OceanStWx
-
Posts
20,222 -
Joined
-
Last visited
Content Type
Profiles
Blogs
Forums
American Weather
Media Demo
Store
Gallery
Everything posted by OceanStWx
-
It's pretty close up here that the 10th percentile is actually my best event of the season so far. Obviously I'm putting it all on red though and hoping for something bigger.
-
Don't forget all the bodies choking Boston Harbor from the Tobin.
-
Who inside 95 wouldn't take that and run though?
-
I'm not sure I could draw it up any better in a textbook. Naked toboggan runs through the neighborhood here.
-
The main issue is that a consensus forecast is almost always better than an individual forecast. And that goes for models too. A single GFS or Euro op run may occasionally beat the ensemble, but more often than not it won't. Even for dynamic situations it will offer a range of solutions more likely to capture the end result.
-
Here's a visualization from Alicia Bentley, who has done a ton of work with the GFS. You can see the GEFS are at least as useful as the op through day 3/4 then it becomes clear the ensembles are better. Obviously an ensemble mean high temp may not be particularly good on day 2, but for the upper levels there's skill still.
-
You can see it pretty well with the QPF density functions from the 00z EPS. At the lower end (yellow arrow) the values have actually gotten lower than previous runs. But at the higher end (red arrow) the values have gotten higher than the last run. So the bad members are getting worse, and the good members are getting better.
-
Just speaking about the EPS, the 10th percentile for you has moved to 0 over the last two days, but the 90th percentile has moved up to 14. Like Ryan has said, the spread has actually increased. I don't know about the Euro/EPS specifics, but at least with the GFS/GEFS the ensembles on average typically provide a better forecast for day 2 even. 24 hours is pretty much op or bust time for the GFS.
-
As others have mentioned it looks like two clusters in the EPS, which could be increasing spread despite there only being two "right" answers. To me it looks like the EPS got a little worse aloft in the mean, so I would expect probabilities and whatnot to come down slightly. Or the day shift was only 50% confident you might get 6 inches.
-
Just gets the mid levels going way too late. Like closing things off near or north of the Cape.
-
There are some QPF queens getting ready to toss their monitors.
-
I mostly stick to the Cobb on there too. Locally (PWM) it looks like the EPS has been steadily trending lower with QPF for the last 5 runs. 90th percentile only changed about 3 inches (17 to 14 inches give or take), but the 10th percentile (basically expect at least that much) went from 5 inches to 0. Shows you how the lead pipe lock has been taken off the table.
-
I would honestly be fine if step 3 was just keep the ratio 12:1 Cobb for those that don't know uses temp, RH, and lift to determine a snow ratio. Not perfect but probably the most scientific.
-
It's honestly fine near freezing, it will melt your face off when the temps are cold.
-
Really deepens the 700 mb low right over the tip of the Cape. Just stacks all the upper lows right there. That would be tremendous forcing, so no wonder it gets precip back to the CT River up here.
-
I might start drinking and I'm still at work.
-
But it matches model runs over the last 24 hours, and that's what these probabilities are based off of. The reality is that we probably aren't able to narrow the range down to something smaller than 0-18" at this point.
-
And I tease Kevin about Kuchera. It can be really useful in situations that include mixed precip. But an all snow, especially cold snow, event is going to be . His forecast high Saturday is 18, that will probably be the max temp below 500 mb. That gives him a Kuchera ratio of 18:1. I wouldn't be banking on that.
-
They used to be, but satellite retrieval techniques and aircraft obs have significantly improved over time. There was a time when 12z might beat 18z consistently, but now that's just not the case. 18z is almost always more accurate than 12z. Again that's for closer range forecasts, as beyond days 3/4 all models are a bit of an unmanned firehose. Not 100% sure, but I believe so.
-
Good thing they are typically under-dispersive. Aside from the more hi-res stuff, all the models clearly ingested up on something today and moved as a result. Now recon starts tonight and satellite retrieval of the northern stream should be better by tomorrow, but if it's still meh at 12z I'm ready to just accept an advisory event to ruin a run at futility.
-
This is demonstrably false. Each new model run is on average more accurate than the last when inside 3 days. Now beyond that models in general just aren't THAT good at nailing down outcomes.
