Saturday, August 27, 2011

A benefit of probabilistic surge forecasts - an example from Irene's impact on NYC

Here's a good site for surge predictions from SUNY Stonybrook.  The latest forecast for the Battery shows a good surge of 1-1.5 m above normal level in the early morning on Sunday.  You can see this forecast below where the dashed line is the expected forecast within a grey zone that represents the standard deviation of all the surge predictions at the Battery.  The red trace represents what the observed surge happens to be and the differences between observations and forecasts is represented in green.  Notice that the forecasts so far have been underestimating the observed water level most of the time.


The forecast surge height is one type of forecast but that doesn't tell how high the water will be relative to mean sea level.  Afterall, if the surge hits at low tide then the impact isn't as high.  So Stonybrook offers this second forecast timeseries superimposing the forecast tide on top of the surge and below is that forecast.


The timing appears bad if you don't want a flood.  The peak surge that you saw above happens to coincide with the morning high tide at the Battery yielding a water level 7 to 9 feet above mean low water.  This forecast is confirming why the New York City government placed mandatory evacuations of all low lying areas in their Zone A.  To see where Zone A exists, check out the city government site's interactive map at http://project.wnyc.org/news-maps/hurricane-zones/hurricane-zones.html

What is nice about these forecasts from Stonybrook is the grey zone of uncertainty.  It allows us to plan for reasonable worst case scenarios because forecasts have uncertainty.  But is the zone broad enough?  Will there be a surge higher than the grey zone?  The answer is quite possibly yes but the probability may be somewhat low.

Another site is available that shows the probability at which a surge is forecast to exceed some height, and also the surge heights as a function of probability.  I like the latter because it allows me to set my threshold probability for taking action and then I can see if the surge height exceeds my elevation.  Getting the forecast is easy.  Just go to http://www.nhc.noaa.gov/ and then click on the storm (Irene) and storm surge probabilities are listed above.  I choose the probability of a 5' surge and you can see the results below.  The forecast shows a 50% probability of a surge exceeding 5 feet in the New York harbour, even more around Staten Island.



  Of course the problem is how cautious do you want to be?  If you were sitting at 5 feet above the water would you move if the probability of you going under is at least 50%?  That's probably playing it a little too aggressive in my opinion.  Fortunately there's a website related to the one at  the NHC that allows you to dial your own probability.  Go to http://www.weather.gov/mdl/psurge/ and select the drop down menus until you find "20% exceedence height".  Now you see that the surge height forecasts are higher, 5 - 7 feet around the Battery to almost nine feet in western Staten Island.


So if you're a little more cautious, like me, perhaps you would dial in your personal probability threshold at 20% and you'll see that if you're below 5-7 feet above the water, you may want to move.  So to put it all together, here's what I'd do to decide for yourself if you need to move.  Note that I'm assuming you have no outside call for evacuation.

1.  Know my altitude above Mean Low Water
2.  Look at the tide forecasts closest to your site at http://tidesandcurrents.noaa.gov/gmap3/.  Click on the forecast point to get a menu and click on 'Tide Predictions'

3.  Determine the height above mean sea level (set at zero) for a set time.  Here I choose 8 am on Aug 28 and I get 5.5 feet.  That's high tide.  

4.  Then look at the probability of exceedence for at least 20%, preferably 10% from http://www.weather.gov/mdl/psurge.  We saw from above that 5-7 feet, so let's choose 7 feet.  

5.  Then add the value you got from 3 and 4.  In this forecast, 12 feet.

6.  Compare what you got from 5 with your altitude above mean low water.  If what you see from 5 is higher, then you make the decision to move.

Better yet, listen to the authorities because I do not espouse taking these steps above as the way to make a critical lifesaving decision when the authorities have so much more information at their disposal.  NYC produced a really nice evacuation map at http://project.wnyc.org/news-maps/hurricane-zones/hurricane-zones.html


BTW, the surge at the Chesapeake Bay Bridge was about four feet at 6 pm today.  That compared to a 20% exceedence of 5-7 feet.  So yes, you may have stayed dry if you were at least 8 feet above the high tide today which was 3.5 feet.  But why take chances?


Sunday, August 14, 2011

The collapse of a stage at the Indiana State Fair 13 Aug 2011

Note:  I've added an addendum at the bottom of this post


Concert stage collapses from sudden wind gusts happen all to frequently and now we hear about another one occurring at the Indiana State Fair last night around 8:50 pm EDT.  The latest one was tragic with the loss of 5 people and multiple injuries, some of them life altering according to the Indianapolis Star.
To understand what happened, at about 8:50 pm, a gust front from a line of thunderstorms struck the stage just to the north of the main grandstand before the Sugarland concert was to begin.
This youtube video shows all the graphic detail of the stage collapse which unfortunately fell beyond the platform and into the VIP seating area full of people.




This tragedy was entirely preventable but a series of missteps aligned to put people in harms way. First, the stage was a 'house of cards', in other words, a flimsy metal scaffolding frame supporting a huge area of fabric facing the wind.   Second, it appears that people in charge of safety at the fair were determining when people should evacuate based on their uninformed interpretation of the meteorological data.  Mike Smith of WeatherData inc. eloquently described the issue with nonexperts serving the role as experts.  Needless to say this night proved how dangerous that can be.  Their hearts were nonetheless in the right direction as they began evacuation procedures.  This leads to a third problem.  Once they initiated evacuation procedures, not everyone was on the same page as the Daily Star reported that the WLHK program director addressed everyone with a mixed message, one that suggested to prepare for evacuation, the second that the show will go on.  Of course the very people closest to the stage would be the same ones least likely to take the first advice and most likely to stay on.

I first explore how strong the winds were likely to have been when the stage collapsed.  Well, that's not an easy question since there are no anemometers around the site.  But there are a few clues leading to a range of likely wind speeds.  One set of clues comes from what was not damaged.  I noticed a flagpole to the right and behind the stage that didn't blow over.  There were even tents nearby that remained.  Assuming a flagpole is similar in strength to a typical powerpole, typically 70 to 80 mph winds are required to initiate damage, sometimes less.  However there was no damage and so it's doubtful the wind exceeded 70-80 mph.  The same could apply to the many light standards around the fairgrounds in the video.  There were also no other reports of significant damage at the fairgrounds, a place full of large open span structures that could easily start to suffer damage at similar wind speeds (this is preliminary of course since I heard there was some minor nondescript damage scattered around the park).  I decided to take a stab at tracking identifiable plumes of dust that speed across in front of the stage platform.  The amount of time some of these plumes tracked across from one end of the platform to the other was about 1.6 to 2.5 seconds.  The stage measured 38 m long according to Google Earth (see figure 1) and that wind speed amounts to roughly 20-25 m/s or about 37-48 kts.


Figure 1.  A satellite view of the Indiana State Fairgrounds courtesy of Google Maps.  The inset shows the grandstand and the stage north of the dirt track.  The length of the stage platform is nearly 38 m and the wind direction was estimated from the WNW.

Could 37-48 kts be realistic?  Well, there is some supporting evidence.  The highest wind speed at the airport was 44 kts at 8:56 pm EDT.  In addition, to the south at the KIND WSR-88D, the highest winds at the radar site only reached to 40-45 kts (figure 2).  
Figure 2,  The KIND four panel display of reflectivity and velocity for 0056 UTC.  The half degree elevation shows up on the top row and the 2.4 deg elevation shows up on the bottom row.
Of course there are some reports that may indicate higher winds occurred in the vicinity.  One report occurred west of downtown about the same time as the stage collapse of a large tree blown down and an estimated wind speed of 70 mph attached.  Perhaps winds to 70 mph were there.  The other wind gust southwest of Indianapolis was considerably higher at a measured 77 mph. Power did go out around the town and by the next afternoon,  and about 1250 customers were still out of power.   Again, I'll wait and see if those winds occurred at the fairgrounds at various times but judging by the video, the wind speeds were considerably less and likely not even 50 kts when the stage collapsed.

figure 3,  A map showing the outline of the severe thunderstorm warning issued by the Indianapolis NWS forecast office, and the three local storm reports received during the course of the warning.

So the stage was set to put people at a high vulnerability level.  Certainly it appears this stage couldn't withstand winds any more than 45 kts and it's likely the stages that failed this year in Tulsa, OK Ottawa, ON, and last year in El Reno, OK also were similarly weak.  Too bad there was no communication as to the vulnerability of these stages from one set of event coordinators to another.

Given this vulnerability and the consequences of its failure,  the onus should be on the event handlers to provide an extra level weather awareness.  Unfortunately there was not since it appeared that the handlers were interpreting something for which they have no expertise.  In addition to in appropriate interpretation, the Indianapolis Star also reported that there was confusion as to when the threat would arrive.  The event handlers that Mike Smith mentioned,  waited till 8:45 pm to consider evacuation plans, four minutes before the stage collapse.  In addition, the Indianapolis Star quoted "According to a timeline issued today by Indiana State Police, at 8:49 p.m. -- about 25 minutes before the storm's forecasted arrival -- a strong gust of wind blew through the fairgrounds, toppling the stage setup onto the those closest to the stage.  Bursten said the early indication was that the "isolated significant wind gust" took authorities and event coordinators by surprise, since the storm itself was still about 30 minutes from arriving. They had been in contact with the National Weather Service for much of the evening." So it also appears that they were expecting a later arrival than what happened.   It's difficult to determine what the nature of state fair official's contact with the NWS had been but there is some evidence that there was some expectation that they had perhaps 25 minutes to prepare for severe weather.  What was the source of this misunderstanding?  

The NWS did issue a severe thunderstorm warning that covered the state fairgrounds and all of Indianapolis from when it was issued at 8:39 pm EDT.  As shown by the text below, the severe thunderstorm warning was issued as shown where an initial location of the threat was labeled and the entire area was covered in a polygon (figure 3 and 4).

AT 835 PM EDT...NATIONAL WEATHER SERVICE DOPPLER RADAR INDICATED A
  LINE OF SEVERE THUNDERSTORMS CAPABLE OF PRODUCING QUARTER SIZE
  HAIL...AND DAMAGING WINDS IN EXCESS OF 60 MPH.  THESE STORMS WERE
  LOCATED ALONG A LINE EXTENDING FROM 9 MILES NORTH OF ZIONSVILLE TO
  GREENCASTLE...AND MOVING EAST AT 25 MPH.
 The NWS marked the line of storms given the town landmarks from nine miles north of Zionsville to Greencastle (figure 4).  However, the text of the warning did not include the position of the gust front which was considerably further east.  
Figure 4 A map depicting the location of the severe thunderstorm warning and the 8:35 pm reflectivity.  Also the convective line (red) mentioned in the text of the warning is shown at the 8:35 pm position mentioned in the 8:39 pm severe thunderstorm warning.  The blue line marks the position of the gust front at 8:35 pm.

This is often the case that NWS meteorologists track the location of the heaviest, most likely severe, portions of the storm (s).  The problem is the way the event handlers may have interpreted the warning, and the location of the threat relative to the confines of the polygon.  The threat that was of the most interest to them should've been the gust front, not necessarily the worst of the severe weather that was located and tracked by the NWS forecasters.  The warning polygon gave the event handlers a lead time of 7-8 minutes before the gust front hit assuming a minute delay from warning issuance to their reception. That's when they should've taken action.  They would've needed all of that time considering that they had poor communication protocol amongst all the players in the warning dissemination, including the WLHK program director.

Large event handlers need a special relationship with expert meteorologists in order to help provide specific warning information.  Canned products are not enough because they don't provide the information that the handlers need nor does it give a tailored product to match the exposure of the event participants.  The severe thunderstorm warning from the NWS was fine but it didn't provide the handlers the timing of the gust front.   The handlers didn't know that they needed the timing of the gust front and so they were confused when the severe weather hit 20 minutes earlier than the actual line of heavy rain which the NWS was tracking.  Lacking knowledge, the handlers should've played it safe and evacuated early.  But the concert-goers, and program directors want to push the envelope and keep running the show till the last minute.  Wouldn't an event handler want a little more detailed information to help resolve these sometimes conflicting needs?

The concert stage was a flimsy structure where upon collapse, proved deadly.  It appears that not even severe thunderstorm threshold winds were needed to collapse the structure.  Thus it's entirely feasible that this tragedy could've happened without a severe thunderstorm warning.  When considering exposure level, everyone has a different threshold at which their safety may be compromised.  Consider figure 5 below that conceptually illustrates the varying thresholds of safety depending on the type of structure or situation for which someone may reside.  Each dot represents a situation where a person's safety is at high risk ranging from being in a small boat when the wind reaches 40 mph (special marine warning) all the way up to being inside a hospital or office building (140 mph).  These values vary from one situation to another but the point is made that each of us has a different exposure level and that official products do not account for this variety.  The event handlers need to know the exposure level of those in their charge and relay that to the meteorological consultants, in this case very low thresholds.  Then the consultants can use that information to make a better warning product.


Figure 5,  A conceptual diagram of exposure level vs. threat severity where personal safety can become compromised.  The threat severity happens to be wind speed.  Several examples are provided and compared relative to the severe thunderstorm wind speed threshold.


Typically when it comes to preserving life, the event handlers can choose to stoke a relationship with the NWS, either by an incident meteorologist (called IMET) physically present, on phone, or on chat.  Or a private sector company can be hired to be the consultant.  Certainly each has its advantages, the former is already paid for while the latter can devote more time and attention.  What should be communicated is what kind of exposure the people are subjected (e.g., under a flimsy stage, outside, in tents, in water), how much time he/she needs for evacuation, and any other special criteria.  The meteorologist can take these constraints and then give a forecast of when the relevant severe weather parameter is expected to strike.  The event handler may ask about the forecaster's confidence, and a reasonable answer should be given that conveys the truth.  Then the even handler can do what he/she has been trained to do.

Let's hope this message goes somewhere.


Addendum:

15 Aug 2011 - Now that a couple days have past, the questions arise as to whether or not the fair directors should've done more to protect the fair patrons.  This article at msnbc summarizes these questions quite nicely.  The questions fall into two camps, one related to who's overseeing the construction standards of temporary buildings, the other about who's responsible for executing a response to the warning of severe weather.  For the latter, I was intrigued to hear that an outdoor concert featuring the Indianapolis Symphony Orchestra at the Conner Prairie Amphitheater was canceled well before the storms hit.  Being that this venue was entirely outdoors with no protection, the concert directors had a plan, a source of expert weather advice, and a conservative attitude.  As a result, there would've been no news about their successful evacuation if it weren't for the tragedy at the fair.

As an aside, I hear some rather intriguing meteorologically-related theories as to why the stage collapsed.  They all have to do with some extreme local wind gust, even called a fluke that nobody could've forseen by the governor Mitch Daniels.   I hope this statement doesn't deflect the need to prevent future tragedies from collapsed stages.  This wind event was not a freak event never to show up again.    What is more likely is that the stage was such a house of cards that it fell where even tents stood less than 50 m away.  I tracked the plumes of dust across the stage platform and came up with rather modest winds of ~ 50 mph (~45 kts).  Perhaps my analysis was wrong and there were higher winds though I doubt I'm far off.   As the story on CBS confirms my thought, there was no freak anomalous wind that struck the stage, what was anomalous was the weakness of the stage.