What's the Road Condition, Really?
tl;dr The MDT sensor I thought was giving road condition isn’t, and I’ve built and will continue to refine something better.
Your car’s thermometer can’t measure the most important thing.
It reads 36 and you relax. Meanwhile the pavement under your tires is 30, and the rain about to hit it will turn to glass. That gap is not unusual. Roads are often significantly different in temperature than the air.
Now that there are enough drivers using the app to produce real data I can start getting it ready for fall and winter conditions…
The dry chip
Until recently “How’s the Canyon?” showed a small chip on the drive screen that read MDT: Dry. It came from Montana DOT’s road condition feed. I put it there because it seemed obviously useful. Official source, current conditions, perfect.
Then I checked the data I’d collected because all I was ever seeing myself was dry.
I have thousands of readings of the feed covering back to when I started working on this idea. All four segments covering our corridor read dry in every single one. Not most. Every one. Statewide, across 544 segments. Even when I watched strong squalls dump rain on the golf course out my window and head into the 191 corridor. (The feed produced exactly one weather statement in the whole month, on a piece of US-191 inside Yellowstone, for one day.🤷♂️)
Over the same 31 days, MDT’s own precipitation sensor on MT-64 recorded 29 separate rain episodes. The chip above it said Dry through all of them.
I know this state is real because I’ve relied on it in the winter. But it seems this condition doesn’t get set in summer? I was rendering it as though it were a live reading. D’oh. My error, not theirs. So it’s gone until and unless it starts showing useful data.
That leaves a hole where the road conditions used to be. Happily what I figured out while looking for a replacement should be way better.
MDT is already measuring the useful thing
Montana runs 118 road weather stations. They report every fifteen minutes and they are public. I have been pulling the whole list down for a month and archiving it for future alert modeling.
95 of those 118 have a working sensor buried in the pavement.
One of them sits on MT-64 at milepost 7.2, up Lone Mountain Trail at 7,290 feet. Air temperature, humidity, wind, precipitation type and rate, and road surface temperature: 2,997 readings out of 2,997, no gaps. It is a perfectly healthy road weather station on our corridor.
There’s a second one 20 meters off US-287 at Norris Hill, on the Madison side. That’s the road that matters on the days the canyon closes.
Why the two numbers come apart
Great home made pizza is better with a stone. You set the oven to 500 and the air gets there in about ten minutes. The stone takes an hour. And it’s the stone that decides whether the crust chars or comes out soggy, because air is light and changes fast, and mass is slow and sucks up energy.
Asphalt is mass. It’s black ‘rock’ lying on the ground, soaking up sun all afternoon, then radiating its heat straight out to a clear night sky while cold air drains down the canyon walls on top of it.
So the two temperatures wander apart, and they wander apart the most on the days and nights we care about most. On a clear, calm night the surface runs 2 to 5 degrees colder than the air (University of Washington, roadway icing: Roadway Icing and Weather). Ice can be forming while your dash says 35.
In the other direction it can get extreme. Across those 2,997 readings, the pavement at Big Sky Road ran a median of 14 degrees hotter than the air, and at the high end so far, 46 degrees hotter. This afternoon while I was writing this: air 78, road 112.
Four combinations
Once you have both numbers, they stop being two facts and become one. There are four cases and they are not the same problem.
Air above freezing, road above freezing. Wet if it’s raining. Normal.
Air above freezing, road below. Rain freezes on contact into clear glaze ‘black ice’. This is the worst one and most likely to catch people off guard.
Air below freezing, road above. Snow melts as it lands and the road is just wet. We see this all the time. Eventually pavement catches up with the air and water becomes ice and snow sticks.
Both below freezing. Snow bonds and builds. Dangerous but obvious.
There’s a fifth state that is rare at altitude because of the dry air, but needs no precipitation at all. When the pavement drops to both freezing and the dew point, in still air, water vapor deposits directly onto the road as frost, and traffic polishes it into black ice. Clear sky, dry forecast, nothing falling, and a slick road. That’s a really useful time to hit the wet/ice button on the drive screen of the app so others are forewarned.
And of course we get lots of days where it warms just enough to melt, and then refreeze. What was a road in reasonable condition becomes slick and dangerous again. Last fall it felt like that happened every day for six weeks.
Ice is most slippery between about 26 and 32 degrees, and gets grippier down in the single digits, because the thin film of liquid water on top of it freezes solid. So the coldest night is not the worst night.
What’s on the screen now
The pavement temperature is in the app, for everyone, today. You don’t need to update anything. (And an improved display will be on the next release.)
It reads as what it is: a measurement, from a named instrument, with the time it was taken. Not a forecast and not a classification. When the road and the air are doing something you can’t see from inside the car, it says so plainly and leaves the driving to you.
I also will be able to determine the freezing level, interpolated across five weather stations that run from 5,391 feet up to 9,200, against a road that climbs from 4,705 at Four Corners to 7,290 at the top of MT-64. When they bracket freezing I can tell you roughly where on the road the snow line falls.
With all this working correctly the app can provide current modeled state of the road, updated every five minutes.
The 22 miles I can’t measure (yet)
Here’s what I can’t do, and hopefully this is fixed before it gets cold.
There is exactly one MDT instrument physically inside the canyon dead zone. It’s at Karst, milepost 55.3. The pavement sensor there has never reported a single reading while the app servers have been monitoring it.
Its air temperature and humidity probes went dark on July 27th and came back at 10:30 in the morning on August 11th. (I’ll be asking MDT. It isn’t critical until we hit our first day with freezing weather this fall.)
So: two working road thermometers on the corridor, and neither is in the 22 miles where your phone doesn’t work. Filling that gap means modeling it, and modeling it honestly means training against real sensors through a real winter first. I have three and a half years of hourly weather history for the corridor and a growing pile of readings from all 95 working pavement sensors in Montana, most of them on mountain passes with the same physics as ours. That’s the work for this winter.
Spread the word about the app, the more people using it the better the real time information will be for all of us. I still have a couple months left to finish modeling road conditions so the system can push alerts out before you start driving!