Network News

X My Profile
View More Activity
The new Washington
Post Weather website
Jump to CWG's
Latest Full Forecast
Outside now? Radar, temps
and more: Weather Wall
Follow us on Twitter (@capitalweather) and become a fan on Facebook
Posted at 11:00 AM ET, 03/ 2/2010

The varying predictability of snowstorms

By Steve Tracton

* Some wet and white tonight: Full Forecast | Weather Wall *
* Bob Ryan's farewell at 4 | New England "bomb" | CWG T-Shirts! *

In a recent post, I reviewed the story behind the headline from just over 10 years ago, "Surprise Snowstorm Ambushes Washington." The "Surprise Snowstorm" of Jan. 24-25, 2000, was aptly named given heavy snowfall was not predicted for the D.C. region until about 6-8 hours before it began to accumulate.

Any snow lover (or hater) can probably recall surprise "no snowstorms" -- false alarms such that heavy snowfall is virtually guaranteed by forecasters within 24 hours or so of a storm that turns out to be a complete bust. Needless to say, whether a surprise snowstorm or surprise no snowstorm, the inevitable results are red-faced forecasters and media outlets eating "humble pie" and blaming busts, not always justifiably, on "THE models."

Given the tremendous advances in observational, data analysis and weather modeling systems run routinely on supercomputers, especially the development of ensemble prediction methods, there are now relatively few complete surprises in forecasting "big ones" affecting the mid-Atlantic and Northeast coastal regions a day or so in advance. "Complete" here means near certitude with a yes or no call.

Keep reading for more on advances in the predictability of snowstorms...

As described in my recent post, ensemble prediction systems provide estimates of the varying degree of the uncertainties in forecasts, by re-running the same model multiple times using slightly different initial conditions. The uncertainty can be conveyed, for example, as statements in the level of confidence and/or in terms of probabilities. It's just the nature of the beast that some forecasts are more predictable (high confidence) than others (low confidence), even in predictions just 12-24 hours ahead.

Be aware, however, that even when forecast confidence is exceptionally high, as was true for Snowmageddon and Snoverkill (to a somewhat lesser degree for the latter), precision in forecasting the details of snowfall amounts and variability across the D.C. area is not possible (and may never be -- blame it on the butterflies).

With Snowmageddon, while forecasts might have been off by 4 inches out of 2+ feet at any given location, it's remarkable -- and unique in my experience -- to see forecasts calling for a storm to almost certainly enter the list of "top 10" snowstorms on record, even if only a day ahead, and actually do so.

With Snoverkill there was very high confidence that the storm would not measure up to Snowmageddon in total snowfall, but less confidence in the accumulation details, given the greater complexity of the storm development. Of importance was that a day ahead there was a solid likelihood of the nine inches or so of snow necessary to challenge D.C.'s seasonal snowfall record.

Much more problematic is predicting a significant snowstorm with a high level of confidence more than a few days in advance. Cases in this category are extremely rare. Probably the best example -- before this winter -- was the Superstorm of March 1993 when a huge storm with heavy snow over an extraordinarily large area was successfully forecast with high confidence 4-5 days in advance. Snowmageddon and (to a lesser degree) Snoverkill can now be added to the list of these rare occurrences.

Does this mean we've moved into an era of increased capability to predict major East Coast snowstorms several days out?

I doubt it, if for no other reason than the first "big one" this winter in December was not foreseen as a significant possibility until about two days before the storm struck (see recap of that storm's evolution here). More generally, forecasts beyond a few days ahead are less predictable (greater uncertainty) than shorter-range predictions, especially when considering the critical aspects of storm track, precipitation amounts and type.

The longer-range predictions provide greater opportunity for the butterflies -- and the "chaos" they iconically represent -- to do their mischief. They have not, nor will they ever be completely stilled. A "big one" affecting this or any region reflects a tremendous coincidence of contributing factors (moisture, cold air, etc.) coming together just right in both time and space ("sweet spot"). Given the extreme complexity of the relevant atmospheric processes it is not unexpected that forecast models have considerable difficulty in getting it right.

The questions then become what were the primary contributing factors that set the stage for development of Snowmageddon and Snoverkill, do they provide any clues on why these storms were intrinsically more predictable than most, and what enabled forecasters to provide exceptionally high-confidence predictions unusually early in the forecast process?

The essentials of the meteorological features that set the stage for these storms appear identifiable, though (in my experience) rather unusual. The (relatively) short story is that the basic features appear largely a function of a coincidental convergence in time and space of the strongest El Niño since 2002-2003 with specific episodes of the Pacific North American (PNA) and Arctic Oscillation (AO) patterns of the atmospheric circulation.

A composite of these three elements, as compared to the nominal configuration of El Niño, is shown in the figure below.

Top: Typical U.S. temperature, precipitation and jet stream patterns during El Niño Winters. Source: NOAA. Bottom: Hand-sketched schematic (by author) of composite El Niño, PNA and AO.

El Niño, defined by warmer-than-normal waters in the central and eastern tropical Pacific, is generally associated with a higher frequency of heavy-precipitation events and increased storm activity along the East Coast (see related study here), which reflects intensification and extension of the subtropical jet and its transport of moisture from the tropical Pacific.

A positive PNA -- characterized by a ridge (high pressure) over western North America and trough (low pressure) over eastern North America -- establishes a "highway" for atmospheric disturbances ("energy") to propagate southeastward from the Pacific Northwest to the East Coast where storm redevelopment and/or intensification may occur, enhancing the likelihood of heavy precipitation or "banding."

The unusually strong and persistent negative Arctic Oscillation, as manifest in unusually strong high pressure over eastern and northern Canada, provides the mechanism to maintain the unusually cold air over the eastern and southeastern U.S.

(It should be noted that periods of positive phases of the PNA generally recur over El Niño winters as dynamical response to El Niño. On the other hand, the negative AO is almost certainly independent of El Niño and, therefore, a purely coincidental element in the crucial mix of ingredients we've seen this winter.)

Why are some cases more predictable than others? Suffice to say, the basic science underlying this question is not yet adequately understood. Nevertheless, although no definitive conclusions can be drawn, the composite does provide some insight why Snowmageddon and (to a lesser degree) Snoverkill might have been exceptionally predictable.

First, the atmosphere over the north Pacific is typically more "stable" during El Niño winters. Second, the rate at which errors grow in forecast models tends to be less during the positive phase of PNA and, consequently, forecast skill is higher over the North Pacific, the North American and the North Atlantic regions. Think of the more stable atmosphere and slower error growth rate as trimming the wings of the butterflies somewhat.

Keep in mind that while these relationships might have been at play, it's likely that other unidentified (perhaps unidentifiable) factors were operative. It follows that while the composite El Niño, PNA and AO enhances the odds of high-confidence predictions, there is no basis to assume it's a guarantee. In the case of the Dec. 18-19 storm (not nailed down until two days out) the negative AO and El Niño moisture feed from the subtropics were in play, but the positive PNA pattern was not (see here for an explanation of how that storm happened).

Whether or not a given storm might be expected a priori to be relatively predictable, the most practically useful approach is to evaluate the level of uncertainty through comparison and consistency of the individual forecast model runs and the much larger and more representative sampling of alternative outcomes with ensembles. In Snowmageddon and Snoverkill the models and ensemble members were converging steadily beginning 4-5 days ahead in prediction of storm track, distribution and amounts of precipitation, and temperatures cold enough for snow.

Because there remains considerable room for improvement in models and ensemble systems, there remain uncertainties in predicting the level of uncertainty, i.e., confidence levels and probabilities. It's not unheard of that, especially beyond 2-3 days in advance, models and ensembles may point toward the same, but ultimately wrong direction. This is where the experience and expertise of forecaster judgment really counts when deciding whether to go with or to make adjustments to the purely model-driven predictions and degree of uncertainty. CWG forecasters strive to be reliable and trusted leaders in this area (to varying degrees of success).

By Steve Tracton  | March 2, 2010; 11:00 AM ET
Categories:  Education, Local Climate, Science, Snowmageddon, Tracton, Winter Storms  
Save & Share:  Send E-mail   Facebook   Twitter   Digg   Yahoo Buzz   StumbleUpon   Technorati   Google Buzz   Previous: Forecast: A mix of wet & white from ocean storm
Next: PM Update: Storm gets closer. Close enough?


Steve: Great article. Should be required reading!

Any way to make articles like this 'sticky'? I am thinking something like a permanent link from the front page with a library of 'education' articles explaining how weather and forecasting works. Other informative posts could go there as well, even a glossary of terms. Just an idea since we have had so many new people logging on this winter.
This kind of article is one of the reasons I began visiting CWG. Great work.

Posted by: dprats21 | March 2, 2010 11:28 AM | Report abuse

I was posting on this board the last two days about how this could job NW and what do you know the latest GFS 12Z gives I-95 from DC to PHL a couple inches of snow and NYC to Boston a bit more than that.

CWG you guys are great and I also like how you take people's posts seriously and incorporate what others find out about a storm into your overall forecast.

Posted by: ajmupitt | March 2, 2010 11:33 AM | Report abuse

And y'all do a damn good job of it too. The weather expertise here is very impressive.

Posted by: AdmiralX | March 2, 2010 11:39 AM | Report abuse

Both in the main posts and in the comments.

Posted by: AdmiralX | March 2, 2010 11:41 AM | Report abuse

Agree with all of the comments. There are so many posters who add value to this site. With a few dozen extra sets of eyes looking at models and other variables, I always get a better feel for what is coming our way.

This site is kind of like "Open source weather forecasting". LOL Thanks to all for the ongoing education!

Posted by: dprats21 | March 2, 2010 11:49 AM | Report abuse

I've noticed this season that the model and ensemble projections have been rather accurate with storm placement as far as 7 to 14 days in advance. In fact the late-January projections for Snowmageddon were hinting at possibly exceeding the total for the 1772 Washington/Jefferson storm, let alone the Knickerbocker storm.

I suspect that the detail lies in the location for where weather records are kept. Reagan National is NOT a representative weather station for Washington as a whole; it tends to under-report snow totals and also skews the official temperature data due to the low elevation and proximity of the Potomac River. Similarly Dulles tends to exaggerate in the other direction. Back in 1922, the weather records were NOT made at DCA. My hunch is that the Knickerbocker total if observed at DCA would have been closer to 18 inches than to 28 inches...hence no greater than 1996, 2003, Snowpocalypse or Snowmageddon.

In sum, official weather records should also be observed at locations other than the airports, which tend to skew the observations by their special characteristics. WGN often uses "Cubby Bear", which refers to Wrigley Field, and I've noticed that Channel 7 is beginning to use Nationals Park data. Perhaps the athletic stadiums [Nationals Park, Fed Ex Field, Verizon Center, RFK] rather than the airports would be better sites for precipitation data in particular as well as official weather data in general. The choice of airports for official weather data seems to introduce a bias of its own into the general weather data picture.

Posted by: Bombo47jea | March 2, 2010 12:14 PM | Report abuse

Latest GFS 12Z ensembles bring 0.5-0.75 QPF near DC tonight. Along with over an inch QPF for Delaware.

CWG I think it is safe to say that DC will get more than a trace out of this storm tonight.

Posted by: ajmupitt | March 2, 2010 12:25 PM | Report abuse

Posted by: ajmupitt | March 2, 2010 12:26 PM | Report abuse

Add to my previous statement that at airports [1] Paved tarmacs create their own microclimates: [2] Aircraft far larger than butterflies are continuously inputting chaos into the weather data.

At the stadiums, only the occasional fly ball, field goal, punt or center kick adds to the chaos factor; in addition far more people reside near the stadiums than near the airports. Thus the stadium is just as good an official weather station as the airport.

BTW, the presence of the Kettler Iceplex should make Ballston Mall an official weather station for west Arlington.

Posted by: Bombo47jea | March 2, 2010 12:28 PM | Report abuse

American University in DC should made official weather station or Naval Observatory.

Posted by: ajmupitt | March 2, 2010 12:32 PM | Report abuse

For those worried about the measurement station at DCA....IT WON'T CHANGE! I understand people's anger/frustration over this but that should NOT be a subject that is most commented on in this blog. CWG gave their insight and I agree with it but changing it (again) would skew future and past events.

In summary....Let. It. Go!

Posted by: DaLord | March 2, 2010 12:45 PM | Report abuse

I apologize in advance for my last's been boiling for a while.

Posted by: DaLord | March 2, 2010 12:57 PM | Report abuse

I understand DCA needs to be the official measuring station for scientific consistency, but may I suggest that CWG adopt a reliable location more representative of DC for its own "official" reporting? Over time, this will provide comparative records and who knows, maybe local media will eventually recognize the CWG site when reporting DC snow events. Just a thought to keep everybody happy.

Posted by: DOG3521 | March 2, 2010 1:26 PM | Report abuse

It might not change...but it should. Keep measuring at DCA. That is fine, to maintain the consistency of their microclimatological data that is completely unrepresentative of the area. But why can't we report basic observations like snowfall totals from somewhere else that is more representative of the larger area. I have a hard time letting go of stupid.

Posted by: rocotten | March 2, 2010 1:40 PM | Report abuse

ajmupitt, others,
.5 - .75 of RAIN, right?

Posted by: walter-in-fallschurch | March 2, 2010 1:41 PM | Report abuse

I am not sure yet about how much of that 0.5-0.75 QPF will be rain and snow. I think at least half could be snow which means 1-3 inches.

Posted by: ajmupitt | March 2, 2010 1:47 PM | Report abuse

The difference in this season's snow between the data at DCA, IAD, and BWI are ample indication (although there is a lot of other historical data) that DCA under-reports snow accumulations for the Washington, D.C. region.

Multiple news stories/TV pieces have mentioned Washington recorded an all-time record snowfall of 56" this winter, when most of us have seen 70-plus".

The argument that moving snow measurements to a more representative location will impact historical data is specious and reactionary and has been used countless times to preserve an untenable status quo. Start a new set of data. Yes, it will take awhile to get a representative statistical base to work from. But so what? It's better than perpetuating DCA's snow measurement misrepresentation.

Posted by: JerryFloyd1 | March 2, 2010 2:03 PM | Report abuse

Light rain in Manssas VA.

Posted by: FIREDRAGON47 | March 2, 2010 2:08 PM | Report abuse

When is everyone going to admit that this storm was underfocast as well. I mean just look at radars. All that precip coming north tonight when it is near or below freezing.

They have reported up to 7 inches of snow in GA and NC and it is snowing heavily in Lynchburgh and VA Tech and snowing in Charlottesville.

Posted by: ajmupitt | March 2, 2010 2:13 PM | Report abuse

Light snow falling outside of Sperryville, VA (about 30 miles west of Warrenton).

Posted by: NorthFork1 | March 2, 2010 2:16 PM | Report abuse

Come on CWG when is the new update with some accumlation maps. TWC of all sites already has 1-3 inches for DC.

Posted by: ajmupitt | March 2, 2010 2:17 PM | Report abuse


Our update will come out at 3:30 p.m. We're mulling some changes to the forecast...but not drastic ones at this point.

Posted by: CapitalWeatherGang | March 2, 2010 2:23 PM | Report abuse

Let's see what 18Z runs show and then we can tell everyone the bad news.

Posted by: ajmupitt | March 2, 2010 2:24 PM | Report abuse

Thanks CWG. I understand this is not going to be a big storm but will definitely bring accumulating snow to DC which could cause travel problems.

Posted by: ajmupitt | March 2, 2010 2:25 PM | Report abuse

CWG I really respect you guys and know you have a hard job presenting forecast to entire DC area. I am not trying to sound like I am undermining your expertise just trying to help you out and report the latest developments on a tricky storm.

Posted by: ajmupitt | March 2, 2010 2:28 PM | Report abuse

NorthFork1 - glad to have a reader in Sperryville! Thanks for your local report, too. Ps- I sure do miss the Emporium there in the Ville, from years past!

Posted by: Camden-CapitalWeatherGang | March 2, 2010 2:34 PM | Report abuse


Dude, what do you do for a living? Geeeeeez. Getting bored in that 1 room Dupont Circle Apt?

Posted by: seasejs | March 2, 2010 2:38 PM | Report abuse

Light rain in Ballston.

Posted by: mcaicedo | March 2, 2010 2:40 PM | Report abuse

Just so you know I am off today. Just got back from a two week long verficiation in asia. So before you start assuming that someone is lazy ask them. Why does it bother you that I am posting about weather on CWG on my day off. I don't even live in Dupont Circle. I hate that place. You dont look any busier than me if you are on here to criticize me. At least I am offering so intelligent information.

Posted by: ajmupitt | March 2, 2010 2:43 PM | Report abuse

I'm happy to see there are at least a few folks out there who are interested in a post that is not exclusively focused on the current forecasts, especially since it's extremely unlikely that the rain/snow tonight/tomorrow will amount to much around here. Then again, maybe there's a surprise in store.

Posted by: SteveT-CapitalWeatherGang | March 2, 2010 2:48 PM | Report abuse


Alright, calm down. It's not so much that you are lazy, it's just the constant pounding about this storm. I am totally tired of snow, so the snow cheerleaders are wearing on me...I am hoping for all rain. Yeah, Dupont Circle is scary.

Posted by: seasejs | March 2, 2010 2:51 PM | Report abuse

So you are still assuming I am lazy. You got some real nerve man. If you dont like snow then get out of the north man. You cant control it. If you are not going to offer any intelligent weather analysis and just complain about the weather then get off the board. You don't see CWG complaining about weather.

Posted by: ajmupitt | March 2, 2010 2:54 PM | Report abuse


Get out of the north, thought this was the mid-atlantic. Anyway, the people on this board are getting a little bent out of shape cause they are running out of snow!

Posted by: seasejs | March 2, 2010 2:58 PM | Report abuse


This isn't the north. It's the mid-atlantic.

Posted by: ThinkSpring | March 2, 2010 3:01 PM | Report abuse

I'm not going to argue with you. Hope you enjoy complaining and criticizing eveyone. Defintiely shows a lack of intelligence and character. I'm done wasting my time and yes it looks like you ran me off the board and CWG has lost someone that can provide analysis and gained you.

Posted by: ajmupitt | March 2, 2010 3:03 PM | Report abuse

Here in Reston its a rain snow mix right now.

Posted by: justin-N-sterlingVa | March 2, 2010 3:04 PM | Report abuse

I smell tweakage in the form of more snow being added to the 330 update....models look too juicy! 8-)

Posted by: GMorg11 | March 2, 2010 3:12 PM | Report abuse

If you go to Cecil County, I definitely feel like it is the "north". DC could be the north. I consider the south beginning in southern Maryland and from about Fredericksburg and points southward. I base that on geography, since it is where the familiar southern landscape of pitch pines, Virginia pines, and loblolly pines, mixed with areas of sandy soils, begins to appear.
In cultural terms, it is harder to say. But I do know that people from the deep South wouldn't even consider someone from say, Richmond, a southerner. Either way, who cares. This bickering is silly. Let's all get along...and hope for more snow (or not...that's cool too). And I'll stop talking about anything non-weather related too...

Posted by: BH99 | March 2, 2010 3:14 PM | Report abuse

Interesting article. I second the "sticky" suggestion made by poster dprats21. And if you do start a sticky feature, at the top of the list should be Andrew's (IMO) classic piece on the global variations in weather while we were having a cold early January. I found it absolutely iconic in pointing out that the world is bigger than your region, and pointed many people to it -

Posted by: B2O2 | March 2, 2010 3:15 PM | Report abuse


snow lover/snow hater - this is about weather including discussion of the latest models. No need attack people. If the discussion of snow is so distressing to you, perhaps you should take a break.


hang in there some of us enjoy observations on developing weather. Rain, snow or whatever comes.

Posted by: Rmjw | March 2, 2010 3:18 PM | Report abuse

walter, ajmupitt, others...0.5-0.75? Interesting..would think some of that might be snow. But hopefully that falls quick and can knock temps down a bit. Interesting that even Atlanta is getting snow, seeing how they're not much above sea level either. Only thing is that it's definitely coming down a lot harder there, which we would of course need to see happen. I'm going with an inch or two of slush on the grass around DC. It's of course no expert opinion- just fun to make a guess..any more guesses anyone?

Posted by: BH99 | March 2, 2010 3:21 PM | Report abuse

ajmupitt - Like how on this board you say you respect CWG and on another bust them for being too conservative. Little transparency please.

Posted by: bodypolitic1 | March 2, 2010 3:24 PM | Report abuse

Snowing mildly and 38.8 degrees in Hamilton VA.

Posted by: worldtraveler83 | March 2, 2010 3:29 PM | Report abuse

@ajmupitt, don't abandon CWG. You add value to this blog. Just blow it off.

Lighten up folks. What someone does for a living is not germane to any discussions about weather forecasting (unless they are a meteorologist). Remember, you're here, too. If someone adds value, who cares what they do or who they are?

Light rain/snow in the Laytonsville, MD area. Looks like it cannot decide what type of precipitation it wants to be.

Posted by: dprats21 | March 2, 2010 3:29 PM | Report abuse

BH99, re .5 - .75 qpf: ajmupitt said "snowing heavily" already in blacksburg, right? radar looks like it's coming towards us, but...what do i know?

is this the kind of thing that COULD "blow up" over us? or is that possibility reserved for philly/ny/boston etc...?

guess we'll see.

Posted by: walter-in-fallschurch | March 2, 2010 3:32 PM | Report abuse


I wasn't questioning either what ajmupitt does or where he lives - what he says. I don't think it's cool to imply you are only here to help and then criticize CWG elsewhere. And using the same screen name at that.

Posted by: bodypolitic1 | March 2, 2010 3:33 PM | Report abuse


Thanks for backing me up...people forget we are not New Hampshire(normal year). Look at Garrett County - 300 inches of snow manure! What a hardship this winter has imposed on so many people, I forgot what is was like to go out on the weekends. I am with you and we need to start our own board.

Posted by: seasejs | March 2, 2010 3:34 PM | Report abuse

NWS posting at 3:14 p.m. still going with mabye an 1"

Posted by: steske | March 2, 2010 3:38 PM | Report abuse

Steve - thanks for the discusion. Do you have any thoughts on whether putting out probablilities for specific snowfall accumulations has any legitimate scientific basis or genuine value? As you note "precision in forecasting the details of snowfall amounts and variability across the D.C. area is not possible." Is a "forecast" that gives a range of possibilities any more precise? For example, if the science "isn't there" to forecast 2" to 4" of snow with any degree of confidence, is the science any better at forecasting a 10% chance of 2" - 4" of snow? I would suspect the level of uncertainty would be even higher when you attempt to attach a specific probability to a range that can't be forecast. By contrast, I can see how varying model parameters to produce a measure of the robustness of the prediction would have value as a reflection of the overall confidence of a forecast. If small changes lead to dramatically different results, one could determine a probability that reflects the liklihood of the forecast outcome. That doesn't, however, appear to support assigning specific probabilities to accumulation potentials. The former seems to be a quantifiable reflection of uncertainty, the latter nothing more than someone's guess. Doesn't the assertion that both are reflections of "probability" based forecasting simply confuse an objective basis for communicating uncertainty with a subjective point of view?

Posted by: manatt | March 2, 2010 3:41 PM | Report abuse

Looks like dewpoints have risen above 32 in most locations.

So, does that mean that snow will have to be heavy to stick?

Posted by: ThinkSpring | March 2, 2010 3:44 PM | Report abuse


Not to worry. This is it for the year. I'm positive.

Posted by: ThinkSpring | March 2, 2010 3:45 PM | Report abuse

@bodypolitic1: I in no way meant to group you in there. I actually was writing my comment before you posted, so I only saw it after I hit the post button. What you pointed out is a valid criticism, if that is what he/she does. Sorry for any miscommunication.

Posted by: dprats21 | March 2, 2010 3:51 PM | Report abuse

with all that discussion i forgot to tell you what a great article that was. it was nice and technical, but not too technical - it was juuust comprehensible to me, which was perfect.

it also shows how many things went right this year.

Posted by: walter-in-fallschurch | March 2, 2010 4:22 PM | Report abuse

Excellent analysis, Steve. Very interesting on relationships between El Nino, PNA and AO, particularly AO being random with respect to the other two.

Equally interesting (to me at least) would be a piece containing your thoughts on recent (past 20-30years) and likely future advances in predictive abilities overall.

You seem to be saying in this article that complete precision in long range forecasting is not possible.

"precision in forecasting the details of snowfall amounts and variability across the D.C. area is not possible (and may never be -- blame it on the butterflies)." and "The longer-range predictions provide greater opportunity for the butterflies -- and the 'chaos' they iconically represent -- to do their mischief. They have not, nor will they ever be completely stilled."

These comments seem a bit offhand, probably because this is not the subject of the article.

If it remains true that advances in scientific knowledge are, by definition, exponential, and that successful prediction of the effects of even the fluttering of Lorenz' butterflies wings requires only much greater scientific knowledge (i.e. more powerful computers and more data), then why is it not only a matter of time, and not really all that much time if you apply "exponential" to the actual advances we've seen in meteorology over the last 3 decades, until great storms can be predicted in great detail a great many days into the future?

Also, on a more immediate and equally interesting matter, it is now snowing with 3/4 mile visibility and 32.6 degrees near Wolfsville, MD! Dare we hope that total predictability for this storm remains far in the future and we will actually get one last big hit this winter?

Posted by: dustygroundhog | March 2, 2010 4:24 PM | Report abuse

Hey Camden,

Actually, the Emporium is still open (slightly, but not for long)...Faith Mountain has closed.

We're now getting a light rain/snow mix.

CWG is terrific. As someone who lives in a very rural area, in a "holler" between two fingers coming off of the Blue Ridge, with our own little microclimate, I find the discussions here helpful and interesting, although I wish some of the "bring-it-on" crowd would simply lighten up. I get the fact that if you don't like snow, move to somewhere near the equator...but if you live where I do, and you have a critically ill relative who depends on you to get to NVA for health care, sometimes enough is enough.

Posted by: NorthFork1 | March 2, 2010 5:18 PM | Report abuse


Your questions are a bit difficult to interpret, but here's my best take:

First, by all means probabilities of specific snowfall accumulations have a legitimate scientific basis and can have genuine value. The issues, it appears, are the degree of specificity and what constitutes useful.

Weather forecast models are limited in their "precision" to a large extent by the model resolution. The NAM, for example, cannot differentiate features smaller than about 20-30+ miles. Hence, assuming there is not a rain versus snow issue, the model can barely differentiate snowfall within a few inches between DC versus Dulles, let alone across just across the immediate DC region. And, even if the model predictions were perfect, the details of snow accumulating on the ground will depend upon local conditions such as elevation and closeness to the Potomac (like DCA). On the other hand, within these bounding limitations, it is possible to reasonably estimate the probabilities of snowfall in categorical intervals of +/- a few inches. It is certainly justifiable to discriminate probabilities between a dusting, a few inches, and what I've called a "big one", > 8-10 inches.

Whether or not this information is useful is up to the user of said information. If you could care less whether a storm produces a light dusting with just wet roads or a raging blizzard, the weather forecasts have no value. If you are an emergency manager who must decide what actions must be taken to ensure life and property, the value is genuine and potentially invaluable.

The probabilities, as opposed to just a yes or no, are a tool for use in decision-making in a cost/benefit sense. Given a certain probability for snow greater than X", for example, city managers must judge whether it's better to prepare and have the snow not occur, or not prepare and pay the piper if it does snow. The decision, of course will depend on whether "X' is 2-4" versus a foot or more.

Posted by: SteveT-CapitalWeatherGang | March 2, 2010 6:11 PM | Report abuse


The atmosphere and it's interaction with land and ocean surfaces are incredibly complicated and interdependent. Every parameter on every time and space scale interacts with the others with a mix of positive and negative feedbacks. Suffice it to say that modeling the relevant physical and dynamical processes is a massive challenge. There has been considerable progress over the years in understanding the relevant processes, but at best they can only be approximated in numerical models.

Even if we had a "perfect" model, predictability would be limited. Yes, the butterflies due to uncertainties in formulation of this model would be killed off. But, it is not possible to perfectly observe and define the initial state from which the models run, and errors in inital conditions (butterflies) would eventually limit predictability.

Also, all models must be constructed on a finite grid which limits their resolution. Even if the resolution were down to the molecular level - and we could describe the inital state perfectly - there would still be sub-molecular scale processes operative in the real, but not model atmosphere (tiny, but not insignificant butterflies) which would grow and ultimately lead to error in the forecasts.

Posted by: SteveT-CapitalWeatherGang | March 2, 2010 6:59 PM | Report abuse

I recall the March 1993 snow storm which hit on Friday night into Saturday was starting to be forecasted on the previous Sunday and Monday as a huge storm, and it never changed.

Posted by: irish031 | March 2, 2010 7:37 PM | Report abuse

Thanks Steve - I agree useful planning information can be gained from resources like the NWS precipitation probability outlook that highlights, for example, areas with a greater than 70% chance of seeing a foot or more of snow. I see much less value in a forecast that claims to put probabilities on specific ranges from zero to infinity, and even less when there is little difference between the ranges. A 25% chance of less than 4", 4-8", 8-12" or 12"+ has no practical value. By comparison, a forecast that "4-8" of snow is "expected" implies a 50% liklihood the accumulation amount will fall in that range and seems reasonable. I think one goes too far to suggest they can predict a "10% chance" of more than 12" of snow. It's a guess, not a forecast, and implies a level of precision that doesn't exist.

Posted by: manatt | March 4, 2010 4:28 PM | Report abuse

The comments to this entry are closed.

RSS Feed
Subscribe to The Post

© 2012 The Washington Post Company