You may know Nate Silver: he was the data analysis guru on the New York Times who had an amazing track record calling recent elections. He is big on sport's statistics as well. Then he started his own web site: 538.com.
Anyway, he and Reuben Fischer-Baum looked at the weather statistics at 120 cities supposedly to learn how unpredictable the weather was around the country. In reality, they looked at the variability (root-mean-square-deviation from climatology) at these stations, which is NOT necessarily the same as how predictable the weather is. These folks are not meteorologists, so you can excuse them.
Let me show you their maps. First, for an average of maximum, minimum, and daily average temperatures:
Variability (their predictability) is least along the West Coast and Florida. Variability is huge in the upper Plains. It all makes sense: the temperatures on the West Coast are mellowed by the Pacific and the cold air from the interior is blocked. On the other hand, the Plains are buffeted by surges of cold air during the cool half of the year. Southern Florida generally does not get winter storms and is surrounded by warm water.
For precipitation, the Northwest still does ok. Variability is low in the DESERT Southwest. Hey, that is why it is a desert....it doesn't rain. Much more variability in the east where they get lots of thunderstorms during the warm season as well as wintertime fronts and heavy snow near the Great Lakes.
According to Silver's article, Seattle has one of the most predictable (least variable) climates in the country. We do tend to be mild and drizzly much of the year.
But let's do this right! How does the skill of National Weather Service forecasts vary around the nation? In fact, I did a paper on this with Jeff Baars, a research scientist at the UW. But instead of presenting my tables, let me show you some nice graphics produced by ForecastAdvisor.com for 2013 (see below) for high temp, low temp, and precipitation. Other years would probably be similar.
First, high temperature. Maximum temperature looks similar, but not identical to Silver's results. Less accurate in the Upper Plains and more skillful in Florida and the interior southwest. But in this case the Northwest and the California coast are less accurate, possibly due to forecast busts due to the proximity of the coast, where temperature varies rapidly.
Low temperature is a different story. More skillful along the West Coast, which makes sense since even on warm days during summer we tend to cool back to roughly the same temperature. The same is true of the Southeast, which tends to be uniformly warm and humid during the summer with minimal variation in the low temperatures.
For precipitation, forecast skill (rain versus no rain) is high in the Southwest and even in our area (not surprising considering how reliable our rain is). Less reliable near the Great Lakes and over Florida where they are lots of thunderstorms. It looks like the Puget Sound convergence zone is hurting skill in our area!
The National Weather Service has also analyzed precipitation skill regionally. Here is a sample. The West Coast (green bars) has the most skillful precipitation forecasts in the nation! You can be proud of that. I know that I am.
I read the article yesterday on fivethirtyeight.com and instantly did not like their definition of predictability. At least they clearly stated how they were defining it. They were just wrong.
ReplyDeleteCliff - Is it possible to take this a step further and analyze predictability by season as well? It seems like spring can sometimes be an unpredictable time in the Puget Sound area, but I wonder if the data would play out like that.
ReplyDeleteCliff- It's only fair: Your next post should detail your predictions for the 2016 Presidential race.
ReplyDelete