Research in NOAA
Interview with David Stensrud
November 8, 2007
The following may contain unintelligible or misunderstood words due
to the recording quality.)
BARRY REICHENBAUGH: This is Barry Reichenbaugh with the
NOAA Research Communications Office, and I'm in Norman, Oklahoma, at
the National Severe Storms Laboratory with Dave Stensrud. Dave, can
you tell me what it is you do here?
DAVID STENSRUD: Technically I'm a research meteorologist.
My main focus of my research is to improve forecasts of severe weather
events, including things like damaging winds, perhaps hail, tornados,
even winter weather, in terms of blizzards or freezing rain events.
BARRY REICHENBAUGH: Well, let's build on that a little
bit. My first question has to do with the types of research done here
to improve forecasts and warnings. Can you talk a little bit about that?
DAVID STENSRUD: Yes. Forecasts -- perhaps you could
say they're fundamental tool that we use for producing forecasts of severe
weather, is what's called a numerical model. In its most basic essence,
a numerical model is just a computer program, but it happens to be a
very sophisticated computer program because it has equations in it that
let us know how the fluid -- because the atmosphere is a fluid -- evolves
But those equations have parts that we have to approximate.
And approximations introduce errors. And so one of our tools is basically
trying to find out how best to use these imperfect models to give better
and better forecasts for severe weather events.
BARRY REICHENBAUGH: Could you describe NSSL's work regarding
ensemble modeling, and what is ensemble modeling?
DAVID STENSRUD: Ensembles are just a group of forecasts
created by a computer that are valid over the same time period. Ensembles
are very helpful because we've known for about 40 years now, in particular,
that the atmosphere is very sensitive to errors in our initial picture
of the atmosphere. So to start a model up, you have to actually have
a picture of what the atmosphere looks like right now.
And so we, you know, launch weather balloons at Weather
Service offices across the country. We have satellite data. Other countries
also launch balloons and we have surface observations. Even when you
fly back home, your aircraft may actually have instruments on it that'll
sample the temperature and moisture of the atmosphere.
And so first thing we do is collect all these very diverse
and different data sets and try to mold them into a picture of what the
atmosphere is like at this moment in time. But that picture's imperfect,
and those imperfections lead to basically forecasts being incorrect at
some future point in time.
And so an ensemble will take our initial picture of the
atmosphere and will sort of adjust it, or you could say shake it a bit.
We know it's an error by a certain amount on average, and so we can adjust
those conditions and get a variety of different initial pictures that
all are pretty close to what we think is true.
And from each one then we can make a forecast using a
model. And that gives you more of a probabilistic approach to actually
looking at what's going to happen for tomorrow.
BARRY REICHENBAUGH: Can you go a little bit into what
you mean by a probabilistic approach? I understand you're talking about
ranges here, but for the benefit of the nonscientist in terms of probabilities,
what are you getting at with that?
DAVID STENSRUD: Well, you can look at it perhaps from
a gambling perspective, where you know the odds may be of -- if you're
holding two tens and you're playing blackjack, you might know the odds
of the dealer actually getting 21, for example. And so probabilistic
forecasts are dealing with the odds of different events happening.
I think we're most familiar with precipitations forecasts
being done on probabilities. 20 percent chance of rain for tomorrow.
But there also might be a 20 percent chance that tomorrow's weather is
going to be sunny and highs may be in the 70s, and there's an 80 percent
chance that it's going to be cloudy and maybe highs only reach the low
And of course if you forecast like we do nowadays on TV,
you might get, The high tomorrow is going to be 60. But you're not telling
people it also might be very nice and it might be 70 degrees. And so
we're trying to give the full information, everything that we know, because
then people can make decisions.
It's always more helpful if you know the odds of something
happening. And in particular, for example, if you're a power company
in Phoenix, Arizona, and it's summertime, and so a demand for power for
air conditioning is pretty high. And maybe on a given day there's a
50-50 chance of it being 110 versus 95. And that's a big difference,
then, in power demand for air conditioning, those two temperature forecasts.
But if they know the odds of each one happening, they can make a decision
how to save money, in essence, depending on betting on the odds.
BARRY REICHENBAUGH: The National Severe Storms Lab has
long been known for radar research. And I'm wondering if you could talk
a little bit about how radar data may be used in these models that are
DAVID STENSRUD: Yes. NSSL has a long history in looking
at radar data from a perspective of pretty much real-time operations
and making warning decisions. So you look at the radar data. You diagnose
it. You look for various patterns and then you can tell, based on what
the radar tells you, if you have severe thunderstorms. And then you
can make a persistence forecast, which is that this storm will last a
while. And so based on track, you can predict what town it'll go over
in the next 20 minutes or so.
What we'd like to do is be able to make a bit more a short-range
forecast. So if we can get the radar data into the model, then we can
actually use the model then to not only just look at what the storm is
right now, but how it might change over the next 20 or so minutes. And
by doing that, we actually might extend lead times for warnings for tornados
or hail or damaging winds, because we not only take a persistence view
of it, we can also then put it in the model and hopefully the model will
give us information to improve the warning.
BARRY REICHENBAUGH: Can you describe briefly what this
Hazardous Weather Testbed is and then how has that been helping NSL's
overall research efforts?
DAVID STENSRUD: The Hazardous Weather Testbed is a collaboration
between all the NOAA units in Norman. It's designed to test out new
forecasting and warning techniques in a real-time environment, so we
actually have forecasters participating in this, but it's not actually
in their operations yet.
And so it's a way to introduce these new ideas to forecasters
and get their impressions on how useful it is, how valuable it is, and
how excited they are about this technology and what it might do to make
their job hopefully easier, but also produce better forecasts and warnings.
And so what we do is every year we get together with various
forecasters and we look at some of the new ideas that are out there and
decide on a few to bring in and then test. And then each spring, in
a six- to eight-week period, each week we bring in people from across
the United States. They can be from different forecast offices. They
could also be from various universities or from other national centers.
And we bring them in and we sit them down and we go through and start
using these new tools and have input into how they're doing and try to
then use them actually as they were intended.
So if they're a forecast tool, we'll go ahead and make
a forecast in the morning based upon what the tool is telling us. And
then we'll evaluate it the next day to see how well it did in actually
helping to produce a good forecast. And through this eight-week period
then you get a whole range of opinions, new ideas that come in, and it's
actually been so well received that we're actually getting international
visitors as well. We've had forecasters from Canada and from the United
Kingdom in the last couple years that have participated regularly.
BARRY REICHENBAUGH: Let's get big picture. Can you talk
to me a little bit about the ultimate goal of NSL's forecast and warning
DAVID STENSRUD: Our ultimate goal is to improve the forecasts
of severe weather for the public. And part of that involves improving
these tools that we have, like numerical models. Part of it involves
ways to actually take model data, because models provide a load of data,
a huge amount of data. And nowadays it's harder and harder for forecasters
to digest all of the data. So you have to find smart ways to pull out
the important pieces of information and then let the rest of it just
fall to the floor because you just can't handle all of it.
And we would love to be able to make day-two or day-three
predictions, so predictions for tomorrow and the day after as accurate
as we do predictions for today. And we always know that they're going
to be imperfect because that's just the way the atmosphere is. It's
a very hard system to predict with any degree of perfection. But I think
we can get better and better as time goes by.
BARRY REICHENBAUGH: What got you into this field? Did
you have an early interest in science in your life, or can you talk to
that a little bit?
DAVID STENSRUD: I've always loved science. As far back
as I can remember, I've always been intrigued by it. Initially I started
off as an undergraduate in college as a Physics major. But physics,
I found, when it got into quantum mechanics, it just wasn't my area of
I like the classical physics, what we would call Newtonian
physics or physics of fluids and in motion. I always had a love of the
atmosphere, and it just so happened that the university I went to, University
of Wisconsin, Madison, has a meteorology program. And so once I realized
that my love was really classical physics, I shifted into meteorology
and have loved it ever since.
BARRY REICHENBAUGH: And how did you end up in this particular
DAVID STENSRUD: Well, there's a bit of serendipity involved.
I was a Master of Science student at Penn State, which is where I got
my advanced degrees. And I met a nice young woman who I ended up marrying.
And basically we were in a place where staying there for a Ph.D. at that
point wasn't really a good choice for us. And so I started looking for
a job, and I ended up sending a resume here to NSSL and got a response
and came down, interviewed, and they were nice enough to offer me a job.
And since then it's just been a wonderful fit that I really do feel that
NSSL is a great place to work. The opportunities and the excitement
here for meteorology are quite outstanding. And I've been very blessed
by being here.
BARRY REICHENBAUGH: Now, you're located here on the university
campus, and I'm wondering if you could talk a little bit about what you
say to someone who's interested in a career path like yours.
DAVID STENSRUD: Well, I would say you need to have pretty
good skills in math and in science to be a meteorologist. A lot of people
don't realize the amount of math that's involved, in particular. And
so having backgrounds in mathematics and in physics, and nowadays even
chemistry is getting more important. And of course, there's always this
foundation that goes on of computer programming that most of us have
to deal with because it's there every day. To really make any kind of
prediction, you end up using computers quite a bit.
But if you have those skills -- I mean, meteorology, I
think, is a very exciting science. And one of the things I love about
it is the clear public benefits that you get. I mean, you're helping
people to make decisions and certainly forecasters are helping to save
lives by what they do every day. And that's quite a reward for any career.
BARRY REICHENBAUGH: All right. Well, thanks very much.
DAVID STENSRUD: You're very welcome.