Tuesday, February 7, 2017

Messaging Significant Events

Over the past few days, a part of our CWA  (NWS Great Falls) has been pounded with heavy snow (here's a PDF from our office with some of the impressive snowfall amounts, records, and pictures). And, when I say pounded, I mean 30-50+" of snow, avalanches, roads nearly impassable or closed, roofs/buildings endanger of collapsing, emergency declarations, people stranded and/or unable to travel, etc. Perhaps pounded isn't a strong enough word, which brings me to the challenge of the day. How do we approach events like these from a forecasting and messaging standpoint?

I am a firm believer that a good message starts with a good forecast and good forecast starts with sound science. For me personally, my approach to forecasting (and this is by no means the only way to do it) has been to pour over the data, then build a forecast that best represents what I am seeing from model data, current observations, input from other forecasters, experience, etc. From there, I step back, look at the finished product, and then try to determine what, if anything, needs messaging. This is all well and good, but what happens when the finished product shows the potential, or even likelihood, of an anomalous event?

Perhaps more experienced forecasters than I don't struggle with this as much, but while working this recent event, stepping back and looking at the forecast prior to the event got me wondering if the ridiculous snowfall amounts I was coming up with were valid. So much so, as soon as I calculated my snowfall amounts, me and the other forecaster I was working with immediately picked up the phone and called a neighboring office that would also be impacted. When they answered, their response was the same. They also were, somewhat dumbfoundedly, looking at the amounts they were coming up with. In fact, in their words, they said, "We [the two forecasters at the other office] have been staring at our computers trying to figure out if these amounts are real."

Turns out, those amounts were real. But, hindsight is always 20/20 and, for many of us in the moment, I think there can be a hesitation to put out such a forecast. For me (and I assume I'm not alone in this), there is that concern of, well what happens if this doesn't actually pan out? This points to a larger question. What is more helpful/harmful? To see the potential and not put the forecast out, or to put it out? If you forecast the high-end event and it happens...awesome! If you forecast the high-end event and it doesn't happen...not so good. But, what if you don't forecast the high-end event and it does happen? Which is worse...forecasting the event and it not happening or not forecasting the event and it happening? Both have negative consequences and this is where the challenge comes in.

Sometimes, I think there is a tendency to back off from higher-end events. Let's face it, oftentimes at least one or two models show a worst-case scenario, only to back off as the event draws near. Or, maybe they never back off, only to be way off in the end. This isn't always the case, but it does happen. And, then, there is the issue of over-hyping an event or the boy who cried wolf syndrome.

But, at the same time, when even one model indicates the potential for a higher-end event, it probably isn't prudent to immediately write it off (outliers do verify at times). And, while never 100% perfect, I think this is where sound science, experience, and collaboration comes in. If after all of this, there is above average, or maybe even average, confidence, perhaps it is time to go ahead and pull the trigger on messaging the event as a higher-end event. Afterall, that is part of our job as Meteorologists. To inform people of the weather. If the weather is going to be really bad, people need to know so they can prepare. We can't make people prepare, but we can at least stress the expected magnitude of an event.

If there was a Messaging Hall of Fame, two statements/events come to mind to be included (and, of course, this is in now way an exhaustive list!) 1) The statement issued by NWS New Orleans prior to the landfall of Hurricane Katrina. That's not a statement any NWS office issues on a regular basis. 2) The higher-end wording by SPC, NWS offices, and the media during the April 27, 2011 tornado outbreak. Both of these events fell outside of the norm, in my opinion, and forecasters had to make the not-always-easy go/no go call on these higher-end watches, warnings, and statements. Should all events be messaged just as hard or do higher-end events need stronger wording?

At the moment, I would tend to lean towards higher-end events need to be messaged harder. A 40" snowfall in a non-mountainous area probably deserves higher-end wording/statements than a 10" snowfall. Winter Storm Warnings, for example, would likely be issued for both, but 40" has the potential to cause more significant impacts/disruptions to everyday life than a 10" snowfall. Both are significant, but again, the challenge is how to approach these higher-end events.

In our case, wind wasn't much of a concern for much of the hardest hit areas, so we felt that a Blizzard Watch/Warning wasn't warranted. Granted, it would certainly grab people's attention. I would imagine, and this is purely speculation, that a Tornado Watch also grabs a lot of attention and probably even more so than a Severe T-storm Watch. But, do you put out a Tornado Watch if you are not expecting any tornadoes? I hadn't thought of this before, but could it be that we approach winter storms like convective events and have PDS Winter Storm Warnings in addition to regular Winter Storm Warnings? For our CWA, I think this event would have justified such a warning. 10" of snow in a snow-prone area is definitely warning-worthy and causes significant issues. But, even here, 40" basically shuts everything down. The local DOT here mentioned in a statement that they are going to have to bring in heavy equipment to remove all the snow and that it would take multiple days to get through. Cars are buried, roads impassable, roofs are struggling to hold the amazing weight of the newly fallen snow. That's not your average winter storm. One of the shifts after mine added wording into the original warning, stating this is a dangerous situation. I thought this was a great idea and reminded me of the kind of enhanced wording SPC uses in PDS Tornado/Severe T-storm Watches.

I think it is always important to look back at an event as an office and as individuals to see how things played out and what, if anything, could be improved on. For me, I put out the higher-end forecast, but wasn't bold enough to go all in with the messaging, even after collaboration with national centers and a neighboring office. Basically, I used the process I mentioned above, but didn't fully apply it, unfortunately. I think the fear of being wrong and the fear of over-hyping or messaging too strongly got the best of me. It's important to make sure you're not missing something, but at the same time, if after a careful scientific/collaborative process, confidence is still above average, I think at a certain point you just have to pull the trigger. I think the lesser of two evils is people preparing for an event that doesn't happen as opposed to not being told of an event and it happening. The latter case, using a snow example, has the potential to leave people stranded with no food, water, power, etc, because who is going to prepare for an event they don't know is coming? Having extra food, water, candles, etc doesn't hurt anyone, but NOT having those things and then NOT having a way to get those things is much worse in my opinion. Perhaps these types of events only happen a few times in any of our careers, but when they do, don't fall into the trap I did of letting fear prevent a stronger-worded message from going out.

On a quick side-note, this event has reminded me that, no matter how good the models become in the future, I still hold firmly to the idea that an understanding of the science behind what the models spit out is vitally important to our messaging. Knowing WHY a model is showing something can help us explain HOW an event is going to play out and even know what to stress that even the best models might be missing out on.