6 July 2015

“Weather by Icon” Is A Bad Way To Get an Accurate Forecast

Posted by Dan Satterfield

A group called CustomWeather made this forecast for timeanddate.com. It's basically worthless. It's mainly icons and forecasts beyond 5-7 days have little skill.

I found this forecast on timeanddate.com. It’s basically worthless,since It’s mainly icons, and forecasts beyond 5-7 days have little skill.

Every meteorologist will tell you the same thing. The most often heard comment they hear when they tell someone what they do for a living is something along the lines of “It must be nice to get paid to be wrong most of the time!”. My usual response is that my batting average is far above that of most major league baseball players, and they make a lot more than I do! In reality, I am right most of the time, and my batting average (and that of most forecasts around the world) is quite high for a forecast out to 48-72 hours.

SUNNY WITH A 60% CHANCE OF CONFIRMATION BIAS
There are several reasons why many people think otherwise, and the main one is that people are far more likely to remember when the forecast is wrong than they are the 90% of the other times when it was right! Older people likely also base their erroneous belief in the fact that forecasts 30-40 years ago were much less reliable than they are now, and make no mistake, forecasts have improved dramatically in that time. A recent paper published in the Quarterly Journal of the Royal Meteorological Society looked at forecasts for Melbourne Australia, both by numerical weather model and by forecasts produced for the public. This is from the Abstract:
The accuracy of the current official Day 5–7 forecasts is found to be similar to that of Day-1 forecasts from 50 years ago. The accuracy of experimental Day 8–10 forecasts is comparable to that of the Day 5–7 forecasts, when they were first officially provided 15 years ago. Some overall skill, albeit limited, is evident out to Day-14 and significance testing indicates that it is unlikely that this apparent skill arose by chance. 

Numerical weather prediction is getting more accurate, with Europe continuing to lead the U.S. From Stern, H. and Davidson, N. E. (2015), Trends in the skill of weather prediction at lead times of 1–14 days. Q.J.R. Meteorol. Soc.. doi: 10.1002/qj.2559

Numerical weather prediction is getting more accurate, with Europe continuing to lead the U.S. From Stern, H. and Davidson, N. E. (2015), Trends in the skill of weather prediction at lead times of 1–14 days. Q.J.R. Meteorol. Soc.. doi: 10.1002/qj.2559

I suspect that you would see a similar result for most of the U.S.,Canada and Europe, and perhaps nearly worldwide, and much of the increasing skill is due to faster supercomputers and more sophisticated numerical weather prediction models. Globally the best weather model continues to be the European Centre for Medium Range Weather Forecasting’s ECMWF model which has on average substantially better skill than NOAA’s Global Forecast System (GFS) model. NOAA’s new 3km resolution HRRR model is very good though for short-range, and at a weather conference I attended last month, an overwhelming majority of forecasters present indicated they would gladly give up one of the 4 daily runs of the WRF model (4 daily runs to 84 hours) for a longer run of the HRRR (it currently runs out to 15 hours on an hourly basis).

A perfect forecast is still likely to be misinterpreted by many, and some of that is their fault, while some of it is our own doing. Frankly, some folks do not take the time to look closely at what the forecast actually is, and in some cases we broadcast meteorologists do not make it easy for the to do so!

Screen Shot 2015-07-05 at 10.39.49 PM

These are some of the new NWS icons that will be introduced this week. Notice also the yellow “caution” color for the flood watch. Europe has used color coded warnings for some while now, and it needs to become widespread here. These icons were produced from significant research by NOAA.

WEATHER BY ICON
The 7 Day Forecast is a main-stay of TV weather reports and there is good reason, the public loves them. It’s by far the most important graphic I show, and a 7 day forecast graphic is also posted online on almost every TV station website. The NWS has developed a new set of weather icons for their website (that will debut this week) as well, and this is just fine if you want a basic idea of what to expect over the next few days. If you need something better though, it may not be so easy to find. The NWS provides a text forecast with the icons online and I produce the same for the TV station I work for here in Maryland.

I did a random browse of several TV station weather pages and found some are quite good about putting a text forecast online with more information, while others were seriously lacking. If you present a 7 Day forecast after a 3 minute chat about the weather, then the public has some context to interpret it by, but many times, we get blamed for a bad forecast by someone who based that judgement on a few weather icons they saw on an app or on a web page.

So with that in mind, here are some rules to remember when looking for a decent accurate local forecast:

1. IGNORE hour by hour rain chances you see online. We do not have the skill and the model reliability to do this, in spite of the fact that places like weather.com do so. We can give you perhaps an average rain chance over a 4-6 hour period out to 3 days or so, but that depends on the day.

Hour by hour forecasts (like this one by weather.com) of rain chances are worthless most of the time.

Hour by hour forecasts (like this one by weather.com) of rain chances are worthless most of the time.

 
2. Read the text forecast, and if there isn’t one, watch the video. If there is neither for your LOCAL area, go somewhere else.

3. Forecasts beyond 7 days are almost always worth less than you paid for them.

4. Beware of computer generated forecasts, they are almost always lousy. Forecasts of temperature, humidity, rain chance and winds on an hour by hour basis are almost certainly raw guidance from a computer that has not been interpreted by a meteorologist. weather.com and accu-weather are two sites that use this data online, and you are wasting your time looking at them.

5. Do use sites that have a 7 day graphic with a text forecast. That gives you the best online forecast, but beware sites that just put a word or two, that may be computer generated. The image below shows a text forecast I posted with the 7 Day graphic last week:

Screen Shot 2015-07-05 at 9.55.57 PM

6. Get the forecast for the right location. Weather can vary greatly, especially due to elevation and along the coast! This is where computer generated forecasts often go badly wrong. I’ve often seen computer generated forecasts show the same high temperature for two places 25 miles apart, but one was on the coast and the other inland. The coastal forecast is often badly wrong by over 10 degrees F.

Icons are great, but they cannot tell you that the rain chance is only 30% after 4 PM, and that a strong sea-breeze will develop around 1 pm and drop the temperature 12 degrees!