Communicating Winter Weather Uncertainty

Dr. Shepherd, former president of the American Meteorological Society, tweeted this as a catastrophic winter storm, slated to impact most of the Gulf and East coasts, began coming together in the forecast:

Most of us have struggled to explain model uncertainty and weather prediction to our friends outside of the science. I have blogged on this topic before. This time, I’ve noticed a few different viewpoints coming from professionals across the forecasting spectrum:

NWS Forecasters

Understandably, NWS forecasters seem to be struggling with sounding the alarm prematurely (as the Sterling office did not issue a winter storm warning for the D.C. metro area until late last night, 24 hours before the storm was forecasted to begin) versus not sounding the alarm loudly enough (the Atlanta forecast office began discussing the storm last weekend, likely in an attempt to make the government take notice).

Non-NWS Forecasters/Academics

People in the private sector and academia seem to be playing up the uncertainty most unapologetically. I think this is understandable, as their audience/clients may be more interested in a big picture of the event. Rather than needing to issue a public weather advisory/watch/warning or tell an entire community to buy bread and water, they can be more liberal in adjusting forecasts as new information comes in without confusing the general public.

TV Meteorologists

Many of my friends and social media followees (not a typo!) are TV meteorologists. While I think they are also doing a good job of communicating uncertainty, they are doing it at a price. Their name and face goes with their forecasts, and, often, they alone will receive the brunt of a blown forecast. While the other two groups tend to work in teams and are often anonymous (forecast discussion names and meteorology journalists notwithstanding), an on-air meteorologist is often held responsible for his or her forecast by the community. Being more in the public eye than the other groups, their forecasts cannot change drastically from day-to-day (or hour-to-hour!) without raising red flags with the audience.


Some of us are not working as forecasters or otherwise involved with meteorology in an official capacity right now. Nonetheless, friends, family, and coworkers still turn to us as the “resident meteorologist.” While I try to keep up with the data as much as possible, I, personally, find myself parroting others’ forecasts when caught off-guard and behind on the updates. Other than my personal reputation, however, I have little at stake than some after-the-fact ribbing for a busted forecast.


Of course, these are generalities, and your mileage may vary.

I’ve appreciated that the local D.C. TV meteorologists have, overall, done a decent job introducing the storm at an early enough date ahead of time and explained that different models have shown different data. The Capital Weather Gang is always under fire during the winter in D.C., as they live-blog model runs and give regular updates on how they are adjusting their forecasts. Many mets in all categories have taken to social media, either through professional or personal pages, to show and explain model data to their friends, family, and followers.

How much is sticking?

You would think that if we were properly communicating this uncertainty to the public that we wouldn’t hear the same tired complaints about forecast accuracy every week. Where is the disconnect? I see the updates throughout the event from tons of people, but, in the end, the public remembers what went wrong.

Is this a case of people wanting to complain, or are we really not reaching laymen the way we need to?

With current technology, we can reach more people than ever, so why is this an ongoing problem?

Weather Models and the #Snowquester Snowstorm

There is finally snow on the horizon.

Meteorological spring kicked off on March 1st (this should not be confused with astronomical spring during your regularly-scheduled equinox), and with it came the faint promise of snow in the weather models. The further out in time we are attempting to predict something, the more noise we are going to have in our data. Part of a meteorologist’s job is finding the signal within the noise in the information we have available (weather balloon data, numerical model output, ensemble forecasts, and other things).

You may be familiar with the “butterfly effect,” which suggests that the formation of a hurricane may be dependent on a butterfly flapping its wings weeks earlier. While an over-exaggeration, Edward Lorenz, a mathematician and meteorologist, used this simple example to demonstrate the universe’s sensitivity to initial conditions. In other words, a seemingly small (perhaps overlooked) detail may have a profound effect on the end result of a mathematical/scientific/meteorological prediction model.

If our starting conditions are wrong, even at a very small level, the predictions our weather models make get increasingly worse with time. But it gets even better! If the equations you use in your models are approximated at all, those small differences add up over time as well. Let’s use a familiar example:

2 + 2 = 4

Now what if that was a simplification of a slightly more complicated problem?

2.4 + 2.4 = 4.8

With these values, most third graders could tell you that it’s typical practice to round the value of 2.4 down to 2. But there is a big difference between 4 and 4.8, right? What if we continued adding these approximations together over time?

2 x 1000 = 2000

2.4 x 1000 = 2400

Now we’re up to a 400 [unit] difference!

That’s how weather models work, in a nutshell. We can’t start with perfect data and we can’t use perfect numerical values within perfect mathematical equations. The very minute changes we have to make to account for computing power limitations mean that our predictions further out are less certain.

Now that I’ve tricked you all into a science lesson, back to #Snowquester. This storm is looking to be the best chance for snow that the D.C. area has had in awhile. There is still some uncertainty in the models (as was just explained), but the closer we get to the event, the more the different models come into agreement. There is a chance for over a foot of snow in the higher elevations west of D.C., and the Capital Weather Gang currently has the 12”+ bullseye around Winchester, Virginia.

The temperatures west of the District should remain low enough to keep the precipitation falling as snow throughout almost the entire event, while parts of D.C. and east will begin with rain before the changeover to snow overnight Tuesday into Wednesday morning. Because the temperatures will be hovering around the freezing level, this will be a very wet snow with potential to down power lines and trees.