Busts happen. We’ve all seen it. Maybe it was that much-anticipated beach day in July when a sunny and 90 forecast turned out rainy and 68. Or it might be the disappointment when a forecast for snow and 30 degrees ended up being partly sunny and 42, dashing dreams of a run down the local sledding hill.
From a business perspective, a blown forecast is much more problematic. A forecast that’s far off the mark could cost energy traders millions of dollars, not just ruin weekend plans.
What actually constitutes a “bust” when it comes to weather forecasts? For our purposes, and as defined in our “Analysis of High Temperature Forecast Accuracy of Consumer Weather Forecasts from 2005-2016”, a busted forecasted is when the forecast high temperature differs from the actual high temperature by more than 10°F.
It’s no secret that weather enthusiasts and others take note of a busted forecast more than an accurate forecast. That speaks both to the increase in accuracy we’ve come to expect, as well as the impact of forecasts that are wildly off the mark.
Fortunately, forecast busts are less common than you might think — and becoming more and more rare. In 2005, one in thirty (approximately 3%) one-day-out forecasts (forecasts made the night before the day in question) were busts. By 2016, busts for the same forecast period occurred only once in seventy forecasts (1.5%).
The same trend was noted through all forecast periods. For example, the percentage of high temperature forecast busts for four-day-out forecasts dropped from 11.5% in 2005 to 4.3% in 2016. Likewise, in 2005, nearly 22% of seven-day-out high temperature forecasts were at least ten degrees off. Twelve years later, that same percentage was just 13.3%.
The assessment of forecast busts is part of a larger, more comprehensive analysis of changes in high temperature forecast accuracy from 2005-2016 that includes a review of specific providers and an examination of seasonal variation in forecast accuracy. The analysis was based on forecast data provided by ForecastWatch and includes almost 200 million high temperature forecasts for nearly 800 locations in the U.S. over a 12-year period. The data included forecast high temperatures for up to ten days into the future.