Verification of forecast snow total composite maps

[Andrew Pritchard]

With all the buzz regarding each model output of forecast snow totals in advance of the blizzard early this week, I decided to save a few of the maps and compare them with reality after the storm had passed. I overlaid the actual totals and added a little insight to the inaccuracies. Any comments on glaring errors, and any factors I didn't touch are welcome. The write up focuses on northeast Missouri into the Chicago region.

http://prairiestorm.blogspot.com/2011/02/verification-of-composite-snow-forecast.html

 

[Robert Dewey]

I have to say, the HRRR model pretty much nailed the event once it was within the 15 hour range. The dry slot was well depicted, as was the area of heaviest snow & winds. By 15z, it became quite obvious to me that the 10-15 inch totals for our area were in serious jeopardy with the dry slot punching northward, and I figured we'd probably be looking at closer to 10-12 inches. Once it started snowing - and seeing the continuation of the HRRR's solution - I was even having doubts about hitting 10 inches. Once it started sleeting around 4z with 3-4 inches of snow on the ground, I knew we weren't going to realize much past 8 inches.

This is the 2nd or 3rd time the HRRR has absolutely nailed it. I think from now on I'm going to give it some serious weight rather than treat it as an outlier.

 

[Andrew Pritchard]

Good point. I think I only checked the hrrr twice during the event, so I don't have anything archived but do recall it haven't a decent handle on things. Of course, I was more interested in the handling of models 12-36 hours out so the hrrr couldn't be included.

 

[rdale]

This is the 2nd or 3rd time the HRRR has absolutely nailed it. I think from now on I'm going to give it some serious weight rather than treat it as an outlier.

Thanks for the notes... I had given up on it last year after it became a waste of monitor ions in fall convection, maybe it's worth taking another look at.

And certainly would have helped with a forecast office near you that had up to 22" falling in their zones

 

[Robert Dewey]

I'm kicking myself for not archiving it... but from now on, I'll throw it in my bag of tools when making a decision regarding the dry slot. The radar depiction was also pretty accurate, although it tended to smooth out the highest reflectivity in the snow region leaving you with a massive blob of 35 dbz that depicted the heaviest snow.

 

[Mike Smith]

Guys,
I believe the first clue that the MONSTER snows in OK were not going to verify was the relatively low PW's in north Texas. PW's of 0.6 to 0.9" do not translate to 20+ inches of snow in 18 hours.
Mike

 

[Tom Stefanac]

Unfortunately there has been quite a backlash that forecasters have been fighting off. This storm was considered by many to be a huge dud in Southern Ontario and Western New York. Once the storm began to track northeast of Chicago an apparent dry slot opened up. This was not resolved well by any of the models, RUC/NAM-WRF & GEM REG both had significantly more precipitation.

Part of the problem looked to be the amount of deep convection forcasted along the 925mb inversion where the IP/ZR mix was supposed to originate. There was a jet-max that was probably underdone in the WNW quad and this eventually to the sfc low tracking farther north than originally anticipated. With the track of the low shifted the deep convection was not as deep seeing that it had become removed from the best PW's running right along a theta ridge associated with a deep LLJ. So in essence the entire foreseen moisture transport systems was in shambles. There was still some deep convection that produced thundersnow all across Southern Ontario but Western New York was left with next to nothing.

The second problem was two-fold in nature. Firstly, the models suggested upwards of 20 inches of snow based on Omega values for Hamilton, Niagara and Buffalo NY. It's not often that 1.5 inches of liquid will amount to that much snow but -9C temps are conductive to low packing ratios. Secondly the strong easterly winds which prompted blizzard warnings and a cold snow-growth region not necessarily conductive to big fluffy dendrites created an environment of fine snow crystals that were compacted and blown about. This led to a significant reduction of the snow ratio that was not foreseen and highly over forcasted. Then of course instead of having 1.5 inches of liquid only 1 inch or less fell and presto 20 inches of snow becomes 6 or 7 inches.

The media outlets in Southern Ontario, especially Toronto, were hyping this up and for the first time since 1993 official blizzard warnings were issued. So when people awoke to what looked like a dusting there was a significant amount of dismay. Ironically after a several hour gap the TROWAL dropped a quick 3 or 4 inches on what had already fallen and eased the criticism but Buffalo NY had no such good fortune and there it was hyped up possibly worse than the hype coming out of Chicago.

The Toronto met office received calls from the Buffao office asking what totals were like so that media outlets would understand it tracked a little north and the 20 inches fell but just not in WNY. I don't think Buffalo was happy when even the Toronto office only reported half-foot totals.

Well whatever the case the snow was still hard and dense as cement.

 

[Patrick Marsh]

At least for the Oklahoma portion of the event, the NAM snow algorithm was predicting snow (at alarming rates!) while the point forecast soundings clearly indicated sleet. This lead to the 18-20" snow totals across a large part of Oklahoma and SW Missouri. Simply put, the algorithm's precipitation type (and thus the conversion to snow totals) did not accurately include the sleet we received -- it was simply all converted to snow. [Granted, some locations (Owasso, OK) did crack the 20" mark...]