Chase season predictors: urban legends or useful?

[Dan Robinson]

It's about that time of year when everyone starts looking at the drought monitor and CPC climate outlook for spring, and subsequently either freaking out or jumping for joy. But how reliable are those for predicting the quality of the chase season?

I decided to create these graphics showing side-by-side images of commonly-cited predictors of chase seasons and how the seasons actually turned out.

Graphic A:
2000-2017: UNL Drought Monitor as of early April/late May, Palmer Drought Index for April and May, Oceanic Nino Index, the April-June CPC climate outlook issued in March, SPC tornado reports for April, May and June and my subjective opinion on whether a May and overall season was above average, average or below average in quality. My opinion on "quality" takes into account the quantity of good chase events and availability of chaseable, photogenic storms/tornadoes during the peak season.

This is a huge image (6.5MB in size), so you will need to use this Dropbox link if you want to view it in full size (it's too big for the forum):

season-prediction-v3.png

Shared with Dropbox

www.dropbox.com

Graphic B:
Palmer Drought Index maps, Oceanic Nino Index and April, May and June tornado reports from 1999 through 1980. The "chase season quality" was assessed using the number of events chronicled in the Stormtrack magazine archives (veterans who chase during this time, please correct me if I am off on any of these!):



1980-1999 graphic full:

seasons-1990s-v2.png

Shared with Dropbox

www.dropbox.com

Spoiler: I see no reliable correlation between drought conditions, precip/temp outlooks and the quality of chase seasons. I think all the talk about dry conditions leading to poor seasons or "everything east of I-35" are urban legends that have entrenched themselves in the chase community. What do you think?

Sources:

UNL drought monitor archives:
Map Archive | U.S. Drought Monitor

SPC monthly reports:
New Rufsum Page - NOAA/NWS Storm Prediction Center

SPC SeverePlot:
SPC Historical Severe Weather Database Browser (SeverePlot 3.0)

CPC outlook archives:
Climate Prediction Center - Monthly Verifications

Palmer drought index maps:
Historical Palmer Drought Indices | National Centers for Environmental Information (NCEI)

 

[Jacob Hernandez]

Thank you Dan for this well thought out post. I've heard a coupe of times that the tornado season, at least for Oklahoma, may have a below average tornado count and number of severe weather days, but the events that do occur may be bigger than normal events. In your (or anyone else's) opinion, does that tend to be true with a La Nina like we have now?

 

[Warren Faidley]

Nice job Dan. One question: I believe the drought outlooks are just that and do not represent the actual conditions that occurred? Or maybe you are just suggesting the outlooks are misleading?

 

[Dan Robinson]

The drought monitors show the current state of droughts at a given time, that is, they aren't forecasts but realtime charts. For this graphic I pulled the first April chart plotted for each year (usually between April 1 and April 7).

 

[Greg Blumberg]

Interesting method of analysis. There are two issues I'm seeing though.

1.) "east of I-35" is a very big area and your maps span over the entire US. I think most people who mention that are KS or OK chasers and their statements are state-level assessments. I think you'd need to control for that in your analysis. It might help if you zoom into the Great Plains and denote I-35 on the maps. You could calculate the number of reports east and west of I-35 to better quantify these differences.
2.) As you mentioned above, the drought monitor isn't a forecast. It's only a forecast if you consider it a persistence forecast, which particularly for 2015 was a really bad forecast. 2015 had a lot of rains after early April that greatly eliminated the drought in Oklahoma. I think you need to include information about the trend or composite several drought monitor maps together.

There is evidence in papers that I've read that drought (soil moisture gradients) in Oklahoma does affect where the dry line sets up. I'd hope you'd get some sort of signal indicating that if you looked at drought over time instead of just early April. I don't think a drought map in early April has any predictive skill of season quality as it's a single data point and has no information about the variability of that map over the season. Try testing this using the trends in the drought map from April to June or breaking up the "east vs west I-35" analysis by every 3 weeks or so. You might be able to get an idea if the statement about "drought being an independent variable for tornado location" actually is useful at smaller timescales than 3 months.

I think your test is a good one, but I'm concerned that this experiment doesn't account for the various things that people actually mean when they say "drought influences chase seasons".

 

[Warren Faidley]

Another way to look at it...... The dust is good for photogenic landspouts and tornadoes.

 

[Brett Roberts]

I've probably been among the biggest believers in the idea of drought's impacts on chase season, so the long and short is that I disagree with some of your analysis.

A methodical study would be ideal to address this subject. Short of that, the conceptual idea I have of this phenomenon (largely from anecdotal observations over the past 10-15 years) shows up over the time period you addressed, from my point of view. It's just more nuanced than saying "drought years suck" or "everything is east of 35."

A few key points, from my subjective observations:

1. Plains drought years certainly don't have to suck for CONUS tornadoes. There's no reason activity in Dixie Alley, the Midwest, or other regions can't be prolific. If anything, the pattern encouraging Plains drought in the first place could easily favor this.

2. I would focus solely on *current* drought monitor status during a severe weather day, and even then, solely on *short term* drought (affecting vegetation, as opposed to hydrological concerns). This is really about evapotranspiration, most likely. The early April images in your rundown are not necessarily relevant to the heart of the season. In fact, both 2015 and 2016 are prime examples where drought was rapidly erased during April, leading to lush conditions out west in May/June (and plenty of tornadoes occurred both years in the late season).

3. Local drought conditions seem to have less effect on big synoptic systems with strong moisture advection ongoing during the event. There's no reason an event like April 26, 1991, or May 24, 2011, couldn't happen during severe drought.

4. As far as the geographical distribution of the "problem," I'd really hone in on the southern High Plains. If you look exclusively from I-70 southward and 99 W westward, the association between short-term drought and tornado days seems fairly convincing to me, over the years in question. The crude chase season scoring method I developed shows the "good" years for this region were: 2004, 2005, 2007, 2008, 2010, 2015, and 2016. Of those, only 2004 and 2005 had some mild, localized drought by late May; the others had virtually none. Meanwhile, 2006 and 2011-2014 were horribly dry and also terrible seasons.

The reason I feel some level of confidence about this is one I imagine other frequent Panhandle chasers share: the tendency for moisture to mix out during drought years on virtually every setup. In other words, it's not just some loose correlation in the data; the apparent consequence of drought, if it's real, is plainly visible (and very frustrating) when you keep driving out there and finding 88/54, setup after setup, in years like 2011-2014. Then, all the sudden, we get inundated with well timed April/May flooding rains in 2015-2017 and regularly see 65+ F dew points during May and June.

I would say factors like land use/vegetation, elevation, and geographical proximity to various airmass source regions could all modulate whether a given part of the Alley is prone to having their season ruined by bad drought. A place like the TX Panhandle with few trees, high elevation (and thus, typically, shallower moisture even in "good" conditions), and close proximity to desert could be more prone for physically meaningful reasons. Plus, those are simply the areas that have actually seen extreme drought most frequently over the period in question. I could probably ramble about this topic for another page, but will leave it at that for now.

 

[Quincy Vagell]

I've wanted to chime in on this topic as well, but decided not to until I looked at the data a bit more. I think it's a nuisanced subject and largely agree with most of what Brett said.

Just eyeballing the data, look at the early April periods that saw substantial drought conditions across the High Plains.

• 2014 - Very quiet across the Plains in April. May was relatively quiet too, but aside from May 11th, May was well below average for tornadoes as well. June was saved by the tornado outbreaks from the 16th to 18th.
• 2013 - Unusually quiet over the Plains in April (only one tornado in Kansas). May was a bit better, but even that month was heavily clustered near/east of I-35, aside from a few strong tornadoes in Kansas. (May was basically near average for tornadoes, overall)
• 2012 - The whole season from April - June was essentially dead, although the one decent event was a major outbreak. As Brett mentioned, a powerhouse storm system can overcome many challenges and still perform.
• 2011 - Only three tornadoes west of I-35/135 in April and all of those were within a few miles of that corridor. Dixie obviously had an historic month here... May saw zero tornadoes in West Texas/panhandles and aside from May 24th, there was well below average tornado activity as well across "prime" chasing terrain in the Plains.

Other years of note:

• 2015 - The drought was quickly eradicated later in April...
• 2006 - Drought conditions were broad, but not particularly extreme, but the chase season was another below average one. (not as bad as some years, as the High Plains did see a few notable tornadoes)
• 2004 - A mixed case. Note that there was virtually no drought whatsoever across Oklahoma and Texas in early April, although the central High Plains was amidst a significant drought to start the period. April only saw two strong tornadoes west of the Mississippi and those were both in eastern Missouri. May only saw three tornadoes in Oklahoma, although the month was a big performer from Nebraska into Iowa.

Aside from 2015, I didn't even take a deeper dive into what happened with the drought after early April.

The general theme appears to be that big events can happen during significant droughts, but that overall tornado activity is either:

• Displaced to the east of the drought area
• Delayed until later in the season
• Sporadic (a few events, but many dull periods/states with well below average activity)

 

[Dan Robinson]

I updated the graphic to add the drought monitor maps for late May. The new Dropbox link is here:

https://www.dropbox.com/s/xymwg293zenkwmy/drought-climate-v2.jpg?dl=0

 

[JF Massicotte]

I am still among the ones who would think dryer conditions would favor easier dryline mixing to the East and could also potentially keep LCL from lowering to an ideal level. However, I have to admit that it is not clear that we can conclude that it would waste a season.

I've only been chasing since 2006 but my experience from season to season was that synoptic weather pattern drives the season quality more than anything. For example, 2009 was wasted by that death ridge and eastern throughing sent 2012 to the Shitty Chase Seasons Hall of Fame. If we take the good seasons, most of them were well timed with a persistent West through and clean East coast ridge that had the GOM moisture pumping into the Plains right at the climatic maximum for tornadoes.


While I think drought conditions can certainly make the difference between a marginal day becoming magic and the same day staying crappy, but I don't think it is what makes a chase season.

 

[Randy Jennings]

Here are links to a couple of papers written on this subject:

The Effects of Antecedent Soil Moisture Anomalies on Tornado Activity in the United States
Ryann Wakefield, Esther Mullens, Derek Rosendahl, Harold Brooks

https://www.caps.ou.edu/reu/reu15/FinalPapers/Wakefield.pdf
https://ams.confex.com/ams/96Annual/webprogram/Paper284671.html

Seasonal Predictability of Tornadic Activity Using Antecedent Soil Moisture Conditions
Theresa K. Andersen and J. Marshall Shepherd
https://earthzine.org/2011/06/10/seasonal-predictability-of-tornadic-activity-using-antecedent-soil-moisture-conditions/

As for me, my gut tells me that drought is more a symptom of the synoptic pattern than it is causation of the resulting spring weather season (or lack there of). That doesn't mean we can't use it as an water cooler indicator, but as they point out in the Wakefield paper there are anomalies that are hard to explain.

 

[Dan Robinson]

Thank you everyone for the input! I would like to see a study on how a deep moisture plume from the Gulf is affected when it moves over a large drought-stricken area. I looked through a bunch of 2013 setups on the SPC events page (during a severe drought year in the High Plains), there are plenty of days where the moisture plume advects over the dry soil areas, with the dryline setting up in the Texas panhandle/western Kansas. The Rozel day is one such example.

http://www.spc.noaa.gov/exper/archive/event.php?date=20130518

Loading up the hourly HPC surface charts on those days, it doesn't look like dewpoints are cratering consistently during peak heating (I see it on a few of the days dropping ~5 degrees between 16z and 00z) but presumably a deep moist layer with a good supply fetch is relatively unaffected.

The scientific literature seems to point to at least some correlation between soil moisture and tornado counts, but I'm still a little confused/skeptical. Evapotranspiration surely impacts surface obs, but how much does it impact overall moisture depth? I'm not seeing any reference in the literature to soundings measuring how deep the ET effects go - for all we know, it could only be the first 50 feet above the ground. Presumably a week of a cornfield in the Midwest cooking in the sun would have a noticeable impact as opposed to a few hours in the Plains.

I think it's plausible that ET might insulate a setup a little from mixing effects and give a 1 or 2 degree boost to an already good setup (as JF mentioned). I'm having a hard time, though, understanding how a good fetch from the Gulf is going to be impacted by dry soil to the point of trashing a setup and/or pushing it 100s of miles east.

Very interesting subject!

 

[Greg Blumberg]

Hi Dan. I can speak to some your recent ideas as my dissertation is on the moisture evolution in the Southern Great Plains, particularly during the sunset hours. I've spent a lot of time looking at moisture depth using remote sensors that give you high-resolution soundings.

From what I can tell, ET doesn't play a huge role in actually putting moisture deep into the boundary layer during the daytime. In fact, it's very hard to see how deep the influence is beyond the surface layer (lowest 50-m or so) due to the strong mixing present in the daytime. However, when solar radiation heats the surface of the Earth the incoming radiation gets broken up into radiative fluxes (the Earth emits radiation), soil fluxes (energy goes into the soil), sensible heat fluxes (heat energy goes into the atmosphere), and latent heat fluxes (energy goes into changing the phase of water vapor...evaporation). The ET argument is typically associated with the latent heat fluxes, and it's assumed that the cause of larger dew points is that the ET is putting moisture into the atmosphere. However, the latent and sensible heat fluxes are linked through a variable called the Bowen ratio, and can be used to describe which of the fluxes is larger. The sensible heat flux magnitude is linked to entrainment at the top of the boundary layer and the depth of the mixed layer that develops during the daytime. So, in places where ET or latent fluxes may be strong, the sensible heat flux is much smaller and therefore entrainment at the top of the boundary layer is less than in places where ET is small. This suggests that increasing ET not only puts moisture into the boundary layer, but also helps reduce the rate at which moisture is removed at the top of the boundary layer by entrainment.

If you want, you can run the WRF model to test some of these ideas. WRF allows you to edit a lot of things in the simulation (e.g., land surface properties) so you can test to see how important land-surface characteristics are to the forecast. It's like The Sims for the atmosphere. There's a learning curve, but I know people who have gotten it running when they were in high school.

Some links:

https://en.wikipedia.org/wiki/Bowen_ratio
https://www.e-education.psu.edu/meteo300/node/790
https://www.colorado.edu/geography/class_homepages/geog_1001_f10/lectures/Lecture_5_surface_energy_balance_Molotch_guest lecture.pdf

 

[Greg Blumberg]

Regarding the WRF...here's how I'd do it:

Compile WRF. Pick a severe weather event (dry lines will be best...maybe May 23rd, 2011, there was a evening supercell in south central OK then that formed from a dry line) from the last 5 years or so. You should be able to find NAM, RAP, or GFS model data on the WRF website you can use to initialize a small nested domain (a smaller high resolution domain of the model in a larger one) of 3-4 km grid spacing around Oklahoma (can simulate convection there). Aim to start the simulation at 00 UTC the night before to give the model time to "spin up". Generate the WRF input files using real.exe (interpolates and initializes the data for the model to run.). Before you run wrf.exe (the executable that runs the model), edit the soil moisture parameters to dry out parts of the soil in the state (I don't remember what the variable names are, but it shouldn't be too hard to look up). Run the model with the soil moisture adjusted and without. The runs might take a little while depending upon your computer architecture. Compare the dry line positions from those two and/or convection that developed. WRF output files can be read and plotted easily using Python's netCDF reader and plotting packages.

It's a long process, but it's a more controlled test to get at what you're trying to find out with just observations.

 

[Brett Roberts]

Hi Dan. I can speak to some your recent ideas as my dissertation is on the moisture evolution in the Southern Great Plains, particularly during the sunset hours. I've spent a lot of time looking at moisture depth using remote sensors that give you high-resolution soundings.

From what I can tell, ET doesn't play a huge role in actually putting moisture deep into the boundary layer during the daytime. In fact, it's very hard to see how deep the influence is beyond the surface layer (lowest 50-m or so) due to the strong mixing present in the daytime. However, when solar radiation heats the surface of the Earth the incoming radiation gets broken up into radiative fluxes (the Earth emits radiation), soil fluxes (energy goes into the soil), sensible heat fluxes (heat energy goes into the atmosphere), and latent heat fluxes (energy goes into changing the phase of water vapor...evaporation). The ET argument is typically associated with the latent heat fluxes, and it's assumed that the cause of larger dew points is that the ET is putting moisture into the atmosphere. However, the latent and sensible heat fluxes are linked through a variable called the Bowen ratio, and can be used to describe which of the fluxes is larger. The sensible heat flux magnitude is linked to entrainment at the top of the boundary layer and the depth of the mixed layer that develops during the daytime. So, in places where ET or latent fluxes may be strong, the sensible heat flux is much smaller and therefore entrainment at the top of the boundary layer is less than in places where ET is small. This suggests that increasing ET not only puts moisture into the boundary layer, but also helps reduce the rate at which moisture is removed at the top of the boundary layer by entrainment.

If you want, you can run the WRF model to test some of these ideas. WRF allows you to edit a lot of things in the simulation (e.g., land surface properties) so you can test to see how important land-surface characteristics are to the forecast. It's like The Sims for the atmosphere. There's a learning curve, but I know people who have gotten it running when they were in high school.

Some links:

https://en.wikipedia.org/wiki/Bowen_ratio
https://www.e-education.psu.edu/meteo300/node/790
https://www.colorado.edu/geography/class_homepages/geog_1001_f10/lectures/Lecture_5_surface_energy_balance_Molotch_guest lecture.pdf

This is really fantastic information. It's awesome to see these topics being addressed in formal research! Even though this is not my area of expertise at all, my experience as a chaser and armchair observer over the years has made me a firm believer that there's something here, and it's just a matter of filling in the details. I'll be anxiously awaiting the final results of your dissertation and any related publications. In my personal opinion, having a firm understanding of how this all works -- and disseminating that information widely in the operational community -- could lead to a surprising improvement in short-medium range severe weather forecasts in the Plains.

I think your point about sensible vs. latent heat fluxes is absolutely crucial, and the way you describe it modulating entrainment at the top of the PBL could easily be the dominant reason for the phenomenon I was trying to describe in my post yesterday. When I see a setup in the High Plains and get worried because they're in drought, it's because I expect to see the PBL deepen and T/Td spreads to skyrocket during the mid-late afternoon period, from experience. If the vegetation is lush and they got lots of rain in March-April, I don't expect to see that happen, to nearly the same extent. All else about the synoptic environment and parcel trajectories being equal, of course.

As a chaser, what I most hate to see is a weakly forced setup in the southern High Plains with severe drought. It *seems* like setups associated with big cyclones and strong S/SE flow in the PBL could be less impacted vs. subtle days where air in the PBL is more "stagnant," but that's just speculation. Maybe one of the many interesting details to be explored as work on this topic evolves.

 

[Dan Robinson]

Thanks for that Greg, I will try that sometime.

The other factor to this question that comes to mind is that good tornado seasons in the High Plains and the Texas panhandle are relatively infrequent to begin with. We really haven't seen a season like 1995 that was so intense and focused in those areas in the modern era. Even in the 90s, most of the bigger events were closer to the center of Kansas/Oklahoma (thinking of April 26, 1991 and May 3, 1999). I wonder if there is a bit of a bias in chaser expectations for those type of seasons vs how relatively rare they are.

It would be interesting to look at drought data for the 1990s, but the UNL archives only go back to 2000. Anyone know of a source for older data? EDIT: I found this site, I'll try pulling some of this data.

https://www.ncdc.noaa.gov/temp-and-precip/drought/historical-palmers/

 

[Dan Robinson]

I added a second graphic showing Palmer Index drought maps and tornado reports from 1980-1999:

https://www.dropbox.com/s/4bivagn8tbaia66/drought-1990-v1.jpg?dl=0

If I can find some time this weekend, I'll go back and add a column to the original graphic with the Palmer maps just to see how they compare to the UNL drought monitors.

 

[Warren Faidley]

Excellent discussion. I for one wish I had more time to do the kind of data gathering everyone has done here. A lot of my chasing observations over the last 30+ years are done so from memory as I am terrible about keeping actual records. I will say there has been a shift of tornado activity in the western parts of TA over the last 10+ years. This is mainly because it's one of my favorite regions to chase given the visibility and road networks. The biggest change in tornado activity has to be in the LBB to MAF areas. I remember that was once an area I would travel to at least twice a year. Now it's rare to go once a year. I would love to see someone put together a moving map of tornado activity over the last 50 years to see if indeed the active regions have changed. I attended a lecture by Ted Fujita once where he theorized tornado activity shifted in circle every so many years.

 

[Tim Supinie]

I'm kind of interested in where this leads, but I'd like to point out some issues I see with Dan's analysis. Brett and Greg have touched on these a little bit, but I'd like to add my take.

One problem I see is with the subjective nature of what we're trying to look at. A "good" chase season is subjective, and one can have a good chase season with few opportunities or have a bad chase season with many opportunities. Additionally, an active Southeast season and a dead Plains season has fewer opportunities for me (based in Oklahoma), but way more opportunities for someone based in Alabama. You could try to use tornado reports as a quantifiable proxy, but lots of tornadoes doesn't mean lots of good chase opportunities (as you've alluded to). Maybe daytime tornado days within x miles of a point is a better proxy for a "good" chase season? I'm not sure. It won't account for visibility, but that's going to be really hard to objectively evaluate, especially before WSR-88D's.

Speaking of the time before WSR-88D's, another issue with taking this analysis back to 1980 is the tornado record. You'll probably want to consider only F1+ tornadoes because I think the F0 tornadoes are the most susceptible to being undercounted in the tornado record. Looking at only F1+ tornadoes definitely doesn't solve all your problems, but it's something.

Another item: the CPC outlooks I'm not sure are useful here. My understanding is that the CPC outlooks aren't incredibly skillful over a 3-month period, even when that period starts at the time of forecast issuance. And even then, "above normal precip" doesn't necessarily translate to "good chasing." I'm open to data, though.

Full disclosure: I'm of the opinion that, while drought may have a small effect on how the spring goes, the other large-scale-pattern-related factors tend to overwhelm the drought signal. Though, again, I'm open to data (which I'm not sure exists in this case).

 

[Paul Knightley]

This is great information - my thoughts have been that decent influxes of moisture from the Gulf won't really 'care' if the ground is dry or not. Perhaps thoughts about drought years being poor have also come from other kinds of thoughts - e.g. lots of dust making the viability poor, and hence giving the notion of a 'bad' year. Of course, if a link were found, then maybe it would have more to do with the prevailing long wave pattern which caused drought through the winter persisting into the spring and being unfavourable for severe weather. However, I do think there is something of an urban myth about it.

Personally speaking, as someone who has to pick their chasecation months in advance, the immediate synoptic set-up of those two weeks is much more important than anything else - I'll wager for many other people who aren't always on tap to chase, this is true too.

 

[Dan Robinson]

I updated the original graphics to include Palmer Drought Index maps for April and May as well as the Oceanic Niño Index plot through the years shown on the graphic (shows El Niño and La Niña state). The ONI plot is continuous (start of each year is at the bottom of the cell, end of the year at the top) so chase season is roughly 1/3 to 1/2 of the way from the bottom of each row.

These are much larger graphics that exceed the limits of JPEG compression - to preserve clarity, the export is in PNG format -

2000-2017 graphic full 6.5MB image is here:

https://www.dropbox.com/s/szlzrb2cv5oytvy/season-prediction-v3.png?dl=0

a lower-colorspace version is here, about 2MB:

https://www.dropbox.com/s/znv3udsg8x9klat/season-prediction-v3-2.png?dl=0

1980-1999 graphic full:

https://www.dropbox.com/s/dmcmo2v8uaf8rlm/seasons-1990s-v2.png?dl=0

lower colorspace version:

https://www.dropbox.com/s/imjej3c24j170fx/seasons-1990s-v2-2.png?dl=0

I also edited the title of the thread since this is evolving more into a general "chase season prediction" exercise.

 

[JF Massicotte]

I updated the original graphic for 2000-2017 to include Palmer Drought Index maps for April and May as well as the Oceanic Niño Index plot through the years shown on the graphic (shows El Niño and La Niña state). The ONI plot is continuous (start of each year is at the bottom of the cell, end of the year at the top) so chase season is roughly 1/3 to 1/2 of the way from the bottom of each row.

This is a much larger graphic that exceeds the limits of JPEG compression to preserve clarity, so the export is in PNG format - full 6.5MB image is here:

https://www.dropbox.com/s/szlzrb2cv5oytvy/season-prediction-v3.png?dl=0

a lower-colorspace version is here, about 2MB:

https://www.dropbox.com/s/znv3udsg8x9klat/season-prediction-v3-2.png?dl=0

I also edited the title of the thread since this is evolving more into a general "chase season prediction" exercise.

Not sure that I like what we see if we try to make analogues out of the oceanic nino index. However, the A-M-J CPC outlook this year is very similar to the 2013 one.

Current:




2013:




So many contradicting signals. I guess that we'll have a better idea in one month!