Ponderings on communicating a sharp-gradient forecast

I’ve come to the conclusion that, as the difficulty of a weather forecast increases, the challenge in effectively communicating that forecast increases not linearly but exponentially.

On November 25, 2018, southern Iowa experienced heavy snow and, in some cases, blizzard conditions. Other places, literally miles away, got nary a snowflake. There was a seven-inch difference over about eight miles. We knew that there would be a very sharp gradient in snowfall amounts on the northern edge, but it ended up being tighter than we imagined and roughly 25 miles south of expectations. Unfortunately, that meant that Cedar Rapids, the largest city in our television market, got zero snow when our initial forecast was 4-6″, then revised down to 2-4″, and an eleventh-hour update from one of our meteorologists conceded that there might be zero snow at all. Meanwhile, Iowa City, a quick drive south, was digging out from 6-8″ of snow. There was a narrow band some 30 miles tall of people who were upset with reality versus forecast, and everybody on either side (hopefully) was pleased that things turned out as we said they would.

A few days have passed since this event and I’ve had some time to reflect on the communication challenges. Thankfully, events like this are unusual, but there are lessons to be learned and applied to other situations.

Pre-storm build-up

Meteorologists had reason to believe that there was a high likelihood of a strong winter storm affecting Iowa on Sunday days beforehand. Unfortunately, some meteorologists – as well as weather enthusiasts – chose to illustrate that potential by posting computer model snowfall forecasts, down to the tenth of an inch for specific locations, four days before the first snowflakes fell. Four days is an eternity when it comes to such details. First off, I hate snowfall forecasts that are that precise. We do not have the ability to accurately predict with that kind of precision one day beforehand, let alone four days. It suggests that we have abilities that we do not have. It’s practically lying, in my humble opinion, and no amount of “it’s just a model, not a forecast” excuses it. Now, one could argue that these maps are useful because they are showing people that there is the potential for heavy snow and sharing the information is a public service and is helpful. This alleged altruism is a bunch of bologna, though, because although giving people that heads-up is important, there are far less hype-y ways of doing it. One easy one is to have contours but no numbers, because you better believe that if there’s a 12.7″ over a city, people will see that and will run with it. No numbers, one less unnecessary thing to fixate on. Another, my preferred one, is to simply show confidence contours on a map illustrating where areas of heavy snow are more likely based on current expectations. But those wouldn’t get the same volume of clicks, likes, and shares, because such a map isn’t as sexy and exciting as outright saying a foot of snow is on the table in your backyard.

Those who aren’t in the know may wonder why we care about what other people are showing or doing. After all, we can’t control them, so why does it matter? Unfortunately, it still affects us. It is not infrequent for a weather forecast from one source to be attributed to another source or some kind of nebulous catch-all meteorological realm (“they say we’re supposed to get a foot of snow” – who’s “they”?). Because of this, our hand is forced – damage control is on for something we didn’t say or do. In fact, my station even put a video out on social media prior to the event telling people to calm down because there were still many questions on what was going to happen.

But hype wins. As much as people say they don’t like hype, they consume it as fast as it can be churned out. Not everyone, of course, but enough that it makes our job a lot more difficult.

The edge problem

From the very start, we knew that there would be a sharp gradient in snowfall amounts on the northern edge of this system. We even had a box of text stating as much on every single iteration of our snowfall forecast. As I said earlier, it ended up being even tighter than we expected, but this was something we hammered every. single. time. But these transition zones are where the biggest forecast bust potential is, and that’s no secret in the weather world. I’ve seen meteorologists out east lament nor’easters in which the location of the rain/snow line (or sleet/snow line, or whatever) would mean the difference between little to no snow or a foot of snow. But how do you get that point across but also give a useful forecast? It’s impractical to give a forecast of 0-12″, even if that’s really what the potential is. So, how do we tell people what we expect, but also help them truly understand that there is a reasonable chance of a range of possibilities that is nowhere close to what our expectations are? The edge is often where the greatest uncertainty is, so I guess that dovetails nicely into…

Honesty about uncertainty

“I don’t know.” Words that I suspect some would say a meteorologist should never utter. I disagree.

Am I an expert? Yes, I certainly like to think so. Does that mean I know everything about what weather will happen? Absolutely not. In an era of people exclaiming “fake news” as well as distrusting experts, and skepticism toward science, meteorologists (especially broadcast meteorologists) are at quite the disadvantage, it seems. In general, people trust weather forecasts… otherwise they wouldn’t be mad when it’s wrong; they’d simply shrug their shoulders and then be stunned when it was right. We all know that it takes time to build trust but that trust can be broken quite easily, and I think for the reasons outlined above, that trust is a little more fragile than we’d like it to be.

There has been more than one event in which I look back and feel that there’s some aspect of the forecast process I could have been more honest and open about with our viewers. There’s been a growing trend in weather communication to be more upfront in the confidence of the forecast, either in general or specific aspects of it. I tried to do that with this event: stating that we had high confidence that southern Iowa would be getting much more snow than northern Iowa, that there would be a sharp cutoff in snowfall amounts, that the snow would fall during the daytime on that day, and that it would be very windy with gusts over 30 mph. But I also stated what I was not confident in – what I did not know – specifically, where that cutoff would occur, that I felt it would be in such-and-such area (which, again, was off by about 25 miles), but any small shift in the track of this weather system would greatly affect who got what. So while I technically conveyed the potential for less snow in Cedar Rapids than forecast, was it as explicit as it could have been? In hindsight, I don’t think so. (An aside: I would have never gone so far as to say Cedar Rapids could get nothing, because I felt the possibility was so remote that it wasn’t even worth mentioning.)

I feel that expressing where there is real uncertainty can build trust because it shows honesty… but on the other hand, can it be detrimental if not done the right way? There’s the obvious argument that, by admitting one doesn’t know, it can undermine one’s credibility. I have a feeling that’s pretty insignificant. I’m more concerned about it leaving people not knowing what to do/expect. My job is to give people information so that they can make informed decisions, but if my expert analysis involves enough uncertainty/alternate scenarios, are viewers left throwing their hands in the air not knowing what to do? If, at the end of a forecast, somebody is left with a blank stare and thinks “so…?” then I haven’t done my job as effectively as I need to.

Unintended interpretation

My stepmother-in-law has a master’s degree and worked in education for a long time. She is an intelligent individual. Right after I finished my 10 p.m. weathercast the Friday before the event in which I had Cedar Rapids in the 6-8″ band, I receive the following text message: “Predicting around 6″ now?” I respond with the same thing I’d just gotten done saying – that Cedar Rapids is in the northern part of the band and any small shift would result in a big difference in amounts. I was baffled as to how “four to six inches” equated to “around six inches.” To me, “four to six inches” means “four to six inches.” I’m not sure how to get any more explicit than that. That said, there are other places that we have ambiguity that can lead to interpretations that aren’t intended or desired. Of course, sometimes that ambiguity is intentional because it offers some buffer, e.g. “northern Iowa” rather than “north of Highway 20.” That’s convenient for us because it gives some leeway, it may not be helpful to somebody who’s left wondering “am I considered northern Iowa or not?” We had a Facebook comment from somebody asking what exactly we meant by “the southeastern half of Iowa.” After an explanation that, if you drew a line from the southwest corner to the northeast corner, it would encompass everywhere southeast of there, there was a little pushback saying that some people in that area would not consider themselves in southeastern Iowa. So, something that is technically true (such as Dubuque being in the southeastern half of Iowa) may not actually be interpreted as such (most people would probably say Dubuque is in northeastern Iowa, even though that doesn’t apply because we’re talking northwest half/southeast half).

Final thoughts

I’m certain there are things that crossed my mind between the event and now that have slipped my mind. And by no means are the things I’ve written about here anything new. It’s stuff that’s come up before and will keep coming up. I wanted to get it written down so that I could at least come back and look at it and maybe get a little feedback to create some action items or even get just some food for thought. It was a little rambly at times, but this was me just kind of typing my thoughts out as they came.