January 2016 Pattern Disco

[OceanStWx]

I have a hard time relating probabilities to any given LR situations when Ma nature is involved, you know what the probability of BWI getting a 30 inch snow storm was going into this winter?. Point being stats are great post mortem. Predictability? I don't know if they are any use for individual years.

 

Well obviously because ORH may have a 15% chance of reaching normal snowfall doesn't mean we'll stop forecasting snow there for the rest of winter. I don't consider that a predictive statistic for day to day forecasting purposes. Likewise BWI's chance at a 30" snowfall during winter. All that tells me is there is a chance in any given winter of a 30" snowfall.

 

But I do think there is utility to probabilistic forecasts for individual storm or events. Certain set ups lend themselves to higher or lower confidence, and a snowfall distribution can be fit to that for instance. Or how SPC does it. Say your chance at a 30" snowfall on any winter day is 0.05%. If you put out a 5% chance of 30" of snow for a forecast, that tells you your risk has gone up 100 fold. A big deal. We're currently not quite there yet outside of severe season though.

[Organizing Low]

Need at least 45" in both February and in March, and say 10" in April to get to 125" average snowfall here at the house.

Going to be a tall order. Mountain top needs about 100" per month to get close.

Let's hope it gets stormy and real fast.

 

 

well below normal snow and mild temps (although not the blowtorch some media had portrayed)

 

incredibly tepid winter,  especially after the last 2 ice bucket winters.

 

I don't know how much snow we have received exactly in saint albans this year as I don't know how to get official measurements, but its not much, cant be much more than 24 inches or so.

 

 

Montreal has seen 29.1 inches of snow so far this year

 

Ottawa has seen only 21.6 inches this year (normal 102 inches).

[ORH_wxman]

I have a hard time relating probabilities to any given LR situations when Ma nature is involved, you know what the probability of BWI getting a 30 inch snow storm was going into this winter?. Point being stats are great post mortem. Predictability? I don't know if they are any use for individual years.

 

It's a good starting point though.

 

The M.A. also had a higher probability than simple historic frequency to get a very large snow...considering almost all of their big ones come in El Nino winters with just a couple exceptions, and this is an El Nino winter.

 

Looking at past El Ninos and using the Euro ensembles of the pattern going into February, I'd have a hard time saying ORH has a better than 20% chance of getting to climo...perhaps as high as 25% if you really want to stretch it. Now a 1 in 5 shot is not the same as zero. it means that 1 in 5 times you actually do make it to climo. People tend to view these probabilistic arguments in the lens of a deterministic forecast. It isn't. What it means is that the chance is relatively low. Whatever the result happens to be doesn't debunk the theory of those odds.

[powderfreak]

I have a hard time relating probabilities to any given LR situations when Ma nature is involved, you know what the probability of BWI getting a 30 inch snow storm was going into this winter?. Point being stats are great post mortem. Predictability? I don't know if they are any use for individual years.

 

Wow that's a surprising post from you Ginxy.  That's like tossing most of DonS's posts out the window as he's all about long range probability, especially concerning a current base-line pattern and what the future chances of something occurring are.

 

So all those posts of "when the AO reaches -4.0 there is an increased probability of XYZ pattern"... we don't want to discuss that with LR situations because Ma Nature is involved and she doesn't care about probabilities?

 

I mean 99% of the long range stuff we talk about is based on probabilities and what has transpired previously (looking at statistics) based on an initial condition.

[powderfreak]

It's a good starting point though.

 

The M.A. also had a higher probability than simple historic frequency to get a very large snow...considering almost all of their big ones come in El Nino winters with just a couple exceptions, and this is an El Nino winter.

 

Looking at past El Ninos and using the Euro ensembles of the pattern going into February, I'd have a hard time saying ORH has a better than 20% chance of getting to climo...perhaps as high as 25% if you really want to stretch it. Now a 1 in 5 shot is not the same as zero. it means that 1 in 5 times you actually do make it to climo. People tend to view these probabilistic arguments in the lens of a deterministic forecast. It isn't. What it means is that the chance is relatively low. Whatever the result happens to be doesn't debunk the theory of those odds.

 

I bet BWI has a higher probability of a 30" snowstorm than say Saranac Lake, NY in any given year.

 

And good point on the 1 in 5 thing.  That's why its not unheard of, just not what you would say is "likely."  Heck we did manage it in NNE in 2006-2007, as every system broke right for us that second half.  And it was active and we didn't miss anything.  It was like a home run derby in Feb/Mar/Apr. 

 

So say that's the 1 in 5 year recovery.  Definitely could happen again, but if you are a betting person, you play the odds.

[Ginx snewx]

Wow that's a surprising post from you Ginxy.  That's like tossing most of DonS's posts out the window as he's all about long range probability, especially concerning a current base-line pattern and what the future chances of something occurring are.

 

So all those posts of "when the AO reaches -4.0 there is an increased probability of XYZ pattern"... we don't want to discuss that with LR situations because Ma Nature is involved and she doesn't care about probabilities?

 

I mean 99% of the long range stuff we talk about is based on probabilities and what has transpired previously (looking at   

 

Wow that's a surprising post from you Ginxy.  That's like tossing most of DonS's posts out the window as he's all about long range probability, especially concerning a current base-line pattern and what the future chances of something occurring are.

 

So all those posts of "when the AO reaches -4.0 there is an increased probability of XYZ pattern"... we don't want to discuss that with LR situations because Ma Nature is involved and she doesn't care about probabilities?

 

I mean 99% of the long range stuff we talk about is based on probabilities and what has transpired previously (looking at statistics) based on an initial condition.

LR forecast are meh.I don't think probabilities are predictable for months. What were these probabilities Jan 26th last year for Boston to exceed its normal. Sorry don't see the value .Chaos reigns

[Ginx snewx]

It's a good starting point though.

 

The M.A. also had a higher probability than simple historic frequency to get a very large snow...considering almost all of their big ones come in El Nino winters with just a couple exceptions, and this is an El Nino winter.

 

Looking at past El Ninos and using the Euro ensembles of the pattern going into February, I'd have a hard time saying ORH has a better than 20% chance of getting to climo...perhaps as high as 25% if you really want to stretch it. Now a 1 in 5 shot is not the same as zero. it means that 1 in 5 times you actually do make it to climo. People tend to view these probabilistic arguments in the lens of a deterministic forecast. It isn't. What it means is that the chance is relatively low. Whatever the result happens to be doesn't debunk the theory of those odds.

Means nothing really when it comes down to it. Probability is not predictability for the 2 months left PF is referencing

[Ginx snewx]

Well obviously because ORH may have a 15% chance of reaching normal snowfall doesn't mean we'll stop forecasting snow there for the rest of winter. I don't consider that a predictive statistic for day to day forecasting purposes. Likewise BWI's chance at a 30" snowfall during winter. All that tells me is there is a chance in any given winter of a 30" snowfall.

 

But I do think there is utility to probabilistic forecasts for individual storm or events. Certain set ups lend themselves to higher or lower confidence, and a snowfall distribution can be fit to that for instance. Or how SPC does it. Say your chance at a 30" snowfall on any winter day is 0.05%. If you put out a 5% chance of 30" of snow for a forecast, that tells you your risk has gone up 100 fold. A big deal. We're currently not quite there yet outside of severe season though.

this I agree with. The probability of a BWI 30" snow fall was ? because it never happened. Individual events 100% agree

[OceanStWx]

 

LR forecast are meh.I don't think probabilities are predictable for months. What were these probabilities Jan 26th last year for Boston to exceed its normal. Sorry don't see the value .Chaos reigns

 

So why do any of this?

 

Clearly it's possible to add some skill to climatology. You can absolutely use probabilities (whether in a specific pattern, time of year, etc.) to shift the numbers to your advantage.

 

Your example for BOS last year. Sure chances were low they exceed normals. But don't you think given the pattern predicted for February 2015 that your chances of exceeding normal was higher than it otherwise might have been? Were we not forecasting above normal snowfall for February?

 

You're never going to capture extremes, that's why they are extreme, but there is definitely value.

 

[powderfreak]

Means nothing really when it comes down to it. Probability is not predictability for the 2 months left PF is referencing

We'll have to agree to disagree then. We will reference this discussion though at future times when stats and probability are given (like in a Don S post about likelyhood of a given outcome based on initial conditions).

Pretty much every long range forecast is based on probability whether it acknowledges it or not. It's looking back at past patterns and running the most probable outcome and using that as a basis/starting point. Even as simple as saying "we have a weak El Niño so I like our chances this winter" is using past history and probabilities to make a statement about the future.

And extreme patterns will and can occur...no one is ever going to nail those spot on months out.

[ORH_wxman]

Means nothing really when it comes down to it. Probability is not predictability for the 2 months left PF is referencing

 

I disagree strongly that it means nothing.

 

Obviously we don't know exactly what will happen, but probabilities gives us an idea of what is more likely. Just like we don't know that BOS will see 100" of snow next winter...we have a pretty good idea that it is unlikely based on historical records of their climo. Of course we hope it is wrong and 2015 or 1996 walks through that door, but pretending that past climo doesn't matter when calculating odds is downright silly to me.

[CoastalWx]

Maybe the Mashantucket scrolls will tell us.

[BostonWX]

Sample size is probably the biggest detractor to the probalistic forecasting idea. We have a tiny sample in the grand scheme of things, in all likelihood too small to account for variance (i.e whether we've been performing better or worse than we should based on a longer term sampled average or any given parameter like a super el nino or nina)

[Damage In Tolland]

Too much anger in here

[CoastalWx]

I guess it depends on the metric. We can have a high probability of below normal temps in the east with a +PNA. We have a large sample and understand the physics behind it. However, perhaps saying we have a high probability of below normal snowfall in Boston in a strong El Nino is more risky. The sample size is small...but the thing that offsets that, is the physics behind stronger Ninos. However, this assumption also may go back to the small sample size. 

[Ginx snewx]

Too much anger in here

Seems like good discussion , lots of learning stuff. Perhaps over your head

[Ginx snewx]

So why do any of this?

 

Clearly it's possible to add some skill to climatology. You can absolutely use probabilities (whether in a specific pattern, time of year, etc.) to shift the numbers to your advantage.

 

Your example for BOS last year. Sure chances were low they exceed normals. But don't you think given the pattern predicted for February 2015 that your chances of exceeding normal was higher than it otherwise might have been? Were we not forecasting above normal snowfall for February?

 

You're never going to capture extremes, that's why they are extreme, but there is definitely value.

 

I went away from trying to do LR stuff. I agree why do it. The probability of Boston exceeding climo was not predictive though Jan 25th. What was the probability of the Pats winning yesterday. I saw 80% on the Vegas line, chaos reigned. Same reason I don't see a bookie anymore.

[BostonWX]

I guess it depends on the metric. We can have a high probability of below normal temps in the east with a +PNA. We have a large sample and understand the physics behind it. However, perhaps saying we have a high probability of below normal snowfall in Boston in a strong El Nino is more risky. The sample size is small...but the thing that offsets that, is the physics behind stronger Ninos. However, this assumption also may go back to the small sample size.

Yeah that's true for seasons that have more atmospheric profiles. And I would think snowfall adds another layer of complexity to it with a lot of moving parts and shortwaves etc. Temperatures must be easier (still not easy per se).

[Ginx snewx]

We'll have to agree to disagree then. We will reference this discussion though at future times when stats and probability are given (like in a Don S post about likelyhood of a given outcome based on initial conditions).

Pretty much every long range forecast is based on probability whether it acknowledges it or not. It's looking back at past patterns and running the most probable outcome and using that as a basis/starting point. Even as simple as saying "we have a weak El Niño so I like our chances this winter" is using past history and probabilities to make a statement about the future.

And extreme patterns will and can occur...no one is ever going to nail those spot on months out.

LR is a crapshoot at best.

[powderfreak]

LR is a crapshoot at best.

Agree with you there.

And lol about "the anger"....you're right, some good discussion and debate that isn't getting personal.

[Ginx snewx]

Agree with you there.

And lol about "the anger"....you're right, some good discussion and debate that isn't getting personal.

Well its probable I will have dinner cooked for me when I get home but ya never know lol. Great stuff man

[CoastalWx]

LR is a crapshoot at best.

 

I wouldn't say crapshoot. We have reasonable skill in temps. 

[OceanStWx]

I went away from trying to do LR stuff. I agree why do it. The probability of Boston exceeding climo was not predictive though Jan 25th. What was the probability of the Pats winning yesterday. I saw 80% on the Vegas line, chaos reigned. Same reason I don't see a bookie anymore.

 

I guess we just don't agree.

 

Most forecasts I saw last winter argued for good potential for above normal snowfall. We talked for weeks on here when the pattern failed to deliver right away that patience was required. Now trying to use probabilities to predict absolute numbers at the end of a season is a fool's errand, but they argued for above normal snowfall and that's what ended up happening.

 

So there's value in that.

 

It's like Scott just said about the PNA. Chances of above or below normal temps is 50/50 any given day. But in a +PNA or a -EPO, we know our chances for below normal go up. It would be pretty bold to forecast above normal temps then.

[Ginx snewx]

I wouldn't say crapshoot. We have reasonable skill in temps. 

But but, we are talking snow.

[Ginx snewx]

I guess we just don't agree.

 

Most forecasts I saw last winter argued for good potential for above normal snowfall. We talked for weeks on here when the pattern failed to deliver right away that patience was required. Now trying to use probabilities to predict absolute numbers at the end of a season is a fool's errand, but they argued for above normal snowfall and that's what ended up happening.

 

So there's value in that.

 

It's like Scott just said about the PNA. Chances of above or below normal temps is 50/50 any given day. But in a +PNA or a -EPO, we know our chances for below normal go up. It would be pretty bold to forecast above normal temps then.

The discussion was on Jan 25th are probabilities predictive enough to say the next two months we are able to reach normal climo snow. I say no they are not predictive enough, PF says yes they are.

[OceanStWx]

I wouldn't say crapshoot. We have reasonable skill in temps. 

 

That's my take. I think in here anyway the thing most care about is snowfall, which ultimately depends on two variables (temp and precip). Snowfall is hard to predict 48-72 hours out, let alone 48 days.

 

Also, if people are searching for hard numbers long range is a crap shoot. Everyone was on board with a warm December. We had a warm December. Nobody was forecasting a +10 December. But does that make them wrong? Does it mean that the probabilistic guidance we had when it came to the pattern didn't provide value?

[Ginx snewx]

That's my take. I think in here anyway the thing most care about is snowfall, which ultimately depends on two variables (temp and precip). Snowfall is hard to predict 48-72 hours out, let alone 48 days.

 

Also, if people are searching for hard numbers long range is a crap shoot. Everyone was on board with a warm December. We had a warm December. Nobody was forecasting a +10 December. But does that make them wrong? Does it mean that the probabilistic guidance we had when it came to the pattern didn't provide value?

wow did this discussion get twisted,

[OceanStWx]

The discussion was on Jan 25th are probabilities predictive enough to say the next two months we are able to reach normal climo snow. I say no they are not predictive enough, PF says yes they are.

 

Well for planning purposes a reasonable worst (or best) case scenario is typically held to the 10th percentile standard. So having a 1/5 chance for normal means it too early to throw in the towel right now.

 

1/10 isn't likely, but it's likely enough that decision makers feel you should be prepared for it.

 

Besides, if we're specifically talking snowfall then you're asking for way too much from long range forecasts. We talked about the potential with this latest pattern. New Haven ends up with over a foot, and Litchfield less than an inch 30 some odd miles away. Does that really mean we were wrong about the pattern. It supported a good chance of above normal snowfall. Above normal snowfall occurred for parts of the area. There were just some backyards that were left on the outside looking in.

[Damage In Tolland]

Seems like good discussion , lots of learning stuff. Perhaps over your head

Everyone just angry over your arguing

[OceanStWx]

wow did this discussion get twisted,

 

When you say long range is a crap shoot I assume you mean in general, not snowfall.

 

Snowfall is much, much harder to predict on the monthly scale. I think Will said it too, people tend to relate probabilities to deterministic forecasts. If we say good chance at above normal snowfall, people want to know X number of inches above normal for their backyard. It doesn't work that way.

 

Temps are fairly evenly distributed, QPF even less so, and snowfall even less so that QPF.