The Grand Illusion
I’ve noted earlier that weather forecasts for Cedar City tend to be hit or miss, possibly because Cedar City is roughly fifteen miles north of Black Ridge, and Black Ridge is the southern end of the plateau on which Cedar City is situated. South of Black Ridge, the ground drops close to three thousand feet in less than thirty miles.
I understand the difficulties this poses for forecasters, especially since Cedar City is not exactly a major metropolitan area, but as I write this, it’s been snowing consistently for the past six hours, and we’ve gotten about seven inches of snow, and it’s still falling.
All the forecasts say it’s partly cloudy and that we’ll have scattered snow showers.
I’ve lived in New England at the foot of the White Mountains, in Colorado at the foot of the Rocky Mountains and here in Cedar City, essentially between three mountain ranges, and in none of those places would seven inches of snow be considered intermittent snow flurries or showers.
Last night, Cedar City was supposed to have flurries. We got about three inches of snow.
I understand that the location of Cedar City makes forecasting difficult, but still stating that it’s partly cloudy with possible snow flurries as the snow continues to fall strikes me as either a continuing reliance on unreliable algorithms or incompetence, if not both. It’s one thing to miss a forecast; it’s another to report the current weather wrong – continually.
Or perhaps it’s just that none of those so-called meteorologists even bothered to check with any of the 50,000 -60,000 people who are experiencing those “scattered snow showers,” because algorithms are so much more accurate than real people, not to mention, cheaper.
The problem is not with the algorithms for weather prediction. It’s with the algorithms for dealing with customer feedback.
That’s why it’s impossible to point out to the weather predictors that a look out the window might be useful.
It’s very difficult to have an good automated system for customer feedback. But as you say, a lousy automated system is so much cheaper than paying a person. So it’s an easy target for penny-pinching.
Most automated customer feedback systems are like the “suggestion box” in comic strips… they’re positioned on top of a trashcan.
My guess is that much of the reporting and forecasting is automated and not reported by ‘eyes on’ at all. You could try some different weather sites. “Weather underground” can be good as they use many combined local small home-owned weather stations and combine the data. Windy.com might also be useful but would require you to interpret some of the voluminous data yourself.
I now live along the Oregon coast and see that weather predictions here are equally challenging, with micro-climates being packed tightly due to the terrain. There’s a lot of moisture coming from the ocean and it hits abrupt elevation changes within a couple miles or less due to the coastal range. Given that I like to cycle and that this activity is impacted by the weather in a strong way, I would also like to have predictive forecasting (as I had for many years in the midwest). But alas, I’m coming to accept the vagaries and unpredictability of this local weather and am now emebracing it. Happy days. Variety is the spice of life.
Weather Underground is great. I use an app for San Francisco and the Bay Area called ‘Mr Chilly’ that pulls data from WU stations and predicts temperatures for the different parts of the city. (Believe it or not, but you can have a 10 degree difference across parts of this tiny city, and a more than 30 degree difference a half an hour outside the city.)
Maybe there aren’t enough (any?) automated sensors nearby; or maybe they have trouble measuring snowfall.
There are some weather measurements at KCDC (Cedar City regional airport). I can’t tell whether they’re automated or human made or verified (a regional airport might not operate 24//7). You could call them or the area weather service office in Salt Lake City to find out what measurement practices apply in the area.
https://forecast.weather.gov/MapClick.php?lat=37.70675&lon=-113.09695#.Y_nvNi-B08Y
Dark Sky used to be great for hyper-local weather; but Apple bought them, and you won’t see their data anymore except with the app on a Mac, iPhone, iPad, etc.
Most algorithms are fine. The problem is they get applied to situations for which they are not meant.
Algorithms are only as good as the underlying assumptions and follow-on data gathering can make them – I don’t think people knowingly design bad algorithms (I won’t claim to know anything about meteorology and I won’t claim that insurance companies have anything other than their bottom line at heart) – the problems are mainly on the user end.
In medical algorithms, they frequently get bent into the shape that the user needs. An algorithm for ‘chest pain’ generally assumes that you’re looking for heart attacks – using it for pain caused by trauma, pneumonia, pulonary embolism, etc. is not going to give you the information or outcomes you need.