Well the end of the first week was pretty tough. I think we almost have all the bugs out of the system, managed to pull off a live-blog on Thursday and now here we are with a weekly summary. We managed to learn a few things relating to our goals already, and some of them will not surprise you.

1. 00 and 12 UTC ensemble Convection allowing model guidance are equally skillful. I use that term loosely. In some areas, 12 is better and in others 00 UTC is better. A frustrating fact of forecasting on the mesoscale. Of course it would be super easy if the 12 UTC guidance was always better, so “It is what it is”. So we will evaluate this skill and see where that takes us.

2. Updating our forecasts for the 3 hour periods is frustrating. The guidance doesnt change all that much and we dont get wildly new observations. In fact, we arent getting the kind of observations we need. We always knew that, but it becomes painfully obvious at these time/space scales.

3. There is information in the models that we simply cannot (or have not*) extract in summary diagnostics. You simply have to look at it in all its glorious detail. And when you have an hour to make a robust probabilistic forecast, you cant look at everything!

*Sometimes you have to stare blindly at enough data that you can summarize it. That is why we call this research. We have to look before we can see.

4. We had quite a few days with messy mixed modes and ongoing convection. This will require better resolution, perhaps even better models, better data at the scale of the phenomenon we are trying to predict. Thankfully storms like supercells, MCS, and derechos are all relative large enough that we can address these in models are coarse as 1km. This doesnt make them reliable, just capable. we still need to input data that is of sufficient quality and resolution that we capture the environment and critically how that environment changes with the storms. Documenting this feedback process will actually be attempted in the upcoming field experiment called MPEX.

5. We have been using ensemble probabilities and ensemble maximum fields which seem to provide good information. But there is something aloof about this approach. We try to anchor ourselves with our experience, but sometimes we need to anchor to the most plausible model solution, or at least successfully merge our human intuition/experience with that of the models. This is an art form and the very act of trying to hold on to this perspective can yield valuable insight into the forecast. This is of course very difficult under the time pressures and amount of guidance to sift through.

Stay tuned as we enter week 2.