Weather Ready Nation: Tornado Warning Frequency

Posted by on 13 Dec 2011 | 23 comments

Today kicked off the first day of the Weather Ready Nation: A Vital Conversation. (OU is recording and posting the presentations on the web.) Dr. Harold Brooks of the National Severe Storms Laboratory really got things going with a presentation on our known challenges. One of his main take-away points was that the number of tornado warnings issued has dramatically increased in the recent era. To illustrate this point, I provided two county-level heat maps (below). The top figure is the average number of tornado warnings per county per year from 1986-2007. The bottom figure is the same figure except for 2008-2010. As you can see, the average number of tornado warnings per county per year has increased almost uniformly across the county, although the increase is much larger in some areas and almost non-existent in others. Whether or not this is an improvement in National Weather Service “service” is one of the topics open for discussion in the following days.

I’m sure I’ll create more figures in the coming days.

1986-2007_tors_by_county_0 2008-2010_tors_by_county_0
  • Lenholliday

    Patrick, You are doing a great work! NOAA has to be proud of you! Len Holliday

  • http://twitter.com/nsj Nate Johnson

    So, obviously, one thing changed in 2007. But did anything else? For example, are offices issuing more warnings now, either because they’re smaller (but may still touch multiple counties), because of an increase in actual tornadic events, or something similar? Do we have any stats on the amount of mile-minutes under warnings (e.g. 100 mi^2 * length of warning) and whether that has changed?

    I don’t doubt that when SBWs are reduced to a list of counties, the number of counties goes (way) up. Actually, this was discussed at one of the early SBW workshops back in 2006 as a known issue. What’s surprising is how poorly some channels (e.g. TV, radio) have handled the polygons in some situations.

  • http://www.facebook.com/profile.php?id=100002612903860 Kelton Halbert

    I’m curious about whether or not there is a proportional (or maybe not so proportional) increase (or maybe a decrease) in tornadic activity in the same time frames? In other words, is this increase justified quantitatively by the number of tornadoes?

  • http://www.patricktmarsh.com pmarshwx

    In my opinion, there are several factors contributing to the increase. Yes, you might now issue 2 storm-based polygons in a single county, whereas a single county would work in the past. However, I have seen enough subjective evidence to cause me to think there is something else there. I’m working on teasing that out…

  • http://www.patricktmarsh.com pmarshwx

    We have to be careful here. The number of reported tornadoes is increasing, however, that does not mean the number of actual tornadoes are increasing. We do a much better job identifying F/EF0 and F/EF1 tornadoes now than we did in the pre-88D period. This does play a role, but as I replied in another comment, I have enough subjective evidence to suggest an increase in warnings without an appreciable increase in the threat. I’m working on objective measures now.

  • http://www.patricktmarsh.com pmarshwx

    Now if I can only convince someone in NOAA and/or the NWS to hire me… =)

  • James Correia

    Ok, so what is the percent difference, by county, between the periods 86-07 and 86-10? I know what you are showing but directly comparing a 3 yr average to a 22 yr average is not enough.

  • Gegco

    These maps tell one story – for county-based waring dissemination systems (e.g., sirens, Wx Radio, tv bugs), they are being exercised more frequently in the storm-based warning era (2008-2010). Number of warnings equates to number of times the radio alerts, or the sirens sound. But there is some underlying data here that needs to be exposed that Nate mentioned. Can you add three more sets of maps?:

    1) Create similar maps that show *number of minutes* under tornado warnings.
    2) Show the actual polygon area coverage on the 2008-2010 map, instead of the full county filling.
    3) Repeat #2 but also showing number of minutes under warning. Account for removal of warning area via SVSs too.

    -greg

  • http://www.patricktmarsh.com pmarshwx

    Greg, all of these maps are under development, as well, as other kinds. The problem is that the amount of CPU and memory required for these computations is more than I can do on my MBP without blowing it up. I’ll need to port all these datasets to the NSSL process machines.

  • Lenholliday

    Give them my name and #. After I talk to them, they will hire you. Sen. Lindsey Graham from S.C. is a very good friend of mine. I’ll be glade to help if you need me! Len Holliday lenholliday@yahoo.com or 864-245-3587 THANKS!

  • http://www.facebook.com/profile.php?id=100002612903860 Kelton Halbert

    Very interesting. I failed to think about the pre-88D detection.
    And I agree with what you are saying – I’m looking forward to what you can come up with objectively, as well as subjectively.

    Maybe if there was a way to quantify the false alarm rate for each county and compare it to the increase? Might be lengthy but interesting. Could also be interesting to do with the tornado emergencies.

    I’m even more curious where you are trying to go with this research – first the tornado emergencies, and now this :) . Count me as intrigued.

  • Dave Whiteis

    Ok, I know that I am just a weather layman, but is there similar data available of confirmed tornado sightings or damage? Is the bump 2008-2010 a function of better meteorology, communications or other factors, such as concerns of ‘failure to warn’. Just asking…

  • Ben C

    Looking at the bottom figure, it’s interesting to see the pink county just to the east of Fargo. It seems like it’s been a busy couple years up here for tornadoes but I don’t have any numbers handy.
    I think there could be a higher number of warnings issued for that county due to a combination of 1) Storm locations (obviously) and more importantly 2) Distance from radars. While it’s closest to the KMVX radar, the beam height would be up there a bit. Same story and even worse for other radars in range, KABR and KMPX.
    I guess it would be interesting to know whether the tornado warnings were radar indicated or confirmed by spotters/general public. If majority were radar indicated, then I wonder if strong rotation aloft in that county would trigger a “benefit of the doubt” warning (if you will). Better safe than sorry when you can’t see the lowest levels of the storm. Then again, I have no clue what the warning criteria is, just a thought really.

    In the end it could be a coincidence as well. Interesting stuff and I look forward to figures to come.

  • http://www.patricktmarsh.com pmarshwx

    The data do exist and I’ll work on those in the coming days. I firmly believe that the increase in tornadoes isn’t meteorological. In other words, the number of tornadoes isn’t increasing — they’ve always been there, we’re just now seeing them. The increase in warnings is most likely an attempt to try and “get” all these weaker, until now, unseen tornadoes.

  • MNWx

    I highly doubt the significant numbers of warnings in Otter Tail county is related to radar coverage or a meteorological reason, but more due to the verification and warning culture of that specific area.

  • Seth

    Very interesting, Patrick. A couple of observations:

    1) 48 TDWRs have come online in the NWS since 2007 – we see a whole lot more detail/circulations with those (who have them) than with the 88-D (and lower to the ground too)
    2) I would like to see this broken out more. Pre-88D. Post 88-D. Post TDWR. Lack of TORs (detection) in pre-88d era likely bringing those top graphic numbers numbers down.
    3) I wonder how the % of warnings has increased with mode of communication – eg what has been the % increase of cellphone usage since 2007 in our population. In the past, many may have seen a funnel or strong rotation but had no way to immediately call it in. Now you’re the odd man out if you don’t have a cell phone. If environment is good, and radar neutral, that spotter/public report may be the tipping point to warn.
    3) What is going on on the GA/AL state line? That’s probably the most interesting thing off that bottom graphic.

    Glad to see these issues being looked at…

  • Seth

    Also super-res velocity products were deployed 2008. Probably lots going into the increase – technology (radar and comms), SBWs, better knowledge of environments/mesoscale analysis etc and applied research? Just guessing…

  • Gegco

    We need to be careful about making any conclusions on how warning frequency is tied to climatology or changes in radar data until we see what these maps look like for actual coverages of the warning polygons, rather than “counties intersected”. My 2nd and 3rd suggestions below. -greg

  • Jeff Duda

    The theme of my comment has already been brought up, but I wanted to point out something different about it. A few years ago I did a very similar thing for a project in a GIS class. I found that the only way (given my knowledge of the software) to create a map of warning density was to make a count for each time a storm-based warning overlapped a county. What I really wanted to do with the project was to simply show the counts of the number of times each specific point was under a storm-based warning, regardless of county. Unfortunately I was not able to find a way to do this. Can you? I’m thinking of either creating a grid (1km by 1km) and applying the counts on that grid, or simply overlay ALL of the storm based warnings in a period on top of each other, but get the software to color in pixels based on the number of warnings that lay over that pixel.

  • http://www.patricktmarsh.com pmarshwx

    Hi, Jeff,

    I already can do the differing color based on pixel by pixel. I have those maps created for each year and will work on getting those posted in the next day or so. I’m always currently working on code to grid each warning on a 1km-by-1km grid. However, that’s a non-trivial computational task. I’m working on ways to try and speed up this process a bit.

  • Brad Barrett

    Patrick,

    I’m late to the discussion- sorry! When you compute the avg # of tornado warnings per year, I assume that pre-2008, you just counted the # of times a tornado warning was issued for County “X” during a year. Simple enough.

    What about post-2007 though? Example: A polygon is issued for McClain County Okla for a tornadic storm moving hard NE, and the northeastern part of the warning just clips the NW corner of Cleveland Co. Forty-five minutes later, for the same storm that has now turnedhard east, a tornado warning polygon comes out that’s now entirely in Cleveland Co. Does that count as 2 “hits” for Cleveland Co.? If so then I expect your 2008-2010 numbers will be significantly affected by that operational change – whereby prior to 2008 the NWS forecasters would have been more careful in their county warnings. (I.e., wait to warn Cleveland Co. after it was clear the storm was actually going to enter Cleveland Co., rather than miss it to the NW). Your thoughts? (And again sorry if you answered this earlier- I’m still catching up on these posts. Finally can you be so kind to ping me on FB if you reply- I don’t know if your website will notify :) Thx!!)

  • http://www.patricktmarsh.com pmarshwx

    Hi, Brad,

    For the figures here, each county is color coded based on the number of times the county itself is warned. So, yes, you are correct in that Cleveland county would be warned twice in your scenario. Undoubtedly this would play a role in the counts when comparing post-2007 and pre-2007. Unfortunately, there isn’t too much we can do to get around that at the county level. In subsequent posts I’ve attempted to do somethings at the WFO level — and still have more to show, including areas, duration, area-hours, etc.

  • Greg Stumpf

    One way to interpret these maps is a count of the number of times your weather radio will alert. So yes, in the above scenario, my weather radio tuned for Cleveland County alerts, will alarm twice.