The weather forecast accuracy, or rather the persistent lack of it, has been bothering me for a long time. I had wanted to get an appreciation of what I had suspected and what is virtually never mentioned, namely that weather forecasts consistently miss their mark and by how much. That is a fairly important issue, it seems. After all, if forecasters cannot accurately predict what local weather conditions will be a few days or even only one day in advance, what hope is there that climate change for the end of the century can be predicted with a credible or even only reasonable degree of accuracy for the end of the century?
Many people love the dog-and-pony show presented by various weather forecasters, who ultimately all get their meteorological information on which they base their forecasts – indeed, even the forecasts themselves – from the same sources, satellite measurements. Local variations and circumstances are provided to some extent by local weather stations, whose measurement data is fed back to national weather services. The latter feed all of those data into their computers, do the number crunching, and then send the results of the calculations back to local weather forecasters.
The forecasters put those results into graphic format that they make available to the public in the presentations during the news hour and on the Internet. The members of the public who are the consumers of that information are eager and happy to have the latest information on the weather – never mind that what they get to see of the forecast components are the results of calculations on data that is a few hours old and is never or at best rarely and even then only coincidentally accurate.
The following three screen shots of weather forecast results are a case in point. They are for Edmonton, Canada and for nearby Elk Island National Park (the latter is identified in the graphs by the designation of its weather station, CWFE). They cover a few of the attributes of forecast and actual attributes of weather conditions that weather forecasters and their fans relish and revel in.
Forecast and actual conditions for Edmonton and nearby Elk Island National Park:
2018 05 04:
2018 05 07:
2018 05 10:
Data sources are indicated at the bottom of the web page at WolframAlpha from where the screen shots were taken.
Four attributes are indicated in the images: temperature, cloud cover, conditions (rain), and precipitation rate per hour. Only one of those, temperature, was persistently forecast with reasonable accuracy. The forecasts for the others were consistently proved wrong by reality.
That performance is nothing to be proud of. Don’t take bets on the weather forecasts, especially not on whether it will be cloudy, raining or raining much or little. It appears likely that the opposite will happen from what the forecast called for. Even if the forecast was made just one or two days earlier.
Regardless of how wrong or useless weather forecasts are, they do serve a purpose. They attract large viewing audiences and are a sure-fire method for conveying many TV commercials to a captive audience.