On social media (i.e. “Weather Twitter”), there has been a lot of discussion this morning regarding several potential upcoming severe episodes. Today, rather than my usual style blog post, I was going to take a look at a few of these upcoming potential setups.
Before we proceed: this is a medium to long range forecast. Take this with a tiny grain of salt. All usual forecast disclaimers and “this is my own personal opinion” disclaimers apply now as they always do. This discussion should in no way take precedence over any official sources. This post is merely meant for interest and educational purposes!
With that disclaimer out of the way, on to the forecast!
There are two main factors I like to consider in any type of extended range forecast: (1) Does climatology favor a certain scenario or type of hazard? (2) Do ensemble forecasts support said scenario or hazard type? To further expound on #1, if I see a deterministic model (I’m looking at you, GFS) showing snow in Texas in April, I immediately raise eyebrows knowing it does not make climatological sense. Indeed, snow has and can happen in Texas in April, but it is exceedingly rare. Related to this is the impact of various teleconnections (i.e. El Nino, Arctic Oscillation, etc.) on a region’s weather, but I will concede that I am not super knowledgeable in this area. To expound on #2, beyond about five days, I tend to not trust deterministic forecasts. Sure, I look at them, as they are another tool, but ensemble guidance is a much better tool at these time ranges.
For the uninitiated, ensemble forecasts are forecast models run using one dynamical core (for instance, the GFS or ECMWF), but are run using slightly different initial conditions in order to account for initialization errors. The idea is that since we do not know (and will probably never know) how to perfectly initialize a forecast model, we run the model with a varied set of plausible initializations in order to get a spread of potential outcomes. Ideally, the verification should lie somewhere within the ensemble spread. The larger the spread, the higher the uncertainty in the outcome. Finally, these ensemble forecasts allow us to come up with probabilistic forecasts and alternative scenarios. In short, they are incredibly useful, especially at longer ranges where small initialization errors can have major impacts on the forecast.
Tackling these two questions
1. Climatology: SPC has a fantastic webpage where one can view climatological severe weather probabilities by day. Their webpage has a more detailed explanation, but in short, it is based on the number of days in which severe weather occurred within 25 miles of a point during the time period from 1982 to 2011 (return frequency) on a particular day. If we look at these graphics, we can see that indeed, the Southern Plains and Deep South are climatologically favored areas for severe weather this time of year. What this tells me as a forecaster is that it is totally plausible to see potential severe weather setups in model guidance. In other words, climatology can be used to “sanity check” the models.
In this example, Paris, TX is in the 3% area, meaning severe weather occurred within 25 miles of Paris 3% of March 25ths from 1982 to 2011, or once. If we assume a 3% chance of severe weather each day the week of March 25 (i.e. a 97% chance of no severe weather each day), we can raise 97% to the 7th power to get an 80.8% chance that severe weather will not occur, or a 19.2% chance of severe weather (climatologically speaking!) during the upcoming week. Put a final way, Paris can expect severe weather, on average, every fifth year during the last week of March. Obviously, there are problems with this approach: Paris could have severe weather affect it two straight years in the last week of March, then have several years pass with no severe weather. Climatology in this case is really only used to give us an idea of “is it plausible?” Not “will it happen?” Myself as well as most meteorologists are annoyed at the discussion of whether a place is “due” for a particular weather hazard. Each season is (for all intents and purposes*) independent. Whether Paris saw severe weather in 2016 should have no bearing on whether is sees it in 2017
2. Ensemble guidance: Now we will finally look at the models, knowing that it is plausible to see severe weather in Texas (and the Deep South) this time of year. The main thing I will be assessing as I look at the ensemble members is how active of a pattern do the show (lots of troughs/ridges vs. zonal flow), and the timing (phase) of these troughs and ridges.
First, looking at Friday, the GEFS members are in strong agreement that a trough will move through the Great Plains on Friday. But as we get further into time…
We can see that most (all?) members so some kind of trough digging into the Southern Plains, but there is a lot of disagreement about timing. The “eastern camp” of members would place the greatest severe weather threat in the Deep South, while the “western camp” would place the greatest threat in the Southern High Plains. This to me says severe weather will be possible somewhere next Tuesday, but there is significant uncertainty on where exactly that will be. Nevermind the details (i.e. mode of severe weather and intensity).
Finally, looking 10 days out, we can see that the ensemble members are all over the place, with some showing ridges and some showing troughs over the Southern Plains, meaning I have no idea what we will be dealing with late next week. This is not useless however. One thing to note is that the individual lines all show varying ridges and troughs as opposed to a generally zonal (straight west to east) pattern. So this tells me, we cannot say anything about the weather late next week, other than that an active pattern may be continuing.
So what does this all mean? Climatology favors the Southern Plains and Deep South for severe weather this time of year. Ensemble forecasts show a deep trough in the Great Plains late this week, and another trough somewhere between the High Plains and Deep South by the early to middle part of next week. This means that we are likely to see at least a couple severe weather episodes (the term “episode” does not intend to suggest anything about intensity/severity) across the southern tier of the country during the next seven days. The extended range ensemble guidance tells me that we are likely to stay in an active pattern beyond Week 1, but the details are highly uncertain.
*One could possibly make some kind of Chaos Theory argument here to say that the probabilities are not independent, but for our purposes, it is a close enough approximation.