In today’s ultra-sophisticated, data-rich environment, where forecasts are generated at the push of a button, it can be hard to imagine that weather prognostication was once a new and rudimentary science. Yet, it wasn’t that long ago when forecasting was in its infancy.
An article in The New Yorker detailed the fascinating evolution of forecasts and forecast accuracy. Here are a few notable developments that we found interesting:
- Telegraph lines, which were well established in the United States by the middle 1800s, often provided suggestions for how the weather might turn out. How? When it rained, the lines often had problems. Thus, when cities to the east of the affected locations experienced delays in receiving communications, they reasoned that a storm might be headed in their direction.
- In 1862, a Norwegian physicist and meteorologist, hypothesized that weather resulted from the movement of air masses. His work ultimately laid the foundation for the modern practice of weather
- In the early 1900s, Lewis Fry Richardson incorporated mathematical models of the principal features of the atmosphere and made calculations by hand to try to predict weather. His methods were later utilized by the first modern electrical computer in 1950, which produced a 24-hour forecast; however, it took approximately 24 hours to produce the forecast.
Fast forward to the current era of weather forecasting. The European Centre for Medium-Range Weather Forecasts (ECMWF) produces a 10-day forecast in just over two hours. Its accuracy is renown; it famously predicted the landfall of Hurricane Sandy in October 2012 seven to eight days in advance while the prevailing American weather model had the hurricane going out to sea.
Advances in computing power and an increase in the number of data points have fueled continual accuracy improvement. ForecastWatch’s study of accuracy for high temperature forecasts <link to study> over the 12-year period ending in 2016 underscores how forecasts have improved. Some highlights of that study include:
- The average error for 3-day-out forecasts is as accurate as 1-day-out forecasts 12 years ago.
- Five-day-out forecasts are as accurate as 3-day-out forecasts were 12 years ago.
- The rate of improvement for Day 1 forecast was nearly one degree over the 12-year period (from an error of about 4 degrees to 3 degrees).
- Five-day-out forecasts have improved at nearly double the rate of 1-day-out forecasts over the 12-year period.
As we look to the future, there are two things that can be said with certainty about weather forecasts. The first is that increased accuracy is not a question of if but one of how much and how quickly. The second is that the pursuit of perfection will probably never end.