Improved Forecasts in Uncertain Situations
With the weather outcome on so many days far from certain, how do forecasters create improved forecasts in uncertain situations?
A little-recognized ingredient adding value to every forecast is human judgment and experience. Experienced forecasters know when to believe the forecast models and when to discount the computer-generated predictions due to characteristic model errors. Just such a situation arose with the Upper Midwest forecast at the start of April. The image below shows the 4-day Quantitative Precipitation Forecast (QPF) for April 4th and 5th (based on an average of the April 1st forecast models) and was created by the NOAA Weather Prediction Center — with moderate to heavy rain amounts for southern Minnesota.
However, the forecasters at both the Twin Cities NWS and Sioux Falls NWS forecast offices were convinced that (as seen several times in recent months) the forecast models were too far north, and were forecasting little to no rain in our local area. Sure enough, the edge of the heavier rains ended up about 250 miles farther to the southeast than the picture painted by this forecast map!
Statistical expert Nate Silver wrote an article whose title which may give you a chuckle. The Weatherman is Not a Moron describes the way in which weather forecasters accept and even utilize the uncertainty in their profession for benefit. Most notable to me is the way Silver’s article documents the advancing skill of weather forecasts over time….
“In 1972, the (National Weather S)ervice’s high-temperature forecast missed by an average of six degrees when made three days in advance. Now (the error is) down to three degrees. More stunning, in 1940, the chance of an American being killed by lightning was about 1 in 400,000. Today it’s 1 in 11 million. This is partly because of changes in living patterns (more of our work is done indoors), but it’s also because better weather forecasts have helped us prepare.”
So, the probability of being killed by lightning has been reduced by 96% over an 80-year time span. That’s an impressive success story surrounding weather forecasts – if the average person is outdoors even 20% as much as in 1940.
Whether people notice or even care about weather forecast accuracy often has more to do with the individual’s activities than with the skill of the prediction. In an era when it is more common for children to be indoors playing video games and less commonly involved in outdoor sports the weather may be perceived as less of a concern than it was one or two generations ago. The level of interest in weather forecasts will also depend on the season of the year and the level of development of gardens and crops. In a blog post from October of 2016, I gave examples of forecast challenges during a month when a small error in temperature forecasts (near the end of the growing season) can make a big difference for those hoping to prolong garden harvests. In the end, some weather situations still defy the model tendencies and day-to-day trends—one reason forecasts seldom come with guarantees. That is why I encourage an attentive and thoughtful approach to the weather which usually makes for happier forecast users.