Monday, November 12, 2012

Hurricane Sandy: Life Takes an Ironic Turn

November 15, 2012: The Sandy Service Assessment, discussed below, has been abruptly cancelled by the NWS as of today. Details here


ORIGINAL POSTING BELOW:


Even at the age of 60, I'm often surprised by life's turns.

On November, 2, I wrote a piece on this blog titled, "The National Weather Service Should Not Investigate Themselves." I talked about the NWS starting the process of doing a self-assessment of its service during Sandy and thought it would be better handled by an outside agency.

Last week, I received a surprising phone call from NWS HQ in Washington: Would I be willing to co-lead the service assessment?

After some consulting with Kathleen and AccuWeather, I agreed to do so. My co-leader is Ms. Nezette Rydell, the meteorologist-in-charge of the National Weather Service in Denver.

Hurricane Sandy is a historic event both for our nation and for weather science. We need to know what worked, what didn't, and how to improve our services in the future.

We officially begin our work tomorrow. The team imposes silence on itself until the final report is released. So, you will not be seeing fresh postings about Sandy for the next few months.

The team certainly welcomes your good wishes as well as your input. If you have suggestions for things you would like us to look into, please post them in the comments below.

/s/ Mike

15 comments:

  1. Congratulations, Mike! Your writing and thorough analysis of the success and failures of the whole warning system makes you the ideal person for this role!

    I'll be looking forward to the release of the report next summer (at least that's what I'd expect, about 6 months).

    ReplyDelete
  2. Mike,

    Congratulations on this. This is great news for a lot of us hoping to get deep into what happened here.

    A couple of the things I'd like to know...the NWS WFO's in the Northeast (OKX, Mt. Holly, Boston, etc.) all seemed to do an outstanding job both communicating the risks and making the forecast. With that in mind, besides the obvious question of why, technicalities aside, hurricane warnings were never issued, why were so many people caught offguard? Was it how people received the communications and info? Was it fatigue from Irene not hitting as hard as it was made out to be and making people complacent? Was Mayor Bloomberg at fault for downplaying it over the weekend and waiting too long to evacuate Zone A? Were the TV meteorologists not communicating the info clearly enough?

    In my work/forecasting for my job in the days leading up, a lot of people took it like Irene and their concerns were a.) inland rainfall and b.) wind gusts. Surge wasn't considered the main concern until I made it one. I suspect Irene complacency may be a large part of what happened here, but we'll see. I guess the logical follow up would be even if Sandy were classified as a hurricane, would people still have compared it to Irene and assumed it wouldn't be that bad?

    I don't know where the breakdown occurred or if it was just human instinct ("the last one wasn't so bad, so this one won't be either") in the way. But I wish you good luck with this and hope the answers you all find help the community going forward.

    ReplyDelete
  3. Hi Mike,

    I'm new to your blog and was cross-referenced from a Facebook user group. First, congratulations on the assignment! Having worked with Ms. Rydell for a number of years around the turn of the decade, you'll be teaming with one of the best in the business.

    I've posted in a number of areas during and after Sandy on the following topic: Regardless of whether hurricane warnings should (or shouldn't have) been issued, I've wondered why the following experimental tropical cyclone impact tools were not used: 1) Tropical Cyclone Impact Graphics (TCIG; http://weather.gov/tcig); 2) Probability of Hurricane Inundation from Surge Height (PHISH; http://www.nws.noaa.gov/mdl/phish/ . TCIG uses algorithms that consider deterministic, probabilistic, and forecast confidence to produce potential impact levels (none, low, moderate, high, extreme) for wind, storm surge, inland flooding, and tornadoes. For storm surge, tide information can be added to increase or reduce potential impact. Output is provided in real world language based on water depth above ground level. PHISH does not include tide, but converts storm surge probabilistic information from vertical datum or MLLW into water depth above ground level.

    While it is impossible to say at this time whether conveying storm surge in terms of inundation (i.e. "8 feet of water will flood above the level of a one story home") would have improved the message being communicated, it is a question worth asking. Would some fence-sitters have decided to evacuate based on real-world language? Or was the power of "confirmation bias" so strong, based on lifetime(s) with no similar experience, to overcome even the best visualization and conveyance of the threat? Past studies (Morss and Hayden, 2010) from Hurricane Ike suggest that at least some "fence sitters" moved to evacuate based on potential impact wording.

    Thanks for any consideration the assessment team gives to this question. Looking forward to answers as the process moves ahead.

    Best regards,

    Barry Goldsmith
    Warning Coordination Meteorologist
    NWS Brownsville/Rio Grande Valley, TX

    ReplyDelete
  4. Barry and Matt,
    Welcome to the blog. Thanks very much for the input. This will be shared with the entire team.

    ALL READERS:
    You don't have to be a meteorologist. If there is something about which you are curious w/r/t Hurricane Sandy, please feel free to post it here.

    Mike

    ReplyDelete
  5. Mike, I recommend that your team talk to Amanda Ripley, author of "The Unthinkable - Who Survives When Disaster Strikes - and Why." As it applies to your task, she has some good insight into how people react to risk and warning, including the people responsible for issuing warnings. One of the points she makes is that government officials tend to "panic about the public panicking" and under-warning. She also explains why a percentage of people, despite all warning and evidence, will paralyze.

    Having worked at a response agency for 20 years, I have witnessed this sort of thing often. Ripley's book was a great synthesis and explanation for all that I had observed.

    ReplyDelete
  6. Hi! This sounds related to Sassenach's post above, talking about the extreme pressure not to over-warn. I've been reading about how large organizations behave under threat, specifically with reference to the USS Cole attacks and the Columbia disaster. Here's a blog link: http://waynehale.wordpress.com/2012/11/12/after-ten-years-a-few-words-from-admiral-gehman/

    The speech the quotes are drawn from, given by Admiral Gehman, who led some of the investigations: http://www.usna.edu/ethics/publications/documents/GehmanPg1-28_Final.pdf

    ReplyDelete
  7. I have a few thoughts:

    1) On the warning process, I believe that the status - tropical or post-tropical - should not be as heavily considered in operational advisories. In this case, it is likely that the warnings - although technically accurate - were not perceived by the public as "serious" when in reality they were. At the same time, extreme extratropical storms can and do occur elsewhere (i.e. west coast, Alaska) but do not impact nearly as high of a population.

    Recommendation: Consider renaming High Wind Warning as "Tropical Storm Force Wind Warning" (39-73 mph sustained, or 58+ mph gusts) and "Hurricane Force Wind Warning" (74+ mph sustained), used for both tropical and non-tropical systems. Evaluate whether such would create confusion in non-tropical systems though.

    ReplyDelete
    Replies
    1. That's a real good idea, Craig. There are marine headlines called Hurricane Force Wind Warnings already for extratropical systems, so I can't see why these wouldn't work over the land as well.

      Delete
    2. Yeah thanks. The only issue I see is whether public confusion might arise that a REAL tropical storm or hurricane is approaching (as opposed to another storm of similar intensity), although the impacts would be similar regardless. That confusion reigned after the Northwest storm of December 2007 (it is posted in that assessment). At the same time, it could also strengthen the call to action for such, since the fact of people ignoring them has caused problems in the past (look back to the Hurricane Ike inland windstorm where like 30 people died far from the coasts in the high winds).

      For areas normally prone to very high winds, the TS force wind warning could be ignored during marginal events.

      Delete
  8. 2) Likewise, Hurricane Local Statements were not issued despite the tropical origins. That left it up to Public Information Statements and Special Weather Statements. A lot of data was not included or spread over multiple products. The transitioning status complicated things significantly.

    Recommendation: Introduce a new product, Major Storm Local Statement, to be similar to the HLS and issued in major storms. It would be issued when certain criteria are met - examples include widespread wind events, major flood events (i.e. Nashville 2010), major snow events and tornado outbreaks.

    3) The hand-off from NHC to HPC seemed quite orderly, but there was one issue on timing. Tropical Cyclone Updates had to be issued as Sandy was approaching to maintaining 2-hourly advisories. HPC advisories are 6-hourly.

    Recommendation: When winds are tropical storm force and the storm is over or near land, HPC advisories should be 3-hourly. When winds are hurricane force, HPC advisories should be 2-hourly. Intermediate advisories would be much less detailed than regular advisories.

    ReplyDelete
    Replies
    1. I'll add this to 3:

      Best Practice: Despite not being operationally accurate as no coastal tropical warnings were in effect, the NHC used 2-hourly advisories in the lead-up to landfall using Tropical Cyclone Updates.

      Delete
  9. 4) Very high winds and storm surges were also reported on the south shores of the Great Lakes. Tropical warnings cannot be issued there, and storm surge products do not really exist other than the Lakeshore Flood products. No NHC or HPC products mentioned flood levels on the lakes.

    Recommendation: Allow the use of tropical marine products on the Great Lakes (i.e. use Tropical Storm Warnings, or on unlikely occasions Hurricane Warnings, if necessary.) Also show storm surge levels on the lakes if warranted.

    5) NWS Mount Holly clearly had the right idea when they issued strongly worded statements leading up to Sandy (including this blunt one: http://hnbbs.net/?p=5036), and also issued online briefings starting 5 days out from the storm. No other office, from my observations, did so. Very few fatalities were reported in the Mount Holly CWA despite severe wind and flooding.

    Best Practice: NWS Mount Holly went on their own initiative and gave a clear indication of the threats starting 5 days before the storm, and clearly identified the threat level, in a very personal tone.

    Recommendation: In the lead-up to major events, briefings should be issued by all offices, initially once daily but more frequently as the event approaches.

    ReplyDelete
  10. Hi Mike, congratulations to you! I look forward to the report you and Ms. Rydell produce.

    Many of the thoughts I would have expressed have been voiced already in the comments above (particularly Matt's and Berry's).

    Complementing (IMO) Thomas Smith's post above, I found the rationale for the lack of hurricane warnings for the landfall area shared some similarities to those provided to the Tropical-Storms community (http://tstorms.org/tropical-storms/) after landfall of 93L 09Oct11, suggesting a growing comfort within the NHC with leaving the warning on complex (nontraditional) systems to local WFOs (unnamed 1991 hurricane, 24-45Sep08 unclassified STS, and the 09Oct11 event for a few key storms). What is further troubling is that from the outside at least, it appears this decision then influenced the classification of the system upon landfall more so than did meteorology.

    I don't know if this second point will fall within the scope of your work, but the past four years have illustrated repeatedly the weakness of our current classification scheme for the purpose of warning the public of the magnitude of land-falling storms. Ike, Irene, Issac, and now Sandy all were massive systems whose power (IKE if you will) and destructive potential (both wind and surge) were far greater than was suggested by their SS scale classification using only max sustained winds. People are repeatedly dying for lack of understanding the true threat such approaching storms are presenting.

    ReplyDelete
  11. Craig and WxJAK, thanks for the excellent input.

    Readers: Feel free to keep the suggestions coming.

    ReplyDelete