Thursday, April 24, 2014

Analogs to Saturday's potential for severe weather

As many people are now aware, this weekend looks to have some potent severe weather in the southern Plains.  The SPC has a day three slight risk for a wide area of Nebraska, Kansas, Oklahoma and Texas on Saturday:

 The model forecast fields are pretty good---a strong upper-level trough digging across the central Rockies with a powerful jet streak entering the base of the trough on Saturday evening.  Divergence downstream from the trough axis looks to provide broad lifting support across a very moist warm sector, in addition to enhancing strong lee-cyclogenesis in eastern Colorado.  Strong directional wind shear is forecast with southerly flow at 850 hPa to near westerly flow at 500hPa. All classic ingredients for severe weather.  Here is the NAM 60-hour forecast from this morning's run:
We're looking at surface pressure and precip in the upper-left panel, 850hPa heights and winds in the upper right, 500 hPa heights and vorticity in the lower-left and 300hPa heights and winds in the lower right.

One method that I've occasionally mentioned for predicting the eventual outcome of a particular forecast is to look for analogs in the historical record.  That is, to go back in time to when the atmosphere was in a similar state (or, better yet, when the model forecasted a similar state) and see exactly what ended up happening.  There are groups that do this sort of thing in real-time.  One is the Cooperative Institute for Precipitation Systems (CIPS) group at St. Louis University.  With every model forecast they go back through the North American Regional Reanalysis archives and try to find days from the past when the forecasted pattern was similar.  We can then look at what actually happened on those days to get a sense of what may happen now.

Take this Saturday, for instance.  If we go to the CIPS southern plains sector, we see that their number one analog for the NAM forecast for that evening is 00Z, April 26, 1984.  Here's the same four-panel plot as above, but showing what happened at that time:
You can see a lot of similarities in these patterns! Both had similar strong lee cyclogenesis at the surface in Colorado underneath a broad upper-level trough over the intermountain west. It makes sense that these are a close match.  So what ended up happening on April 25-26, 1984?  Here are the severe weather reports from that day:
Five tornado reports, mostly in Nebraska.  Lots of hail reports in Nebraska with a few scattered elsewhere.  At least we know this pattern is capable of severe weather.

Let's look at the second-best analog identified by CIPS: 00Z, April 10, 2011.  Here's the pattern:
Still similar, though the differences are more apparent.  The surface low in eastern Colorado is not as intense.  The upper-level trough is also much narrower and more positively tilted.  Still the same "overall" pattern, but there are noticeable differences.  What happened on this day?

There were still hail reports in eastern Nebraska with tornado reports in western Iowa.  There's another whole area of wind and hail (with a few isolated tornado) reports located across the southern Appalachians.  It seems that subtle shortwave in the upper-level analyses over West Virginia-Virginia may have played a bigger role than one might have thought.

Finally, the third closest analog---00Z, May 4, 1999.  That's the evening of May 3, 1999.  You know what's coming.  Here's the pattern:
Again---there are definitely differences.  The 500hPa trough for this event isn't as sharply defined as in the current NAM forecast.  There's also a jet streak across the central Midwest in the current forecast that was not there on May 3rd, 1999.  These differences are pretty important.  With stronger and sharper large-scale dynamics being forecasted for this weekend, I feel like the severe weather may organize more upscale more quickly than it did on May 3rd, 1999.  Such strong synoptic-scale dynamics can work against maintaining discrete storms for very long.  Anyhow, here's the storm reports from May 3-4, 1999:
This was the day of the first Moore F5 tornado.  Lots of severe weather throughout Oklahoma and, generally, in the same corridor that the SPC has outlined.

The CIPS site also breaks out the analogs by individual variables and how closely they "match" the forecasted pattern.  For instance, if we're interested in moisture return, the May 3-4 event has the highest correlation of these three in the surface dewpoint field.  The 1984 event, however, has the higher correlations for overall upper-level patterns and winds.  So it's a mix and match sort of deal...

How much can we trust these analogs to tell us what's going to happen?  You can see that in just the top three closest events there are already notable differences between the patterns on those days and the current NAM forecast.  Plus, as each new forecast run gets out, the closest analogs change.  We also have only a limited record (some 40-50 years) that we can go back to look for close matches.  Furthermore, in just three top analogs we've identified very different regions as areas of high impact.  What we CAN say is that this sort of pattern definitely is conducive to severe weather, and in roughly the area the SPC has outlined.  Beyond that, you still need your expert forecasting skills to help narrow down the exact areas favorable for severe weather with this particular setup.

1 comment:

  1. This pretty neat. A reported tornado in Nebraska from the april 1984 model is almost smack dab where a tornado just landed about 30 minutes ago.