July 2nd, 2024

Weather forecast accuracy across US cities based on Weather Service gridded data

The Washington Post analyzed U.S. weather forecast accuracy, showing Miami's week-ahead precision contrasted with Paonia's 6°F errors. Forecasts impact agriculture, transportation, and safety, with coastal and arid areas proving more predictable.

Read original articleLink Icon
Weather forecast accuracy across US cities based on Weather Service gridded data

The Washington Post analyzed the accuracy of weather forecasts across the U.S. based on government data. The study revealed that in Miami, forecasts are accurate up to a week ahead, while in Paonia, Colo., even the one-day temperature forecast can be off by almost 6 degrees Fahrenheit on average. The National Weather Service's forecasts power emergency alerts and various commercial weather products, impacting sectors like agriculture, transportation, and public safety. The analysis highlighted that coastal regions and arid areas like the southwest tend to have more predictable weather patterns, while regions like the Great Plains experience more variability due to air mass interactions. The Weather Service blends data from multiple computer models to generate forecasts, with recent advancements in computational technology improving forecast accuracy over the years. Despite the rise of AI-powered forecasts, human forecasters still play a role in refining predictions based on local knowledge and conditions. The study emphasized the importance of accurate forecasts in saving lives and guiding various industries reliant on weather information.

Link Icon 8 comments
By @oldmariner - 10 months
I think there was a HN post a few weeks or maybe a month or two ago that showed which weather forecasting service was more accurate in general by comparing predictions to actual. I think the winner was Microsoft's with about a 70% accuracy (with one or two others only a percentage point or two behind). Does anyone else remember this or have the link?
By @ibejoeb - 10 months
The "weather" discussed in this piece is just temperature.
By @mathsmath - 10 months
I find their error model to be off for my location. I'm in a mountain valley in Colorado, and forecasting the weather is like throwing darts at the wall (especially with precipitation).

We have no radar coverage, and the only "official" sensor I know of is at our local airport. It was recently upgraded to be somewhat accurate, but isn't operational about half the time. The most accurate forecast we have is the text synopsis from a forecaster in Pueblo talking about how it "might be stormy in the afternoons this week."

Our nearest city is a 2.5 hr drive, so we're in a pretty big gap of weather coverage!

Anyway - hope someone finds this interesting. I envy those of you with accurate weather! It has been interesting moving out here.

By @therabbithole - 10 months
By @therabbithole - 10 months
The author of that article also makes the data and code for some of the interactive data visualizations available via Observable notebooks: https://observablehq.com/@climatelab/map-temperature-forecas...
By @pxx - 10 months
my initial reaction: wait who cares about the temperature forecast? it seems like the easier part and the not particularly-important part.

> The Weather Service also assesses its precipitation forecast, but the patterns are a little harder to unpack — especially with just one year’s worth of data — so we stuck with temperature.

that seems like flimsy justification. I'd love to see the lead plot with precipitation.