8/1/26 Saturday 4:30 am
I was not crazy about the overnight model runs. I am feeling like I might have set myself up for the forecaster’s curse after I wrote this yesterday:
“The model’s target is Thursday, with Wednesday being the pivot day between dry and wet. It definitely could change, but several days of model runs indicate this exact timing.”
Several days of model runs is very good run-to-run model consistency. That being said, the model went off script the last two model runs. Delaying the return of moisture to our area.
I also remember back in June, when we were so desperate to see the rain at the end of the tunnel. I wrote the same mantra that I have been saying for years. “Dry likes to stay dry, wet likes to stay wet.”
This is true in reality. But it is always more pronounced with the weather models. It is a well-documented feedback loop and one of the biases of the operational models. There is one problem. You don’t dry out soil moisture in two days.
This bias may be taking over already after two days. In other words, the model is still trying to resolve the effect of going from high-humidity days and abundant moisture to dry conditions so abruptly.
Let’s go back to the issue of how the bias works. When we talk about a feedback loop, we are referring to a model deficiency that extends the current (at that time) conditions.
This is difficult to explain, but let’s say you could hear the model processing data. It would say. “I see all of this dry weather across the western US, so I can’t possibly be seeing results that show a pattern change to wet; I need to ignore that for now and extend the hot, dry weather.” Then in the next run, the model would see that extended period of hot, dry weather (that the model created in the prior run), and come to the same conclusion. Each time extending the dry weather. The feedback (loop) of the hot and dry from the model locks in the next solution (forecast), which perpetuates the hot and dry forecast. The model (for now) does not say, “hey, I didn’t acknowledge the change back to wet in the last run, maybe I should consider that pattern change.” No, every six hours the model starts fresh, then it sees the hot, dry weather and forecasts more hot, dry weather.
I feel like I have done a horrible job explaining it, but that is the best explanation I can come up with this morning.
It occurs to me that this is a lot to soak up without falling asleep reading it. If I go on to my next topic, I will lose the reader’s attention. So I am going to stick with that and not go further down the rabbit hole. My goal was to demonstrate that we might not be getting accurate model information at the moment.
I will post again on Sunday, hopefully after a full night’s sleep!