eduggs

You are indirectly supporting a point that I have been trying to make for years that long-range averaged anomaly charts are poor tool for identifying regional snow threats and likelihoods. They work OK for 10-14 day regional temperature forecasting. In northern Japan and Hokkaido, you can often see snowy periods coming 2 weeks in advance. But for us, snow is primarily correlated to 500mb features that are not identifiable at a time and space scale resolvable at 2 week lead times. This leads to a hype and disappointment cycle based on pretty looking anomaly charts.

Yup. But this is true of most climate indices and telecons, especially during peak winter climo. The nuances of the 500mb evolution and configuration are key. Each combination is unique, and a minor feature can occasionally lead to local snow (particularly as you gain latitude and elevation) regardless of longwave pattern and telecon state.

Cold relative to last winter, sure. Even compared to the 10-yr average. I guess the statistics say compared to the recent 20 or 30 year average, the metro area has been slightly below normal cumalatively (using simplistic max plus min divided by two for daily departure). But compared to any longer term reference this has not been a cold winter. And I think there's a good chance we end up above normal when all is said and done. Take Newark as a local example. There was a 2-3 day cold stretch in January and 1 day in December. All other days this winter have been near or above freezing.

Yup. Medium to early-long-range longwave pattern recognition using ensembles is pretty good for airmass forecasting, okay for anticipating general regional storminess, and poor for local snowstorm forecasting. The averaging effect of ensembles especially towards the end of the usable range masks the critical shortwave details that drive sensible weather outcomes. And superimposed model error further compounds the difficulty.

Yeah Sapporo snow is crazy. If you look at the same number of days it took Boston to get 110", Sapporo probably has 15-20 or so such periods since record keeping began in 1953. And the really crazy thing is there are places very close to Sapporo that get 2 or 3 times as much snow. Sapporo City is not in a heavy snow zone.

Boston does not feel like a big city. But yes, EMA has a large and dense population. Sapporo City on the other hand feels bigger - with a city center that feels more like Manhattan than Los Angeles.

Sapporo City gets very cold and frequently frigid. But the city itself is warmed by the Sea of Japan on NW winds, so it's a bit warmer than inland/mountainous areas. Overall, Hokkaido is colder than SNE.

FWIW, Sapporo is a bigger city than Boston. Obviously metro sizes depend on where you draw the borders, but Sapporo also feels significantly bigger than Boston. And although it isn't particularly snowy compared to the immediate surrounding area, it has received significantly more snow during many months than BOS did that winter. But Japan is kind of in a category of its own.

Yeah it really does hurt. The storm looks like a sure bet down in the Gulf. That's why I mostly look at 500mb and sometimes 250/300mb. The surface is just a response to the upper levels. It's easier to anticipate things falling apart when you look at the upper level charts.

Yup. And it's usually a bad forecast. I'm okay with forecasts of 1-2" or 2-3". Winter QPF is generally forecastable with a higher precision than 0.1 - 0.3". The cases where a 1-3" range occurs locally is usually with high liquid to snow ratios where the fluff factor in isolated bands can significantly increase the local snowfall variability. Regardless, I think it's bad practice to forecast a snowfall range that spans the gap between nuisance and solidly plowable. Sorry for the off-topic...

3" feels like a snowstorm, especially if it's cold and accumulates easily on all surfaces. Most people will be happy with that. 1" usually feels like a nuisance event if hoping for a moderate event and will leave most people disappointed. 1-3 is not a thing. It's an old-fashioned, lazy, range forecast.

Really pretty daytime snow today. But I was surprised it got a little warmer than I expected. A little above 30F now with full clouds. Some melting going on.

Should be a good burst for most of NYC metro over the next 30 minutes or so.

The 1/9 - 1/11 event seems to still be in play despite being a bit of a long shot. I like the shortwave diving south through the upper mid-west midweek. The end of run ICON looked workable but I'd sure like to see some improvement to the GFS & CMC this run. Trof axis is okay but obviously too much confluence as modeled through New England and the Maritimes. We've seen that resolved favorably on occasion through the years.

Every day presents a new configuration of weather features. It's not like we have a perfect standing wave jetstream creating a static "pattern." Snow doesn't come from "patterns." At the local scale, snow is loosely correlated with a set of global ocean-atmosphere-climate indices that are themselves somewhat vaguely defined. It is the nuances of the features, their evolution, and interactions; the details... that determine whether we get snow. It's the "pattern" obsession that leads to so much disappointment. Especially for the people who use long range 500mb height anomalies to try to sniff out favorable periods. Those charts always look better than reality because of the averaging effect that smooths out the critical details. They can sometimes portend temperature trends, but are much less useful for long range snow prediction. Most of us are still in the game for Monday.

Totally agree about single-run caution to avoid possible windshield wiper effect. But this was a really good cycle (plus 18z ECM). All medium range models and ensembles are improved - and it's a meaningful improvement. At this time range I think holding serve two more cycles or so would signal a really good chance of a light to moderate snowfall locally.

I'm mildly intrigued by the 3rd - 4th on the GFS, but minimal ensemble support and lots of disagreement between the other models. No clarity in weenie range when day 7 still isn't close to being resolved.

Brooklyn posts a lot in several subforums and has a red tag. Few people provide a counterpoint to his posts despite their sometimes repetitive and simplistic nature. Non-red-taggers get constant push back so I don't feel it would be constructive to police all their posts.

Really nice plot. The correlation between PNA+ and NYC snowfall jumps out. But one thing that a stickler would say is missing from the analysis to make it statistically robust is an accounting of the relative frequency of time spent in each quadrant. For example, hypothetically if 95% of winter days occurred with an AO- PNA+ combo, observing a majority of significant snow events with these parameters would not be telling. What you're getting at is the relative likelihood of significant snow on any given day with each combination. And for that, some kind of normalization of the data is required.

Gotta marvel at the 12z CMS with 3 rain events and another just past the end of the run as we approach peak winter climo. Fortunately other mid-range guidance starts to differ early in the run.

Yes please occasionally post something other than toggled 500mb anomaly charts that you think show a favorable "pattern" for cold and snow. Better yet, spend some time on analyzing why so many of your highlighted patterns fail to portend cold and snow. Once you have a better handle on that, you will be a much better mid- and long-range forecaster. I'm not sure if this applies to you, but I really think younger forecasters are at a disservice having grown up in the era of tropicaltidbits, pivotweather, and various plotters to view historical data and "patterns." Without a deep understanding of statistics and a more intuitive feel for forecast uncertainty that comes with age and experience, the ease of data access gives youngsters the impression that they understand things that they really don't understand.

Well Brooklyn primarily posts about "favorable" signs - usually mid or long range ensemble mean anomaly charts... over and over again. It's been like that for a few years now if my memory servers me correctly. That seems kind of weenyish to me. Sometimes I appreciate it... but other times a bit more objectivity is warranted. Presumably he will acquire more balance with more experience.

Yeah exactly. In hindsight it's easy to identify what went wrong. But model ensembles 10-15+ days in advance miss key details.

That's slightly more optimistic than I would mentally lock in. But statistically it's not unreasonable.

Nothing on the 12z ECM out past 10 days either. Well other than rain.