Testing the Absurdity of AccuWeather’s 90-day Forecast

I love snow.

In fact, some may even call my love for snow (and weather in general) mildly obsessive. During the winter months, and especially when white stuff is in the forecast, I check my go-to weather sources frequently with fingers crossed, hoping that the models weathermen use for their forecasts put the Washington, D.C. area in the bulls eye for some high snow totals.

When my favorite local weather source, the Washington Post’s Capital Weather Gang, mentioned not 1 but 2 opportunities for snow in their forecast early last week, I felt giddy inside. The first chance of snow was for Thursday night from a storm that traversed the continental United States. The forecast called for an inch but D.C.’s famous “snow hole” (caused when the nearby Blue Ridge Mountains weaken moisture starved storms) gave the area only a few flakes. Meanwhile, forecasters were growing weary of the region seeing any snow at all during the weekend, but at the last minute, a shift in the models gave snow lovers joy as the city was blanketed with about an inch of snow from a coastal storm.

The moral of the story? Weather forecasting is hard. And the forecasters are not entirely at fault. The models they use to predict the weather are constantly changing and differing from each other, even when a weather event is a few days away.

So when I saw several months ago that one of the largest weather forecasting companies in the United States, AccuWeather, started releasing 90-day (yes, 90!) for cities around the world, I instantly knew the feature was bullshit.

Why release a 90 forecast then? In a press release, AccuWeather  said that the forecasts are “a valuable tool for planning the best time of the season for road trips, vacations and outdoor activities, letting users follow the forecast and make more detailed plans as the date approaches.”

Sure, so I’m going to change my beach plans in 62 days because the weather forecast calls for a 50% chance of rain. Right…

People were quick to call AccuWeather out on their outlandish idea. Former TV meteorologist John Bolaris said it was “ridiculous”. Tech site Gizmodo called it “misleading as hell”. Random Twitter user @Stap_Jr proclaimed that it would be “a great tool for people who love to be disappointed. But AccuWeather held steady, noting that their predictions were based by science, even while acknowledging that their long range forecasts should not be used a as a “strict guide”.

And while criticism has been pouring in, there is not too much information about the true accuracy of these forecasts, which is why starting tomorrow, I plan to take on Big Weather and show the world how absurd their 90 day forecasts are using all the knowledge I obtained.during my one semester of college statistics!

While I do not intend to run a deep statistical analysis, I do plan to try to answer the following questions:

  • What are the differences between predicted and actual high/low temperatures?
  • What was the accuracy rate of predicting days in which there would be precipitation in the forecast?
  • What trends can be observed over time? Do the forecasts grow worse and worse by each day? Are they accurate at any point at all?

Starting on Monday, January 9th, I will start compiling daily weather observations for Washington D.C. zip code 20005 (collected from AccuWeather) and compare them to the AccuWeather 90 forecast. In about 3 months, I will report back  with what I hypothesize will be fairly crazy data.

Stay tuned…



2 thoughts on “Testing the Absurdity of AccuWeather’s 90-day Forecast

    1. Raj: As of now, I don’t plan on using specific software, but I’m open to ideas! As of now, I plan at looking differences between predicted vs. actual temperatures (maybe I’ll make some cool graphs), and maybe look at some medians and modes as well (actually, this takes me back to elementary school stats hahaha). I also plan to examine predicted vs. actual precipitation amounts and weather conditions (eg. was it sunny on a day they predicted it to be cloudy?)

      And I actually saw that link when researching for this blog! It’s interesting because he did the test way back in the day (2013) when AccuWeather only did 25 day forecasts.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s