Rob Murdoch: NIWA deputy chief executive on the agency's new super-computer worth $20 million

NIWA has unveiled its new state of the art super-computer - worth $20 million.

It's the fourth the agency's bought since 1999, with up to three times more power than its predecessor.

Rob Murdoch, NIWA's deputy chief executive, says the technology will help model potential weather disasters.

"Which is important when you have to think about things like Cyclone Gabrielle, for example - where it's very isolated and the hills can have a massive impact on the amount of rainfall."

LISTEN ABOVE

See omnystudio.com/listener for privacy information.

Episode Transcript

Speaker 1 (00:00):
Evening. Coming up to the next hour, we've got that
Berg official cash rate call. Tomorrow we're going to talk
to Brad Olsen about that. Shortly New Plymouth's mayor on
his challenge to labor for how labor can actually fix
this power crisis. We're in and Jamie Mackay on the
Canterbury Amp Show at seven past six. Now, KNEEWA has
got itself a fancy new toy. The state owned Wheather
Research Organization has brought itself a twenty million dollar supercomputer. Now,

(00:23):
once this thing is installed in two data centers in Auckland,
it will be the biggest research computer in the country.
Rob Murdoch is niea's deputy chief executive.

Speaker 2 (00:31):
Hey Rob, Hi, how are you?

Speaker 1 (00:32):
I'm very well, thank you. Does this mean we're going
to get more accurate weather forecasts?

Speaker 2 (00:36):
Well, hopefully we will, yes in the future, and hopefully
it'll be used for a whole range of the fruit applications.
Though we'll hopefully get better at modeling our future climate
as well as our oceans and a whole range of
other environmental Thanks.

Speaker 1 (00:52):
How much more accurate, like, for example, will it actually
have if we had had it in January last year,
would it have been able to forecast those rains that
we got at the start of the year.

Speaker 2 (01:03):
Well, I think the key in having a bigger and
better supercomputer means that we can actually model things at
a higher resolution both in time and space. So it
means we can model things down to smaller scales, which
is important when you have to think about things like
Cyclone Gabriel for example, where it is very you know,

(01:24):
it's very isolated and the hills can have a massive
impact on the amount of rainfall, and the better we
can model those hills, the best of the forecasts will be.
But it also means we can do things faster. So
instead of at the moment our weather models, for example,
our high resolution where the models can predict whether in
the location around one point five square kilometers, but we

(01:47):
can actually get down to three hundred meters. But to
do that we need more compute power. And what this
new system will nable us to do is to be
able to instead of having models that take six hours
to run, we'll be able to run them in three hours.
So that means instead of having four forecasts today, we
can have obviously double that number.

Speaker 1 (02:09):
So taking the example of cycling, Gabrielle. Would it have
avoided the situation where the Esk Valley was called unawares
they would have been told in advance if you had
this computer at.

Speaker 2 (02:18):
The time, Well, I think I suspect not. I think
we needed more in place than what we currently have.
And you know, that was a very severe event and
it was very difficult to predict, and certainly there was
a lot more rainfall, although we did tend to get
the rainfall levels somewhat right. I think the other issue

(02:39):
in all of this is that we're not only going
to predict the rainfall. We've actually got to predict what
happens once that water touches the ground, and in that
case it causes a flood. We've got to be able
to really have good models around flood forecasting as well,
and that's what will enable us to do.

Speaker 1 (02:55):
I mean, what everybody in Auckland would love to know
is basically what I was asking you at the very start,
is could we would we have known that those that
that water hitting the ground in January last year was
going to cause the trouble it was going to cause,
both by the sounds of things, not necessarily even with
this computer.

Speaker 2 (03:10):
No, it depends on the type of rainfall it is
and some some some types of rainfall are easier to
predict than others, and when it gets these flash flooding
events aren't quite so easy. But there's no doubt that
the higher compute power will enable us to better develop
the models that we've got and be able to work
at a higher resolution both in time and in space.

(03:33):
And yes, we should be better at forecasting these things.

Speaker 1 (03:36):
Does met Service have anything like this?

Speaker 2 (03:39):
No?

Speaker 1 (03:39):
So are you going to put them out of business? Then? No?

Speaker 2 (03:42):
Well, hopefully not. I mean there's a review at the moment,
and hopefully we'll see met Service and there's obviously we're
looking across both NEEWA and net met Service about how
we can make sure we've got a forecasting system that's
fit for purpose for the country.

Speaker 1 (03:56):
Where is that review at because it was supposed to
have wrapped up in May.

Speaker 2 (04:00):
Yeah, I think it's as far as I'm aware, there
has been a final report completed. So it's just we're
just waiting on.

Speaker 1 (04:08):
Time now to sitting with the minister.

Speaker 2 (04:10):
Sitting with the ministers and see what happens.

Speaker 1 (04:13):
Are you still competing with met Service for weather forecasting contract?

Speaker 2 (04:18):
I don't think we've ever competed. We work very closely
with the MET Service. We have different clients and certainly
our modeling capability enables us to deliver some products for
clients that are different to the MET Service, but we're
very complimentary.

Speaker 1 (04:32):
But either way, if I have to look at a
new with forecast or a MET Service forecast, you've got
the supercomputer. So you're going to be more accurate, now,
aren't you.

Speaker 2 (04:40):
Well sometimes in some considerations, some not, but I think
you know, costing US.

Speaker 1 (04:47):
Twenty million bucks, the answer is, yes, we're going to
be more accurate. It's money working, yes, yes.

Speaker 2 (04:51):
Yes, yes, it's definitely in advancement. And the other element
that we need to think about is that this new
supercomputer is going to enable us to apply AI to
a lot more science questions. AI is changing the way
we do science, the type of science we can do
with the sort of information that's been collected in the
vast data sets that we're getting today. It's enabling us

(05:14):
science to questions that we wouldn't have thought were possible before,
thanks like satellite in the jury and all those sorts
of things. So it won't be just weather and climate.
It's going to be oceans and a whole range of
other things, everything from greenhouse gas production to you know,
and it won't be just new. There's opportunity for other
science organizations to have access to this compute, things like

(05:34):
seismic modeling for earthquakes. The list just goes on and on.

Speaker 1 (05:37):
Yeah, very cool stuff. Hey, Rob, thanks very much for
your time. Rob Murdoch Nee is Deputy chief executive for
more from Hither DU plus Yellen Drive. Listen live to
news talks. It'd be from four pm weekdays, or follow
the podcast on iHeartRadio