For most of my life, I’ve heard weather forecasts in the news be described as infamously unreliable. Has there been any serious advance in this field? Is the unreliability of weather forecasts just another of those too-popular memes about things we love to hate?
For some of my life (till the 70s or 80s), weatherman jokes were a staple of humor, and the joke was simply that the weatherman was wrong. I haven’t heard a weatherman joke for a long time.
The bizarre thing is that I’m sure the jokes existed, but I can’t remember any of them.
Nate Silver (of 538) has some space that he’s dedicated to this effort in The Signal and the Noise. Randal Olson’s reproduced some of that related to current-day abilities, which show that we’re currently able to give better-than-random results for a few days in advance, but not much better after that. And, unsurprisingly, data beats expertise when it comes to accuracy.
A good deal of the data-collecting tools have been developed or implemented relatively recently, and that seems to correlate with improvements to short-term forecasting, to the point where a five-day forecast in 1991 was roughly as likely to be accurate at a three-day forecast in 1981.
They’ve improved enough that they can probably be trusted to determine whether you should bring an umbrella tomorrow, but the historical numbers and especially expertise-based numbers were inaccurate enough to explain the origin of the meme.
Nate Silver’s The Signal And The Noise has a chapter about this. The short answer is yes,, weather forcasting has gotten better, but comerical forcasts have a known “wet bias” in favor of predicting rain. The reason for this is that people get more upset at forcasters when they say it won’t rain and it does than when they say it will rain and it doesn’t. Acording to Silver, the National Weather Service’s forcasts are the most reliable, followed by various large comercial services (e.g. weather.com. etc.), with local news forcasts being the least reliable.
Disclaimer: I am not a meteorologist, but a friend of mine is and I had discussed this with him some time ago. Paraphrasing from memory, I might make mistakes.
Only in recent decade or two we got enough computational power to run huge weather models (such as Aladin) comfortably, with live updating on incoming data. However, the model is only as good as the input data—if the meteorological weather stations are positioned every few kilometers, the prediction is extremely good, but if they are more sparse, the errors increase. The complexity of terrain plays a role too—on a flat plain, weather stations might reasonably be spaced tens of km’s, but a small hill means they have to be spaced much more closely to get reasonable results.
There is also sometimes very helpful “local knowledge”—e.g. whatever the abrupt weather change in Germany, you can be reasonably sure in 3 days it will happen in Slovakia.
Taking this into account, professional weather forecast is very reliable for a day or two—enough to leave your umbrella at home. However, interpretation by news forecasters glosses over finer details, such as “sunny, high temperature with 10% chance of rain” will be interpreted as “sunny, high temperature, a little of rain” and give false signals.
The weather forecasts really were rubbish in the past.
To give one notorious example, a weatherman on Britain’s main television station once mocked a member of the public’s claim that the country was about to be struck by a hurricane. That very night, Britain was hit by one of the most severe hurricanes of modern times, killing 18 people.
For most of my life, I’ve heard weather forecasts in the news be described as infamously unreliable. Has there been any serious advance in this field? Is the unreliability of weather forecasts just another of those too-popular memes about things we love to hate?
For some of my life (till the 70s or 80s), weatherman jokes were a staple of humor, and the joke was simply that the weatherman was wrong. I haven’t heard a weatherman joke for a long time.
The bizarre thing is that I’m sure the jokes existed, but I can’t remember any of them.
I’ve heard jokes to that effect and I was born in 1990.
Nate Silver (of 538) has some space that he’s dedicated to this effort in The Signal and the Noise. Randal Olson’s reproduced some of that related to current-day abilities, which show that we’re currently able to give better-than-random results for a few days in advance, but not much better after that. And, unsurprisingly, data beats expertise when it comes to accuracy.
A good deal of the data-collecting tools have been developed or implemented relatively recently, and that seems to correlate with improvements to short-term forecasting, to the point where a five-day forecast in 1991 was roughly as likely to be accurate at a three-day forecast in 1981.
They’ve improved enough that they can probably be trusted to determine whether you should bring an umbrella tomorrow, but the historical numbers and especially expertise-based numbers were inaccurate enough to explain the origin of the meme.
Nate Silver’s The Signal And The Noise has a chapter about this. The short answer is yes,, weather forcasting has gotten better, but comerical forcasts have a known “wet bias” in favor of predicting rain. The reason for this is that people get more upset at forcasters when they say it won’t rain and it does than when they say it will rain and it doesn’t. Acording to Silver, the National Weather Service’s forcasts are the most reliable, followed by various large comercial services (e.g. weather.com. etc.), with local news forcasts being the least reliable.
Disclaimer: I am not a meteorologist, but a friend of mine is and I had discussed this with him some time ago. Paraphrasing from memory, I might make mistakes.
Only in recent decade or two we got enough computational power to run huge weather models (such as Aladin) comfortably, with live updating on incoming data. However, the model is only as good as the input data—if the meteorological weather stations are positioned every few kilometers, the prediction is extremely good, but if they are more sparse, the errors increase. The complexity of terrain plays a role too—on a flat plain, weather stations might reasonably be spaced tens of km’s, but a small hill means they have to be spaced much more closely to get reasonable results.
There is also sometimes very helpful “local knowledge”—e.g. whatever the abrupt weather change in Germany, you can be reasonably sure in 3 days it will happen in Slovakia.
Taking this into account, professional weather forecast is very reliable for a day or two—enough to leave your umbrella at home. However, interpretation by news forecasters glosses over finer details, such as “sunny, high temperature with 10% chance of rain” will be interpreted as “sunny, high temperature, a little of rain” and give false signals.
The weather forecasts really were rubbish in the past.
To give one notorious example, a weatherman on Britain’s main television station once mocked a member of the public’s claim that the country was about to be struck by a hurricane. That very night, Britain was hit by one of the most severe hurricanes of modern times, killing 18 people.
Given not complete reliability I would really like to get weather data via an app that gives me confidence intervals of various sizes.
Such as SHMUdroid (for Android)? Unfortunately, these apps tend to be country specific.
I don’t speak that language so it’s hard for me to tell but probably.
I’m from Germany