Snowfall Forecasting Suggestions

[Amped]

As most of you know, there was a snowstorm forecast to bring 6-10" to the DC area on Wednesday. In reality it wasn't quite able to produce 0.3".  This bust effected millions people across the Washington  was a huge embarrassment to the meteorological community that is not going to fade from the general public memory any time soon.

 

 

I had thought that  the snowfall forecasting strategies of the NWS had improved after March 2001.  However large over-forecasts that left millions laughing at the weatherman occurred again in January 2008 and  again last week. I am sure there have been many others that have occurred  in recent years in other parts of the country.

 

Do you think forecasts busts are the fault of models, or how they are interpreted? 

 

I personally think there are issues with both.

 

 

TV Mets,

Do you think current snowfall forecasts and uncertainty is communicated well to the public?  

 

Is social media helpful or is it a distraction?    

 

My eyes still burn from what I saw on FB and twitter Tuesday and Wednesday.

 

 

 

Weenies and non mets feel free to chime in here too.

 

 

 

 

[Roger Smith]

Maybe a system of percentage confidence in outcomes -- there are some storms where you could have a narrower range, like last week's snowstorm in northern IL where it was pretty obvious that 6" would fall and perhaps more. This recent DC storm was perceived to be more uncertain in outcome as shown by the wide range of forecasts on the forum here. That probably reflects the same range of uncertainty among actual forecast providers, so if they had said something like this:

 

for DC region 20% chance 12" greater

30% chance 8-12"

30% chance 4-8"

20% chance trace to 4" or even 2-4"

 

the public would have been more attuned to the range of possible outcomes.

 

I'm sure that many TV mets stressed the gradient expected from the shore to the mountains and so perhaps not everyone was that surprised when DCA and BWI came in with low totals, after all the people would have seen hours of melting snow falling and heard reports of nearby heavier falls, so the "bust" was a relative thing, not like a storm that just never happened.

 

There's a pride taken in "nailing" the exact amounts for various reasons (economic impact being the main one) but this science should be realistic and, as with other risk-reward activities like sport and business, identify the degree of difficulty.

 

(also pay attention to the moons of Jupiter) -- in joke

[Amped]

Maybe a system of percentage confidence in outcomes -- there are some storms where you could have a narrower range, like last week's snowstorm in northern IL where it was pretty obvious that 6" would fall and perhaps more. This recent DC storm was perceived to be more uncertain in outcome as shown by the wide range of forecasts on the forum here. That probably reflects the same range of uncertainty among actual forecast providers, so if they had said something like this:

 

for DC region 20% chance 12" greater

30% chance 8-12"

30% chance 4-8"

20% chance trace to 4" or even 2-4"

 

the public would have been more attuned to the range of possible outcomes.

 

I'm sure that many TV mets stressed the gradient expected from the shore to the mountains and so perhaps not everyone was that surprised when DCA and BWI came in with low totals, after all the people would have seen hours of melting snow falling and heard reports of nearby heavier falls, so the "bust" was a relative thing, not like a storm that just never happened.

 

There's a pride taken in "nailing" the exact amounts for various reasons (economic impact being the main one) but this science should be realistic and, as with other risk-reward activities like sport and business, identify the degree of difficulty.

 

(also pay attention to the moons of Jupiter) -- in joke

 

Yeah some probability would be nice. Maybe highlight some uncertainty.  Also widening the ranges might help. 1-10" instead of 5-10" might imply the uncertainly.

[famartin]

I don't want to Monday morning quarterback here...  I didn't forecast for DCA for this storm.  (I will say I thought I had the best forecast for TTN, highlighting the high probability that little to no accumulation would occur that far north, which turned out correct).

 

But a lot was made before the storm of the belief that "oh, models show 2 m temps of 33-34 but heavy precip will bring it down to freezing". 

 

That works great where there is no advection of milder air.

 

But this storm did have a strong easterly push of milder air from the east.

 

And of course, QPF forecasts were less than accurate.

 

A misinterpretation of the surface temperature output (assuming the models were wrong there) while going with the high QPF output (assuming they were right there) combined to create this DCA bust.  I'm sure plenty has been made of this on the DC subforums.

[thunderman]

Yeah some probability would be nice. Maybe highlight some uncertainty.  Also widening the ranges might help. 1-10" instead of 5-10" might imply the uncertainly.

I got laughed at for using 8-16 down here.  Apparently that is too large of range.  I don't even want to think about what kind of a reaction 1-10 would get.

[thunderman]

I don't want to Monday morning quarterback here...  I didn't forecast for DCA for this storm.  (I will say I thought I had the best forecast for TTN, highlighting the high probability that little to no accumulation would occur that far north, which turned out correct).

 

But a lot was made before the storm of the belief that "oh, models show 2 m temps of 33-34 but heavy precip will bring it down to freezing". 

 

That works great where there is no advection of milder air.

 

But this storm did have a strong easterly push of milder air from the east.

 

And of course, QPF forecasts were less than accurate.

 

A misinterpretation of the surface temperature output (assuming the models were wrong there) while going with the high QPF output (assuming they were right there) combined to create this DCA bust.  I'm sure plenty has been made of this on the DC subforums.

^This.

Even here (west of the BR) we got into the mid and upper 30s after precip stopped.  Luckily I had most of my precip during the morning and nighttime hours or things may have wound up differently here too.

[Amped]

I got laughed at for using 8-16 down here.  Apparently that is too large of range.  I don't even want to think about what kind of a reaction 1-10 would get.

 

I don't understand the public too well but I promise 1-10 would still  not get you laughed at close to as hard as DC mets got laughed at this week.

[BlizzardWx]

I didn't really look at it too much myself but at least from a practical perspective it is important to let people know how certain you are or are not about something. Here in the mountain west because of the complicated terrain small differences can result in hugely different outcomes. Some storms we get one inch here and it snows 10-15 a few miles north or south. So just be honest about what you know and don't know.

[Ed Lizard]

This is actually half way boasting.  Bear with me.  Forecasts can bust when models give excellent guidance, and, BTW, on model guidance, the HPC prob guidance for snowfall amounts (never heard of it before lurking New England subforum as I do lurk in non-home subforums for potential big ticket) and best I can tell, a multi-model (including several global models and their ensemble means, and meso models and ensembles) a multi-model ensemble probability forecast works well, not for predicting amounts, but predicting odds of a bust.

 

Back to models working.  Some red tagger I should thank but can't remember gave a primer on top down forecasting, I discovered NIU sounder page can make a version of the skew-T that is not logarithmic, ("anagram") so I can see vertical isotherms, and knowing crystal growth regions, temps, and saturations, GFS and NAM were both predicting freezing drizzle, a large subfreezing column but never colder than about -5C, and local NWS and local TV station forum pro-mets see sub 540 and below freezing 850 mb, and we're due 2-5 inches, and when we get freezing drizzle, the models busted and if only HGX could launch balloons and give us a local sounding.  The models, if read right, gave excellent guidance.  In a place where snow is a twice a decade occurrence, magic thickness and sub-freezing 850 mb is only p-type tool known.

 

Models can agree, be correct, and humans can bust.

 

There is a degree of self back patting, but I am also making a point.

[Amped]

This is actually half way boasting.  Bear with me.  Forecasts can bust when models give excellent guidance, and, BTW, on model guidance, the HPC prob guidance for snowfall amounts (never heard of it before lurking New England subforum as I do lurk in non-home subforums for potential big ticket) and best I can tell, a multi-model (including several global models and their ensemble means, and meso models and ensembles) a multi-model ensemble probability forecast works well, not for predicting amounts, but predicting odds of a bust.

 

Back to models working.  Some red tagger I should thank but can't remember gave a primer on top down forecasting, I discovered NIU sounder page can make a version of the skew-T that is not logarithmic, ("anagram") so I can see vertical isotherms, and knowing crystal growth regions, temps, and saturations, GFS and NAM were both predicting freezing drizzle, a large subfreezing column but never colder than about -5C, and local NWS and local TV station forum pro-mets see sub 540 and below freezing 850 mb, and we're due 2-5 inches, and when we get freezing drizzle, the models busted and if only HGX could launch balloons and give us a local sounding.  The models, if read right, gave excellent guidance.  In a place where snow is a twice a decade occurrence, magic thickness and sub-freezing 850 mb is only p-type tool known.

 

Models can agree, be correct, and humans can bust.

 

There is a degree of self back patting, but I am also making a point.

 

Yes, scapegoating the models is easy though. We all trusted the models the NWS and eachother,.  Mets have to think about what the public hears when they forecast 6-10" which is essentially

"I promise the roads will be to dangerous to drive on tomorrow, there is no way there will not be 6-10" on the roads tomorrow, bare pavement is impossible, but if it happens feel  free to cuss at me"

Models don't output this so some interpretation is required by the mets, they have to look at a lot of variables instead of a few variables on a lot of models. They also need to stop the Facebook pimping prior to a major event leave it for noncritical weather days.

 

Edit: That  last point wasn't really the gist of my post though. Minor issue.

[Amped]

A week has gone by and the thread is dying. so i guess I Hopefully the NWS at least issues a statement apologizing and announce plans to fix long standing forecast issues in marginal situations. This is not the first time a seemingly sure high impact event turned into nothing.   So for now, I'll tackle this one.

 

I apologize if this is a rant, I am not out to get the NWS and throw them under a bus, but I also don't like taking "oh well the models were wrong" for an answer on this issue.

 

 

Lessons Learned  and things to keep in mind:

 

Not every low tracking 100 miles to your southeast during winter is 8-12" with a 20"+  Jackpot death band on the northwestern edge.

 

It doesn't always get colder as you go northwest

 

The rain snow line is not always a sharp line.  Some storms can have a very large marginal area 100 miles northwest of line line doesn't guarantee all snow.

 

540 thickness is useless if the surface is marginal

 

850mb 0c line is useless if the surface is marginal

 

Partial thicknesses are useless if the surface is marginal.

 

Verbatim model snowfall output is useless if the surface is marginal

 

If it's warm the day, before snow will not accumulate at all if the surface is marginal

 

Soundings plots are deceiving for PTYPE usage, They  make a 400 meter 1.8 C warm layer  at the surface look like not a big deal.  It's rain!!

 

Only surface temperatures will tell you if the surface is bad. Look at them carefully.

 

The days where models significantly under do CAD are over with the new hires models. Do not assume it will be 5f colder because than what the EURO shows verbatim.

 

Look at the air-mass to your northeast, north and northwest. it will be over you the next day.

 

Look at more variables and less models!!! The HRRR isn't going to help.

 

Do not assume the NWS bandwagon is always right and has thoroughly examined the event.

 

Snowfall does not equal snow depth.

 

Make sure it's foretasted to get much colder on the back-end of the storm before saying rain to snow.

 

Tone down snowfall forecasts for the major cities during the day if there is no sub 30f surface temp foretasted within 50 miles of them. 1-3" will suffice for most cases, there is rarely a high impact event.

 

Recommendations:

 

 I promise not to mention this again if someone agrees with me. 

 

  The general public cares about one thing during a snowstorm, "The Roads" not how many inches of snow there is even if they don't ask the question that way. We  are the only ones who care how much snow is on some weenie snowboard on a 1000ft hill 50 miles northwest of the cities.  The general public outnumbers us 10,000 to 1 so I think it is about time we start forecasting for them. Currently the GFS, NAM and Euro do not output a snow depth on a paved surface variable (correct me if I'm wrong) NWS doesn't issue warnings or forecasts based on snow depth on a paved surface, and they don't use snow depth on pavement  to verify their forecasts. Neither do most media outlets from what I can see. This will have to change someday, might as well start now

 

 

 

 

All right, I've about said everything on my mind. If you have criticism please be respectful.

[famartin]

NWS doesn't issue warnings or forecasts based on snow depth on a paved surface

 

We are taking baby steps in this direction. 

 

Also keep in mind, for a very heavy wet snow the trees and powerlines are also a factor, even if the pavement is wet.

[MGorse]

A week has gone by and the thread is dying. so i guess I Hopefully the NWS at least issues a statement apologizing and announce plans to fix long standing forecast issues in marginal situations. This is not the first time a seemingly sure high impact event turned into nothing.   So for now, I'll tackle this one.

 

I apologize if this is a rant, I am not out to get the NWS and throw them under a bus, but I also don't like taking "oh well the models were wrong" for an answer on this issue.

 

 

Lessons Learned  and things to keep in mind:

 

Not every low tracking 100 miles to your southeast during winter is 8-12" with a 20"+  Jackpot death band on the northwestern edge.

 

It doesn't always get colder as you go northwest

 

The rain snow line is not always a sharp line.  Some storms can have a very large marginal area 100 miles northwest of line line doesn't guarantee all snow.

 

540 thickness is useless if the surface is marginal

 

850mb 0c line is useless if the surface is marginal

 

Partial thicknesses are useless if the surface is marginal.

 

Verbatim model snowfall output is useless if the surface is marginal

 

If it's warm the day, before snow will not accumulate at all if the surface is marginal

 

Soundings plots are deceiving for PTYPE usage, They  make a 400 meter 1.8 C warm layer  at the surface look like not a big deal.  It's rain!!

 

Only surface temperatures will tell you if the surface is bad. Look at them carefully.

 

The days where models significantly under do CAD are over with the new hires models. Do not assume it will be 5f colder because than what the EURO shows verbatim.

 

Look at the air-mass to your northeast, north and northwest. it will be over you the next day.

 

Look at more variables and less models!!! The HRRR isn't going to help.

 

Do not assume the NWS bandwagon is always right and has thoroughly examined the event.

 

Snowfall does not equal snow depth.

 

Make sure it's foretasted to get much colder on the back-end of the storm before saying rain to snow.

 

Tone down snowfall forecasts for the major cities during the day if there is no sub 30f surface temp foretasted within 50 miles of them. 1-3" will suffice for most cases, there is rarely a high impact event.

 

Recommendations:

 

 I promise not to mention this again if someone agrees with me. 

 

  The general public cares about one thing during a snowstorm, "The Roads" not how many inches of snow there is even if they don't ask the question that way. We  are the only ones who care how much snow is on some weenie snowboard on a 1000ft hill 50 miles northwest of the cities.  The general public outnumbers us 10,000 to 1 so I think it is about time we start forecasting for them. Currently the GFS, NAM and Euro do not output a snow depth on a paved surface variable (correct me if I'm wrong) NWS doesn't issue warnings or forecasts based on snow depth on a paved surface, and they don't use snow depth on pavement  to verify their forecasts. Neither do most media outlets from what I can see. This will have to change someday, might as well start now

 

 

 

 

All right, I've about said everything on my mind. If you have criticism please be respectful.

 

The NWS were not the only ones who were wrong.

[Amped]

We are taking baby steps in this direction. 

 

Also keep in mind, for a very heavy wet snow the trees and powerlines are also a factor, even if the pavement is wet.

Good to hear. At least there looking into what can be done.    In the mean time I personally am not telling people there will be over 4" of snow. If they ask me how much snow there is going to be i'll either say 1-3" or "I don't know, check The Weather Channel"   Let them get it wrong.

[Quincy]

TV Mets,

Do you think current snowfall forecasts and uncertainty is communicated well to the public?  

 

Is social media helpful or is it a distraction?    

 

My eyes still burn from what I saw on FB and twitter Tuesday and Wednesday.

I think overall, the uncertainty in on-air forecasts is not communicated very well. It could be a combination of factors, including that an on-air meteorologist may be "stubborn" and wants to lean one way or another. There's a fine line between having confidence and explaining the different scenarios. I know it gets tricky when you're close to an event and if you don't give specifics, the public gets annoyed.

 

I've been experimenting with adding a "confidence meter" to some of my forecasts. I know that some markets have toyed with the idea, but that most don't do anything like it. It's one thing to verbally explain confidence, but graphics and maps really seem to stick in the minds of most viewers.

 

Social media is helpful, but we're now noticing a lot of speculation coming out many days in advance of storms. It adds to the hype, but it's also a medium by which incorrect information and unscientific "forecasts" can go viral. There are pros and cons, but I think that in most cases, it is helpful. I try to personally stress if a Tweet or post I am making is an actual forecast or if it's simply one potential scenario or output from a specific computer model run.

 

I think forecasting using/referring to climo is something many forecasters lack in. If it's March and we're talking about a highly anomalous weather event, you have to have very high confidence in such a scenario actually happening. The forecast may have not been clear cut, but there were signs that there was high bust potential in the mid-Atlantic region with the event being discussed.

[MGorse]

I think overall, the uncertainty in on-air forecasts is not communicated very well. It could be a combination of factors, including that an on-air meteorologist may be "stubborn" and wants to lean one way or another. There's a fine line between having confidence and explaining the different scenarios. I know it gets tricky when you're close to an event and if you don't give specifics, the public gets annoyed.

 

Time has to play a role as well.

[Amped]

I am much happier with the conservative bust on today event. Maybe 1-2 was a bit conservative for I95 for the 3-5" that fell but roads were mainly wet and slushy, It was an advisory criteria 3-5".

 

3/8/2005 in the NYC metro was 3-5" as well but probably deserved a warning.

[weathafella]

Psychology inevitably "infects" physical science.

http://www.americanwx.com/bb/index.php/topic/39657-2-major-busts-in-the-same-stormdid-the-first-one-cause-the-second/

[Amped]

Psychology inevitably "infects" physical science.

http://www.americanwx.com/bb/index.php/topic/39657-2-major-busts-in-the-same-stormdid-the-first-one-cause-the-second/

 

Yes it will, wherever there is room for subjective judgement or there are mixed indicators these types of things will happen.  Most professions checklists or operating procedures to hopefully minimize the use of subjective opinions. . The NWS was able to issue a forecast of 4-7" with a low of 35f in January 2008 for NYC, so I doubt such a checklist existed then,  I don't know if things have changed now.

 

DT has a checklist of course, but he gets too nuts when a big storm hits to actually apply it.

[Justin]

A couple on air met friends here in the Midwest have explained they wish they had more time to explain things on big and marginal events, but its all about the advertising dollar.

They also have facebook pages where they put model graphics up and stress its not a forecast just guidance they're using. Then when a storm doesn't lay down what was shown they take a lot of grief. On camera meteorologists need to use facebook carefully.

These storms that are marginal or unsure right up to when they get underway need to be emphasized that little or lots may fall there is just to much uncertainty. Honesty works better than constantly being wrong...and this isn't really a constant thing anyways, maybe once or twice a season.

[Amped]

A couple on air met friends here in the Midwest have explained they wish they had more time to explain things on big and marginal events, but its all about the advertising dollar.

They also have facebook pages where they put model graphics up and stress its not a forecast just guidance they're using. Then when a storm doesn't lay down what was shown they take a lot of grief. On camera meteorologists need to use facebook carefully.

These storms that are marginal or unsure right up to when they get underway need to be emphasized that little or lots may fall there is just to much uncertainty. Honesty works better than constantly being wrong...and this isn't really a constant thing anyways, maybe once or twice a season.

 

Forecast to go with in a marginal case is probably. (Not enough snow to cancel school) to (Just enough to cancel school.)   Around here it's 2-4", but in some areas it might take more.

 

I think I will start a short range forecasting thread (Mostly for non-mets) early next winter so everyone can pool their thoughts together. Obviously we have to look at different variables in the short range then we do in the medium range where precision doesn't matter.

[weatherwiz]

A couple on air met friends here in the Midwest have explained they wish they had more time to explain things on big and marginal events, but its all about the advertising dollar.

They also have facebook pages where they put model graphics up and stress its not a forecast just guidance they're using. Then when a storm doesn't lay down what was shown they take a lot of grief. On camera meteorologists need to use facebook carefully.

These storms that are marginal or unsure right up to when they get underway need to be emphasized that little or lots may fall there is just to much uncertainty. Honesty works better than constantly being wrong...and this isn't really a constant thing anyways, maybe once or twice a season.

 

I've noticed a great deal of mets now using their facebook/twitter pages to post model forecast graphics and while they do add a disclaimer mentioning that is only a piece of guidance, not a forecast people don't seem to comprehend that.  Many of the graphics posted also are usually a model output showing some extreme solution and the met mentioning that...once you do that and explain what it shows that's all people will remember and care about.  

[famartin]

I've noticed a great deal of mets now using their facebook/twitter pages to post model forecast graphics and while they do add a disclaimer mentioning that is only a piece of guidance, not a forecast people don't seem to comprehend that.  Many of the graphics posted also are usually a model output showing some extreme solution and the met mentioning that...once you do that and explain what it shows that's all people will remember and care about.  

I don't have a "following" but I've definitely done that... because we get wowed too

[weatherwiz]

I don't have a "following" but I've definitely done that... because we get wowed too

 

I don't think there is anything wrong with it!  It's just you know how some people are and their lack of reading skills...but I guess that is their problem!  

 

As long as the disclaimer is there stating this isn't a forecast and only a piece of guidance than people can't really say anything negative.

[Amped]

Just FYI I saved the Ewall map from 5/1.The NWS did a good job keeping amounts low on this one in Iowa and Western MO. In February or March this type of map may pass as somewhat believable.