Winter Science in the PNW

[pipedream]

(others seem to be reacting well to it, so maybe it's just me...i'm no scientist)

As a scientist (of computers), I also find it somewhat confusing. Perhaps separate plots for the predicted elements (temp, SWE, etc.) would increase clarity?

Also where are those sexy "running total precip" graphs from?

[::: :::]

I want to add that I'm excited and interested in the proposed winter weather prediction site cmor is working on. I don't want to derail the thread, so please ignore me if this is unwanted advice. A suggested read for folks who present information for a living: Edward Tufte's "The Visual Display of Quantitative Information." Great book that talks about how presentation of data is almost as critical as the data.

[cmor]

I want to add that I'm excited and interested in the proposed winter weather prediction site cmor is working on. I don't want to derail the thread, so please ignore me if this is unwanted advice. A suggested read for folks who present information for a living: Edward Tufte's "The Visual Display of Quantitative Information." Great book that talks about how presentation of data is almost as critical as the data.

It's good feedback. I have invested thousands of hours into the underlying data, but the UI was a single two day push to get something together. There's a number of limitations with parts of the framework and I had to invest in the easiest solution for a number of cases.

The SWE 72-48-24 etc. takes up entirely too much space. The charts are a good start but I may try separating, there's a few ideas I have to combine the precipitation and temperature in other ways. I can add more information such as those running totals, but I don't want too much on the page. Maybe I need another verification page for each site or something.

[cmor]

I made some changes to the charts which should make them a bit easier to read, still need to redo the SWE blocks.

Crystal Mountain stepped up and funded a significant increase in the processing power being used. We're now able to go out to a 4 day forecast and updates are coming out faster. Super stoked to have them on board, especially considering my processing bill far outpaces my rent bill every month.

[John_B]

I'm liking the website. Still working on getting it into my daily list of things to check. AT least on my computer I have a hard time getting the Wx Map to show the northern part of the state. I'm in Winthrop, so really interested in our area and Washington Pass. Maybe the map would be improved by having a few more landmarks on it? Maybe just adding county lines or major highways?

Unfortunately for me I can't really use your forecast due to the Washington Pass weather station being down. Would it be possible to still have a weather forecast there despite the station being down? Thanks for the work on the website.

Let me know if you start collecting money for your processing bill.

[cmor]

Thanks for the feedback.

I made significant changes to the layout that hopefully ease reading. Added cloud cover to the maps. Made a host of other improvements. I agree the fx should not go down when the weather station goes down. It's just a monster amount of data transformation to even get to the point it's at. Once I have a minute I'll rejigger to fix.

We've also had enough days of output to compare the machine learning with actual forecasters. It compares very favorably.

There's still plenty of work to be done on the UI. The machine learning still falls down at times, especially in a few specific scenarios that require further research. I do have updated models that will go live shortly and improve further. That said, I believe it to be the most accurate mountain forecast made, human or machine.

[Norseman]

Right on.

[John_B]

Awesome. I keep forgetting to bookmark this but I'm glad to hear it's working better and better.

[Norseman]

Been using the site a fair bit recently.
A couple of suggestions:

1. Animate the time steps with a toggle play/pause button, and allow users to set the time window and animation speed.
2. Add numbers to the gradations on the color scales to aid in quickly understanding the quantities represented.
3. Include one more map scale, zoomed out a touch. The current one is great for detail in the Oly's and WA Cascades, but cuts off the northern Oregon volcanoes, plus Whistler and Baker... all common areas of interest for the skiers of the region.


Thanks

[cmor]

Been using the site a fair bit recently.
A couple of suggestions:

1. Animate the time steps with a toggle play/pause button, and allow users to set the time window and animation speed.
2. Add numbers to the gradations on the color scales to aid in quickly understanding the quantities represented.
3. Include one more map scale, zoomed out a touch. The current one is great for detail in the Oly's and WA Cascades, but cuts off the northern Oregon volcanoes, plus Whistler and Baker... all common areas of interest for the skiers of the region.


Thanks

1. The slide bar is pretty cheeseball. I will try to update that into something better.
2. Colorbars are also cheeseball. Will update.
3. This one is tough. I'm already producing over 25,000 images every day just for the maps on that page. I keep searching for alternatives to make the maps more interactive, tiling rasters or rendering data in browser, but neither have been able to get the quality I would like. It's hard to scale spatially and temporally at the same time. I don't have an answer yet.

[::: :::]

any chance at tracking the snow elevation?

[cmor]

I can pull the freezing level off the raw WRF and make the simple assumption that the snow level is 1000 ft below that. Unfortunately there's nothing to learn off to go further than that. At any given point in time we don't really have any instrumentation in place that can directly measure the snow level. Machine Learning is stuck in kindergarten on this one.

I have worked on getting better information for the passes. It's very difficult to give a point forecast during an east flow event with such huge gradients over short distances. Sometimes the overall pattern is more meaningful. I've been interpolating vertical cross sections along the DOT mile markers to achieve this. These aren't straight lines, it's as if you stretched out the highway. The below is what I-90 looked like last week for one of the forecast hours, should be easy to see the east flow and inversion. The purple dots are freezing rain, yellow is snow, yellow-green a mix. Green is rain. Vectors are wind. I'm running White Pass, Stevens Pass, 14 through Hood River, and the North Cascades highway right now in a similar fashion. Should be public in a week or so when it's a bit more cleaned up.

[Norseman]

Outstanding. Thank you for your work.

[Zeedashbo]

Holy shit. You are an awesome human

[wyeaster]

that is really cool

[cmor]

Cross Sections are live:

http://winterscience.com/cross/Cross_Section_of_2/

They are going to evolve into something a bit more mobile friendly, but I wanted to get up what I currently have. I forecast avalanche hazard for BNSF rail at Stevens and Stampede pass and these have proven invaluable to have day to day. I was at Snoqualmie yesterday and was pretty happy with the accuracy. One caveat on the i90 cross-section, the west side isn't interpolating far enough down to road level. Therefore it's given snow too far down the road at this point, as it's really forecasting for up to ~700' higher elevation. I'll also include a how to read on the page eventually.

Been using the site a fair bit recently.
A couple of suggestions:

1. Animate the time steps with a toggle play/pause button, and allow users to set the time window and animation speed.
2. Add numbers to the gradations on the color scales to aid in quickly understanding the quantities represented.
3. Include one more map scale, zoomed out a touch. The current one is great for detail in the Oly's and WA Cascades, but cuts off the northern Oregon volcanoes, plus Whistler and Baker... all common areas of interest for the skiers of the region.


Thanks

Haven't gotten all of your suggestions, but the overall nav is much better. Give it a look

[::: :::]

Outstanding

[abraham]

Seriously. This is awesome and so helpful.

[John_B]

Better and better! What unit are the profiles in? I was having a tough time understanding them.

[cmor]

John_B, I added some colorbars with units. Also blew it out to the volcanoes and made some rad wind and cloud visuals. For roads the x-axis is mile-markers, these are stretched out representations. For the volcanoes they are straight lines and are kms. Check out the wind over Rainier through Crystal:

http://winterscience.com/cross/Cross...ystal/cross_w/

[Norseman]

Man, this is killer!

Question about the cross sections... R-xtal screenshot for example...



How are we to interpret the surface precipitation graphic? I think I understand that the colors represent precip types as with your "bumps" beta version, with yellow as snow, green as rain, etc, and this lines up with the dashed line freezing level. What about the thickness of the colored surface layer? Does it represent an hourly rate?

[cmor]

Man, this is killer!

Question about the cross sections... R-xtal screenshot for example...



How are we to interpret the surface precipitation graphic? I think I understand that the colors represent precip types as with your "bumps" beta version, with yellow as snow, green as rain, etc, and this lines up with the dashed line freezing level. What about the thickness of the colored surface layer? Does it represent an hourly rate?

Yes, an hourly rate. I switched from a scatter to a raised area, I thought it was a bit more intuitive. The minus is that it looks more like accumulation which it is not. The thicker the faster the rate, so in your pic there is heavier snowfall at 3PM west of Rainier.

As of the last run it was too aggressive at showing precipitation on the cross sections, essentially you need a cutoff for this view (else it will show .00001 inches), and I don't have it set quite right. The next batch completes in 4 hours, I made some changes to clean it up.

[Norseman]

Ok, thanks.

The choice of map scale between W coast, WA/OR and Crystal is great, too. Do you plan to put out a focused map for the other main zones? I imagine Mt Dweebus had to come first, being a spancer, eh? 😁

The site is really getting good. Bravo.

[cmor]

Ok, thanks.

The choice of map scale between W coast, WA/OR and Crystal is great, too. Do you plan to put out a focused map for the other main zones? I imagine Mt Dweebus had to come first, being a spancer, eh? ��

The site is really getting good. Bravo.

Crystal is my home mountain, hence the inclusion. The maps solution is not scalable. I mentioned the issues previously, but long term I need to find a better option to make it more flexible.

[Norseman]

Just a teasing jab.