I know its been another hot minute. It’s not often we can talk about sever potential out near day 7 with any degree of confidence, or even skill beyond climatology, but we have such a chance today. Of course, at this range, we have to rely on global models (and their ensembles of course). Reminder, it’s never too early to prepare! Stay up to date with forecasts from the SPC, and follow the preparedness advice from the NWS and ready.gov! Let’s dive in!
*** Note: Weather forecasts are highly perishable. This was written mainly with 00Z or 12Z 29-Mar-2023 forecast cycle products. Please use the links to see the most up-to-date information. ***
Let’s take a look at the synoptic pattern, since that is what is going to be most predictable by global models at this range. We’ll start with the very tired, rather non-scientific “GFS vs ECMWF” approach, using only the deterministic runs. For this first example, the 12Z 29-Mar-2023 cycle.
As we can see it looks like there is decent agreement between the two model systems there is going to be an anomalous trough coming out into the central plains during this time. But, we only have two data points, which, in the information age, is laughable. Let’s try to dig up some more data just to make sure we aren’t biting on a fringe solution that isn’t supported by the ensembles. Staying with pivotalweather.com, let’s zoom in a bit and look at both ensemble system means to see how these deterministic runs compare.
In this case, the means don’t exactly tell a different story than the deterministic runs; that is, we can infer the deterministic runs are well within the envelope of the ensemble solution, and that the ensemble system has fairly high confidence in the general pattern. So let’s take that a step further with the EPS/ECMWF and look at something like spread.
The plot above might take some getting used to, but is very valuable. We’ll start with the right side, as I think it becomes easier to interpret the shading that way. The right side features the operational/high-resolution/deterministic forecast, with the standard deviation (in meters) derived from the ensemble forecasts shaded. While useful, we would expect fairly high uncertainty at this lead time. So, ECMWF uses the standard deviation over the last 30 days at this forecast time range to normalize this field in the left plot. Green shading – that is, normalized standard deviation less than 1, indicates lower spread/higher confidence than usual (last 30 days) at this time range. And as we can see, there is a lot of green on that map, once again affirming we should have pretty elevated confidence in this forecast.
Another important tool in the arsenal at this range is WPC Clusters. Let’s take a look at those below.
Above are the current (as of this writing) Day 6 and 7 WPC 500mb clusters for April 4th and 5th. Notice the really tight clustering on Day 6, with 4 clusters really “slicing the onion” pretty thin. We can confirm this by switching over to the EOF view on the same page (not shown here), and noting minimal shading, with minimal magnitude. We get a little better separation come Day 7 (April 5th), with exactly how this energy ejects into the plains. And that is a pretty important condition for severe weather. While the majority of EPS members fall into cluster 1, the remainder of members are pretty well spread out between the other three clusters, so we can likely use interrogation of EPS data as fairly representative of the range of possible solutions. So let’s dive into more EPS data.
One of the first things I think of to look at is CAPE. Even coarse, global models can give us a good view of the ingredients. Yes, we could look at mean CAPE, or some threshold probabilities (like probability of CAPE > 500, 1000), but this is an excellent chance to introduce something you might not have seen before: EFI (Extreme Forecast Index). Since the ECMWF keeps the reforecast of the EPS current, and uses that data to further post-process the ensemble, we can calibrate the model to itself (thereby effectively mitigating a lot of bias, with no obs!). We’ll look at ECWMF’s CAPE-Shear EFI from ECMWF Charts below.
Once again, we get two panel plots from ECMWF that provide us with lots of context. In the slides above we have plots for both April 4th (ending 00Z April 5th) and April 6th (ending 00Z April 6th). On the left side of the plots is EFI (shading) and Shift of Tails (contour). We won’t get into the SoT here, but both documentation and experience has shown that values of EFI around/greater than 0.8 are noteworthy, and indicate high ensemble support for a (model) climatologically unusual event. On the right side, we are given the 99th percentile of model climatology to help put that EFI into context. We can see a broad area along the Mississippi highlighted for April 4th, with that area beginning to lift to the northeast into the Great Lakes for April 5th. Remember back to our clusters and other comparisons – there was quite a bit of uncertainty as to how that upper-level trough ejects, and the ECMWF operational run gave us a glimpse of a more progressive, deeper, and more northern trough than the mean. This should definitely raise some alarm bells for these regions.
While we don’t have a similar product that I am aware of for the GEFS, we arguably might have something better: Colorado State Machine Learning Probabilities. This is a machine learning algorithm trained on the GEFS reforecast, run on the current GEFS forecast. There are links to their published papers on that page if you are interested.
As we can see here, for April 4th, the machine learning model has both greater than 30% (actually a smaller area of greater than 45% underneath the max marker), AND is highlighting significant severe potential (hatching). This continues to the north and east the next day. The value of these is they are tuned and can be interpreted similarly to the SPC outlooks. So, a similar alarm bell coming from the GEFS as well.
Okay, wrap it up!
It is important to note that although I have chosen to focus on April 4th and 5th, timing can, and likely will change at this time range. Even with this exact timing, precedent severe weather is possible further upstream on April 3rd. April 4-5th were chosen because of both the forecast range (Day 7,8) and remarkable ensemble support for severe weather. As always, I highly recommend following the authoritative source for severe weather predictions, the Storm Prediction Center, and if you happen to be in the regions that may be affected, it’s never too early to start preparing! I’d recommend starting at the NWS Severe Preparedness and Ready.gov pages!