EMForum Presentation — May 14, 2014
Storm Surge Forecasting
At the National Hurricane Center
Robbie Berg
Hurricane Specialist
National Hurricane Center
Amy Sebring
EMForum Moderator
[This transcript contains references to slides which can be downloaded from http://www.emforum.org/vforum/NHC/StormSurge.pdf
An audio recording of the live session is available at http://www.emforum.org/pub/eiip/lm140514.mp3]
[Welcome / Introduction]
Amy Sebring: Good morning/afternoon everyone and welcome to
EMForum.org. I am Amy Sebring and will serve as your Host and Moderator
today and we are very glad you could join us.
As you know, the official beginning of hurricane season is almost upon
us and as we have done in the past, we have invited the National
Hurricane Center to brief us on what’s new. Please note, there is a
link on today’s background page to the complete update for 2014, but
today’s program will focus on storm surge forecasting.
[Slide 1]
Now it is my pleasure to introduce today’s special guest: Robbie Berg
serves as a Hurricane Specialist at the NHC in Miami where his work
involves the issuance of track, intensity, and wind radii forecasts as
well as associated watches and warnings for tropical cyclones in the
Atlantic and eastern North Pacific ocean basins.
Please see today’s Background Page for further biographical details and links to related resources.
Welcome Robbie, and thank you very much for taking the time to be with
us today. I now turn the floor over to you to start us off please.
[Presentation]
Robbie Berg: Thanks, Amy. I want to thank everyone
for allowing me to give this talk today about storm surge forecasting at
the National Hurricane Center. The way the talk is going to go is
I’m going to go a little first into storm surge forecasting in
general—some of the older parts we’ve used. Towards the end of the
talk we’ll get into the new stuff coming along the way—some of which is
experimental beginning this year.
[Slide 2]
To begin let’s just jump into how we model storm surge. There are
different ways to tackle the problem. First off you can look at
statistical methods. The problem with statistical methods is that
they do use only historical data and a lot of that time the data is
nonexistent. We have had hurricanes make landfall in the United
States but not every location along the coast has been hit, let’s
say by a category four hurricane.
We can’t use the historical data to extrapolate what kind of storm surge
we might get at any location on the coast. This method isn’t very
helpful for us when we’re doing forecasting.
[Slide 3]
The second way we can do it is deterministic numerical models.
This is for forecasting storm surge solely based on physical equations
of how things move such as how the wind from the hurricane pushes water
along. This is strongly dependent on accurate meteorological
input—how accurate is the hurricane forecast coming from the hurricane
center?
Unfortunately because of the uncertainty in the forecasts—we know our
forecasts aren’t perfect—this particular method can be inaccurate.
I’ll have a great example coming up on why we can’t use this particular
method.
[Slide 4]
If we can’t use that, what do we use? We like to use what we call
the numerical model ensemble. This is looking at many different
runs of the same model but with different conditions, in other words
like a family of storms. This is the best approach for determining
storm surge risk and vulnerability for an area since it does account
for forecast uncertainty. We’ll have illustrations coming up that
will show this in a more vivid description.
[Slide 5]
What do we use at the Hurricane Center and the National Weather Service
for forecasting storm surge? The model we use is called SLOSH
(Sea, Lake and Overland Surges from Hurricanes). This was
developed by the National Weather Service several decades ago—probably
back in the seventies—and it is used to estimate storm surge height
resulting from historical, hypothetical or predicted hurricanes.
[Slide 6]
Even though it is old, SLOSH is pretty good at what it does and it does a
lot of things you would be concerned about. It knows about flows—water
flows through barriers, gaps and passes along the coast. It knows
those passes can be deep between different bodies of water. It
knows about inland inundation or wetting and drying of cells.
There are some surge models that will raise the water level along the
coast but they won’t bring that water inland in areas that are normally
dry. SLOSH model is able to do that so it can take normally dry
areas along the coast and make them wet from the storm surge.
It can overtop barrier systems, levies and roads. It can account
for astronomical tides. Unfortunately two important factors that
SLOSH cannot include which are not storm surge related are breaking
waves along the coast and normal river flow and rain. While SLOSH
cannot account for those two parameters by itself the Weather Service
has other ways around that and in addition to that we are putting a lot
of research into putting those parameters into the storm sured forecast
and allowing SLOSH to be able to see those things within its
simulations.
We are not there yet. It is one of the challenges we are working
on. We hope within five to ten years we will be able to include
breaking waves and river flow in the SLOSH model itself.
[Slide 7]
The way SLOSH is arranged is that you run the model in a grid.
What you see on your screen are the many different grids you can choose
from. In a real time storm we are only going to choose the grids
that are affected by the storm. What that does is cut down on the run
time that is required to run the storm surge simulation.
If we had to run the simulation for every grid it would slow the whole
process down. When we’re making a forecast we need the information
very quickly. In order to cut down the run time we only focus on
the areas that are at risk from the surge in the hurricane itself.
[Slide 8]
We know there has been talk about SLOSH being old and comparing it to
higher resolution and higher performing storm surge models. One of those
models is ADCIRC. ADCIRC has its plusses and we don’t want to say
that we don’t like ADCIRC but we don’t really think it is useful for
our operations. One of the reasons is because it is such a high
resolution model it takes a long time to run.
You remember before we were talking about we need to run a family of
storms when we are looking at simulations and not just one run.
With SLOSH we can run one storm through the model in minutes on a
desktop computer. ADCIRC on the other hand takes a super computer,
a much higher performing computer and it would take that one storm up
to several hours to run.
If you’re thinking about running a family of storms we don’t have the
luxury to waste that amount of time to get a storm run through the
ADCIRC model. That is one of the main reasons we have stuck with
SLOSH—because it is a very fast model that gets us the information we
need in a quick amount of time.
One thing I would note and you can see it on the screen—if we take the
same storm and run it through both models, the flooding pattern isn’t
much different between the SLOSH and ADCIRC models. You could zoom
in really close and see some differences but in an overall,
community-wide sense, you are still seeing the same types of flooding in
both models.
In some sense the resolution of the model can only get so high for our
purposes when it comes to operational real time forecasting.
[Slide 9]
To sum up how we forecast on surge, I think this is the point we really
need to take away. All storm surge models, no matter how
high resolution they are, are strongly dependent on the accuracy of the
meteorological input. If the hurricane forecast is wrong by any
little bit, whether it be the track, the intensity or size of the storm,
it is going to have huge implications on how well the storm surge model
runs and what type of flooding pattern you will see.
The meteorological uncertainty in that hurricane forecast will always
dominate over storm surge model specifications and that includes the
physics, resolution and other types of parameters. Different
vertical datums are not important here. In reality storm surge is
only one component in the real water rise in the storm.
We have touched on this already. If you live on the coast you are
worried about how high the water will rise from this storm and that
includes the component of the storm surge, the tide, waves and the fresh
water flow. The SLOSH model does a good job with the surge and
tides. We have other models that can forecast the wave action
along the coast and a model that does the river stages but right now the
three models don’t talk to each other very well.
As I mentioned some research going on now is going into how to make
those models talk to one another and in the end what we are really
getting is the total water that will rise. Fortunately in
most scenarios it is the surge and tides that are the biggest
contributors to the water that will rise. While we are not
explicitly including the other two factors in most cases those are not
the two that are most important.
[Slide 10]
I talked about an example of why the family of storms is so important so
let’s take a look here. This is a case where we have a hurricane
in the Gulf of Mexico to the southeast of Louisiana. You can see from
the black line this would be the actual forecast track given by the
Hurricane Center and in the white is the cone of uncertainty.
In this case we have the storm to the southeast of Louisiana and
the twelve hour forecast point is right along the northern gulf coast,
just to the west of Mobile Bay. Keep that in mind. We are only
talking about twelve hours out from potential landfall. That is a
short amount of time.
[Slide 11]
If we were to take the forecast for that storm and run it through the
SLOSH model this is the flooding footprint you would get. Now you
are seeing we are getting a lot of flooding in Mobile Bay and the
barrier islands to the south of Mobile Bay. We’re seeing values
close to twelve to fourteen feet above the datum NGVD twenty-nine.
In this same figure look further to the east and you see in Pensacola
you are not getting nearly as much flooding in that area as you are
getting in the Mobile area. You are still seeing some, about two
feet or so but that is not as significant as what you are seeing in the
west.
The question is—here is our forecast for only twelve hours from
potential landfall—is this the type of guidance you want to see to
analyze the risk for this storm?
[Slide 12]
Here is what actually happened with this storm. It continued
moving north and although the black line was the twelve hour forecast
the storm made a slight shift to the east. This is not a huge
shift. This is a pretty good forecast. You are only
looking at about thirty miles to the east of the twelve hour forecast
point.
Instead of making landfall to the west of Mobile Bay, now we are making
landfall to the east of Mobile Bay. If this is what the storm
actually did, what do you think the storm surge flooding actually
occurred in this storm?
[Slide 13]
Now we can take that track and put it in the SLOSH model and here is
what you get. Now you notice a big difference in the flooding
footprint. Now Mobile Bay, instead of having wind pushing water
into the bay, you have winds blowing offshore and blowing out of the
bay. Instead of seeing twelve to fourteen feet of surge you are
seeing four feet of surge.
In the meantime look to the east to Pensacola. Remember from the
previous diagram we showed two feet of surge. Now in some parts of
the bay near Pensacola you are looking at about ten to twelve feet of
surge. The track shifted a little bit of the east and the size of
the storm grew more than was forecast.
A little secret when it comes to hurricane forecasting—we are pretty
good at track, we are so-so at intensity and not very good at all when
it comes to size forecasting. That is one of the parameters you
have to factor in uncertainty to a large extent. What happened in
this particular storm is that the storm grew in that twelve hour time
frame and the storm extended far away from the actual track so we pushed
more water into the Pensacola area.
With this storm’s evolution, this is the type of flooding we got.
It is a lot different from what was originally contained the twelve hour
forecast. We can’t have this kind of scenario occurring so we
have to figure out a way to analyze the true risk and vulnerability to
the surge of any particular storm.
[Slide 14]
There are perils for not accounting for uncertainty. I like to use
this diagram because it shows what happens if you don’t account for
uncertainty in a hurricane forecast. You are essentially putting
all your eggs in one basket. If something were to go wrong, the
track were to shift a little bit, the intensity were to go up a little
bit, the size to increase a little bit—you could end up looking like
this.
You put all your eggs in one basket and you’ve made a really bad
decision ahead of the storm. We have to use the ensemble
approaches in order to analyze our true risk.
[Slide 15]
How do we do that? There are several ways. We want to look
at alternatives to single runs. One way is to look at a map with
pre-computed storm surge maps that are based on different direction and
motions of the storm, different landfall locations, different
intensities, different storm sizes and different forward speeds.
[Slide 16]
The way we can do that is to look at these three different
products. I’ll go through each one by one. The first is the
MEOWS (Maximum Envelopes Of Water), the MOMS (Maximum of MEOWS) and the
real time simulations which are the P-Surge (Probabilistic Storm Surge
product).
[Slide 17]
Let’s start with the MEOWS (Maximum Envelopes of Water).
[Slide 18]
If you have the SLOSH display program, the MEOWS are available to you if
you open up the program and look for your area. What happens with
the MEOW is that as the user you select the SS category of the storm,
the forward speed, the storm direction and the tide anomaly. The
Radius of Maximum Winds and landfall location are already built into
this product.
[Slide 19]
The best way is to show you an example of the MEOW and how it is
created. Here is as simulation of the storm. You can see at
the bottom of this figure there is a black line moving into the
Charleston, South Carolina area. That is one storm. When I
run that one storm in the SLOSH model in this particular grid you get
this type of flooding.
I want you to notice the grid cell here in somewhere in southern North
Carolina that says 4.8 feet. From that one storm it produced
4.8 feet at that location. I selected a category three storm
moving to the west northwest at fifteen miles per hour at high tide.
Let’s say we’re not so sure that the storm will make landfall at that
exact location so what if it moved to the north of this and made
landfall there? Move it up the coast—and what if we had another
storm and it moved further up the coast and we kept doing that?
Notice when we have storms the flooding changes based on that particular
storm.
Also notice that at our inset that the values are changing as the storm
gets closer. As the storm gets closer you would expect a higher
surge at that particular cell. What if I have all this family of
storms and I hit the coast everywhere with any storm with those
parameters—category three moving to west northwest at fifteen miles per
hour at high tide.
Now what I do is retain the highest value at that grid cell from any one
of those storms. In this storm it was thirteen point eight
feet. The MEOW for that location with those parameters was
thirteen point eight feet. This is one particular MEOW. For
any phase in you can get tens to hundreds of MEOWS depending on the
different parameters and how you choose them.
One thing about the MEOW is that what you see on your screen shows
flooding but this isn’t the actual flooding that would occur from that
storm because we are looking at a family of storms. This is more
or less like a worst case scenario for a storm given those particular
characteristics.
[Slide 20]
Thinking about what MEOW is we can go even further and say what the maximum of MEOWS are—the MOMs.
[Slide 21]
We’ll go back to the MEOW. This is the one I just showed.
With the MOM you only select the category. All the other
parameters are baked into the cake for you. I took the same MEOW I
used and it showed me thirteen point eight feet at that location in
southern North Carolina.
What if the storm makes landfall at different directions into the coast
using the same parameters of the category three storm? Also what
happens if the speed changes or if the tide level is different?
After I hit the coast at all different directions I still retain the
highest value within that particular cell. Now you can see I have a
category three MOM showing fourteen point seven feet at that location in
southern North Carolina.
This is taking all the different uncertainty that exists with a
particular storm or event and composites them to show you what the
actual risk is along the coast from that category of storm.
[Slide 22]
MOMs are important because that is the product that goes into
determining evacuation zones along the coast of the United States.
A category three storm in this area of the country has the potential to
do this sort of flooding along the coast. Because of that
simulation that output is used to determine what type of evacuation
should be called for a category three storm threatening the coast of
North Carolina and South Carolina in this particular case.
It is important to know about the MOMs in that they are the basis for
the nation’s evacuation zones anywhere from Texas up to Maine.
This is an example of what I’m talking about. If you look at this
figure from New York City giving you the evacuation zones from the
city—what I want you to notice the pause between the evacuation zones.
Here I am showing the category four MOM.
Any category four storm could potentially produce this amount of
flooding in New York City. Watch again how the two phase in and
how the evacuation zones line up very close to the category four
MOMs. That is not by coincidence because those MOMs are being used
to create the evacuation zones along the coast. I wanted to show
you an illustration of how that occurs and how municipalities and states
use that data to create their zones.
There are the two on top of one another. The black line is the
category four MOM and the colors would be the evacuation zones for the
city. All the products I showed you—the MOMs and MEOWs are
available to you right now. You can go into the SLOSH program and
look at them.
[Slide 23]
When you are in an actual event we will start producing the probabilistic storm surge product.
[Slide 24]
It is based on the NHC official advisory. We produce a forecast
that says an advisory goes out at five PM eastern time. Based on
that forecast we will run these probabilities and create a storm surge
forecast. It is available roughly forty-eight hours prior to the
arrival of tropical storm force winds along the coast and another way of
looking is when we have a watch or warning in effect for the United
States coastline.
The P-Surge product is beneficial because it accounts for meteorological
uncertainty in the hurricane forecast. It accounts for track and
landfall location uncertainty. It accounts for uncertainty in the
size and forward speed of the storm and in the intensity of the
storm. These uncertainties are based on historical errors that
we’ve had in our hurricane forecasts.
We keep a very good running verification of how good or bad our
forecasts are. We use the past five years of our forecasts to
determine our historical errors. Those errors are input into the
P-Surge product. The latest version (2.0) which will be available
this coming hurricane season also accounts for tide and can be available
above ground level.
That is important because in the past, the previous version did not
account for tide. If you were looking at the P-Surge
guidance you had to find a way to add tide into the guidance. If
you weren’t looking at the same vertical datum you could screw up the
whole process. This is a big improvement to the model because now
the tide is accounted for and you don’t have to manually add the tide to
what you are seeing in the output.
The product is going to be above ground level as well in addition to the normal datum which is typically NAVD 88.
[Slide 25]
Here is an illustration of how P-Surge works. Here we have a
storm. It happens to be Hurricane Irene from a couple of years ago
to the east of Florida just north of the Bahamas. Here is the
actual forecast from the Hurricane System in the black lines and we had
the forecast landfall somewhere in eastern North Carolina and riding up
the Atlantic Coast into New England.
[Slide 26]
We know the forecast isn’t perfect. We have to assume the track
could shift to the east or west. We have to assume the size could
vary. We know hurricanes come in all different sizes and even
though we forecast in a particular size, that size could change.
We have to assume the speed is a little faster or slower than we are
forecasting and we have to assume the storm could be stronger or weaker
than what we are forecasting.
With P-Surge all these differences are centered around the official
forecast. It’s not like we’re ruling out the forecast but it is
centered and weighted heavily on the official forecast.
[Slide 27]
Once you have run P-Surge this is the type of output you receive.
This is an example shot from Hurricane Sandy in 2012. You’ll
notice that P-Surge did a good job of illustrating the areas along the
coast that had the highest risk from the storm surge from Sandy—in fact
New Jersey, Long Island, New York and Long Island Sound including the
Connecticut coast was ground zero for the highest storm surge. It
realistically painted the highest risk from the worst surge.
[Slide 28]
Do you remember the example I showed earlier in the presentation?
Let’s go back to that one. Here is our storm and there is the
original forecast showing it making landfall to the west of Mobile Bay
where we had a lot of flooding possible in Mobile Bay and not quite as
much in Pensacola.
The deterministic SLOSH run sent a limited surge threat to
Pensacola. At twelve hours away from landfall we have a likely
hurricane warning in effect which means P-Surge would be available.
[Slide 29]
Let’s look at what the P-Surge output would have shown for this
storm. What I’m showing his the probabilities that the surge would
be greater than eight feet. Now you can see that the P-Surge was
showing a fifty percent chance of eight feet of surge in Mobile Bay,
which makes sense based on the forecast, but it is also showing at least
a fifty percent chance of eight feet of surge in parts of Pensacola.
The P-Surge product is realistically taking into account the
uncertainties in the forecast and it is assuming that some of the tracks
could shift to the east bringing the storm closer to Pensacola or it is
assuming the storm could grow in size more than what is in the
forecast. When it takes those uncertainties into account it shows
there is a significant storm surge risk in the Pensacola area as well as
the Mobile Bay area.
During an event realistically somebody making a decision on things such
as evacuations should be looking at a product like this and deciding
that whether in Mobile or Pensacola people should be evacuated from the
coastline because there is that amount of risk from that particular
storm.
[Slide 30]
Here is reality of what happened. The storm shifted to the east
and created the storm surge in the Pensacola area. This is
Hurricane Ivan in 2004. We didn’t make it up, this is an actual
storm that occurred and where critical decisions had to be made based on
that forecast. This is showing how using P-Surge can help
immensely when having to make those critical decisions.
[Slide 31]
I’ve thrown three products at you—the MOMs, the MEOWs and the
P-Surge. We like to show what we call the “storm surge decision
support wedge” so you get a good understanding of what you should be
looking at during different points of the forecast. If you start
at the bottom of the wedge in tier three this is your planning and
mitigation stage.
This is more than five days from landfall. You are looking now
when there isn’t a storm out there. You’re planning hurricane
season. At this point you should be looking at MOM. MOMs are
used to create your evacuation zones. When you are well away from
a landfall and more than five days out—more than five days—look at your
MOM to assess your risk along the coast. When you get within two
to five days of a potential landfall, now you are at the readiness
stage.
I would still suggest looking at MOMs but now that you have a forecast
from the Hurricane Center showing the storm approaching your area you
can start to refine some of the parameters associated with that storm
and looking at specific MEOWs because you may be able to pick a
particular category and you may be able to eliminate some of the
different directions that are available to you.
Maybe you know that the storm will move to the north, northwest or
northeast, but not to the south so you can start to eliminate some of
the possibilities and look at some of the MEOWs to further refine the
assessment of risk. When you get within forty-eight hours of a
landfall within that watch warning time frame you are in the response
phase.
We recommend looking at your MEOWs and doing further refinement to
figure out which one is better to look at but P-Surge is available to
you at this stage in a real time sense so you get a better handle on the
risk from the storm.
In addition to those products you’ll also start receiving explicit
numbers from the Hurricane Center within our advisory and also within
the local statement provided by the National Weather Service forecast
offices along the coast. As we get closer to the landfall
everything starts converging to more refined information.
When you are more than two days out there is too much uncertainty within
the forecast to provide specific and extremely accurate information.
[Slide 32]
Those products have been available to people for several years but let’s
look into the future and where we are going—especially coming up in
this hurricane season and we are starting with the experimental
potential storm surge flooding map, or you have heard it called the
inundation map.
[Slide 33]
To add some background to this graphic, the first question would
be—which product will drive the inundation graphic? It is going to
be P-Surge—in fact, P-Surge 2.0 which is coming online this year.
One thing I didn’t mention about P-Surge is you will select the level
you concerned with. For this inundation graphic we are going to
elect the ten percent exceedance.
What that means is it’s a good way of looking at what you should plan
for in this storm. If you were to look at a zero percent
exceedance that would be a worst case scenario—nothing could potentially
top what you are seeing in a zero percent exceedance. The problem
with using a zero percent exceedance is you will end up over-evacuating
people in a potential event.
We don’t want to over-evacuate. If I was to look at a fifty
percent exceedance product essentially what I’m saying is that there is a
coin flip, that is a fifty-fifty chance that the level that is being
shown on the map is higher or lower. When it comes to critical
decisions on things like storm surges which might be a life-threatening
hazard we don’t have the luxury to do a fifty-fifty coin flip. We
have to exclude and not account for the fifty percent exceedance.
For that reason we use the ten percent exceedance. What it means
is that there is only a ten percent chance that the values shown on the
graphic will be exceeded by the real storm. It is not the worst
case scenario but it is a plausible upper bound of what could
happen. It is what we suggest you should plan for in a storm given
the uncertainties in the forecast.
The grids that are used for this product are from the latest SLOSH basin
updates which have been updated in NAVD88 which is the newest
datum. Topography-wise we are using a Coastal Surge Center
sea-level rise DEM. They have been producing this topography
data set to use sea level rise studies but it is being done in a timely
fashion that we can use it as well for our inundation graphic.
It is at such a higher resolution that we have to do some re-sampling to
make it a little smoother resolution because in some of the polling and
focus groups we’ve done when it was too high resolution some people
thought it was not accurate. By doing a little bit of smoothing we
were able to make the figure look at little more attractive to people
when they are looking at it and they put more faith into what they are
looking at.
In some areas when they haven’t finished this DEM we have been
augmenting it with USGS NED data but that is being phased out as we go
along to the higher resolution data. For processing we are locally
using the ArcGIS for server and desktop but we are working toward
leveraging National Weather Service’s program, the IDP, which in the
coming years when they get that set up it we won’t have to do that
locally—it will be done within the larger National Weather Service
framework.
[Slide 34]
Here’s an example. One of the things we have to do is merge all of
those grids I showed you later from the top. This is an example
of the merged grid. On top of that you can also see the topography
data that we are using to create this inundation graphic. You can
only calculate inundation if you know how high the land is.
The storm surge model produces a storm surge height. You know how
high the land is at any of these locations so you use subtraction to get
how much water is sitting on dry land. That is why we use the
high resolution data. You can see from this map that we include
that data for every area along the coastline from Texas to Maine.
[Slide 35]
In a simplistic fashion this is what it looks like. This would be
raw output from the P-Surge guidance. We get a grid of data
showing storm surge height at different locations. The post
processing we do transforms that data into the inundation graphic or the
potential storm surge flooding graphic.
[Slide 36]
Many of you on the webinar may have seen this example. This is an
example of a simulated storm that we’ve done in the Fort Myers, Florida
area. It shows you the potential storm surge flooding that might
occur from a simulated storm. The higher values in red are
decreasing as you go further inland. The lower values are showing
in blue.
This is what the prototype looks like. Again, this is the Fort
Myers area. The red would be extreme or nine feet or greater of
inundation or storm surge flooding above ground level. Orange is
designated as high or six to nine feet. Moderate would be three to
six feet and low would be three feet or less.
[Slide 37]
How is this useful? I haven’t mentioned it yet because this is in
the experimental phase this year. Because it is in the
experimental phase we won’t actually be disseminating the raw data in
any sort of format like the GIS shape file but in the coming years once
we go operational with this product it will be useful when it comes to
making important decisions.
Here’s an example of looking at southwest Florida in the Fort Myers
area. These are the evacuation zones starting with T for tropical
storm, and then A, B and C. I think they have a D and E but they
are not showing on the diagram. What is interesting and what we
think the potential for this product is you notice as I flip back and
forth, I can take this inundation graphic from this storm and overlay it
on my evacuation zones to determine which zones are most vulnerable
from the storm surge and I can make a more educated decision on which
zones should be evacuated.
In this case we definitely evacuate zones T, A, B and zone C could be up
for debate. There are a lot of areas in zone C in a light pink
color that aren’t quite touched by the water being moved inland by the
storm in this scenario. There could be arguments for zone C to not
have to be evacuated or fully evacuated.
In a real time scenario we think this product will go a long way to help
emergency managers make these critical decisions as far as
evacuations. It can be beyond that. It can go into things
like looking at shelters. Are there shelters located in areas that
could be under potential flooding? Other critical facilities such
as hospitals—you could look at transportation routes—which roads are at
risk of being inundated by storm surge from this storm?
We think this product has great potential going forward into helping
make critical decisions before a storm and as a storm is beginning to
make landfall so we can get ready for the response phase after the
storm.
[Slide 38]
We have four examples that we’ve produced leading into the hurricane season. This is the Fort Myers example I showed.
[Slide 39]
This is one from the Houston area and Galveston. We can see the
legends there with storm surge flooding with four categories, blue up to
red. This is a simulated storm. It is not actually
Hurricane Ike. It is a simulated storm we created. It shows
there are some areas that are more susceptible to storm surge flooding
because the flooding penetrates further inland than other areas.
To the north of Galveston there is a lot of potential for storm surge
penetrating well inland even up to I-10 in this scenario and then
showing down further south into the Galveston area.
[Slide 40]
This is an example from Charleston, South Carolina. This is not
Hurricane Hugo. It is a storm we generated as a simulation.
It realistically shows the worst surge could occur along the barrier
islands and the back bays of the islands especially closer to the
ocean. It also shows the surge can get into the creeks and
waterways that extend outward of the Charleston area.
[Slide 41]
The last example we produced is the New York City area. It is not
Hurricane Sandy or Irene, just a simulation we ran. It shows the
areas that would be at risk from surge. We got areas along the
shore to Manhattan, Brooklyn, Queens and into the Jersey side like
Hoboken and Jersey City and further inland into the Meadowlands.
You can notice from the diagrams that the inundation graphic knows where
higher land is.
If you are familiar with the New York City and New Jersey area you know
there is a bit of land extending through Hoboken and Jersey City that is
elevated. It is not going to inundate those areas because it
knows the land is higher. In the meantime it knows in the
Meadowlands it is very low lying and very marshy so it allows the storm
surge to get up in those areas and you can see the type of flooding that
would occur in that kind of storm.
[Slide 42]
The takeaways—this product will be available in this hurricane season on
an experimental basis via the National Hurricane Center website.
What the experimental phase means is that we can’t guarantee it will
always be available. The trial phase is for us to work out the
kinks in this product before we go live in an operational
sense. There may be times when we can’t get it up during an
event.
We want to make it available to people so they can look at it during an
event when it’s available and to assess how useful it might be to them
or if there are changes to be made before we go operational. For
that reason there will be no data dissemination during this phase.
We won’t be disseminating the data in a raw sense like a GIS shape
file.
It will only be available via the Hurricane Center website in a Google type format where you can pan and zoom in on your area.
[Slide 43]
The last thing I want to talk about in brief is storm surge watch and
warning. The National Weather Service has decided that based on
some recent storms we’ve had that we need to move forward with
initiating storm surge watches and warnings. If you look at the
hazards associated with a hurricane, there are four main hazards—the
wind hazard, the storm surge hazard, the fresh water flooding and
rainfall hazard, and the tornado hazard.
If you look at which of those hazards we warn for, we warn for the wind
in a hurricane warning. By definition a hurricane warning means
that hurricane force winds are expected in that area. We warn for
river flooding, flash flooding from rainfall and tornados with tornado
warnings. The National Weather Service does not have an explicit warning
for storm surge.
There’s a problem there because storm surge has been shown to take the
most number of lives in tropical cyclones. You only have to look
back at storms like Katrina, Ike and Sandy to see that the storm surge
in a lot of cases is the hazard that takes the most numbers of lives.
Because of these more recent storms we notice there is a gap in our
warning scheme so the National Weather Service has agreed and decided we
are going to move forward with.
This is something we are looking to go experimental with in 2015 but not
in this hurricane season. One thing you might notice from this
diagram on the screen is that the storm surge warning will be different
from other warnings you’ve received in the past. In the past
you’ve received warning on the county scale.
You might see a hurricane warning for Miami Dade County in
Florida. That would cover the entire county. Some warnings
are done in a fashion like tornado warnings where a polygon is issued by
the NWS and you know if you’re in the warning whether or not you reside
in that polygon.
The problem is that those methods are not really useful when it comes to
storm surge. We can’t use the county based method because in many
cases storm surge only stays on the immediate coast line. In
cases where it penetrates further inland it is not going to cover an
entire county. It doesn’t make sense for us to issue a county
based storm surge warning because we would be over-warning more area
than we really should be.
The way we think NWS is going to go with this is to use the NDFD grid
which is used to make the actual forecast. You can see in this
diagram why it looks boxy—it’s because it is using the grids in that
dataset. What you’ll notice is that this would be a Hurricane
Irene example—the area that is painted onto the storm surge area is the
area that is at risk of potentially life-threatening storm surge.
We are limited to areas along the coast and the barrier islands and the
shores of the Pamlico and Albemarle sounds in eastern North
Carolina. It does a better job of painting the area that is at
risk for that storm surge. The warning itself—a lot has to be
determined within the Weather Service in terms of criteria, the official
definition of the warning. I think we are pretty sure we only
want to paint those areas that are life-threatening storm surge drifts.
We won’t be issuing a storm surge warning for every little flooding that
might occur on the coast. We are probably looking at thresholds
of three to four feet of inundation which for most people would be up to
their waist. Many kids are only three to four feet high so it
would be life-threatening for them.
A lot has to be decided on this internally in the Weather Service—how we
are going to move forward. It will be a collaborative
process. It won’t just be coming from the Hurricane Center.
We will be working with the local national weather forecast offices to
create this storm surge warning. It will be a collaborative
process coming from both offices during a storm.
The experimental P-Surge flooding map will be available this coming
hurricane season. It will be available on our website when there
is a watch or warning in effect for the United States. Because it
is experimental there will be no data dissemination. For the storm
surge watch and warning we are looking at 2015 experimental product.
In that sense it will be the same sort of ground rules—being in an
experimental phase, it may not always be available. We are going
to work kinks out. After a year or two of testing it should go
operational in 2017 or 2018. With that I am finished and I can
take questions.
Amy Sebring: Thank you very much Robbie. We will move to the Q&A portion.
[Audience Questions & Answers]
Question:
John Callahan: Is P-Surge based on SLOSH model runs?
Robbie Berg: Yes, it is. What happens in P-Surge is
the SLOSH model is run for all the different families of storms
centered on the official forecast. So the SLOSH model is the base
model for P-Surge.
Amy Sebring: Is P-Surge available to the National
Hurricane Center or something that will be available to emergency
management people in future perhaps?
Robbie Berg: It has been available on our website for
people during a storm for several years now. It is available in a
Google type of framework where you can pan and zoom. It is also
available in GIS shape file so emergency managers and others can take
the information, download it and use it as they see fit.
One thing I didn’t go into in depth is that you do select the levels you
are interested in. For the height of the storm surge the levels
go from two to twenty-one feet. Based on what you are interested
in, you would select the level and it would give you the percentage or
chance that the surge would reach that level all along the coast.
It is available now and has been available for a couple of years.
Whenever there is a watch or warning for the United States—that is when
you would see it on the website.
Amy Sebring: You can run it or just get the data?
Robbie Berg: You’re not running anything. It has already
been run on the super computer by the National Weather Service but what
you are looking at is the output from that model run.
Amy Sebring: Will people in the future be able to actually have access to the model itself and not just the data with P-Surge?
Robbie Berg: P-Surge isn’t a model. It is running
off of SLOSH which is the model. People can download the SLOSH
display program. That is available to you now. You can look
at the MOMs and MEOWs which I went over. You cannot look at
P-Surge in the SLOSH display program because it is a real time
product. The SLOSH display program itself is available for
emergency managers to download and use.
Amy Sebring: Do you update the P-Surge and inundation map with each advisory?
Robbie Berg: It is updated every six hours with the new
forecast so you could see some changes from forecast to forecast.
With the inundation graphic since we are using a ten percent exceedance
we are already looking at the upper bound of possibilities. You
shouldn’t see too many major changes from forecast to forecast.
That is an excellent question and it will be available every six hours
with the new issuance of a hurricane forecast.
Question:
Peter Slovinsky: Hello, Peter Slovinsky from Maine Geological
Survey. Questions: has there been any feedback regarding the
proposed storm surge depth color schemes? (Intuitively, I think
of darker colors as deeper areas of inundation). The existing
color scheme seems to correlate well with greater risk of storm surge,
not necessarily inundation depths.
Robbie Berg: I guess I don’t quite understand the question
because the storm surge is what is creating the inundation depths so the
highest depths are in red which mean more danger. I should
provide some background into the development of the product. There
was a very large social science outreach campaign in the marketing of
this product.
We didn’t just sit back in our office and create this product
ourselves. We contracted with several social scientists that went
out and produced surveys for emergency managers, the media, other
weather service employees, the public and they also conducted focus
groups with these audience numbers.
We went through many iterations of developing this particular graphic
based on the feedback we got from all these different groups. As
you can imagine as each group had its own particular wants and needs it
was hard to create one product that will fit and satisfy everyone’s
needs. Based on the overwhelming evidence we got from these focus
groups and surveys, this was the final result we got.
As far as the colors go, we tested a bunch of color schemes whether it
was one color blue with different shadings, different colors going from
blue to red or green to red, and what you see was what bubbled up to the
top as what was most liked and most interpreted correctly by the
greatest number of people.
Question:
Peter Slovinsky: Does the inundation depth outputs take into
account inherent model error (and what is the error of P-SURGE...SLOSH
was supposedly 20%).
Robbie Berg: That is a complex question because the
number that had been quoted historically about SLOSH being accurate
within twenty percent is a little broad and hard to extrapolate that to
how things are now. I don’t have time to show you this but we have
some recent studies on how SLOSH performed in the New York area for
Sandy and we got some really good results.
I would say that the errors, comparing it to tide gages and other data
we had, I think it would be less than twenty percent. I don’t know
how accurate the twenty percent is now but whatever models of accuracy
is in SLOSH itself, since that is the base for P-Surge and the
inundation graphic—that is the type of inaccuracy you might get.
I would suspect the twenty percent that has been quoted historically has
gotten a lot better especially since we have made the model much higher
resolution than it was ten or twenty years ago.
Question:
Kris Ludwig: Thanks for the great and informative presentation.
I may have missed this, but how are changes in elevation implemented
into MOM/MEOW calculations? (e.g., is the DEM revised to include new
elevation data after major storms like Sandy?)
Robbie Berg: What happens at the Hurricane Center is every year
we update a handful of basins. You remember the basins I showed
you. There are different ways that those are chosen to be
updated—whether they are the ones that haven’t been updated in a while
or there has been a storm in a basin and there have been changes in the
coastline.
Those basins are updated based on any new data we receive. A lot
of that is LiDAR data that is being taken from the coastline. In a
sense we are at the mercy of the state that has collected the LiDAR
data. If the state hasn’t collected the data then we don’t have
the new data to update the model or the basins.
It all depends on the states that have invested in updating the data and
whether or not their basins get updated in subsequent years.
Question:
John Callahan: Are there plans within NWS for similar products for extratropical storms like nor'easters?
Robbie Berg: Very good question. The short answer is
yes. We recognize that extratropical storms also produce storm
surge events and some of them are quite significant. The way the
Weather Service is tackling this now is—we are trying to finish the
tropical stuff first. There is a process in play right now where
we are trying to think about how we are going to do the extratropical
storms.
One of the problems is that at the Hurricane Center we focus on
hurricanes. That’s our specialty. There is no agency within
the Weather Service that just does extratropical storms. We have
to figure out a paradigm shift as to how to handle that. There are
slightly different models that are used to forecast extratropical storm
surge because of the structure of the storms.
The short answer is yes. The long answer is it will be a couple of
years out before we get to that stage but our hope is to be able to
produce the same sorts of inundation graphics and have a storm surge
warnings for extratropical storm surge events.
[Closing]
Amy Sebring: On behalf of Avagene, myself, and all our
participants, thank you very much Robbie for being with us today and
sharing this information with us. We wish you continued success
with your efforts and a quiet hurricane season!
Our next program will be June 11th when we will be pleased to welcome
back the ever popular Doc Lumpkins who will be providing an update on
the activities of the National Integration Center where he is now
serving as Director.
Until then, thanks to everyone for participating today and have a great afternoon. We are adjourned.