OpenSnow Data: Inflated Forecasts and Incorrect Actuals (24 Hour & History)

[ChrisC]

I have given up on OpenSnow for anything resembling accuracy for Europe, South America, New Zealand, etc. Basically, it gives a rough estimate of whether it will snow.

Most of the time, the following is incorrect:

• Snowfall Estimates - to be expected
• Last 24 Hour Snowfall - no actual resort snowfall from websites is reported, with the exception of advertisers Vail and Alterra
• Historical Snowfall - this data is either 25-50% of actual, or sometimes 300% of actual. I wonder how often it is within one standard deviation?

There is such a heavy reliance on inaccurate data estimates that are fed into and recycled through its model that it is a bit insane. And to ask subscribers to pay 100% more $50/yr to $100/yr for AI-reports? The entire site is one big AI hallucination.

What pisses me off is that they often have the 'real' data fed to them via APIs, likely as a subscriber. OpenSnow has the actual data for most major European weather sites. For example, Gornergrat is a great representative site for ski conditions at Zermatt; this data is reported to OpenSnow. However, they never take a 24-hour new-snow reading or a current snow-base reading. WTH? Zermatt is just one example; there are many ski resorts that they neglect.

The only value of OpenSnow is when one of the Experts interprets snow and weather data from multiple sources and provides a written forecast.


Today's Example: New York State from the late January winter storm for Monday, 1/26/2026 (last 5 days - basically just Jan 25 & 26th snow totals).

Interesting, Mount Peter and Thunder Ridge are reporting big snowfall! Suprised Plattekill did not receive more than its neighbors.

However, when you look at the actual snow report from these ski areas nothing Open Snow reports is accurate:

Mt. Peter - 18"
Thunder Ridge - 16"
Plattekill - 29-34"

So almost all of the highest snowfall areas in the Northeast are inaccurate and over-reporting. And it's not just a transcription or double-entry error.

I used to think there were reporting issues more internationally. However, unless you are a Vail or Alterra resort, the data will most likely be incorrect, sometimes considerably so. Therefore, I would not rely solely on OpenSnow for Snow Report Display functionality. Seems so basic, especially for $50/yr.

This is so bothersome to me because a lot of my career has been making sure SaaS tech companies have accurate, real-time data. I was always looking at various sites and databases to make sure all data was accurate and correct - it was not my job (more NetOps, QA, Account Mgt, etc) - but if you are financing $100 millioins to billions, you could not lose interest calculations/revenue for incorrect data synchronization. So I do not have much patience for a chronic issue.

I generally like NOAA better in the USA.

 

[Harvey]

Suprised Plattekill did not receive more than its neighbors.

As predicted, this was not a lake event:

Of the four Catskill hills, Plattekill gets the most snow from lake enhancement and the least from synoptic or coastal events.

Open Snow is expensive, and totals are way exaggerated, but it's extremely useful.

I want to know where it's going to snow the most, and when. If they say 36" and I only find 18, whatever. You know that going in, so it really doesn't matter.

 

[Tony Crocker]

Everyone should understand that forecasts are just that and may not be accurate. I applaud all of the OpenSnow experts emphasizing that point and even refusing to forecast specific snowfall numbers more than 5-7 days out due to their speculative nature.

ChrisC is complaining about past snowfall reports. Readers will expect that PAST numbers are real, despite the small print caveats at the bottom of the monthly snowfall calendars. When so many of them are wrong, and often by huge amounts that people can call out, that can undermine the credibility of an otherwise great product.

To fix that probably means a hands-on human validating that you are getting the same number that’s on a ski resort’s website, and perhaps custom programming to get that number. I’m guessing that OpenSnow is now a large and successful enough enterprise to make that effort worth considering.

 

[ChrisC]

I would be so happy if OpenSnow could just report actual snowfall data and snow bases - with some OK-ish forecasts.

And not always be off by a factor of 50% or 250% for historicals. Sigh.

... but it's extremely useful.

I want to know where it's going to snow the most, and when. If they say 36" and I only find 18, whatever. You know that going in, so it really doesn't matter.

One would think actual snowfall reporting should not be off by a factor of 2x for perhaps the largest storm of the year.

And it does matter that snowfall reporting is accurate. For many skiers with a multi-mountain pass, a 2x difference in snowfall does affect decisions.
I could see choosing one ski area over another depending on 6 vs. 12 inches of new snow, or 3 vs. 6 inches, or 18 vs. 36 inches. One could result in a good day, another in an exceptional/all-time day.

To fix that probably means a hands-on human validating that you are getting the same number that’s on a ski resort’s website, and perhaps custom programming to get that number. I’m guessing that OpenSnow is now a large and successful enough enterprise to make that effort worth considering.

I worked with a company called Yodlee in the 2000s, which just did financial screen (and web) scraping and data aggregation before RESTful APIs, etc. It went public in 2014 and is now the backbone of most of Fidelity's personal account reporting and other financial institutions.

My point, especially with AI and offshore labor, is that writing a few screen-scraping or, more appropriately, web-scraping tools is an already-solved problem and could be done cheaply.

I might put together a top-10 list of ski reports I am interested in and work with the engineering team to create accurate historical and 24-hour snow reports. I feel like this issue was solved 20-25 years ago.

 

[jasoncapecod]

I have given up on OpenSnow for anything resembling accuracy for Europe, South America, New Zealand, etc. Basically, it gives a rough estimate of whether it will snow.

Most of the time, the following is incorrect:

• Snowfall Estimates - to be expected
• Last 24 Hour Snowfall - no actual resort snowfall from websites is reported, with the exception of advertisers Vail and Alterra
• Historical Snowfall - this data is either 25-50% of actual, or sometimes 300% of actual. I wonder how often it is within one standard deviation?

There is such a heavy reliance on inaccurate data estimates that are fed into and recycled through its model that it is a bit insane. And to ask subscribers to pay 100% more $50/yr to $100/yr for AI-reports? The entire site is one big AI hallucination.

What pisses me off is that they often have the 'real' data fed to them via APIs, likely as a subscriber. OpenSnow has the actual data for most major European weather sites. For example, Gornergrat is a great representative site for ski conditions at Zermatt; this data is reported to OpenSnow. However, they never take a 24-hour new-snow reading or a current snow-base reading. WTH? Zermatt is just one example; there are many ski resorts that they neglect.

The only value of OpenSnow is when one of the Experts interprets snow and weather data from multiple sources and provides a written forecast.


Today's Example: New York State from the late January winter storm for Monday, 1/26/2026 (last 5 days - basically just Jan 25 & 26th snow totals).

Interesting, Mount Peter and Thunder Ridge are reporting big snowfall! Suprised Plattekill did not receive more than its neighbors.

View attachment 49092

However, when you look at the actual snow report from these ski areas nothing Open Snow reports is accurate:

Mt. Peter - 18"
Thunder Ridge - 16"
Plattekill - 29-34"

View attachment 49094


View attachment 49093

View attachment 49095



So almost all of the highest snowfall areas in the Northeast are inaccurate and over-reporting. And it's not just a transcription or double-entry error.

I used to think there were reporting issues more internationally. However, unless you are a Vail or Alterra resort, the data will most likely be incorrect, sometimes considerably so. Therefore, I would not rely solely on OpenSnow for Snow Report Display functionality. Seems so basic, especially for $50/yr.

This is so bothersome to me because a lot of my career has been making sure SaaS tech companies have accurate, real-time data. I was always looking at various sites and databases to make sure all data was accurate and correct - it was not my job (more NetOps, QA, Account Mgt, etc) - but if you are financing $100 millioins to billions, you could not lose interest calculations/revenue for incorrect data synchronization. So I do not have much patience for a chronic issue.

I generally like NOAA better in the USA.


View attachment 49096

LOL...i was in the Catskills today..15" Same as my house..10 miles north of TZ
Mt Peter did not recieve 44...Hard no. same with Thunder Ridge...hard no

 

[ChrisC]

LOL...i was in the Catskills today..15" Same as my house..10 miles north of TZ
Mt Peter did not recieve 44...Hard no. same with Thunder Ridge...hard no

I know, I know: it's snow-reporting. More marketing and art than science and accuracy.

 

[ChrisC]

Again, another reason why I think OpenSnow should not be spending time on AI initiatives, and increasing plans from $20-30/yr to $50/yr, and New-AI plans to $100/yr.

OpenSnow has a simple data collection problem. New snowfall is always incorrect. Year-to-date snowfall is always incorrect. Model estimates are wildly off, ranging from 25% of the actual to 300-400% of the actual new or accumulated snow.

All of its AI initiatives should be suspended immediately, and the R&D should be put to data collection. The numbers fed into its current model are wrong, and accurate real numbers are used to correct or improve the model.


One example, very straightforward today.

Zermatt received 14 cm/ 6 in on 1/28/2026 according to the Swiss Gvt snow-site:

OpenSnow lists 2 inches of new snowfall - about 66% off.

My issue is using a bad model when they even have access to the actuals.

The Gornergrat Zermatt station is reporting 13F.

Even the nearby station at the top of Monte Rosa on the Swiss-Italian border is a good source of actual snowfall data. They could possibly use this as a check on Monterosa ski resort (IT) estimates or Zermatt(CHE) for a sanity check.

Again, OpenSnow has a severe data accuracy issue and should not be building AI models on very poor-quality data.

 

[snowave]

yep, too automated with not enough human input.

As many here know, I've been complaining for a long time about their extremely poor (overestimates) for Brundge Mtn, ID. This year was the first time I've seen one their (new) "local" forecasters acknowledge that fact in the "daily snow". However, it's still had no effect w/regards to forecasting or reporting numbers.

 

[ChrisC]

Yes. I am seeing the Local Experts completely ignore the OpenSnow model and estimates. As if it did not exist.

They look at the raw weather data, the American, Euro, and other models, and add their own expertise. The output - often in contrast with the automation.

The local experts are the best thing about the site at this point.

I cannot rely on forecasts, overnight/24-hr actual snowfall, or historical actual snowfall.

It does not work as a data aggregator or comparison tool; I have to get to each resort website. Or NOAA Weather.

 

[Tony Crocker]

The local experts are the best thing about the site at this point.

They have always been the key selling point, differentiator vs. other forecast sites and the reason I subscribe.

All of that data is elegantly presented, easy to navigate and equally comprehensive on mobile as on desktop. Whoever Joel's IT guy is, he understands presentation and user friendliness. Readers will be attracted to the data and want to use it. So that makes it even more a sore thumb that so much of it is inaccurate.