It has been frequently mentioned that only have GFS/NAM/RAP for which to initialize top-down grids is a limiting factor in the utility of forecast builder. This is especially true in the long term, but also a limiting factor in the short/mid term. Furthermore, GFE has limitations in vertical resolution that can impact the quality of these grids even from the available models. I feel as though we could improve this input if we were able to focus on an element of pre-processing before getting into GFE. Our current method, in this way, is scientifically sound but the tools are ahead of their time in terms of the availability of the data to create a great output with minimal grid editing of these elements.
In other words, within D2D Volume Browser, I am able to load the MaxTw in a layer from at least 8 different useful model inputs (rap, hrrr, canadian, gfs, nam, arw, nmm, ecmwf) and I suspect that this data could be gathered from SREF/GEFS members as well. Is it possible to construct input grids for ProbIcePresent, MaxTwAloft, and ProbRefreezeSleet prior to bringing the information into GFE for FB use. The data are out there...it would be GREAT if we could take advantage for the betterment of our forecast.
Another pair thoughts related to the ProbIcePresent grid:
1) The predominance of the creation of "RA/ZR" as opposed to "L/ZL" when the ProbIce is below critical thresholds. A mention of freezing rain instills a reaction toward actual rain and ice accumulation while in (what I would venture to be) the majority of these events with no ice in the cloud the result will be drizzle. At the same time, we don't exactly want to be blasting drizzle and freezing drizzle in the forecast either. While the science is good, the inputs (and thus the outputs) aren't necessarily good. See 2 below for a specific example.
2) Instances when the GFS/NAM/RAP ProbIcePresent does not match the SuperBlend PoPs. For example, my initialized PoPs may be slow to fall below 15 as a system exits while at the same time the rap/nam/gfs blend moves/interpolates ProbIcePresent down below critical thresholds while PoPs remain >15. This results in that band of freezing rain on the back side of a system that only rarely occurs and almost never in the real-life version of the scenario I mention. Simply assigning 100 to ProbIcePresent is an easy fix but not a good scientific fix. Perhaps we could consider a tool (and I'm honestly not sure if this is a good idea) to match the critical thresholds of ProbIcePresent to the critical thresholds of PoP. This tool would not typically be applicable during a dry slot scenario, but would during the period of decreasing or negative vertical motion with an exiting storm system.
Finally, and this may be bigger picture, but I would like to propose some sort of "wintery mix" wording in the days 4-7 period. There have been instances of very specific wording regarding hourly Wx and p-type in the long term...and while there are targets of opportunity during which we should use more depth than rain/snow wording, perhaps "wintery mix" would be a suitable middle ground. Of course, making this a quality grid with minimal input would require addressing point 1 above...an enhancement of the top-down inputs during this time frame.
Sorry, this got pretty wordy...hard to put down a lot of thoughts at once! Thank you for your time and effort in gathering all of this feedback!
