A look back at forecasting the week's winter weather event

Nick Mikulas is a contributing meteorologist for KALB.

ALEXANDRIA, La. (KALB) - It’s was almost 8 days before the snow started falling that I jokingly took a screenshot of an image of the Canadian model “for posterity”. That image showed a bullseye of 15 inches of snow around the southern end of Toledo Bend, and I sent it to a fellow weather nerd for a good chuckle.

I moved on, but had that general time frame circled on my calendar as a possibility for snow or ice. The next couple of days were chaotic. We had one run with almost 2 feet of snow near Marshall Texas, and some that completely removed the threat. But the sum total of what I saw, the fact that we are near our coldest time of year climatologically, and the upper-level pattern that was forecast showed that we had a decent chance to at least see something frozen in Central Louisiana.

Fast forward a couple of days, and we had that minor threat for a tiny bit of ice on the morning of January 12th. That prompted an update from me on the evening of the 10th where I reluctantly mentioned the long-range threat for wintry weather 6 days from then. I hate doing that, because forecasts like that tend to have little skill. In this case, I went for it because every major global model hinted at something, with some bringing a significant storm. I knew the Canadian scenario was very unlikely since it would’ve shattered records, and by that timeframe, the Canadian model had settled down to something a bit closer to reality.

Friday evening was when we started to sort of commit to the snow and ice threat, and here was what I’ll call my first mistake in the forecast. I said there was a 60-70% chance we would see something frozen in the area, and in retrospect, that’s just too high of a chance that far out, even with all models on board to some degree.

About four hours after my huge commitment of a 60-70% chance of snow or ice in almost four days, as if they were faithful readers waiting for their big chance to say I told you so, the models went from about 1/4 inch of liquid equivalent to about 0.05”. This was not a trend I expected, and every single model showed it. When I went to bed Friday, I was entirely unsure of what direction this trend would take us, and hoped it was a blip along the way, and the consistency of the past couple days would return.

Saturday brought consistency, but not the consistency I was looking for. Models were consistently showing lighter values in that 0.05” range, which when converted to snow and sleet would be a dusting, to maybe 1/2 inch in spots. This is why I don’t give accumulation values, and generally don’t post what models are saying regarding accumulation until around 36 hours out. We’d gone from 1-3 inches on most models to 1/2 inch or less. My saving grace was that most models still showed some snow or ice, albeit very light, and the European model ensembles were really on board for at least light snow. With all that knowledge, I didn’t change my forecast from the previous night, because I hate going from 70 to 40 and back to 70. I left things as is because I hadn’t given accumulation numbers, and chances were still high that we’d see something. Knowing the psychology of the Deep South in snow events, if we see at least some snowflakes, even though I’d know I was wrong in what I thought would happen, the forecast wouldn’t be a total bust, and I could still look people in the eye at the grocery store.

By Sunday, I had a great idea on the timing of the snow, but pegged chances at around 80% in our northern parishes, down to 40% in our southern areas. Models were intent on drying things out as the system moved from northwest to southeast, and my estimates felt a bit too high over southern parts of our area. But experience told me that there was at least a chance the models were over forecasting the dissipation as the snow pushed south. It was more of a hunch than anything, and I didn’t see what was really coming just yet. This is a case where my logic was sound, but the eventual reasoning behind the snow was different than what I was thinking. That happens more than I care to admit. I call it “backing into a good forecast”. I finally mentioned a dusting to an inch around our area, even though I thought the higher totals would stay in our northern parishes.

Monday brought solid agreement with every European and GFS model ensemble member showing at least flurries, so I was comfortable upping my accumulation forecast to the 1/2 to 2-inch range over a large part of the area by evening. Winter weather advisories were issued area-wide, and things were really coming into focus. By Monday evening, models started to indicate a possible secondary snow maximum south of Alexandria as a subtle disturbance “recharged” things over southeast Texas. That’s what really prompted me to expand those accumulations southward, even though it seemed like this was a lower probability event. We were now in the range of all models, including short-range models that only go out 18 hours, so I knew I’d have time to adjust, or “nowcast” as things developed. The general idea had been communicated that we would see white on the ground, and if someone saw 1/4 inch, or 2.5 inches, it wasn’t likely they would consider that 1/2 to 2 inch range a bad forecast. That’s something that’s hard to talk about as an event unfolds, but solid endpoints are important in communicating a forecast. If I said 1/4 to 2.5 inches, that just sounds weird. I honestly didn’t love the look of 1/2 to 2 inches, but it seemed a little more decisive than a dusting to 2 inches, so we ran with it.

Tuesday morning things were coming together just about as expected. Short range models were really on board with the secondary maximum to our south, and the snow/sleet was surging in from the north. There was a little dry air to overcome in the mid-levels in our area, which is why we saw sleet for a few hours. I’d mentioned that possibility, so my bases were covered. The fear I had was that the northern band would dissipate, and the southern band would move just south of the immediate Alexandria area. I mentioned a possible weak spot in accumulations through the middle of the area, but even in that weakness, I expected flakes, and at least some sort of minor accumulation. It wasn’t worth changing the forecast. As radar echoes continued to dissipate as they approached my house, I wondered if that dry mid level air would win out and our biggest population center would miss most of the snow, rendering my forecast bad in the eyes of those around town.

Thankfully, around 9:00 a.m., we saw the snow move in, and large flakes quickly coated the ground. The forecast was saved, and I could shop in peace, and avoid all the comments about how meteorologists can be right 50% of the time and keep their jobs. I was still watching the radar out west, and as it started to fill in, I knew the secondary maximum down south would happen, and boy did it. That’s the area where the forecast probably came closest to busting, but from what I saw, totals stayed below 3 inches. So while I didn’t love that I didn’t expect that to happen exactly as it did, my range of 1/2 to 2 inches was still close enough, and I could live with my forecast. That’s why I like to broad brush things, and not get too cute with the forecast a couple days out.

Looking back on the whole event, this forecast went about as well as it possibly could. We had plenty of lead time, and there were no major letdowns in the area that didn’t at least see something. I learned from this system, as I always do, but what I really take away from these winter events is that communication, and consistency are key. It would be annoying if I told you 2-4 inches of snow were coming, then went down to a dusting, then up to 1-3 inches, and then settled on the midpoint. Trending a forecast from a conservative starting point is much easier than flip flopping with every model run. It’s why it can be frustrating to deal with people that just post single model runs and say “this is a worst case scenario”. It serves no purpose, and makes the job of actual forecasting harder as you have to discredit a model that might not even be wrong, but is an incredibly unlikely event. When I saw the initial run of the Canadian model that showed 15 inches, I knew it was off the charts unlikely. I couldn’t say it was impossible, but I also couldn’t say to you all that there was a low chance of 15 inches of snow. It all comes down to you trusting us as a weather source, and us not betraying that trust because the models look really neat 7, or even 4 days out. We’ll always try to improve both our forecasting and how we communicate it, but at least in this case, things worked out. We still have another 5-6 weeks where winter can throw us another system.

We’ll be watching closely, but probably won’t say anything until about 2-3 days after we start talking about it behind the scenes. If you could only see the texting, and messenger stuff that goes on between the weather team before you hear a peep. Tom, Abby, Tyler, and I really do work hard to get you a good forecast. Of course, we totally love it too! Til next time!

Comments are posted from viewers like you and do not always reflect the views of this station. powered by Disqus