August 27, 2025 • 39 mins
Extreme weather is taking toll on farm output worldwide, with record droughts and wildfires driving up prices in markets for cocoa, cattle and coffee. In this episode of ESG Currents, Bloomberg Intelligence senior climate analyst Andrew Stevenson is joined by Dave Farnham, chief data scientist at ClimateAi, to discuss how better data can help minimize climate costs.
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:09):
ESG is constantly evolving. Over the years, it has shifted
from socially responsible investing to impact to sustainable finance. While
the terminology continues to change, what hasn't changed on the
underlying science, market pressures and tangible physical and financial impacts
on the climate crisis, increasing regulatory scrutiny, and rising consumer expectations.
(00:32):
We aim to filter out the noise by speaking with
experts to identify what is really driving value. Welcome to
ESG Curdits. I am Andy Stevenson, Senior ESG analyst, your
host for today's episode. Today, we're talking about ways to
minimize climate risks with Dave Farnham, Senior director of Data
Science at Climate AI. Dave looks to provide insights to
(00:55):
help clients be more proactive in finding ways to address
climate relation costs across the number of sectors. We'll discuss
climate as a cost of doing business with Dave.
Speaker 2 (01:07):
Thanks for joining us, Thanks for having me, Andy, so great,
Thanks again for coming on. I just wanted to start.
Speaker 1 (01:13):
Maybe you could go through, you know, kind of Climate
AI's mission and ambitions and.
Speaker 2 (01:18):
The kind of the role that you played in that
would be really helpful.
Speaker 3 (01:21):
Yeah, absolutely so.
Speaker 4 (01:24):
At a At a high level, we're trying to climate
proof the economy, and you mentioned that climate has impacts
on a wide variety of sectors. We've we've started with
a lot of our work in the agricultural sector, which
is obviously very exposed to climate and weather fluctuations, but
we have work in other sectors as well. So really
(01:46):
we're trying to bring world class analytics to be accessible
to everyone, right in a scaled platform way. So I
lead an incredible team of data scientists, software engineers, machine
learning engineers, geospatial engineers, climate scientists, meteorologists, hydrologists, many of
which were former professors, and you know, we're deep in
(02:10):
the research area to sort of to push the bounds
not only of our understanding of the physical climate system,
but how we can actually manage climate risk to things
like engineered water, energy, food systems. And a lot of us,
myself included, transitioned from that sort of more purely academic
space into the private sector because we saw this this
(02:32):
sort of gap in terms of what was actually being
used at scale to manage these risks. A lot of
it comes down to, you know, being able to parse
all the data and information we have around climate variability,
climate change, but also how to use that data to
make decisions to mitigate your risk or to hop on
(02:56):
opportunities on the other side of the coin that those
do exist. Sometimes it's it's a tricky game because these projections,
long term climate projections as well as seasonal forecasting, which
is something we do, is inherently a probabilistic game.
Speaker 3 (03:12):
You don't know exactly what's going to happen. But what
you're trying to.
Speaker 4 (03:15):
Do is say, hey, if if this extreme of extreme
heat event, you know, during this time of year, which
which maybe would be an issue for heat exposure for
outdoor workers, if that was a one in four event,
you know, on an average year over the past thirty years,
if that's now fifty percent or one in two event
this upcoming season, can you can you make some modifications
(03:36):
to your plans to be able to mitigate that.
Speaker 2 (03:38):
But you need to be you need to be able.
Speaker 4 (03:40):
To absorb probabilistic information and utilize it. So I guess
the point I'm making is a lot of what we
spend our time doing is thinking about how to translate,
how to translate pretty complex probabilistic information into workflows for
you know, whether it be small time producers or large
multinational corporations who all have different workflows and how they
(04:04):
interact with data and make decisions. And I'll finish up
here because i know I'm probably opining too long here,
but we really we're we're in the adaptation I should
have started with this. We're in the adaptation space. So,
you know, a lot of climate tech companies loosely, you
can group it into mitigation, are you you know, carbon
markets things like that, or adaptation where it's about, hey,
(04:24):
how do we actually manage the variability that goes on
today and will continue and in many cases be exacerbated
into the future. And our tools provide people basically decision
making support for strategic decisions. So these are long term
climate projections, decades into the future. Think people thinking about
where to move their operations potentially to to to limit
(04:47):
their exposure to climate risk or operational so near term,
next couple of weeks, next next few months, and I'm
happy to get into some of the examples of how
people use these tools.
Speaker 1 (04:56):
Yeah, I mean, as you said, it's I mean, it's
becoming more and more clear that climate, you know, and
extreme weather is a cost of doing business, right, we
have to kind of adapt to the fact that our
home insurance premiums are going up. Certain products, certainly certainly
commodities in particular, are influenced all the time, and I've
always been influenced by by whether but the climate is
(05:18):
making those is kind of narrowing in some ways. It's
increasing your probabilistic example there, which is to say that
the chances, as they get closer and closer to fifty
to fifty or higher, become real bottom line issues where
you're you're almost remiss not to have caught them right.
So it's one thing to say it's a one on
(05:39):
one hundred year storm. Who could ever have seen that coming? Well,
you know, obviously you have one chance, it's not impossible.
But if someone said there's a fifty percent chance of
El Nino having this impact on your crop output, and
you decide to completely ignore that kind of information, then
that's that's where you kind of can get fired right.
Speaker 5 (06:00):
So that's that's effectively, you know, almost everything circles back
to how can I get fired right? And so that's
unfortunately one of the more prominent ways to do it.
I want to talk about the data in a second
but let's let's just sort of stick with agriculture, because
you mentioned that right at the beginning. You see things
(06:21):
like and very very clear patterns in the last several
years with what's been happening in Brazil, multi billion dollar
droughts year after year after year.
Speaker 1 (06:30):
Really since twenty twenty two, twenty two, twenty three, twenty
four and twenty five have all been incredibly poor years
from a drought perspective. I'm just really trying to isolate
two crops or two things, right. One of them is coffee,
which we've seen, you know, prices of coffee go up,
and the other is cattle. You know, so when you
(06:51):
think about Brazil, what do they really bring to the
global markets kind of at scale? Those are kind of
two of two examples, and you don't have to stick
with the Brazil example, but I'm just saying that it's
very if you look at the data, we had kind
of sideways to nothing burger drought related events in Brazil
for many, many years, and then last several years have
(07:13):
just been a flurry, And of course that's going to
have an impact. When you're the biggest producer of coffee,
you want to if not the biggest producer of cattle
in the world right up there. Right, So again, you
could pick any example you wanted agriculture, but I just
wanted to kind of flag that as being from our research,
very glaring, you know, cause.
Speaker 4 (07:33):
It effect Yeah, absolutely, Yeah, Brazil's a biggie. I mean
for so many of the big commodity crops, right, you
mentioned coffee, cattle, corn out there.
Speaker 3 (07:43):
So yeah, I mean the what we see there.
Speaker 4 (07:47):
And then I mean the other one that we have
a lot of folks interested in is coco these days, right,
And I'll you know, I don't have that off the
top of my head, but it's something like seventy five
percent of cocoa is grown in West Africa, and there's
been some some large droughts there as well the last
few years, and you saw big volatility and in the
prices there.
Speaker 3 (08:08):
I think it's come down to some degree.
Speaker 4 (08:09):
Over the past couple of months, a lot of the
way that that our customers use our tools. So there
there are folks that that are interested in how do
I actually manage this on the farm. How do I
manage drought on the farm. I mean, to some degree,
drought is a difficult one. If you're not set up
to irrigate if it's a prolonged drought. You know, it's
sometimes for some of our customers, it's just a matter
(08:31):
of knowing is actually valuable, right, even if you can't
physically do.
Speaker 3 (08:36):
Something about it.
Speaker 4 (08:37):
And then a lot of our customers are actually keeping
an eye on the global market, so they're thinking, okay, yes,
Brazil is a major player in terms of of what
let's let's Useles use coffee production globally, and maybe they
source a lot of their coffee as your coffee company
source a lout of your coffee from Brazil, But you
also want to understand how's coffee looking at Indonesia and
(09:00):
the Opion and and many other places where there's where
there's coffee. And so that's again where you sort of
get into that. It's a it's a complex game, of course,
and many folks, you know, commodities traders are well equipped
to in my experience, well equipped to ingest that probabilistic information.
And you know, in their case it's just can I
have an edge over the next person. But the reason
(09:23):
I bring up this sort of global this global view
is for a lot of our customers, what it What
it ends up coming down to is, yes, you want
to understand your primary regions that you're either growing something
or sourcing from, how are they doing? But how are
all the other regions of the globe doing? And so
then you get into these looking at projections for do
we have scenarios where we have concurrent droughts in all
(09:44):
the main you know, corn growing regions of the globe, right,
or we growing because then you really have this issue
where there's not flex in the system where you can
actually go go to the spot market and procure thing.
Speaker 1 (09:56):
So that's really from the demand side, right, I Mean
I had a question this is apply side.
Speaker 2 (10:00):
But you're saying it's good to know about the drought.
Speaker 1 (10:02):
Does that mean that people will like cut their the
amount of you know, fertilizer they're using because they just
know it's a it's a lost cause? Like, is that
the the kind of result that they, you know, are
looking for by understanding that risk?
Speaker 4 (10:18):
Yeah, I would say it depends. It's very sort of
location and crop specific. Just to clarify real quick, I
probably maybe I misspoke earlier. I was thinking more on
the supply side. How much supply is flowing into the
global you know, global coffee coffers if you will.
Speaker 1 (10:31):
But I mean, but the people that care about the
supply side is the demand side.
Speaker 4 (10:36):
Absolutely absolutely, we need to meet x amount of demands.
So totally totally true. So a couple of examples of
how folks actually use it, and actually i'll go here
to actually this is this is different crop, but one
people care about in a lot of cases hops from beer.
So one of the decisions that that people have to
make in that life cycle is when to harvest the hops.
(10:58):
And in a lot of places where they're grown, and
you will be harvesting as it's getting colder, and if
if a frost actually hits your hot crop, it's really
bad news. But every day you wait to harvest, you
get more yield off of the plant. And so then
they're using our analytics to say, what is the probability
of there being a being a significant frost over the
(11:21):
coming weeks, Do we need to schedule harvest or not?
Can we get away with being a little bit more
you know, aggressive, And of course you need a strong
signal there. The other one, a big one is for
planting timing as well. So there's a lot of crops
that need some amount of moisture, especially if you're not
in an area where you're irrigating for the seed to germinate, right,
(11:43):
and so you need to be able to have confidence
that there's gonna be enough enough precipitation occurring in the
coming weeks and so there it's about sort of timing
when you're going to plant. So a lot of it
is about, as you alluded to, moving timing, like what
what degrees of freedom do you have as as a
producer to move your timing. There's only a there's only
(12:03):
a few sort of events per season that you can time, right.
It's like planting, as you alluded to, any inputs like fertilizer,
irrigation and then harvest. But a lot of a lot
of what we see folks using our platform for is
to yeah, play different scenario games with that timing, so
we know what's happened up to the season until now
(12:24):
we know what our forecast is doing. What if I harvest,
then what's my expected yield? What if I harvest you
know at a different time.
Speaker 1 (12:30):
That type of thing and timing is critical because you
have to assemble all the players right to get to
all the labor and everything else.
Speaker 2 (12:37):
That's required to get that done.
Speaker 1 (12:39):
Some people have the rent tractors and you know, like
that kind of stuff, right, so that you need to
get ahead of the queue, not in the back of
the queue, right, if you can target that better. I
also understand, like maybe you don't do wine, but I
understand wine is has this real problem with with wet
weather because the grapes if they're if they're picked while
they're absorbing the water, become like you know, there's less
(13:00):
alcohol in the grapes because they've absorbed the water. Right,
So this is a pretty critical piece of the puzzle
for the for the wine grower.
Speaker 3 (13:08):
Absolutely, we do.
Speaker 4 (13:09):
We do have customers that are vineyard owners and making wines,
and we actually have this sometimes at events will we'll
do this sort of wine tasting and say, you know,
here's I am not a wine connoisseur, so I could
not tell you, you know what wine is. I mean,
I know which ones I like sort of you know,
but okay, this is you know what traditionally you've had
(13:31):
in Napa.
Speaker 3 (13:31):
Now maybe the same sort of flavor is going to
come from up further north in Washington.
Speaker 4 (13:36):
So we actually sort of use some of our climate
projections to say, like, where is an area that was similar,
you know, on a hillside in northern California twenty years
from now, where's the climate actually going to be really
similar where you could expect similar grapes. The other big
one I'll say for wine producers that we've worked with
is smoke taint. So that gets into the wildfire risk.
Speaker 1 (13:56):
And you know, I'm going to ask about that because
it's clearly been a I mean, that just destroys the
crop as far as I can tell, right, it's very
hard to recover from, yeah, smoke taking exactly.
Speaker 4 (14:05):
And in those cases, again, that's one where I don't
have a ton of knowledge on whether there's any any
mitigation they can do once it's happening. It's more about
that long term strategic planning of like you know, if
they're investing in in new lands for vineyards and things,
trying to understand how do we limit our exposure to
(14:27):
things like I mean, in that case, wildfire risks and
you know, and.
Speaker 2 (14:30):
Where the tactical burns are going to happen.
Speaker 1 (14:32):
I mean, that's obviously a big point, right, because that's
you're in some ways the government is really in charge
of that because we do not want consumers.
Speaker 2 (14:41):
Tactical burning of land.
Speaker 1 (14:44):
But the idea is that, you know, if you understand
where how the government thinks about it, and they're just
trying to make sure that no fires are started, or
at least we can limit the reach of the fires
once they start. That that's a pretty critical piece of
the puzzle. And obviously with the wildfires up in Canada,
Canada has a lot I mean, actually do any business
(15:04):
in Canada, but they've got a ton of Agen is
a big wine region.
Speaker 3 (15:08):
So yeah, sorry interrupt.
Speaker 4 (15:10):
We have some smoke here from I believe Canadian wildfires
right now in Indianapolis here today. Woke up and I
was looking at the AQI and I was like, oh,
what's that about? And yeah, so yeah, there's.
Speaker 1 (15:20):
I mean, just to kind of depart a little bit there,
there's all sorts of lost labor and all sorts of
you know, a lot of things happen within the context.
Speaker 2 (15:28):
Of smoke days, you know, what I referred to as
smoke days, and they really.
Speaker 1 (15:33):
Have our having a bigger and bigger impact on you know,
like our ability to work, you know, frankly, like and
health and obviously if you have small children, or older people.
You need to be very careful about that stuff because
more and more it's not burning wood as we saw
in LA it was couches and things like that, which
(15:54):
is horrible for your you know, so as it spreads,
it goes from being a a scent you would be
comfortable within your car to something maybe not so much.
Speaker 4 (16:03):
Absolutely, And just to add to that briefly, I mean,
one of the things that I don't think has necessarily
garnered enough attention over the past you know, decade or
two is just air quality in general, and how many
you know, early deaths. I think I saw a paper
the other day. It was estimated something like four to
six million. You know, this is globally annually early deaths
(16:25):
attributable to air quality, right, So it's it's the scale
of the issue is huge.
Speaker 1 (16:32):
Yeah, And in the United States, which has been basically
on a fifty year train down in terms of air
quality improvement, you know, like we've improved year after year
after year, mainly through the Clean Air Act, and from
nineteen seventy one or whatever it was started, we're starting
to see that tick back up again. Well, so that
is like the first not just deceleration, but reverse reversing
(16:55):
in that trend because of the wildfire smoke, because you know,
it's really really harmful, like from a you know, relatively speaking,
it's much worse than allowing some of those zone kind
of plutants to continue. You know, So this is a
becomes a bigger deal. Absolutely.
Speaker 4 (17:13):
I remember, not so fondly, when I was at Stanford
for you know, a couple It was there for several years,
but for a couple of weeks during one of those
wildfire seasons where it was just funneled right over US
down in you know, Bay area, and it was I
was just sort of living in my n ninety five.
I guess it was early practice for the pandemic.
Speaker 3 (17:32):
But yeah, yeah, no.
Speaker 1 (17:35):
It's a it's you know again, these are these are
costs of doing business. Now, this is how this is
unfortunately part of the game. I would like to ask
a little bit about data, if I may, because obviously
we are seeing I think the USDA is sort of
holding its ground in terms of a lot of the
data that sometimes at one point they were trying to strip.
Speaker 2 (17:54):
Out quite a bit of data out and related to climate.
Speaker 1 (17:56):
But I think they realized that because it's just so
fundamental to crop production. They've kind of put it back,
is there, what are there just overall? What are your challenges?
They could be in the United States or elsewhere in
the world, or you could actually if you might, I
might be helpful for you to explain how kind of
(18:18):
good or bad are data is relative to the rest
of them.
Speaker 3 (18:21):
Yeah, great question.
Speaker 4 (18:24):
So first, just in terms of recent trends, we have
as of as of now, knock on wood, not run
into any issues of data sets that we were depending
on no longer being there and refreshed. There have certainly
been a lot of things taken down in a lot
of cases, you know, if it's if it's more of
a static data set, you know, we have that anyways,
(18:44):
and I know there's also lots of efforts for folks
to create sort of catalogs and other places of those
data sets, which I appreciate. Of course, we have contingency
plans for if some of the data sets that we
depend on operationally go away, but that the case so
far that that hasn't happened. Our biggest I mean, at
(19:06):
a very for very specific example, the biggest data challenge
is getting better and better precipitation data sets for a
wide variety of reasons. One is, you know, we've gotten
a lot of interest in the index insurance or parametric
insurance space, right, so benchmarking. I'm sure your listeners have
(19:28):
some familiarity with that, but benchmarking, you know, an insurance
payout based on whether you got over a certain threshold
of rain, and you know, it cuts down on the
transaction costs of having an adjuster going out and looking
at in the agricultural space, for yeah, it's basically.
Speaker 1 (19:43):
A clarity thing, you know, like if X happens, you
get paid, right, And that's becoming more of an aon
and A and PLC and others have kind of jumped
in the middle here and said, look, I'll be the
middleman here. It's kind of like, what's he wit. I'm
trying to think of the tax people, you know, right right,
it's similar to that where they're hr block.
Speaker 2 (20:04):
That's who I was thinking. It's kind of like, you know,
under that under.
Speaker 1 (20:08):
That umbrella of you know, we did we we believe
what the data and as a result, we're going to
pay you in a hurry. Getting paid to hurry in
some cases is the difference between you know, survival and
not survival, right, because your other people are already you know,
you owe money to people generally speaking in business, right.
Speaker 4 (20:27):
So yeah, yeah, And another example is I believe, I
don't know that this is still the case, but when
I was living in New York, a lot of the
catastrophe bonds in New York where we're indexed by like
the the water level battery city gauge, for example. But
so the reason I bring that up is obviously data
quality becomes extremely important there, right, because that's what's determining
(20:49):
whether something gets sort of paid out or not. And
precipitation is one that's just really really difficult where our
product is global. We we have customers on every continent
except for Antarctica. But the you know, there are dense
rain gage networks in the US, in Europe, in a
(21:11):
lot of Australia some other places, but then there's other
places where we just don't have a long record. We
need to depend on more remote sense, remotely sense precipitation products.
But with any of that, you know, meet satellites and
we ingest tons of information there. It's all great, but
it's usually only as great as you have ground truth
(21:32):
to be able to calibrate too, right, and so in
locations where you just don't really have that, that remains
a problem for us calibrating our forecasting models for us,
you know, working in we were just talking about parametric
insurance space. So that's a very specific one that I
think you know, comes up all the time with colleagues
just in terms of getting better and better data sets,
(21:54):
and there are some that are really good in some regions,
so you know, you can take sort of a mosaic approach.
I think the other data challenge is really the delivery
of data to customers in a way that allows them
to take actions fairly simply. So I was earlier talking
(22:15):
about or mentioning decision making under uncertainty and basically everything
we do you're dealing with an uncertain world on certain
forecasts and on certain projections. That does not mean bury
your head in the sand and throw up your hands, right,
It means, okay, how do I actually extract the information
from this? And actually, as a real side note, just
this morning, I was seeing some former colleagues of mine
(22:37):
published yesterday a paper on how climate models that we've
been using for years may be biased with respect to
their tails. So actually underestimating some extreme heat into the future.
So just to say that uncertainty also is evolving as
the science evolves, that's not great news, right because it
means even if the mean warming signal we think the
(23:00):
are on the right track, some of the extremes might
actually be be larger than we saw.
Speaker 1 (23:05):
Yeah, that's and just as an aside on that front,
I mean, that's if you read those incredibly boring six
hundred page reports from the i p PC, you realize
there's a lot of unknowns, you know, like.
Speaker 2 (23:15):
They they they carve out things. You're just like, what,
how do we not have a handle on this?
Speaker 6 (23:20):
You know, it's it's you know, we are not that
smart and maybe comes this up, so it's, uh, don't
it's not easy to pat ourselves in the back when
it comes to understanding that.
Speaker 2 (23:31):
I mean, it's a complicated problem, don't.
Speaker 1 (23:33):
Get me wrong, but it's a you just can't believe
there are these glaring gaps in the in the data
sets as you're describing right now, with you know, precipitation
for example.
Speaker 4 (23:43):
Yeah, yeah, and and so this is this is bending
your question a little bit, but it the biggest challenge
that I see for us and for many others, is
how to deliver things in a concise way that actually
fit into the workflow of our customers.
Speaker 3 (23:58):
Right.
Speaker 4 (23:58):
That's not making them read some giant report, right, and
not in our case also not. We're trying to do
things more at scale to be able to make the
accessible to a wide group of users. So rather than
you know, if we were a consulting company, it would
be like I'd sit down with you Andy, say exactly
what you know is your operation, go through exactly all
(24:19):
your vulnerabilities, put together a playbook with you in terms
of Okay, if you see this type of event coming,
here's how you could mitigate it. We do that some
with with our enterprise customers, but we want something that
is is able to be translated into their workflow seamlessly.
So touching on the AI portion, we've we've done been
(24:39):
doing a lot of development with agentic systems and in
a slightly different way than I think usually gets talked
about these days. In a lot of cases, you know,
agents are all the all the raven in a lot
of a lot of the business world.
Speaker 3 (24:54):
Often for automating tasks.
Speaker 4 (24:56):
Say hey, I want to just have these four agents
doing these things that I think think are repeatable. How
we're using it is more as an interface for customers
so to be able to say, Okay, our agents have
access to our tools. I'm calling them tools, which is
actually just our analysis that our great team has built
over the years, right, that are repeatable. But the communication
(25:20):
and the way that the customer sort of would ask
questions of that analysis can be quite labor intensive. But
if you can leverage some of the large language models
and some of these agentic systems to give the customer
direct interface to our tools, so it's not them sending
an email to our customer service and then customer service
talking to our analysts and then our analysts doing you know,
(25:43):
putting together report and going back trying to really streamline that.
And what we're what we're starting to find is that
it opens the door a little bit more for us
to be able to provide a valuable information to folks
because it lowers the friction there. It's not something necessarily
extra that they have to do and spend a lot
of time tweaking things, but they can through our platform
(26:05):
go sort of directly to having access to some of
that that world class analysis. This is this is early
days here, right, I mean, sure, it's not trivial to
spin these things up. And we're in a space where
like it's catastrophic for us if if the agent is
providing hallucinations to a customer, right, we have to avoid
(26:26):
that at all costs. So that's been an interesting one.
But but to me, the biggest challenge is communication. Basically,
I guess is the summary or I mean what, especially.
Speaker 1 (26:37):
With commodities, there are ways to actually translate what you're doing.
Your probabilistic assumptions can be translated into probabilistic bets, ye
know what I'm saying, Like, because they can do some
hedging that you know, if you think there's if they
think there's an edge relative to the their peer group,
you can set up a you know, slightly not two
taxing structure around it, right where you sell little bit
(27:00):
of upside against it, just to ensure that you're making
the right you know, like you're covering your yourself in
case it doesn't go as planned.
Speaker 2 (27:07):
But if you're trying to.
Speaker 1 (27:09):
Optimize, and especially when you're like you have to harness
a lot of energy to get some of this stuff
done early, right, I mean, it's not again it's not
there's no button. It's not like a trader where he
just pushes a button and says buy this. It's a
it's a process to actually get your uh you know,
coco out of the ground or whatever it is, right,
(27:31):
you know, you know like that that is like the
decision to go is a lot less like a trading system,
if you know what I'm saying. So it has a
uh because it you know, like that's how bots are
used pretty universally at the moment. But that for what
you're doing, it's almost like, well, okay, you know what's
your confidence level, how can we translate that into some
(27:53):
kind of you know structure you know that are around
trading structure, arounding.
Speaker 2 (28:00):
This kind of uh, you know, this kind of energy,
shall we say around it?
Speaker 4 (28:05):
Absolutely and two two thoughts on that or two sort
of examples are you know, we have customers today who
who use our system just in that way. So one
of the things we do I didn't say this earlier,
but is predicting yield for these for these major crops, right,
and so we'll have customers that either they're procuring all
of their corn or what or wheat or oats or whatever,
(28:28):
on the market, or they have contracts with you know, uh,
with producers in certain regions for fifty percent of what
they need to procure and then they are going to
go to the open market for the other So they
are definitely using our system to say, hey, it's looking
like a bad year in our in the area we
(28:49):
primarily procure from, or a good year. That means we're
going to need to you know, get x amount on
the market. And oh, by the way, and Brazil's looking
really like it's going to be bad potentially in a
couple of weeks. Let's lock in some of those futures
right now. So that's being used. The next step, which
is still sort of early days for us, but where
I'd really like us to go is so we can
(29:13):
give that information in whatever sort of whatever format the
person on the other the customer needs. But in order
to actually give any information on like buy sell, et cetera,
I need to know their cost function, right, I need
to know what the outcome that that decision tree is.
And so we've we've done some exploration with actually allowing
(29:37):
customers to give us that information. Right, it does get
pretty complicated pretty quickly. And then that's where you get
into some of the challenges becoming idiosyncrasies of like how
people have structured their way of making decisions those buy sell,
et cetera. And so the challenge for us is like,
can we do this in a flexible, scaled way that
accommodates that service for a lot of customers.
Speaker 3 (29:59):
Or does it really need to be tailored customer by customer.
Speaker 1 (30:03):
I should Yeah, that makes sense because yeah, as you say,
there's a lot a lot of moving parts. So it's
a it really depends on that person's particular circumstance. Are
they borrowing, you know, are they is the land being
rented to them?
Speaker 2 (30:15):
For example?
Speaker 1 (30:16):
Is a fundamental, fundamental question because that those those pavements
can't be not met, right. Whereas if it's someone that's
been in the been in the business for a very
long time, that's not necessarily.
Speaker 2 (30:28):
Front and center.
Speaker 1 (30:29):
They're literally trying to optimize year to year the best
they can and they understand that other people don't have
that flexibility.
Speaker 2 (30:35):
Right.
Speaker 1 (30:35):
So if you're if you have flexibility built into your
business model by virtue of your great grandfather or whatever
it is, that's a that's not a bad Uh, there's
no real reason why you shouldn't take advantage of it
because others are more constrained, right, they don't have that
kind of luxury.
Speaker 3 (30:52):
Absolutely.
Speaker 5 (30:53):
Yeah.
Speaker 4 (30:53):
I mean we could tell a customer, hey, there's going
to be a great crop of canola or something in
Australia and they're saying, that's great. All my process and
facilities are in Canada, right, yeah?
Speaker 2 (31:02):
Right?
Speaker 1 (31:03):
Can I can I tap a little bit your brain
on supply chains with respect to some of these uh,
you know, because just we can stick with our culture
or really whatever you're kind of comfortable with you want
to go.
Speaker 2 (31:13):
We can switch directions if you want.
Speaker 1 (31:16):
Just how important your kind of you know, the dependence
there is on supply chains, right, Obviously it's a big deal.
It's becoming more of a big deal. Tariffs are all
sorts of a bit. I mean, they're you can you
can talk about tariffs or not, but I mean the
fact is that there is uncertainty out there, and you know,
your supply chains are pretty important these days in terms
(31:39):
of flexibility. Let's just say, based on how the tariff
dialogues go from week to week, month to month, year
to year.
Speaker 4 (31:48):
Yeah, no, great question, I will I don't have anything
to say on tariff that hasn't been said a million times,
I'm sure by smarter people than I. So, but I
mean that does come up in terms of uh, you know,
maybe in this current moment, overshadowing some of the climate
related risks. I mean, not totally, but yeah, yeah, but yeah,
(32:08):
it's it's another layer of uncertainty there.
Speaker 2 (32:10):
But yeah.
Speaker 4 (32:11):
So for supply chains, we have customers I think on
sort of every branch of the supply chain. So I'm
thinking sourcing, manufacturing, transportation, and then on the demand sort
of forecasting side. So I'll give it like a really
brief example for each one. Sourcing regions. We've been talking
about that a lot. You know, what the information we're
providing is, you know, is the cocoa being grown and
(32:33):
ghana under drought stress for example manufacturing there, it's often
about disruptions to manufacturing plants. So is there a signal
of upcoming flooding that is going to limit the ability
for manufacturing plants to operate, workers to come to and
fro the plants, And we have I think.
Speaker 1 (32:53):
That happened recently in Switzerland right with the chocolate factories
or maybe everyone.
Speaker 4 (32:58):
I am not familiar with that specifically, but it's I mean,
it's happening a lot of the time in a lot
of different places. Actually, because this is a fully public engagement.
We work with Hatachi, the Japanese company for resilience on
their supply chain and focusing a lot on on manufacturing
(33:20):
facilities and and they're primarily flood hazard also tropical cyclones
disrupting them. So then then there's transportation. A big an
example of transportation related disruptions that can occur that we
work in is river levels. So Mississippi, for example, a
(33:42):
ton of things get get barged up and down the river.
There are times of year there have been a couple
over the past several years, especially in the fall, when
the river levels get low enough such that you can't
send as fully loaded barges or can't send barges through
and stretches the river. This is a huge issue because
(34:03):
then if you really need to get something down the
river or over somewhere else, you you know, you have
to start trucking or training it, which again at the
the ratio of the cost there can fluctuate a little bit,
but it's a lot more expensive.
Speaker 1 (34:17):
To get wet could get wet, which you know, we
have some very hard to sell as you have to
dry them. Again, there's all sorts of costs associated with
that too, right, right, So so.
Speaker 4 (34:26):
There's there's that transportation link, and then there's the demand forecasting,
which is more of a recent area for US, but
there what we've done so far is more in the
retail space, so things like demand for groceries when there's
a forecast for uh, you know, big white out blizzard
in the Northeast for example. And so what we do
(34:49):
there is we basically quantify what level of demand should
you expect in the days before, then the demand drops
off during everything shut down, and then what kind of
rebound to demand there, I mean groceries also things like
restaurants and lodging and things there. And these are these
are areas where folks know have known there's a relationship
(35:09):
with weather obviously, particularly those extreme events, but being able
to quantify it and being able to use a model
that has seen hundreds of events around the the US
over the past, you know, over a decade to have
sort of more assurance of like here's how much bottled
water we think you're going to need when the hurricane
is bearing down and things like that. That's so that's
(35:31):
that's on that demand side. So I would say, you know,
in some sense, I don't know whether I mentioned supply
chains at the beginning in terms of overall our company,
but that it's it's really about trying to make supply
chains as resilient as possible in the face of weather
and climate.
Speaker 1 (35:48):
Yeah, I mean, they're all inputs, so it's a it's
all part of your part of your math, that's for sure.
Speaker 2 (35:54):
Well, we're kind of widening down here a little bit.
Speaker 1 (35:56):
I did want to ask if you had any kind
of research on the horizon you were excited about in
terms of next steps for some of your your work.
It's all sounds, you know, pretty great guns at the
moments of anything you're working on your like to share
with us.
Speaker 3 (36:09):
Yeah, I was actually.
Speaker 4 (36:14):
I think I mean the big one, as I was saying,
is our sort of foray into agentic systems and seeing
whether it can actually provide that layer of basically communication
and being able to integrate into people's workflows, right, rather
than just you know, automating a bunch of things. Actually,
in some sense it's it's being a translator. The other
(36:37):
I mean I mentioned that paper that I'm curious to
read the read it closely in terms of our projections
on extreme out right because extreme heats driving. I mean
we were talking we mentioned worker safety earlier, we talked
about wildfires. That's a big part of some of the
trends in wildfires can be you know, heat and the
evaporation and the drying out that comes from that one
(37:00):
other thing which is outside of of the research area,
but I mentioned earlier, there are sometimes opportunities from it's
not just about mitigating risk. But we've worked with building
materials companies, for example, where their demand in in the
you know, Southeast US for example, a lot of it
(37:20):
strongly depends on on hurricane activity.
Speaker 3 (37:23):
Unfortunately, you know, people's roofs getting.
Speaker 4 (37:24):
Blown off, and so we actually then provide those forecasts
in terms of how much product do you think you're
going to sell? And we had a success. I should
say that in major quotes because it was you know,
a tragedy for so many but Hurricane Ian we were
showing a really hot signal that year when it when
it went through Florida, and the customer we worked with
(37:45):
was able to supply shingles into that market. Much quicker
than they would have of else wise, because they actually
started producing shingles that were meant for Florida before there
was that, there was that strong signal that was was
known by everyone. And so I know, I I sort
of want rogue there at the end rather than giving
you research.
Speaker 2 (38:05):
But that's all right. It's what you're excited about. That's
that's what we care about.
Speaker 4 (38:10):
It.
Speaker 2 (38:10):
That's great, that's great. I mean, yeah, there's obviously there's
lots to do.
Speaker 1 (38:14):
I mean, the fact that this is a cost of
doing business is becoming pretty clear to almost everybody, right,
I mean, anyone else skinning the game knows it's true.
Speaker 2 (38:22):
That's for sure. Anyone who has a home insurance policy is,
you know, painfully aware of this kind of stuff.
Speaker 1 (38:27):
So with that, i'd like to say thank you so
much for your time, David was It was great great
having you on. I learned a lot And if you
want to find more more information on this particular topic,
you can please go to the Bloomberg function bi Climate
dam dot go, which really looks and tracks the damage
(38:48):
is related to climate that we have on terminal And
if you have an ESG quandary of burning question, you
would like to ask b i's expert analysts. Please send
me an email at A Stevens at ESG, Currents at
Bloomberg dot net. Thank you, Dave, Thanks for your time.
ESG is constantly evolving. Over the years, it has shifted
from socially responsible investing to impact to sustainable finance. While
the terminology continues to change, what hasn't changed on the
underlying science, market pressures and tangible physical and financial impacts
on the climate crisis, increasing regulatory scrutiny, and rising consumer expectations.
(00:32):
We aim to filter out the noise by speaking with
experts to identify what is really driving value. Welcome to
ESG Curdits. I am Andy Stevenson, Senior ESG analyst, your
host for today's episode. Today, we're talking about ways to
minimize climate risks with Dave Farnham, Senior director of Data
Science at Climate AI. Dave looks to provide insights to
(00:55):
help clients be more proactive in finding ways to address
climate relation costs across the number of sectors. We'll discuss
climate as a cost of doing business with Dave.
Speaker 2 (01:07):
Thanks for joining us, Thanks for having me, Andy, so great,
Thanks again for coming on. I just wanted to start.
Speaker 1 (01:13):
Maybe you could go through, you know, kind of Climate
AI's mission and ambitions and.
Speaker 2 (01:18):
The kind of the role that you played in that
would be really helpful.
Speaker 3 (01:21):
Yeah, absolutely so.
Speaker 4 (01:24):
At a At a high level, we're trying to climate
proof the economy, and you mentioned that climate has impacts
on a wide variety of sectors. We've we've started with
a lot of our work in the agricultural sector, which
is obviously very exposed to climate and weather fluctuations, but
we have work in other sectors as well. So really
(01:46):
we're trying to bring world class analytics to be accessible
to everyone, right in a scaled platform way. So I
lead an incredible team of data scientists, software engineers, machine
learning engineers, geospatial engineers, climate scientists, meteorologists, hydrologists, many of
which were former professors, and you know, we're deep in
(02:10):
the research area to sort of to push the bounds
not only of our understanding of the physical climate system,
but how we can actually manage climate risk to things
like engineered water, energy, food systems. And a lot of us,
myself included, transitioned from that sort of more purely academic
space into the private sector because we saw this this
(02:32):
sort of gap in terms of what was actually being
used at scale to manage these risks. A lot of
it comes down to, you know, being able to parse
all the data and information we have around climate variability,
climate change, but also how to use that data to
make decisions to mitigate your risk or to hop on
(02:56):
opportunities on the other side of the coin that those
do exist. Sometimes it's it's a tricky game because these projections,
long term climate projections as well as seasonal forecasting, which
is something we do, is inherently a probabilistic game.
Speaker 3 (03:12):
You don't know exactly what's going to happen. But what
you're trying to.
Speaker 4 (03:15):
Do is say, hey, if if this extreme of extreme
heat event, you know, during this time of year, which
which maybe would be an issue for heat exposure for
outdoor workers, if that was a one in four event,
you know, on an average year over the past thirty years,
if that's now fifty percent or one in two event
this upcoming season, can you can you make some modifications
(03:36):
to your plans to be able to mitigate that.
Speaker 2 (03:38):
But you need to be you need to be able.
Speaker 4 (03:40):
To absorb probabilistic information and utilize it. So I guess
the point I'm making is a lot of what we
spend our time doing is thinking about how to translate,
how to translate pretty complex probabilistic information into workflows for
you know, whether it be small time producers or large
multinational corporations who all have different workflows and how they
(04:04):
interact with data and make decisions. And I'll finish up
here because i know I'm probably opining too long here,
but we really we're we're in the adaptation I should
have started with this. We're in the adaptation space. So,
you know, a lot of climate tech companies loosely, you
can group it into mitigation, are you you know, carbon
markets things like that, or adaptation where it's about, hey,
(04:24):
how do we actually manage the variability that goes on
today and will continue and in many cases be exacerbated
into the future. And our tools provide people basically decision
making support for strategic decisions. So these are long term
climate projections, decades into the future. Think people thinking about
where to move their operations potentially to to to limit
(04:47):
their exposure to climate risk or operational so near term,
next couple of weeks, next next few months, and I'm
happy to get into some of the examples of how
people use these tools.
Speaker 1 (04:56):
Yeah, I mean, as you said, it's I mean, it's
becoming more and more clear that climate, you know, and
extreme weather is a cost of doing business, right, we
have to kind of adapt to the fact that our
home insurance premiums are going up. Certain products, certainly certainly
commodities in particular, are influenced all the time, and I've
always been influenced by by whether but the climate is
(05:18):
making those is kind of narrowing in some ways. It's
increasing your probabilistic example there, which is to say that
the chances, as they get closer and closer to fifty
to fifty or higher, become real bottom line issues where
you're you're almost remiss not to have caught them right.
So it's one thing to say it's a one on
(05:39):
one hundred year storm. Who could ever have seen that coming? Well,
you know, obviously you have one chance, it's not impossible.
But if someone said there's a fifty percent chance of
El Nino having this impact on your crop output, and
you decide to completely ignore that kind of information, then
that's that's where you kind of can get fired right.
Speaker 5 (06:00):
So that's that's effectively, you know, almost everything circles back
to how can I get fired right? And so that's
unfortunately one of the more prominent ways to do it.
I want to talk about the data in a second
but let's let's just sort of stick with agriculture, because
you mentioned that right at the beginning. You see things
(06:21):
like and very very clear patterns in the last several
years with what's been happening in Brazil, multi billion dollar
droughts year after year after year.
Speaker 1 (06:30):
Really since twenty twenty two, twenty two, twenty three, twenty
four and twenty five have all been incredibly poor years
from a drought perspective. I'm just really trying to isolate
two crops or two things, right. One of them is coffee,
which we've seen, you know, prices of coffee go up,
and the other is cattle. You know, so when you
(06:51):
think about Brazil, what do they really bring to the
global markets kind of at scale? Those are kind of
two of two examples, and you don't have to stick
with the Brazil example, but I'm just saying that it's
very if you look at the data, we had kind
of sideways to nothing burger drought related events in Brazil
for many, many years, and then last several years have
(07:13):
just been a flurry, And of course that's going to
have an impact. When you're the biggest producer of coffee,
you want to if not the biggest producer of cattle
in the world right up there. Right, So again, you
could pick any example you wanted agriculture, but I just
wanted to kind of flag that as being from our research,
very glaring, you know, cause.
Speaker 4 (07:33):
It effect Yeah, absolutely, Yeah, Brazil's a biggie. I mean
for so many of the big commodity crops, right, you
mentioned coffee, cattle, corn out there.
Speaker 3 (07:43):
So yeah, I mean the what we see there.
Speaker 4 (07:47):
And then I mean the other one that we have
a lot of folks interested in is coco these days, right,
And I'll you know, I don't have that off the
top of my head, but it's something like seventy five
percent of cocoa is grown in West Africa, and there's
been some some large droughts there as well the last
few years, and you saw big volatility and in the
prices there.
Speaker 3 (08:08):
I think it's come down to some degree.
Speaker 4 (08:09):
Over the past couple of months, a lot of the
way that that our customers use our tools. So there
there are folks that that are interested in how do
I actually manage this on the farm. How do I
manage drought on the farm. I mean, to some degree,
drought is a difficult one. If you're not set up
to irrigate if it's a prolonged drought. You know, it's
sometimes for some of our customers, it's just a matter
(08:31):
of knowing is actually valuable, right, even if you can't
physically do.
Speaker 3 (08:36):
Something about it.
Speaker 4 (08:37):
And then a lot of our customers are actually keeping
an eye on the global market, so they're thinking, okay, yes,
Brazil is a major player in terms of of what
let's let's Useles use coffee production globally, and maybe they
source a lot of their coffee as your coffee company
source a lout of your coffee from Brazil, But you
also want to understand how's coffee looking at Indonesia and
(09:00):
the Opion and and many other places where there's where
there's coffee. And so that's again where you sort of
get into that. It's a it's a complex game, of course,
and many folks, you know, commodities traders are well equipped
to in my experience, well equipped to ingest that probabilistic information.
And you know, in their case it's just can I
have an edge over the next person. But the reason
(09:23):
I bring up this sort of global this global view
is for a lot of our customers, what it What
it ends up coming down to is, yes, you want
to understand your primary regions that you're either growing something
or sourcing from, how are they doing? But how are
all the other regions of the globe doing? And so
then you get into these looking at projections for do
we have scenarios where we have concurrent droughts in all
(09:44):
the main you know, corn growing regions of the globe, right,
or we growing because then you really have this issue
where there's not flex in the system where you can
actually go go to the spot market and procure thing.
Speaker 1 (09:56):
So that's really from the demand side, right, I Mean
I had a question this is apply side.
Speaker 2 (10:00):
But you're saying it's good to know about the drought.
Speaker 1 (10:02):
Does that mean that people will like cut their the
amount of you know, fertilizer they're using because they just
know it's a it's a lost cause? Like, is that
the the kind of result that they, you know, are
looking for by understanding that risk?
Speaker 4 (10:18):
Yeah, I would say it depends. It's very sort of
location and crop specific. Just to clarify real quick, I
probably maybe I misspoke earlier. I was thinking more on
the supply side. How much supply is flowing into the
global you know, global coffee coffers if you will.
Speaker 1 (10:31):
But I mean, but the people that care about the
supply side is the demand side.
Speaker 4 (10:36):
Absolutely absolutely, we need to meet x amount of demands.
So totally totally true. So a couple of examples of
how folks actually use it, and actually i'll go here
to actually this is this is different crop, but one
people care about in a lot of cases hops from beer.
So one of the decisions that that people have to
make in that life cycle is when to harvest the hops.
(10:58):
And in a lot of places where they're grown, and
you will be harvesting as it's getting colder, and if
if a frost actually hits your hot crop, it's really
bad news. But every day you wait to harvest, you
get more yield off of the plant. And so then
they're using our analytics to say, what is the probability
of there being a being a significant frost over the
(11:21):
coming weeks, Do we need to schedule harvest or not?
Can we get away with being a little bit more
you know, aggressive, And of course you need a strong
signal there. The other one, a big one is for
planting timing as well. So there's a lot of crops
that need some amount of moisture, especially if you're not
in an area where you're irrigating for the seed to germinate, right,
(11:43):
and so you need to be able to have confidence
that there's gonna be enough enough precipitation occurring in the
coming weeks and so there it's about sort of timing
when you're going to plant. So a lot of it
is about, as you alluded to, moving timing, like what
what degrees of freedom do you have as as a
producer to move your timing. There's only a there's only
(12:03):
a few sort of events per season that you can time, right.
It's like planting, as you alluded to, any inputs like fertilizer,
irrigation and then harvest. But a lot of a lot
of what we see folks using our platform for is
to yeah, play different scenario games with that timing, so
we know what's happened up to the season until now
(12:24):
we know what our forecast is doing. What if I harvest,
then what's my expected yield? What if I harvest you
know at a different time.
Speaker 1 (12:30):
That type of thing and timing is critical because you
have to assemble all the players right to get to
all the labor and everything else.
Speaker 2 (12:37):
That's required to get that done.
Speaker 1 (12:39):
Some people have the rent tractors and you know, like
that kind of stuff, right, so that you need to
get ahead of the queue, not in the back of
the queue, right, if you can target that better. I
also understand, like maybe you don't do wine, but I
understand wine is has this real problem with with wet
weather because the grapes if they're if they're picked while
they're absorbing the water, become like you know, there's less
(13:00):
alcohol in the grapes because they've absorbed the water. Right,
So this is a pretty critical piece of the puzzle
for the for the wine grower.
Speaker 3 (13:08):
Absolutely, we do.
Speaker 4 (13:09):
We do have customers that are vineyard owners and making wines,
and we actually have this sometimes at events will we'll
do this sort of wine tasting and say, you know,
here's I am not a wine connoisseur, so I could
not tell you, you know what wine is. I mean,
I know which ones I like sort of you know,
but okay, this is you know what traditionally you've had
(13:31):
in Napa.
Speaker 3 (13:31):
Now maybe the same sort of flavor is going to
come from up further north in Washington.
Speaker 4 (13:36):
So we actually sort of use some of our climate
projections to say, like, where is an area that was similar,
you know, on a hillside in northern California twenty years
from now, where's the climate actually going to be really
similar where you could expect similar grapes. The other big
one I'll say for wine producers that we've worked with
is smoke taint. So that gets into the wildfire risk.
Speaker 1 (13:56):
And you know, I'm going to ask about that because
it's clearly been a I mean, that just destroys the
crop as far as I can tell, right, it's very
hard to recover from, yeah, smoke taking exactly.
Speaker 4 (14:05):
And in those cases, again, that's one where I don't
have a ton of knowledge on whether there's any any
mitigation they can do once it's happening. It's more about
that long term strategic planning of like you know, if
they're investing in in new lands for vineyards and things,
trying to understand how do we limit our exposure to
(14:27):
things like I mean, in that case, wildfire risks and
you know, and.
Speaker 2 (14:30):
Where the tactical burns are going to happen.
Speaker 1 (14:32):
I mean, that's obviously a big point, right, because that's
you're in some ways the government is really in charge
of that because we do not want consumers.
Speaker 2 (14:41):
Tactical burning of land.
Speaker 1 (14:44):
But the idea is that, you know, if you understand
where how the government thinks about it, and they're just
trying to make sure that no fires are started, or
at least we can limit the reach of the fires
once they start. That that's a pretty critical piece of
the puzzle. And obviously with the wildfires up in Canada,
Canada has a lot I mean, actually do any business
(15:04):
in Canada, but they've got a ton of Agen is
a big wine region.
Speaker 3 (15:08):
So yeah, sorry interrupt.
Speaker 4 (15:10):
We have some smoke here from I believe Canadian wildfires
right now in Indianapolis here today. Woke up and I
was looking at the AQI and I was like, oh,
what's that about? And yeah, so yeah, there's.
Speaker 1 (15:20):
I mean, just to kind of depart a little bit there,
there's all sorts of lost labor and all sorts of
you know, a lot of things happen within the context.
Speaker 2 (15:28):
Of smoke days, you know, what I referred to as
smoke days, and they really.
Speaker 1 (15:33):
Have our having a bigger and bigger impact on you know,
like our ability to work, you know, frankly, like and
health and obviously if you have small children, or older people.
You need to be very careful about that stuff because
more and more it's not burning wood as we saw
in LA it was couches and things like that, which
(15:54):
is horrible for your you know, so as it spreads,
it goes from being a a scent you would be
comfortable within your car to something maybe not so much.
Speaker 4 (16:03):
Absolutely, And just to add to that briefly, I mean,
one of the things that I don't think has necessarily
garnered enough attention over the past you know, decade or
two is just air quality in general, and how many
you know, early deaths. I think I saw a paper
the other day. It was estimated something like four to
six million. You know, this is globally annually early deaths
(16:25):
attributable to air quality, right, So it's it's the scale
of the issue is huge.
Speaker 1 (16:32):
Yeah, And in the United States, which has been basically
on a fifty year train down in terms of air
quality improvement, you know, like we've improved year after year
after year, mainly through the Clean Air Act, and from
nineteen seventy one or whatever it was started, we're starting
to see that tick back up again. Well, so that
is like the first not just deceleration, but reverse reversing
(16:55):
in that trend because of the wildfire smoke, because you know,
it's really really harmful, like from a you know, relatively speaking,
it's much worse than allowing some of those zone kind
of plutants to continue. You know, So this is a
becomes a bigger deal. Absolutely.
Speaker 4 (17:13):
I remember, not so fondly, when I was at Stanford
for you know, a couple It was there for several years,
but for a couple of weeks during one of those
wildfire seasons where it was just funneled right over US
down in you know, Bay area, and it was I
was just sort of living in my n ninety five.
I guess it was early practice for the pandemic.
Speaker 3 (17:32):
But yeah, yeah, no.
Speaker 1 (17:35):
It's a it's you know again, these are these are
costs of doing business. Now, this is how this is
unfortunately part of the game. I would like to ask
a little bit about data, if I may, because obviously
we are seeing I think the USDA is sort of
holding its ground in terms of a lot of the
data that sometimes at one point they were trying to strip.
Speaker 2 (17:54):
Out quite a bit of data out and related to climate.
Speaker 1 (17:56):
But I think they realized that because it's just so
fundamental to crop production. They've kind of put it back,
is there, what are there just overall? What are your challenges?
They could be in the United States or elsewhere in
the world, or you could actually if you might, I
might be helpful for you to explain how kind of
(18:18):
good or bad are data is relative to the rest
of them.
Speaker 3 (18:21):
Yeah, great question.
Speaker 4 (18:24):
So first, just in terms of recent trends, we have
as of as of now, knock on wood, not run
into any issues of data sets that we were depending
on no longer being there and refreshed. There have certainly
been a lot of things taken down in a lot
of cases, you know, if it's if it's more of
a static data set, you know, we have that anyways,
(18:44):
and I know there's also lots of efforts for folks
to create sort of catalogs and other places of those
data sets, which I appreciate. Of course, we have contingency
plans for if some of the data sets that we
depend on operationally go away, but that the case so
far that that hasn't happened. Our biggest I mean, at
(19:06):
a very for very specific example, the biggest data challenge
is getting better and better precipitation data sets for a
wide variety of reasons. One is, you know, we've gotten
a lot of interest in the index insurance or parametric
insurance space, right, so benchmarking. I'm sure your listeners have
(19:28):
some familiarity with that, but benchmarking, you know, an insurance
payout based on whether you got over a certain threshold
of rain, and you know, it cuts down on the
transaction costs of having an adjuster going out and looking
at in the agricultural space, for yeah, it's basically.
Speaker 1 (19:43):
A clarity thing, you know, like if X happens, you
get paid, right, And that's becoming more of an aon
and A and PLC and others have kind of jumped
in the middle here and said, look, I'll be the
middleman here. It's kind of like, what's he wit. I'm
trying to think of the tax people, you know, right right,
it's similar to that where they're hr block.
Speaker 2 (20:04):
That's who I was thinking. It's kind of like, you know,
under that under.
Speaker 1 (20:08):
That umbrella of you know, we did we we believe
what the data and as a result, we're going to
pay you in a hurry. Getting paid to hurry in
some cases is the difference between you know, survival and
not survival, right, because your other people are already you know,
you owe money to people generally speaking in business, right.
Speaker 4 (20:27):
So yeah, yeah, And another example is I believe, I
don't know that this is still the case, but when
I was living in New York, a lot of the
catastrophe bonds in New York where we're indexed by like
the the water level battery city gauge, for example. But
so the reason I bring that up is obviously data
quality becomes extremely important there, right, because that's what's determining
(20:49):
whether something gets sort of paid out or not. And
precipitation is one that's just really really difficult where our
product is global. We we have customers on every continent
except for Antarctica. But the you know, there are dense
rain gage networks in the US, in Europe, in a
(21:11):
lot of Australia some other places, but then there's other
places where we just don't have a long record. We
need to depend on more remote sense, remotely sense precipitation products.
But with any of that, you know, meet satellites and
we ingest tons of information there. It's all great, but
it's usually only as great as you have ground truth
(21:32):
to be able to calibrate too, right, and so in
locations where you just don't really have that, that remains
a problem for us calibrating our forecasting models for us,
you know, working in we were just talking about parametric
insurance space. So that's a very specific one that I
think you know, comes up all the time with colleagues
just in terms of getting better and better data sets,
(21:54):
and there are some that are really good in some regions,
so you know, you can take sort of a mosaic approach.
I think the other data challenge is really the delivery
of data to customers in a way that allows them
to take actions fairly simply. So I was earlier talking
(22:15):
about or mentioning decision making under uncertainty and basically everything
we do you're dealing with an uncertain world on certain
forecasts and on certain projections. That does not mean bury
your head in the sand and throw up your hands, right,
It means, okay, how do I actually extract the information
from this? And actually, as a real side note, just
this morning, I was seeing some former colleagues of mine
(22:37):
published yesterday a paper on how climate models that we've
been using for years may be biased with respect to
their tails. So actually underestimating some extreme heat into the future.
So just to say that uncertainty also is evolving as
the science evolves, that's not great news, right because it
means even if the mean warming signal we think the
(23:00):
are on the right track, some of the extremes might
actually be be larger than we saw.
Speaker 1 (23:05):
Yeah, that's and just as an aside on that front,
I mean, that's if you read those incredibly boring six
hundred page reports from the i p PC, you realize
there's a lot of unknowns, you know, like.
Speaker 2 (23:15):
They they they carve out things. You're just like, what,
how do we not have a handle on this?
Speaker 6 (23:20):
You know, it's it's you know, we are not that
smart and maybe comes this up, so it's, uh, don't
it's not easy to pat ourselves in the back when
it comes to understanding that.
Speaker 2 (23:31):
I mean, it's a complicated problem, don't.
Speaker 1 (23:33):
Get me wrong, but it's a you just can't believe
there are these glaring gaps in the in the data
sets as you're describing right now, with you know, precipitation
for example.
Speaker 4 (23:43):
Yeah, yeah, and and so this is this is bending
your question a little bit, but it the biggest challenge
that I see for us and for many others, is
how to deliver things in a concise way that actually
fit into the workflow of our customers.
Speaker 3 (23:58):
Right.
Speaker 4 (23:58):
That's not making them read some giant report, right, and
not in our case also not. We're trying to do
things more at scale to be able to make the
accessible to a wide group of users. So rather than
you know, if we were a consulting company, it would
be like I'd sit down with you Andy, say exactly
what you know is your operation, go through exactly all
(24:19):
your vulnerabilities, put together a playbook with you in terms
of Okay, if you see this type of event coming,
here's how you could mitigate it. We do that some
with with our enterprise customers, but we want something that
is is able to be translated into their workflow seamlessly.
So touching on the AI portion, we've we've done been
(24:39):
doing a lot of development with agentic systems and in
a slightly different way than I think usually gets talked
about these days. In a lot of cases, you know,
agents are all the all the raven in a lot
of a lot of the business world.
Speaker 3 (24:54):
Often for automating tasks.
Speaker 4 (24:56):
Say hey, I want to just have these four agents
doing these things that I think think are repeatable. How
we're using it is more as an interface for customers
so to be able to say, Okay, our agents have
access to our tools. I'm calling them tools, which is
actually just our analysis that our great team has built
over the years, right, that are repeatable. But the communication
(25:20):
and the way that the customer sort of would ask
questions of that analysis can be quite labor intensive. But
if you can leverage some of the large language models
and some of these agentic systems to give the customer
direct interface to our tools, so it's not them sending
an email to our customer service and then customer service
talking to our analysts and then our analysts doing you know,
(25:43):
putting together report and going back trying to really streamline that.
And what we're what we're starting to find is that
it opens the door a little bit more for us
to be able to provide a valuable information to folks
because it lowers the friction there. It's not something necessarily
extra that they have to do and spend a lot
of time tweaking things, but they can through our platform
(26:05):
go sort of directly to having access to some of
that that world class analysis. This is this is early
days here, right, I mean, sure, it's not trivial to
spin these things up. And we're in a space where
like it's catastrophic for us if if the agent is
providing hallucinations to a customer, right, we have to avoid
(26:26):
that at all costs. So that's been an interesting one.
But but to me, the biggest challenge is communication. Basically,
I guess is the summary or I mean what, especially.
Speaker 1 (26:37):
With commodities, there are ways to actually translate what you're doing.
Your probabilistic assumptions can be translated into probabilistic bets, ye
know what I'm saying, Like, because they can do some
hedging that you know, if you think there's if they
think there's an edge relative to the their peer group,
you can set up a you know, slightly not two
taxing structure around it, right where you sell little bit
(27:00):
of upside against it, just to ensure that you're making
the right you know, like you're covering your yourself in
case it doesn't go as planned.
Speaker 2 (27:07):
But if you're trying to.
Speaker 1 (27:09):
Optimize, and especially when you're like you have to harness
a lot of energy to get some of this stuff
done early, right, I mean, it's not again it's not
there's no button. It's not like a trader where he
just pushes a button and says buy this. It's a
it's a process to actually get your uh you know,
coco out of the ground or whatever it is, right,
(27:31):
you know, you know like that that is like the
decision to go is a lot less like a trading system,
if you know what I'm saying. So it has a
uh because it you know, like that's how bots are
used pretty universally at the moment. But that for what
you're doing, it's almost like, well, okay, you know what's
your confidence level, how can we translate that into some
(27:53):
kind of you know structure you know that are around
trading structure, arounding.
Speaker 2 (28:00):
This kind of uh, you know, this kind of energy,
shall we say around it?
Speaker 4 (28:05):
Absolutely and two two thoughts on that or two sort
of examples are you know, we have customers today who
who use our system just in that way. So one
of the things we do I didn't say this earlier,
but is predicting yield for these for these major crops, right,
and so we'll have customers that either they're procuring all
of their corn or what or wheat or oats or whatever,
(28:28):
on the market, or they have contracts with you know, uh,
with producers in certain regions for fifty percent of what
they need to procure and then they are going to
go to the open market for the other So they
are definitely using our system to say, hey, it's looking
like a bad year in our in the area we
(28:49):
primarily procure from, or a good year. That means we're
going to need to you know, get x amount on
the market. And oh, by the way, and Brazil's looking
really like it's going to be bad potentially in a
couple of weeks. Let's lock in some of those futures
right now. So that's being used. The next step, which
is still sort of early days for us, but where
I'd really like us to go is so we can
(29:13):
give that information in whatever sort of whatever format the
person on the other the customer needs. But in order
to actually give any information on like buy sell, et cetera,
I need to know their cost function, right, I need
to know what the outcome that that decision tree is.
And so we've we've done some exploration with actually allowing
(29:37):
customers to give us that information. Right, it does get
pretty complicated pretty quickly. And then that's where you get
into some of the challenges becoming idiosyncrasies of like how
people have structured their way of making decisions those buy sell,
et cetera. And so the challenge for us is like,
can we do this in a flexible, scaled way that
accommodates that service for a lot of customers.
Speaker 3 (29:59):
Or does it really need to be tailored customer by customer.
Speaker 1 (30:03):
I should Yeah, that makes sense because yeah, as you say,
there's a lot a lot of moving parts. So it's
a it really depends on that person's particular circumstance. Are
they borrowing, you know, are they is the land being
rented to them?
Speaker 2 (30:15):
For example?
Speaker 1 (30:16):
Is a fundamental, fundamental question because that those those pavements
can't be not met, right. Whereas if it's someone that's
been in the been in the business for a very
long time, that's not necessarily.
Speaker 2 (30:28):
Front and center.
Speaker 1 (30:29):
They're literally trying to optimize year to year the best
they can and they understand that other people don't have
that flexibility.
Speaker 2 (30:35):
Right.
Speaker 1 (30:35):
So if you're if you have flexibility built into your
business model by virtue of your great grandfather or whatever
it is, that's a that's not a bad Uh, there's
no real reason why you shouldn't take advantage of it
because others are more constrained, right, they don't have that
kind of luxury.
Speaker 3 (30:52):
Absolutely.
Speaker 5 (30:53):
Yeah.
Speaker 4 (30:53):
I mean we could tell a customer, hey, there's going
to be a great crop of canola or something in
Australia and they're saying, that's great. All my process and
facilities are in Canada, right, yeah?
Speaker 2 (31:02):
Right?
Speaker 1 (31:03):
Can I can I tap a little bit your brain
on supply chains with respect to some of these uh,
you know, because just we can stick with our culture
or really whatever you're kind of comfortable with you want
to go.
Speaker 2 (31:13):
We can switch directions if you want.
Speaker 1 (31:16):
Just how important your kind of you know, the dependence
there is on supply chains, right, Obviously it's a big deal.
It's becoming more of a big deal. Tariffs are all
sorts of a bit. I mean, they're you can you
can talk about tariffs or not, but I mean the
fact is that there is uncertainty out there, and you know,
your supply chains are pretty important these days in terms
(31:39):
of flexibility. Let's just say, based on how the tariff
dialogues go from week to week, month to month, year
to year.
Speaker 4 (31:48):
Yeah, no, great question, I will I don't have anything
to say on tariff that hasn't been said a million times,
I'm sure by smarter people than I. So, but I
mean that does come up in terms of uh, you know,
maybe in this current moment, overshadowing some of the climate
related risks. I mean, not totally, but yeah, yeah, but yeah,
(32:08):
it's it's another layer of uncertainty there.
Speaker 2 (32:10):
But yeah.
Speaker 4 (32:11):
So for supply chains, we have customers I think on
sort of every branch of the supply chain. So I'm
thinking sourcing, manufacturing, transportation, and then on the demand sort
of forecasting side. So I'll give it like a really
brief example for each one. Sourcing regions. We've been talking
about that a lot. You know, what the information we're
providing is, you know, is the cocoa being grown and
(32:33):
ghana under drought stress for example manufacturing there, it's often
about disruptions to manufacturing plants. So is there a signal
of upcoming flooding that is going to limit the ability
for manufacturing plants to operate, workers to come to and
fro the plants, And we have I think.
Speaker 1 (32:53):
That happened recently in Switzerland right with the chocolate factories
or maybe everyone.
Speaker 4 (32:58):
I am not familiar with that specifically, but it's I mean,
it's happening a lot of the time in a lot
of different places. Actually, because this is a fully public engagement.
We work with Hatachi, the Japanese company for resilience on
their supply chain and focusing a lot on on manufacturing
(33:20):
facilities and and they're primarily flood hazard also tropical cyclones
disrupting them. So then then there's transportation. A big an
example of transportation related disruptions that can occur that we
work in is river levels. So Mississippi, for example, a
(33:42):
ton of things get get barged up and down the river.
There are times of year there have been a couple
over the past several years, especially in the fall, when
the river levels get low enough such that you can't
send as fully loaded barges or can't send barges through
and stretches the river. This is a huge issue because
(34:03):
then if you really need to get something down the
river or over somewhere else, you you know, you have
to start trucking or training it, which again at the
the ratio of the cost there can fluctuate a little bit,
but it's a lot more expensive.
Speaker 1 (34:17):
To get wet could get wet, which you know, we
have some very hard to sell as you have to
dry them. Again, there's all sorts of costs associated with
that too, right, right, So so.
Speaker 4 (34:26):
There's there's that transportation link, and then there's the demand forecasting,
which is more of a recent area for US, but
there what we've done so far is more in the
retail space, so things like demand for groceries when there's
a forecast for uh, you know, big white out blizzard
in the Northeast for example. And so what we do
(34:49):
there is we basically quantify what level of demand should
you expect in the days before, then the demand drops
off during everything shut down, and then what kind of
rebound to demand there, I mean groceries also things like
restaurants and lodging and things there. And these are these
are areas where folks know have known there's a relationship
(35:09):
with weather obviously, particularly those extreme events, but being able
to quantify it and being able to use a model
that has seen hundreds of events around the the US
over the past, you know, over a decade to have
sort of more assurance of like here's how much bottled
water we think you're going to need when the hurricane
is bearing down and things like that. That's so that's
(35:31):
that's on that demand side. So I would say, you know,
in some sense, I don't know whether I mentioned supply
chains at the beginning in terms of overall our company,
but that it's it's really about trying to make supply
chains as resilient as possible in the face of weather
and climate.
Speaker 1 (35:48):
Yeah, I mean, they're all inputs, so it's a it's
all part of your part of your math, that's for sure.
Speaker 2 (35:54):
Well, we're kind of widening down here a little bit.
Speaker 1 (35:56):
I did want to ask if you had any kind
of research on the horizon you were excited about in
terms of next steps for some of your your work.
It's all sounds, you know, pretty great guns at the
moments of anything you're working on your like to share
with us.
Speaker 3 (36:09):
Yeah, I was actually.
Speaker 4 (36:14):
I think I mean the big one, as I was saying,
is our sort of foray into agentic systems and seeing
whether it can actually provide that layer of basically communication
and being able to integrate into people's workflows, right, rather
than just you know, automating a bunch of things. Actually,
in some sense it's it's being a translator. The other
(36:37):
I mean I mentioned that paper that I'm curious to
read the read it closely in terms of our projections
on extreme out right because extreme heats driving. I mean
we were talking we mentioned worker safety earlier, we talked
about wildfires. That's a big part of some of the
trends in wildfires can be you know, heat and the
evaporation and the drying out that comes from that one
(37:00):
other thing which is outside of of the research area,
but I mentioned earlier, there are sometimes opportunities from it's
not just about mitigating risk. But we've worked with building
materials companies, for example, where their demand in in the
you know, Southeast US for example, a lot of it
(37:20):
strongly depends on on hurricane activity.
Speaker 3 (37:23):
Unfortunately, you know, people's roofs getting.
Speaker 4 (37:24):
Blown off, and so we actually then provide those forecasts
in terms of how much product do you think you're
going to sell? And we had a success. I should
say that in major quotes because it was you know,
a tragedy for so many but Hurricane Ian we were
showing a really hot signal that year when it when
it went through Florida, and the customer we worked with
(37:45):
was able to supply shingles into that market. Much quicker
than they would have of else wise, because they actually
started producing shingles that were meant for Florida before there
was that, there was that strong signal that was was
known by everyone. And so I know, I I sort
of want rogue there at the end rather than giving
you research.
Speaker 2 (38:05):
But that's all right. It's what you're excited about. That's
that's what we care about.
Speaker 4 (38:10):
It.
Speaker 2 (38:10):
That's great, that's great. I mean, yeah, there's obviously there's
lots to do.
Speaker 1 (38:14):
I mean, the fact that this is a cost of
doing business is becoming pretty clear to almost everybody, right,
I mean, anyone else skinning the game knows it's true.
Speaker 2 (38:22):
That's for sure. Anyone who has a home insurance policy is,
you know, painfully aware of this kind of stuff.
Speaker 1 (38:27):
So with that, i'd like to say thank you so
much for your time, David was It was great great
having you on. I learned a lot And if you
want to find more more information on this particular topic,
you can please go to the Bloomberg function bi Climate
dam dot go, which really looks and tracks the damage
(38:48):
is related to climate that we have on terminal And
if you have an ESG quandary of burning question, you
would like to ask b i's expert analysts. Please send
me an email at A Stevens at ESG, Currents at
Bloomberg dot net. Thank you, Dave, Thanks for your time.
Host
Eric Kane
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.