EM Forum Presentation — April 24, 2013
A Look Back at the 2012 Hurricane Season
and a Look Ahead to 2013 & Beyond
Daniel P. Brown
Senior Hurricane Specialist
Warning Coordination Meteorologist
National Hurricane Center
Amy Sebring
EIIP Moderator
[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.
We would be remiss if we did not note the tragic events in the past
couple of weeks. Our hearts go out to not only the families of the
victims, but also the public servants who have given their energy, and
in some cases their lives to respond. We are also extremely
grateful to the National Hurricane Center for making today’s program
possible on rather short notice.
Today’s program is a look back at the 2012 hurricane season and look
ahead to changes the NHC will implement for the 2013 season. We
will also hear about what is on the drawing board for the years ahead.
Now it is my pleasure to welcome Daniel Brown, Senior Hurricane
Specialist and Warning Coordination Meteorologist at the National
Hurricane Center. As hurricane specialist, he is involved with 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 Oceans. His role as WCM includes coordinating
NHC’s outreach and training activities.
Please see today’s Background Page for further biographical details and
links to several related resources that Daniel will be discussing today.
Welcome Dan, 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]
[Slide 1]
Daniel Brown: Thank you, Amy. Welcome to everyone
who could join us today. I am very excited to be able to provide
the presentation to you and thankfully I did have a lot of this ready to
go on short notice. It is good to see the presentation on the
screen for everyone to see.
I am going to discuss a little bit about the accuracy of the 2012
forecast and discuss some of the forecast challenges the Hurricane
Center faced during the 2012 season. We’ll focus on the storms
that affected the United States and then later on we’ll discuss some of
the 2013 product changes and lastly some of the in-house experiments
that NHC is doing that could potentially lead to some product changes in
the future.
[Slide 2]
The good news last year for 2012 when we look back at how the Hurricane
Center’s track and intensity forecast did—we set several records for
track, which happens often these days, as we have been improving our
track forecast over the past couple of decades. You can see the
average forecast error two days out was only out sixty-nine nautical
miles.
One thing we did last year was set some records for intensity forecast
especially at long range at days three, four and five. That is
somewhat more unusual. I’ll talk about that in a moment. That we
have not been making as much progress in our intensity forecast in the
past couple of decades but last year we did break some records.
Although some of these records were broken probably because we didn’t
have too many rapidly intensifying systems in the 2012 season so some of
these storms may have been a little easier to forecast.
[Slide 3]
I will illustrate here how the track forecast errors have been reducing
and how our forecasts have been getting better over the past 20 or 25
years. This graph shows our track errors at the various timeframes
are. The red line is our average track error at 24 hours and the
green line is 48 hours.
This starts back in 1990. The average two day forecast error in
1990 was about 200 nautical miles and that has reduced to about 69 miles
in 2012.
We have been making some great improvements. I’m sure many of you
have heard us talk about this in the past. We used to say we had
cut the track errors in half since 1990 but actually it is getting to
the point where it is greater than half.
The errors have been cut nearly by about two-thirds during that
time. It is really a great success story in being able to forecast
a storm’s future path and location.
[Slide 4]
Now on the other hand our intensity forecasts have not improved much
over this same time period. These are the average intensity
errors. The red line is for 24 hours and the green line is for 48
hours. You can see these dotted lines are basically flat meaning
we have not improved our intensity forecasts over this 20 or 250year
period.
I guess there is some good news in that in the very recent time you do
see some of these errors coming down very recently. In the last
couple of years we haven’t seen much in the way of rapid strengthening
so you see these errors have come down some in the past few hurricane
seasons but I think that is because we have had easier storms to
forecast.
Iwill note that if you look at the blue line for our five day and this
line is for the four day intensity forecast—those trend lines do show
some downward motion. I think some of that is probably because we
have improved our track forecast at four and five days so we have a
better handle is the storm is going to be over land or over water which
contributes to the forecast error.
If we can’t get that right we certainly would likely miss the intensity
forecast. Because the tracks have gotten better the longer range
intensity forecasts appear to be improving slightly but there hasn’t
been much change for the short term intensity forecast.
[Slide 5]
This is a look at the forecast errors for each storm. If you look
at the storms such as Beryl that were early in the season, the forecast
errors were pretty low. These were some of the more successful
storms we forecast during the last hurricane season. You can see
the average errors at three, four and five days listed by the line—you
can see there are a few storms we really struggled with last year.
We’ll talk more about these. One of those was tropical storm
Debby. It formed in the Gulf of Mexico early in the season.
That was a real struggle to forecast where that storm was going to
go. I’ll show you the model guidance we had to work with and why
that was such a struggle.
There were a couple of storms that were thankfully out over the open
ocean that were more of a struggle including Kirk and Nadine, which
lasted for several weeks. Isaac, although we did well in the short
range forecast—some of you may remember if you were joining us from the
Gulf Coast or from Florida—you may remember some of the early forecasts
for Isaac that were more up towards the Florida peninsula more than out
into the Gulf and Louisiana where eventually Isaac made landfall.
We will talk more about that.
Sandy on the other hand late in the season was a fairly good forecast
success from a track and intensity forecast perspective but there were
other challenges we faced forecasting Sandy. You all may know the
dilemma on watches and warnings which I will discuss later.
[Slide 6]
I’ll start out with Debby and say that if you are familiar with our
graphical tropical weather outlook which is on the left side of the
screen here—this was a well-forecast system. We knew it was going
to form. The model guidance did a very good job of forecasting
Debby’s formation, but the problem was where was it going to go?
On the top right is the model guidance and how divergent it was and I
will walk us through that in a moment. For those that don’t
remember, Debby ended up affecting Florida. It made landfall as a
minimal tropical storm near Steinhatchee, Florida on June 26. It
caused five direct fatalities.
Its main impacts were from both fresh-water flooding from a lot of heavy
rainfall and also some storm surge flooding. It was just a
tropical storm but it does illustrate the dangers of storm surge
flooding even from a tropical storm along the west coast of Florida
where there was inundation in some areas of one to three feet.
Overall Debby caused about $250 million in damages.
[Slide 7]
Here are these model forecasts for Debby when it first formed.
This was the dilemma we faced in the first advisory. I have to
joke a little bit that I was on vacation. I picked a good day not
to be in the office as this was a very difficult forecast. Two of
our better models we have in the past few years—one of them is the
European model which had its five day forecast point in Texas and the
U.S. Global model, here in black—its five day forecast was east of the
Florida peninsula.
You can see how divergent the models were. A couple of the models
favored Texas and a few favored Florida and this made for a very
difficult forecast. Our first forecast put a little more weight on
the European model since it had been the best performing model over the
previous few hurricane seasons. Our forecast showed a little bit
more of a track toward the west which ultimately did not happen—leading
to those large errors. I’ll forward a little over time.
[Slide 8]
In advisory number two the models were still very divergent with some
favoring Texas and some Florida. A lot of people may hear us talk
about the consensus models in our tropical cyclone discussions about how
we do use the average of many of these models to help us with track
although in this case it really didn’t help.
[Slide 9]
When you average out the tracks over Florida and Texas you get a track
towards the northern Gulf Coast which really did not seem plausible in
this case. Those were not much of a help either.
[Slide 10]
Over time you can see the model guidance gradually shifted towards the Florida peninsula in our forecast.
[Slide 11]
Here are all the Hurricane Center forecasts for Debby and you can see
how we did struggle. It just illustrates that even today with the
improvements we have seen in track over the past 20 or 25 years that
there still occasionally are times when the models very divergent and we
will struggle and have less confidence in the Hurricane Center
forecast. A good way to gauge that confidence is by reading our
tropical cyclone discussion product. Hopefully you are familiar
with that.
[Slide 12]
Just to review a few of the impacts in Florida—Debby was a significant
tornado producer. There were 24 tornadoes that occurred across
Florida. There were a couple of deaths related to the tornadoes
that did occur. Some of these are fairly spectacular pictures that
were taken in Florida of tornadoes that were associated with Debby.
[Slide 13]
The main impacts with Debby were the water hazards or storm surge and
heavy rainfall. This onshore flow that occurred for several days,
because Debby was a slow mover, ended up piling up water along the west
coast of Florida. Basically from southwest Florida through the Tampa Bay
Area all the way up into the Florida Big Ben region ended up getting
this onshore flow and storm surge that occurred near the coast and did
cause some problems and damage.
[Slide 14]
Now we’ll move over to Hurricane Isaac. I’m sure many more people
remember Isaac. Isaac formed east of the Caribbean islands and
ended up making landfall in Haiti and Eastern Cuba and then moving up
across the Florida straits. Then it made a fairly slow track
across the Eastern and Central Gulf of Mexico. It actually slowed
down as it approached the Louisiana Coast.
It became a hurricane shortly before landfall. Its main hazard was
also storm surge. We saw storm surge inundation as much as ten to
seventeen fit across Plaquemines Parish in Southeast Louisiana and
eight to twelve feet across Saint Bernard Parish. Isaac ended up
causing about $2 billion in damage. I’ll talk here about some of
the challenges we faced.
[Slide 13]
The next slide shows some of the pictures of the storm surge that
occurred. Something to remember with Isaac is that Isaac was a
large storm which does have implications on how much surge a storm will
produce. Because of its large size it produced a rather high storm
surge in a storm surge vulnerable area.
Some people seemed to be caught off guard because they ended up with
more surge from Isaac than from Katrina in some areas but the reminder
is that it took a track that was different than Katrina and its large
size helped to contribute to this surge for a storm that was not as
intense as Katrina was. Surge was definitely the story with Isaac.
[Slide 15]
This shows the struggles we had early on in the storm. Some of you
may remember that the Republican National Convention was scheduled to
occur in Tampa Bay the week that Isaac was forecast to be in the Gulf of
Mexico area. The early forecast showed this threat towards the
Florida peninsula.
This is the model guidance we had when it was just entering the eastern
Caribbean. You can see all these models that were pointed up
toward Florida and the west coast of Florida near the Tampa Bay
region. It made the forecast difficult.
Also the intensity forecasts were very tough for Isaac because of the
amount of land interaction. Several of our forecasts—especially
early on we had a high bias in the forecast—because Isaac ended up
interacting more with the land mass of Dominican Republic, Haiti and
Cuba, making those forecasts even more difficult.
[Slide 16]
When Isaac finally exited the north coast of Cuba there was still a
fairly large amount of spread in the guidance for where it might impact
the Gulf of Mexico. This is actually a three day forecast with a
center right along the northern Gulf Coast. You can see that the
model guidance was basically indicating landfall from anywhere from the
Florida panhandle region all the way over into southeast
Louisiana. It was a pretty wide area for three days that the model
guidance suggested could be impacted by the center of the storm.
[Slide 17]
At this point the Hurricane Center forecast was nearly a consensus, or a
middle of the model guidance, and we had to swing the forecast a little
big left as the model guidance came into better agreement towards
southeast Louisiana.
I will point out here that sometimes if you look in white where the
actual track of the storm progressed it actually is outside of the
guidance once you get beyond about two days. That happens on
occasion. The storm actually moves outside the sweep of model
guidance. A lot of people now go to the internet and look at the
model plots they see. I caution people because it does not
always show you the entire range of possibilities for a given forecast
scenario.
[Slide 18]
We’re moving now into Sandy and talk about Sandy’s impacts and also
forecast challenges we faced. Sandy was in late October. A
storm formed here in the central Caribbean, moved northward and began to
quickly strengthen as it approached eastern portion of Jamaica and then
ended up strengthening into a Category Three storm just before it made
landfall in eastern Cuba.
It moved across the Bahamas. It weakened somewhat and actually
became a tropical storm for a brief period of time before
re-strengthening and then taking a fairly unusual track and turn back
toward the mid-Atlantic and northeast United States. Its
transition from a tropical cyclone which gets its energy source from
warm water into more of an extratropical or more winter-time type low
pressure area made for some forecast challenges on how we dealt with the
watches and warnings.
We will talk about that and I’ll end with some changes that because of
Sandy we will be implementing at the Hurricane Center and the weather
service this year.
[Slide 19]
As far as Sandy, it did produce hurricane force winds across southern
parts of New England and Long Island at least in gusts, and those winds
knocked down trees that ended up killing 20 people. The total
death toll in the United States because of Sandy was 72 but a little
more than half of those deaths were from storm surge.
Storm surge was again the primary hazard and what produced much of the
damage with the significant surge occurring basically from the New
Jersey coast through New York and into southern New England. This
map shows the inundation values—really how much water ended up over land
that is normally dry.
The highest values were up here in northern New Jersey and also near the
Staten Island area of New York, here the highest was around nine
feet. I did have an opportunity with a couple of us here at the
Hurricane Center to do an over-flight of this area about a week after
Sandy and the devastation was quite striking. I was an awful event
for the folks up in that region.
[Slide 20]
Here are some pictures. I’m sure everyone has seen various damage
pictures from Sandy or perhaps some of you have been on the ground and
seen it yourself in person. It certainly illustrates the
devastation storm surge can bring to a region and also how widespread
that damage can be.
[Slide 21]
As far as the forecasts for Sandy, I mentioned earlier, that overall the
forecasts were pretty good. Here in black are the track forecasts
throughout the entire life of Sandy. You can see there were a few
times where we had it forecast to be a little more offshore or too far
east but the model guidance picked up on this turn fairly early such
that we were able to shift our forecast a little back to the west and
ended up with some very good forecasts several days in advance of
landfall in the United States.
The intensity forecast also ended up being very good for Sandy
especially for the approach to the United States. We did struggle,
which we often do, with the rapid strengthening that occurred near
eastern Cuba.
[Slide 22]
Here is a slide to illustrate that. Here is the picture of Sandy
as it made landfall in eastern Cuba. You can see in this graph on
the bottom left in white is basically what happened—the intensity of the
storm at every six hours through its life. All the lines in light
blue are the actual Hurricane Center intensity forecasts.
You can see highlighted in this purple circle that the Hurricane Center
forecast did not forecast this rapid change in strength just before
landfall. That is something we still struggle with. We don’t
have the intensity model guidance that can predict these real rapid
changes in strength.
There are several reasons for this. One is that with intensity
forecasting it is really a much smaller scale phenomenon or mechanism
that is causing the rapid changes in strength. We can look at the
overall storm environment to see whether it is favorable or not but it
really is oftentimes these small scale things that are going on inside
the storm, so it is difficult to observe and model that. We don’t have
quite high enough resolution models and there is some understanding we
don’t have as to these processes that go on.
It is something the Hurricane Center and NOAA are working on. As
you see on this slide there is a project called the Hurricane Forecast
Improvement Program. It is trying to tackle some of these issues.
It is a ten year program that hopefully will provide the Hurricane
Center with some better modeling efforts which we are starting to see
now in real time. Sandy really did illustrate this problem we have
with forecasting rapid changes in strength.
[Slide 23]
As far as the warning challenge in Sandy leading up to Sandy’s landfall
in the United States—our model guidance forecast that Sandy was likely
to make this transition from a pure tropical system to an extratropical
system, more like a wintertime low where it gets its energy more from
the contrast of air masses and various fronts that are tied to it.
This led to some difficulty because once the Hurricane Center starts
issuing warnings it is difficult for us to switch between the tropical
cyclone or tropical storm and hurricane watches and warnings and the
non-tropical warnings and watches that are issued by local weather
forecast offices. It left three scenarios that were not very
appealing during the event.
One was that if this occurred, even if it occurred a couple of days
before landfall, that the Hurricane Center, if it were no longer a
tropical system, would cease to issue advisories and transfer the
responsibility to other local weather offices as well as the
Hydrometeorological Prediction Center. At that point we would be
required to drop the hurricane watches and warnings.
That was a solution we really didn’t want to see happen because we heard
from a lot of emergency managers and others that they didn’t want to
see us change the warning type during the middle of the event because
that could be confusing.
The other possible thing was to just continue calling Sandy a hurricane,
even if it really wasn’t one. That could have been more than a
day or two before landfall—the timing of when the transition would occur
was not well known. We felt that could misrepresent the situation also
and damage our credibility as you can only really stretch the science
and also calling it what it was for so long.
The third would be to actually continue issuing advisories on this
system even after it became non-tropical or what we call post-tropical
but we had advertised in years past when we switched to this new naming
convention for post-tropicals that whenever we issued a post-tropical
advisory it would be the final advisory.
Being able to disseminate additional advisories after the one which we
called the post-tropical we felt could end up breaking some
dissemination processes that are in place by the media and other vendors
that use our products. We just didn’t know that the flow of
information would continue after that transition so we felt we shouldn’t
try that in real time during this.
So what we decided during several meetings was to eventually issue these
non-tropical warnings to north with the forecast that it was going to
transition. The local offices issued the appropriate high wind
warnings and covered the event that way.
I’ll talk a little now about what has happened since Sandy’s landfall.
As you all know, Sandy really transitioned late in its life over the
water and transitioned as it made landfall and became post-tropical just
a couple of hours before landfall in New Jersey.
[Slide 24]
What are we doing to meet this kind of warning challenge for this year
and beyond? Based on what happened last year, and I will talk
about this more in a minute, the Weather Service has basically broadened
its tropical storm and hurricane warning and watch definitions to
allow them to be used for post-tropical cyclones that pose a
significant risk to life and property.
Some of you may have heard about this change already. This change,
the Weather Service and NOAA is doing an assessment of its products and
services during Sandy which it typically does for big events, and this
question was posed to some emergency managers and users of our
products. Preliminary information is that everyone thinks this is a
good change and that is why it is being implemented for 2013.
This is going to give us the option to continue issuing advisory
products on post-tropical cyclones that pose a risk to life and property
and also when the transfer of responsibility from the Hurricane Center
to another national office will result in this unacceptable
discontinuity of services. It gives us more options in a case such as
Sandy in the future.
The other news you may have heard about is that the Hurricane Center for
several years has looked at the possibility of issuing storm surge
watches and warnings. We have been looking at this for the past
decade. With many recent storms such as Katrina, Ike, then Isaac
and Sandy this year illustrating that we need to get more public
response, make sure we get appropriate public response to the actual
hazard that poses the greatest risk to life and property. And that is
storm surge. That is why we do evacuate from the coast.
With this and the events that recently occurred we at the Weather
Service have made a decision to move forward with the storm surge watch
and warning. This will be issued independently of where the actual
hurricane wind warning could be issued. Right now the target date
is for this to be operational by the 2015 hurricane season. I
will talk more about this in a few slides.
[Slide 25]
Here is a prototype of the storm surge warning. It would be issued
for areas right along the coast. This would be a collaborative
process between folks in our Hurricane Center Storm Surge Unit and also
the local weather offices around the country. This is something
planned to debut in the 2015 hurricane system.
[Slide 26]
Another potential upgrade to our storm surge guidance and products is a
storm surge inundation graphic. This graphic is based upon some
probabilistic storm surge guidance that is created at the Hurricane
Center which takes the forecast track and looks for average
errors. Then takes storms of varying intensities and sizes into
the coast spots to get the chance of surge being X number of feet at
this location.
Those products are already available at the Hurricane Center
website. This takes that information and provides the most
realistic worst case scenario is from a potential hurricane
landfall. This graphic is still being fine-tuned here at the
Hurricane Center. It has been tested with some focus groups and
most people feel there is a lot of benefit and need for this graphic.
This is something that could potentially be available for hurricane
landfall during the 2013 hurricane season. A final decision has
not been made as to whether it will be available for this year or next
year but it may be available in the 2013 season.
[Slide 27]
Some other 2013 changes that are going to occur here this season—one of
these is that the definitions of tropical storm and hurricane watches
and warnings have been broadened. The issuance criteria of our
advisories have changed to allow us the option to use both for tropical
cyclones that pose a threat to life and property as I discussed a few
slides ago. Now if we had a similar Tropical case like Sandy we
could issue and maintain a tropical storm or hurricane watch and warning
for systems that might transition before landfall. The Hurricane
Center can continue to issue advisory products on that.
I’ll go over the other three on this slide. I’ll talk about the
new use of our National Hurricane Center Cyclone Update product and a
planned extension of the Tropical Weather Outlook up to five days in the
next few slides.
[Slide 28]
Whenever a storm was approaching the coast and could be tracked pretty
easily on radar the Hurricane Center issues advisories every two hours
and in between those two hourly updates there would be a tropical
cyclone position estimate product that would come out.
The test product used to tell you where the storm was located, how many
miles it is from a major city and what the current intensity and
movement is. In order to streamline our product suite we actually
discontinued this product in favor of issuing this in a current existing
product called the tropical cyclone update.
[Slide 29]
The tropical cyclone update is issued when unexpected changes occur in a
cyclone if we need to issue international watches or warnings in
between advisory times. A new use for that will be for these one
hour updates whenever the storm is making landfall and it can be fairly
easily tracked on radar.
We also sometimes use this product to tell people when landfall has
occurred. Oftentimes the media wants to know when landfall has
taken place so it is product we can use for all those reasons. It
is not a huge change for folks out there but it is another place to get a
continuous flow of information when a hurricane is making landfall.
[Slide 30]
One change I think all of you will really like for the 2013 season—this
is something that is not going to begin on June first. Over the
past few years we have been looking at extending the time period covered
by our tropical weather outlook. The current tropical weather
outlook covers a two day or 48 hour period.
In the past several years we have been assigning the chances of systems
forming over the next five day period, kind of behind the scenes.
We have done verifications of that and it has proven to be fairly
accurate. Beginning this season, at some point during the
season—we are going to be doing some in-house testing and we still have
some technical issues we need to work out in the early part of the
season—but I would think by July or August timeframe you will start to
see this in our text product.
That is—you will get a chance of what this system’s formation potential
for the first two days and also the chance of formation over five
days. I’ll give you a little more information about the longer
term of the system becoming a tropical depression or tropical
storm. This is how this will look in the current existing text
tropical weather outlook.
It makes the paragraph a little longer but you can see here we mentioned
the 40% chance of formation in the next 48 hours and later on in the
paragraph it tells you there is an 80 percent chance of it becoming a
tropical cyclone during the next five days. We get both of these
probabilities now in the outlook beginning in July or August.
[Slide 31]
The other good news is that we are working on a graphic to go along with
that. We are not sure if this will be ready to debut in
2013. We still have technical issues to overcome but this is a
prototype of that graphic.
The current tropical outlook graphic that you see with the satellite
image showing the current location will not change but this graphic will
be in addition to that and will show where the disturbance is located
currently by an X and then will show the area of possible formation over
that five day period and also show using low/medium/high categories
what the chances are.
[Slide 32]
If you click on these you will be able to actually see what the
probabilities are as you click. Some of these areas will overlap
and each one will have its own separate graphic as well and you can get
the probabilities right on the graphic. I think this is a good
addition to help people have a little more time for planning when
systems are forming near land or near the East Coast of United States or
Gulf of Mexico.
[Slide 33]
Just to remind folks about the forecast cone—with our improvements in
track forecasting our cone continues to shrink in size. It is
constructed by circles that are around each of the forecast times in
which two-thirds of the verifying forecast should fall within that
circle. This is based on the five year error distribution from our
previous errors in the previous five hurricane seasons.
As our forecasts continue to improve the size of the circles has
continued to shrink. I would remind folks that storm size has not
changed. It just means that you can get more impacts outside of
the cone. We know the cone is a great briefing tool because it
provides where the center of the path of the storm is likely to go but
it doesn’t tell you much about impacts.
[Slide 34]
With the improvements in track forecasting the size of the cone is going
to shrink again by about three to eight or nine percent than it was
last year. It is not a huge change. You probably won’t
notice much but if you do go back and look at cones from several seasons
ago to today you will certainly notice a difference.
[Slide 35]
Here is the cone from Isaac advisory in 2013 if it had the 2008 size
cone on there. If you look at the edge of the cone it is at about
Jacksonville, Florida all the way to extreme southwest Louisiana.
If you move forward and apply what the 2013 cone would look like you’ll
see it is much smaller stopping at the edge of the Florida panhandle and
over central Louisiana.
[Slide 36]
That is not the change from last year to this year—that is the change
over a five year period. The change from last year to this year is
pretty small. It is just to illustrate how the cone has
been shrinking over the past 10 to 15 years since we introduced the
cone.
[Slide 37]
I’ll move now into some other potential improvements beyond the 2013
time period—some in-house experiments we are working on here at the
Hurricane Center. One of those I have already mentioned is the
storm surge warning being planned for 2015 and the inundation graphic
that might be available this season but if not it will be available in
the 2014 season.
Some other things we are working on here are producing track and
intensity forecasts for disturbances—for systems that have not yet
become a tropical cyclone and whether or not we could start to issue
watches and warnings for those systems. The radar image on the
screen is the system Tomas back in 2010 that was formed very near the
Caribbean islands.
It actually strengthened quickly and became a hurricane as it moved over
the islands. It impacted in such short order that that wasn’t
much lead time on the tropical storm or hurricane watches and
warnings. We are looking at ways to issue watches or warnings
before systems form. We are also looking, as our model guidance
has gotten better, at the extension of our forecast out to seven days.
[Slide 38]
This is a forecast we issued in the upper right hand corner before Irene
actually formed. This was a forecast about a day before
Irene. In the black is the forecast we actually laid out.
The actual track is in white. We did a pretty good job of
forecasting where it would go even before it actually formed. We
were making these forecasts a couple of years ago and we have been
making them in-house.
Right now the forecasts are pretty good but not as good as the
forecaster systems that have already formed for normal tropical storm or
hurricane forecasts but it is something we are looking at.
At some point we are hoping to be maybe be able to issue watches and
warnings before a system forms.
[Slide 39]
Here are a couple of prototype graphics of how that might be done.
We are still working out the details on this. We don’t know
whether we would issue a full advisory of products for a system like
this that is still a disturbance or may only issue a watch or warning on
one of our existing products like the tropical weather outlook.
Again this would be in coordination with the local National Weather
Service offices. I suspect this would be two or three years down
the road at the earliest but it is something we are thinking about here
at the Hurricane Center. I look at Sandy and what happened with
the watches and warnings and we are fixing that on the back end of a
system when it transition—this is looking at the watch and warning
problem as it forms or at the front end of a system.
[Slide 40]
Lastly I’ll finish up with a six and seven day forecasts as the models
have improved. We have started looking at issuing six and seven
day forecasts. The good news is last year we did pretty well on
the six and seven day forecasts. This graph shows in black the
Hurricane Center verifying six and seven day forecasts.
You can see the average error at day six was about 240 nautical miles
and at day seven it was about 300 miles. In red and blue are our
two primary best models that we get—the U.S. Global Forecast system and
the European model. You can see we, the forecasters here, actually
outperformed both of those models at days six and seven.
[Slide 41]
We are planning to do an in-house test on this again this year. I
will mention that when you take those 240 nautical miles and 300
nautical miles errors and place them back on the graph I showed at the
beginning of the presentation it shows you the day six forecast is
pretty close to what the day four forecasts were about a decade ago when
we introduced those. The day seven forecast of about 300 nautical miles
is also now equal to about what the day five forecasts were a decade
ago.
It shows that we probably will start issuing six and seven day forecasts
in the near future. We haven’t decided exactly how we will do
that and I think that is some discussion we are still having—whether we
extend the current cone graphics out to day six and seven or display
these in a different manner because there is more uncertainty in the
longer range forecasts than the shorter range forecasts..
[Slide 42]
I will caution folks that even though we did well last year not all
forecasts will be perfect. Here is our six and seven day forecast
for Sandy and you can see it is a large spread. Here is the
European model that was the first to show landfall at day seven.
The USGFS model is out here in black well off the East Coast of the
U.S., and at the time this was the Hurricane Center forecast circled in
red. It would not have been a very good seven day forecast—much higher
than the average I showed on the other screen.
[Slide 43, 44]
A day or so later the day six forecast begins shifting back toward the
U.S. and by day five the forecasts were pretty good. You can see a
little word of caution about the use of six and seven day forecasts and
how we might want to display those forecasts.
[Slide 45]
The Hurricane Center is using social media. I know a lot of
emergency management agencies are also using social media. Our
social media accounts have been pretty popular. We have about
208,000 followers now on our Facebook page and about 110,000 followers
on our Twitter feed. The main Hurricane Center Twitter feed is
actually an automated tweet of our advisory products when they come out,
but I would also mention that the NHC Storm Surge Unit and Rick Knabb,
the Hurricane Center director, are also very active on Twitter if you
like to follow them.
For those of you who have social media accounts in the last couple of
years we did some public service announcements that were recorded before
the hurricane season. This is an example of one—we actually re-recorded
those this off-season to freshen those up. Those should be
released in the next few weeks via YouTube. Please look for that
on our website and through our social media.
I suggest you could use those during National Hurricane Preparedness
Week which is the last week of May or perhaps you could link to them on
your accounts as well. It talks about the various subjects that are
discussed during Hurricane Preparedness Week and also it has some NHC
forecasters and Craig Fugate from FEMA doing some of the recordings.
[Slide 46]
I’m sure every time there is a hurricane threat or major threat you see
this uptick in number of followers or users and that was the case last
year with Debby and Sandy and some of the major upticks in users.
It is interesting to follow that.
[Slide 47]
For those of you that want to see the faces behind the forecasts I’ll
leave this slide up as I answer questions. These are your
hurricane specialists. You’ll see their names at the bottom of
advisories. It is a pretty tight knit group of folks and we
certainly love and enjoy our jobs and hopefully working with you all to
make coastal residents better prepared and help them make better
decisions when a storm is threatening the area.
Amy Sebring: Thank you very much Dan. Great presentation.
I welcome back to this slide in a moment. We will move to the
Q&A portion.
[Audience Questions & Answers]
Question:
Darren Thraen: Can you touch on what is being predicted for the 2013 hurricane season?
Daniel Brown: NOAA’s seasonal outlook is issued jointly
here with the Hurricane Center and also the Climate Prediction
Center. That forecast will be issued on May 23—towards the middle
to late part of May. I’m sure some of you have seen some of the
other season forecasts that have been issued to date.
Most of these are suggesting we will have another active hurricane
season. I will throw a word of caution in—I know that others may
have heard folks in the Hurricane Center reiterating this—seasonal
forecasts at this point do not tell you where the storms are going to go
and where they will make landfall so as coastal residents and emergency
managers there is not much you can do with a seasonal forecast.
You have to prepare each season as if you are going to get hit. It
only takes one to hit your area for it to be a bad year.
The classic illustration of that is back in 1992 we only had seven named
storms—four hurricanes and one major—a very inactive year if you look
at it in terms of overall numbers. That one major hurricane was Andrew
which struck Florida as a category five storm. Then you go to 2010
that had nineteen named storms and twelve of those were hurricanes but
none of those hurricanes impacted the United States.
Use the seasonal forecast to get people to understand and know that the
season is upcoming and as a sign for preparedness. Besides that
there is not a lot of utility in the seasonal forecast on an individual
preparedness basis.
Question:
Nader Mehravari: Can you comment on reports that the European
UK center ECMWF had forecasted hurricane Sandy hitting NY/NJ regions as
early as 8 days ahead of time. Are they using tools and techniques that
are not available to NOAA?
Daniel Brown: If you remember the slide a couple of slides
(slide 42) ago it illustrates that the European model was the first to
predict Sandy’s turn back to the United States. The European model
is a very good model. For the past two or three hurricane seasons
it was the best model. One of the metrics we look at is a two day
forecast and compare all the models and it was the best at two days for
two or three hurricane seasons in a row.
This past year the USGFS model actually edged the European model out at
the two day metric. Yes, the Europeans do have some different
tools and techniques they are do. That model is finer resolution
than the current US model. Really and truly I would say both the
USGFS and the European are the two models we rely on the most here at
the Hurricane Center.
I would just caution folks that when they are looking at some of the
model guidance to trust the Hurricane Center forecast. We may have
reasons why we go with various models at various times but as you saw
with the slide looking at the five, six and seven day forecasts, the
Hurricane Center tends to beat the individual models long range for an
entire season.
Perhaps in individual time periods the models might beat us but over the
long haul the Hurricane Center outperforms most of the individual
models.
Question:
Any Sebring: From storm to storm you may have varying
levels of confidence in the forecast you pick up in the tropical cyclone
discussion. Has the National Hurricane Center considered trying
to quantify the level of uncertainty in a forecast by storm?
Daniel Brown: That is a really good question. Right now
we more or less subjectively talk about forecast confidence in our
cyclone discussion. You’ll see us talk about the model
spread. Our confidence in forecasts is often dictated by the
spread of the guidance. Some of you may be familiar with some
products we issued called wind speed probability products. These
provide you the chances of receiving a tropical storm of 58 mile per
hour and also hurricane force winds at individual locations.
The technique that produces those uses our past errors to help compute
those chances. That technique also takes into account this model
spread. It has varying confidence levels. Typically when the
spread of the models is lower the Hurricane Center forecasts are
better.
We have thought about using that same technique where it looks at the
spread of the models to have varying cone sizes in cases where the
spread of the models is low and we have high confidence the cone size
would be a little smaller. In times where there is a lot less
confidence in a forecast the cone would be larger.
We have not gotten to the point to incorporate that yet. It is
something we have thought about. If we do at that point we would
put something in the discussion that would quantify it such as
confidence in this forecast is low, medium or high. That could be
something that might be done in the future.
I think we want to include that somewhere in our products because there
are others that use our numbers to make their own cones. This
would provide them guidance as to which cone size we are going to employ
for a given forecast scenario. Right now none of this is
occurring. The cone size stays the same for the entire season and
is not based on model spread.
Question:
James Lewis Free: Does NHC have plans to significantly update
its platform of real-time maps and infuse them with GIS? Are there plans
to eliminate all caps text products? Thanks for the great
presentation.
Daniel Brown: Yes to the first question—the Hurricane
Center has a lot of its products already available in GIS format but
yes, we would like to do more and better products in GIS. Some of
that is a resource limitation currently but over time hopefully we would
continue to provide more in GIS. Some of the storm surge
inundation, and other things, we would like to get that not only as a
static graphic but also in GIS.
The use of caps—why we use all caps goes back to the days of teletype
and the World Meteorological Organization and making sure all our
products get down to the lowest dissemination method. That
regulation was recently loosened by the World Meteorological
Organization and the Weather Service has begun to experiment with mixed
case products.
The Hurricane Center was thinking about trying that with our tropical
cyclone discussion for the upcoming season but because of technical
resource limitations that will wait until 2014 because we have higher
priority projects and things like you saw me present here that we want
to get out the door first before we spent the time to go to the mixed
case.
Hopefully in the next few years you will see us start embracing mixed case like other parts of the Weather Service have.
Question:
Ray Pena: It seems you use cyclone and hurricane
interchangeably. Why not stick with hurricane? (I have also
heard cyclone used interchangeably with tornado.)
Daniel Brown: I know that when we do presentations we
sometimes get into the technical jargon. I used the term “tropical
cyclone” because it really is a catch-all for tropical depressions,
tropical storms and hurricanes. All of those are classified as
tropical cyclones. A tropical cyclone is classified using its wind
speed into the depression, storm and hurricane categories.
That is why we sometimes lump the word “tropical cyclone” to mean all of
those—depressions, storms and hurricanes. The post-tropical
cyclone is a term that simply means it is no longer a tropical
cyclone. It was once one but it is no longer. That term, we have
been using for a couple of years, but it got widespread use during Sandy
and then the media really began the coin the term “superstorm”.
Question:
Amy Sebring: You talked about the ten year improvement
program. Is there anything on the horizon that holds promise for
improvements in the intensity forecast?
Daniel Brown: There is one of the early results that
shows promise. When the NOAA hurricane hunters fly out into the
storms those aircraft have Doppler radar on the aircraft that can
collect Doppler radar data. That has been used to initialize some
of these higher resolution models. It shows some promise.
I mentioned during the discussion is that one of the difficulties we
have is to observe the current structure of the storm. Having that
data into the models to accurately depict the current structure of the
storm goes a long way to helping the forecast. We don’t get the
aircraft data, especially from the NOAA aircraft, all that much but this
is something that is starting to hold some promise and could be helpful
in the future. It will probably be several years.
[Closing]
Amy Sebring: On behalf of Avagene, myself, and all our
participants today, thank you very much Dan for joining us today and
sharing this information with us. While we hope that the 2013
season will not be too active, we do wish you the best.
Our next program is scheduled for May 8th when we will feature the
Silver Jackets program, an initiative that provides an opportunity to
bring together multiple state, federal, and sometimes tribal and local
agencies to learn from one another and apply their knowledge to reduce
risk from flooding. Our guests will include the national program
manager, Jennifer Dunn, as well as a number of state participants.
Please make plans to join us then.
Thanks to everyone for participating today and have a great afternoon! We are adjourned.