Model Analysis Thread and Discussion

[baroclinic_instability]

Now that regions are broken up much more--I think it would be useful to have one place to come back to offer analysis and feedback of the numerical guidance and how well said guidance fared with particular events. It would not only allow meteorologists to offer feedback regarding the guidance--but it would offer meteorologists and non-meteorologists the opportunity to truly analyze why, from a meteorological perspective (using actual analysis--not just random comments or guesses), why a particular model suite performed poorly or well for an event. Moreover, talking with model developers such as dtk--this type of feedback is highly useful to them as well--and positive and/or negative feedback helps the developers key in on potential areas of concern regarding the guidance. I have discussed "model issues" extensively with dtk--especially significant performance issues during high impact events--and I think it has generally been useful discussion. If used properly--I would like to see this thread grow into a usable discussion for both meteorologists/forecasters, developers, and even the weather enthusiasts. This is obviously just an idea, and it is only going to be useful if everyone can keep the simplistic model bashing and negative comments to a minimum. Blind criticism with no verifiable evidence will help nobody.

An example of ways to get this started.

We discussed in central/western the ongoing storm event over the plains/OV now heading out east. Here were a few "large scale" synoptic observations and considerations I made.

Question relayed to me:

"Was banking on the nw trend, like we had with the feb 1st storm. Was that wave stronger coming out of the 4 corners? Or was there just a more impressive baroclinic zone with that one that was facing more NE and less zonal (like this storm)."

Me:

"A lot of reasons--but convection played a huge role--so it does seem it was not negative convective feedback that the NCEP (NAM/GFS) were hitting on. It was such a tough forecast for this reason. We knew convection would play a role in strong pressure falls--just how much that would "drive" storm track was unclear. It does seem convection is playing a large role with this storm heading a bit more E though."

The continued discussion:

SnowstormCanuck:

"That was a much more vigorous upper level system that was able to curl the sfc system NW as it intensified. H5 with this system is fairly meager, which has allowed that deep convection developing over the lower MS valley to dictate the direction of the sfc low. That's my layman's interpretation of what b_i was saying yesterday. Maybe he could illuminate you further."

Me:

"No illuminating needed. You pretty much said it. The upper level wave configuration was completely different and that system was a partial phase job with a deep elongated polar vortex intruding into the CONUS. That storm was essentially a deepening negative tilt longwave trough after phasing with a boatload more of potential energy and much deeper tropospheric deep baroclinity. This is mainly an upper tropospheric wave, positive tilt, with a much "shallower" overall depth that the Groundhog storm. The groundhog day storm had the mother of all coupled jet circulations too--historic storm for a reason. The potential energy with this storm--as impressive as it is--is probably an order of magnitude less than the GroundHogs Day beast."

Final post-mortem comments via SnowstormCanuck:

UKMET/SREFS/RUC = unmitigated fail with this storm.

EURO/GFS/GEM = adjusted their tracks accordingly in the few days leading up to the event. EURO not enough though. GFS/GEM probably the winners.

This is just a shortened example with only a select few comments from a brief discussion we had in the storm forum which included comments regarding the effects convection would potentially have on overall storm development and track--and how well said guidance was simulating it (amped Euro/UK, flat NCEP tracks, etc.). I hope we can get meaningful discussion in a central location regarding numerical guidance--I do believe it could be enlightening and a useful tool for everyone.

[baroclinic_instability]

Hopefully dtk does not mind me posting this back/forth regarding the Groundhog Day storm event--but this is another example of model analysis/discussion I hope to see more often in here since they are tools we use so often.

Me:

Personally I believe massive convective feedback. Has done it the last two days. It could not deepen this low and kept trending farther E and weaker when it was obvious what a powderkeg event this was. You could see massive VV's and a ton of "noise" in the vorticity fields. The NAM would not deepen the upper wave and amplify it due to all the convection in the southern plains and the surface low was weak too. Wind fields way too weak. Finally the NAM would deepen and close off late in the run--but after it was too late. GFS suffered from some of these effects on the runs just before the event as well.

Here is a quickie post I made. Hard to actually see it in the small loop though--but the feedback into the vorticity/height fields was horrific. It resultantly took a weak and flat track and really didn't occlude until late. If I have time I can compile some images/loops and a writeup. GFS suffered too. But the reality of the event was there would be a strong PV max ejecting over one of the beefiest low level baroclinic zones I have ever seen. A very moist theta-e regime in the low levels would result in very low static stability and or/static instability with an amazing east coast looking coupled jet pattern aloft. Convective driven pressure falls would aid in the tanking out of the low and occlusion. NAM/GFS later runs suffered big and could not deepen fast enough. GFS had a good run a couple days ago on its 0Z run but then backed off each day only to re-strengthen quite a bit the last two runs trying to play catch up.

http://www.americanw...post__p__402244

DTK:

Interesting and not at all surprising. I'm curious if you see similar issues with the HiRes Window runs (NMM especially, or ARW) since they don't use convective parm. schemes. One of the problems with running a non-hydrostatic model in the 5-15 km range is that it falls into a gray area as to whether or not to even use a convective parameterization.

I've always struggled with this notion of convective "feedback" as an "issue", because there are mechanisms in the real atmosphere to cascade energy upscale from the convective scales to synoptic scales. I guess the issue comes in when the convective schemes over-cook the latent heating in association with deep convective processes. Not to say it isn't an issue (this kind of thing is observed in the models all the time), I just think that people tend to over-use this as an excuse to throw out a model solution (though you obviously need to do so in the most egregious of situations).

I don't have a good answer other than to state the obvious (yes, even the most sophisticated convective schemes today have serious issues). I do not work on the actual models (and I certainly don't work on model parameterizations)...but I'm going to talk to one of the folks that work on the GFS or NAM the next time I run into them.

I've always struggled with this notion of convective "feedback" as an "issue", because there are mechanisms in the real atmosphere to cascade energy upscale from the convective scales to synoptic scales. I guess the issue comes in when the convective schemes over-cook the latent heating in association with deep convective processes. Not to say it isn't an issue (this kind of thing is observed in the models all the time), I just think that people tend to over-use this as an excuse to throw out a model solution (though you obviously need to do so in the most egregious of situations).

Me:

Yeah I agree--I rarely ever use that term since it is tossed around too much. To me it seemed to be crippling the dynamic fields of the NAM but perhaps it was latent heat release that was really resulting in a lack of deepening of the upper wave. NAM was by far always the last to deepen and close it off--and in this high positive feedback environment--being too slow to the party looks worse as the run progresses. The WRF-ARW/NMM in SREF was superb--and I was trying to tell everyone those would be the likely eventual result. RGEM worked great too. NAM was in big fail mode though--I will try and get back to you with more illustrations. Made a difference though as this storm is currently tanking and rapidly occluding with a much larger dryslot and a farther N progged defo band than progged by earlier NCEP runs--and the weak runs had horrible wind fields too. HRRR did well with this storm.

DTK:

Interesting comments.... and FYI anything you are able to put together can be directly passed onto the relevant developers. I'm not surprised that the models without convective parms. turned on (NMM/ARW in 4km windows) were able to do better in this type of situation. Any insight as to how the EC performed in this instance (it's been sub-par for our area for several cases, but overall I think it's been doing admirably elsewhere, as usual). I just wonder if it too struggled with the rapid deepening/convective component to the development (I've seen it WAY over-amp some EC storms and also a handful of tropical systems this past summer....though there issues seem to be more in the 3 day and beyond range and not < 48hrs lead time).

I know that some of my collaborators will be thrilled to hear about the HRRR success (and it's been noted for some of the NE storms as well). Has the actual / coarser rapid refresh had similar success (at least compared to the RUC). I know people tend to focus on the high-res stuff so much, but I'm curious about the RR since it's actually supposed to be replacing the RUC in a few months (and I have a close friend that will be interested to hear any negative/positive feedback on it).