Hailing from Buffalo, NY, Chris Morin has been skiing at Crystal Mountain Resort since 2005, worked on Crystal’s Ski Patrol team for over a decade, and loves to ski powder. Chris eats, sleeps and dreams skiing. But now that he has a 9am-5pm desk job as the Director of Analytics at Eddie Bauer, Chris created a system to predict the best powder days, so that when he “comes down with a cough,” it is a well spent “sick day.” Better yet, the forecasting system allows Crystal’s operations team to better predict the forecast. We see it as a win-win.
The Crystal operations team has decided to close the upper mountain for the day due to expected high winds. You’re thinking to yourself, how can they predict the high winds so far in advance? Well, thanks to Chris, we now have access to some of the most detailed and near perfect forecast predictions available for our area.
We got down to “brass tacks” recently with Chris about his forecasting system and Crystal specific forecast called The Brass Tacks. We discussed how the forecast helps Crystal Mountain determine operations, and in turn how that helps Crystal’s guests.
What inspired you to make your own forecasting system?
Chris: I used to work on Crystal’s Ski Patrol and many times I would have to make decisions on whether we should go out for avalanche control at 6am, 7am, or not at all. During the day we had to decide if we could get Northway, Southback, or even Mountain Top open. These decisions were driven by the weather, and I would look at 10 different sources of information trying to recreate the wheel. I wanted to provide a powerful system with a simple output that could reduce the uncertainty and time spent deciphering the weather. Despite the heavy tech involved, the goal is to give its users more time in the mountains and less time staring at screens.
How long have you been providing these forecasts to Crystal’s operations team?
Chris: The system was first operational 3 seasons ago (winter of 2014/15), but in the first season it was still a work in progress.
How does this system help Crystal Mountain determine operations?
Chris: The two big impacts are on avalanche control and lift operations. Crystal can be more effective in their avalanche control, open terrain faster, and reduce the avalanche risk for their guests with better weather forecasts. Wind hold is not very common at Crystal, but there are times when they can’t get the upper mountain lifts spinning safely in high winds. While I can’t reduce the winds, I can provide a better trend report. If I can accurately forecast a downtrend, Crystal can keep the staff on the mountain to open the lifts later in the day. On days where it looks super bleak, Crystal can communicate to guests what to expect.
How can this system, that is specifically focused on Crystal Mountain, help Crystal’s guests?
Chris: Being transparent with guests is increasingly rare but highly valued in any industry. Crystal is leading in trying to give it’s guests the most accurate information available. There is no bias in the forecast, it is all math and machines. With a better forecast, you can hit more good days and shred more pow, which is the whole point.
Alright, well how do you create the forecast?
Chris: I pull in weather data created by a model called the global forecasting system (GFS). This is the numerical model that many of the point forecasts you see today are based on. I use this data as a starting point to run a separate weather model, called the WRF, over the northwest. This runs at 100 times the resolution of the GFS, producing 10s of billions of forecasted values every day. A machine looks at these values and compares them to the output of a single sensor at a specific weather station at Crystal. Over years of looking at the models output and the sensor, the machine learns to forecast for that sensor. This learning is then scaled to every sensor at each weather station. The machine learns and forecasts weather just as a human would.
You receive readings for your system from the NWAC telemetry stations at Crystal. If those aren’t reporting accurately, how does that affect your forecast and how to you combat that?
Chris: We had some issues early in the year with this. We moved the base weather station and the precipitation can stopped working, so we put in a new one, and of course that one also wasn’t working. The third time was the charm. During the struggle, I built in processes that will keep the forecast running even if weather stations go down, however it will still impact the text forecast during the closest time frame. The computer reads the weather in real time to continually evolve the forecast for the day. If the precipitation measurement is broken and it’s reading zeros, it may start decrementing the expected snowfall even if it’s dumping out. If the precipitation sensor blows up and starts reading thousands (which it did at one point), the current day forecast is going to be pretty messed up. In the future we’ll hopefully build in better controls to cover this as well.
On the topic of lot’s of snow, Mt. Rainier receives some of the largest amounts of snowfall each winter, how does being so close to Mt. Rainier affect your forecasting?
Chris: It’s difficult to forecast in complex terrain. At Crystal, Mt Rainier heavily impacts all weather systems coming from the SW, which is the dominant direction. A change of a few degrees in wind direction can be the difference between a block with no snowfall and a convergence with heavy snowfall. Global models are not high enough resolution to capture these small-scale effects, but the high resolution models I run are starting to resolve some of this detail. Unfortunately, they still struggle with the finer points. The machine learning makes up for some of these deficiencies, and the combination of strategies produces the best possible output.
Why is your forecast better than the other forecasts for the Crystal Mountain area?
Chris: I think “better” can be a bit subjective. Statistically it is the most accurate. It is updated the most often (hourly). It has a highest time resolution (hourly). The data view is powerful in that it combines many variables from both historical measurements and forecasts, but may be more difficult to understand if you aren’t used to looking at snow-water amounts in that manner. The text forecast is easier to understand, and is more detailed than other text forecasts, but again this detail is a little more involved to read than looking at a snowflake icon. I believe it’s the best skiing weather forecast for most users.
Why does the weather of this region fascinate you so much?
Chris: “Because when it snows the skiing is good and when the winds are light the flying is good.”
Spoken like a true powder hound that wants to make the most of his days in the mountains. For more info on Chris’s forecast visit http://winterscience.com/brasstacks/. To contact Chris about questions or to just say “That’s rad, thanks!” please email him at email@example.com.