is known as CFSv2.  I like to see the slow, usually subtle, forecast changes the model picks up.  But today I noticed a big shock – a dramatic one week change in August’s outlook.

Below is the forecast from July 17.  The top left image is the raw computer output.  The upper right is a statistically “normalized” version of the raw data.  The bottom images are probabilities of occurrence.  Notice the eastern U.S.  Forecast for an above normal surface temperature (2 meters) is quite evident.

That was July 17.  Here’s the same image from today’s run…

An amazing flip-flop!  If correct, we’re looking at a very nice August.  But, if it flips once, it can flip again.  This is a perfect example why I don’t like forecasting beyond 3 days.