How Sofar Ocean Is Reshaping Weather Prediction and Shipping

In this episode, Chuck sits down with Tim Janssen, co-founder and CEO of Sofar Ocean, a company working to close that gap by collecting ocean data at scale and translating it into information people can actually use. 

Tim breaks down why this is such an important unsolved problem, how AI is making this data exponentially more valuable, how shipping companies are using this technology today, and what a real-time intelligence layer across the entire ocean could unlock.

Chuck Templeton: I am really excited to be here with Tim Jansen, co-founder and CEO of Sofar Ocean. Tim, welcome to the podcast.

Tim Janssen: Thanks, Jack. Happy to be here.

Chuck: Awesome. Let’s jump in. Take us back to the beginning. What was the moment you realized the ocean data gap was a problem and worth dedicating your career to?

Tim: Yeah, I don’t think that’s where it started. It probably started where I think I wanted to learn to understand the ocean as something that I was just deeply interested in for many different reasons. Part of it is that I grew up in a country that’s mostly below sea level, so there was a pragmatic, I guess, driving towards that.

But from that, I think there are two ways of understanding the world around us, and one of those is like observations, like the sensing of our reality around us. And then the second part, if you have a bunch of observations, you try to build a model for how does this system actually work.

And I think in the ocean, we’ve had the benefit of having many observations of similar systems, like water in the atmosphere, and we’ve been able to transplant a lot of those models into the ocean with limited observations. But then, once you start learning about the oceans and what the models are and what we’re trying to predict, you at some point realize that you can’t really validate the improvements to the model anymore because there’s not enough information.

And that leads to problems like severe under-sampling of physics. And that means, effectively, you could get the whole system completely wrong and have no way of knowing it. And so when, as a modeler, you realize that basically improving the model no longer leads to unambiguous improvement of understanding, then you know the problem is observations.

Getting observations from the ocean has historically been very hard, very expensive, and to a large extent, that’s still the case. But also in terms of more the operational implications of observations, if you’re thinking about the uncertainty of the environment as something that is incredibly costly to society, both from a safety perspective, like protection of coastal communities’ assets at sea, and just from an efficiency perspective, like making decisions that are suboptimal, resulting in having to burn more fossil fuels and otherwise, the only way to reduce uncertainty in any system is to measure. And so, on the sort of like more near-term pragmatic perspective, if we want to help make better decisions across the maritime space, we just need more information. And that sort of led to the development of, okay, how do we do that?

And we ended up with the idea that information is incredibly valuable for every industry tied to the ocean space, and it is probably the most efficient way to accelerate data density by creating value from that information for the industry so that they get the direct benefit of the progress we’re making.

Chuck: Yeah, that’s amazing. And so you’ve talked about in the past how, each year that we don’t measure, we fall behind, and it compounds on itself.

And so how do you think about, to the extent the data gap is out there? Where should we be? What are some of the opportunities? What aren’t we seeing at this point?

Tim: Yeah. I think that is correct, that basically if we’re not measuring, we’re basically creating a compounding problem. But there are two different parts of observations that are important. One is compounding, the other one is less. If you think about how we build models in general, like we have a pretty good neural network in our heads, and the way we as humans build models is just observing our environment.

And so the scientific development and the understanding of like broad, like how does the world work requires massive amounts of data, which compounds, right? So if you do not observe the reality, you’ll have fewer observations as a sum, and as a result, that will make it harder to either train a model in the modern sort of AI, ML world or just build a model from, say, a physics-based understanding of the world.

The other part of it, and this is less like compounding, but this is more about the real-time nature of observations. If you’re thinking about the reason why we like to observe the environment, ultimately, we do that to reduce uncertainty in the prediction of the environment.

The data we collect is mostly used to try to understand and reduce the uncertainty in what the future state of something might be. Partially, we build a better model, which allows us to relate observations to the future state of the environment. And secondarily, we need to constrain the initial condition, like the latest state of the system, to propagate it forward from that state forward.

And so from a weather forecasting perspective, for instance, if you were to measure the now state of the atmosphere and the ocean, and you do that really excessively in terms of like we have lots of observations, you could create a very accurate initial condition, and with a good forecast model, you can predict that forward.

Now, past observations are less valuable there. So I think that’s less a compounding problem, but more a real-time observing problem that we need to tackle, and we need to do both. So if we have something that captures a lot of data in real time, and we do that all the time, then we’re also addressing the compounding nature of data collection in the ocean.

And basically, we should do both. We should just be collecting information all the time so that we both have the better initial conditions to make predictions of real-time processes like weather and long-term observations to look at longer-term changes like climate and understanding the climate and weather systems better.

Chuck: There’s gotta be hundreds of different things you could measure in the ocean and all of the parameters around it. How do you figure out what is data and what is noise, or is it all really data that’s valuable? And what do you collect currently, and why?

Tim: Yeah. The good thing is there are lots of opportunities to measure stuff that is pretty valuable and unique in the ocean because there’s not a lot happening right now. So there is a big opening and opportunity for doing so. As in many places, what is noise to one person is signal to somebody else.

Generally, we try to be very pragmatic about what we measure, how we can create, you know, the most value in terms of understanding the system, working our way up to increasing complexity because ultimately we want things to scale. And so, for actually improving the predictability of a system, you need lots and lots of data to create a significant critical mass to actually make a difference.

So we try to start with the things that are very scalable, and because obviously we need to build a business around it, we have to look at what the unit economics of both the value of the data and the cost of collecting it. And so we started out on the surface, and the most valuable data points that we are collecting or started collecting were surface weather and surface waves, so understanding the wave spectrum across the ocean and where the waves are coming from, and what frequency they have. So these are your wind-generated waves that you can see every day when you walk along the coast. Understanding barometric pressure, which is basically the weight of the atmosphere from the surface of the ocean, the wind speed, direction, and currents. Those are basically everything that constrains the surface weather of the ocean.

And from that part on, we work our way down. And so once we have the surface dynamics, which is incredibly valuable for anybody operating on the surface, which is where most of the economic activity takes place, you can start looking at what happens below the surface because the surface dynamics impact that severely.

So, from here, basically, we start measuring underwater acoustics, which is basically the way in which information is transferred underwater as different from above the ocean surface, where most of the information is communicated through electromagnetic waves, which is not something that propagates underwater.

So everything becomes acoustics underwater. And then you need to understand the container in which that information is transferred. So you need to understand stratification. So how does the density of the water change while you go down, and what is the depth of the container to fully understand, like what actually the information you’re picking up means.

Chuck: So, as you’re collecting this information, how do you figure out what’s the right package to sell it? Like, how do you productize it and sell it to a shipper or another customer out there? What’s the process that you go through to figure that out?

Tim: Yeah. So we start with understanding the pain point. So if you think about maritime shipping, ultimately, they have no interest in looking at oceanographic data necessarily. What they’re trying to do is operate a vessel in an opportunity space that is captured by the market, the performance of the specific vessel, so how efficiently it can move through the water, and weather, which is basically your resistance field and the thing that impacts your safety significantly.

So all we do basically, in like the context, say, of maritime shipping is like, look, we’re adding the weather information to vessel information that they already collect and market information that they already use to operate their business to create an optimization opportunity that allows us to translate the ocean information into increased efficiency and safety.

And that is generally how we look at what we’re building. We’re basically building a maritime intelligence platform for protection and efficiency. So we’re translating the data into something that somebody, an end user, can use directly to improve safety, to better understand what is going on now and in the future, around an asset or a location, and improve the efficiency of operations.

Chuck: If nothing else, the oceans are global, and you have a product called Wayfinder that is essentially a Google Maps as you think about sort of commercial ships.

How do companies engage with Wayfinder, and what does that product look like?

Tim: Yeah. In the simplest explanation, it really feels like a Google Maps for the ocean. It runs on the bridge of a vessel. It’s a web application. You can log onto it from anywhere in the world. So a shoreside operator would see exactly the same thing as the captain on the ship, and the crew on the ship would see.

But ultimately, it’s about taking the combination of weather information, the performance information of the ship, and the market information and translating it into the best strategy for the ship at any given time. And in an ideal world, it would almost be imperceptible from how the company works.

It is just a tool that sits in the middle to say, ​“Hey, you’re trying to adjust your strategy for sailing. This is the best thing you can do from an economic and safety perspective. Do you want to follow this? And if yes, great. If not, tell me why not, because then we can learn from that.” And so it is really like an interface to a lot of complicated data, which before Wayfinder was maybe like dead-ended in the sense of like in any of these shipping companies, this information is somewhat used, but never in this way.

So you have, in any typical shipping company that operates, commercially operates ships, you typically have a group of people that decide which cargo to move.

You have a group of people that operate the movement of the cargo, so actually, running the ship, like there’s both a master on the ship and a shoreside team to support that. And there’s a group of people that look at the performance of the ship itself, like how efficient is this vessel?

Should we clean it? And so that already means that there is performance information, there is weather information from how do we execute the voyage, and there is information on the market. And so Wayfinder basically enhances the weather information with our proprietary forecast, combines these sorts of like other parts of the information, like the vessel information, the market information, and translates that into a strategy for the ship so that they don’t have to go figure that out themselves, and you just basically get dynamic updates like, ​“Hey, the weather has changed. We should go a little further south,” or ​“The market has changed. We should speed up or slow down.” 

And so that’s really what it is. It helps companies You know, execute on voyages in the best possible way without, without a lot of work. And you could take that one step further and say, for instance, there are marine protected areas or like sensitive areas where maybe a company would want to take the effort of we don’t wanna go through that region, or we wanna take certain precautions in that region.

In Wayfinder, you could draw a line around it and say, ​“Hey, we wanna have a certain behavior of our ships in this region.” And instantly, your entire fleet would start following that, and we can calculate exactly how much that would cost or benefit you so that you can make a decision whether or not you wanna do it.

That’s really like the whole purpose of this. It’s about removing the complexity of fleet management from both economic and potentially other factors that go into that. And that could be safety, it could be around emissions to make that incredibly efficient and also actually optimal because that is not, it’s not straightforward.

And with increasing complexity in regulatory frameworks, those decisions only get more and more complicated. And with a relatively simple optimization platform fed and powered by the best information in the world, both on weather performance and market, you can make that objectively optimal.

Chuck: And that’s gotta be an ROI sale to these shipping companies. So what’s the story, and how would you pitch that to them?

Tim: Yeah. It’s a simple story. And it’s really around embracing the idea that if you tap into a data-driven ecosystem, not only are you gonna get the benefits from it today, but every time you get better information, you’re gonna get more benefits. And just to put it, make it very specific, some of our customers have gone public with the benefits that they’ve seen.

For instance, Mitsui O.S.K Lines published that they saw a 6% efficiency gain from using Wayfinder. Now, these gains or efficiency improvements are always very difficult to calculate, so we prefer our customers to calculate. We also calculate them. We much prefer the numbers that the customers see themselves.

But the thinking is very straightforward. You have all this information. We’re adding the best weather information. We’re making sure that all of the information is as accurate and as real-time as possible, and we basically optimize for the best economical outcome, taking into account the market and the safety requirements that you set for the vessel.

If you do that consistently, even if you get small benefits every day, that compounds. And whether that difference is 2% or 6% or 8%, it is in many ways irrelevant. It’s free there for the taking. And the benefits in terms of like emissions and cost to these companies is astronomical.

Chuck: Yeah, and things like safety are hard to measure. I’m sure you can track incidents and whatnot, but just the risk of something happening, right? Being able to highlight that and point that out. Speaking of safety, so obviously interesting times now, the Strait of Hormuz. How have you been seeing activity around that, and what have you helped these shipping companies do relative to some of the challenges that are happening there?

Tim: Yeah, the Strait of Hormuz obviously is big news now, and everybody’s thinking about that, but this is not new, not to the maritime industry, not to us. Maritime shipping, there are a lot of choke points in maritime transport. There are narrow straits or canals like the Panama Canal, Suez Canal, that are highly sensitive to geopolitical changes or weather, even to make any sort of transit through them impossible.

And as a result, you’re gonna, one, have a massive impact on the global economy, but you also have a massive impact immediately on routing decisions. And so for the Strait of Hormuz and in particular also like around the challenge around the Suez Canal, we’ve done a lot of rerouting down the Cape of Good Hope, the tip of Africa.

And that basically means that you’re gonna have more weather exposure, longer routes, and you’re gonna benefit more from having a good optimization system as part of your fleet operations. And that is what we automatically offer to our customers.

Whenever there are changes in the geopolitical situations, the ability to operate in certain regions, we dynamically and automatically reroute the vessels as best we can to maximize efficiency in the new situation.

Chuck: Yeah. And as a ship travels along, something happens, it changes the algorithms or the machine learning in your systems, and then you can, within minutes, hours? Like, how long does it take to help guide a vessel?

Tim: Ultimately, there is a risk assessment that is driven by the companies themselves, right? So, for instance in war zones, you can operate, some war zones, depending on what the situation is, but the cost of insurance and otherwise goes up astronomically.

So it is the call of the company to say, ​“Look, we don’t wanna operate here anymore. The risk is too high.” That basically translates into a no-go zone, or this region we cannot operate in. And then, yeah, the system will instantaneously reroute the vessel, and so the next best optimal solution.

Just in Google Maps, you were to say ​“I don’t wanna go through that route,” and it will find you the next best one. And that’s exactly what happens here. And this is instantaneous, so there’s no waiting time for that. And then yeah, the other things are changes in the market, like the Strait of Hormuz had a massive impact, which is maybe less appreciated.

It’s appreciated now by most people, I think, that fuel prices for cars have gone up. But for vessels overnight, they were tripled. And in many places, there was uncertainty whether they even could get bunker fuel. So imagine you’re operating a very large vesselin the middle of the Indian Ocean, and you don’t know if you’re gonna be able to get bunkers in the next port because you don’t know what the price is going to be and whether or not you can get the amount you need.

So that level of uncertainty adds to the complexity of operating a fleet. And that is, again, not new in the industry, although this was an extreme example. But that is the reality for maritime shipping when these sorts of geopolitical changes occur.

Chuck: So it’s just factored into the recommendations.

Tim: Factored in. A lot of companies have their own acquisition channels for getting fuel. So their estimate of what they’re actually gonna pay for the fuel is the most important for optimization.

So generally, we recommend tying into their system. So if they have an internal system, we will connect through an API to that, or they can manually input it into our system. The same thing for the opportunity cost of the vessel, like what is the time value of the vessel?

That’s the other sort of big market component of optimization. Again, whoever commercially operates the vessels, should probably have the best understanding of how this vessel is trading in the market. We can tie this to any index in the world that they’d like, but generally, companies will use their internal information, forward curves for opportunity cost to drive the best possible outcomes.

Chuck: Yeah, Tim, one of the things I’ve appreciated about this call and certainly our conversations in the past is you are so customer-centric in this. You keep referring back to what is the customer’s cost, what is their perception of that cost. I think that’s such an important thing, ​’cause oftentimes we operators and entrepreneurs, don’t understand that or don’t pay enough attention to that, in my opinion.

So kudos to how much you guys are paying attention. And you have a product, Spotter Scout, that is somewhat new. So talk about, like, how you came up with the product design for that, all the data that’s in it, and who buys it, and what they buy it for.

Tim: Yeah. And so basically, if you look at the ecosystem we’ve built, it’s been focused on maximizing the efficiency of data acquisition. So we started with the simplest platform, which was the Spotter buoy. Basically no active positioning except for if you tie a mooring to it, and obviously your active positioning you’re grounded to the sea floor.

That works really well if you’re either in shallow water where a mooring is the most efficient way to maintain position or in the deep ocean where moving around is totally fine, and you can actually work with a drifting network. So it’s a very efficient, low-cost way of scaling information quickly.

Now what it leaves open is like intermediate depths and places where you have strong shear currents, like for instance, in the equatorial régime where either currents move to the east or to the west, where a drifter would just get lost very quickly. This is mostly what Spotter Scout does.

It fills in a gap in geographies where either a free-floating or a moored system would not be able to operate quite as efficiently. So the way to look at it is to think of it as a Spotter with active positioning control. So basically, you can now tell it to go somewhere, maintain its position, even if there is a countercurrent or a strong tidal flow, regardless of the water depth, because you’re no longer tied to what the water depth is, and I need to use the water depth to actually maintain my positioning.

But other than that, it actually integrates with exactly the same sensors as our Spotter systems do. So through Bristlemouth, which is our open source standard, you can think of it as the equivalent of USB for maritime applications. You can tie any sensor we’ve ever integrated into the Spotter system directly into the Scout system, and the only thing different is effectively the positioning capability.

So basically, you build an unmanned surface vehicle that has all modern positioning capabilities. It’s not a fast-moving vessel. It’s a very slow-moving vessel, around four knots. It’s cruising speed. But it can be out there for a very long time.

So it’s a long dwell system, just like a Spotter buoy. And it can collect exactly the same information. In its basic configuration, it will collect waves, winds, and anything water temperature related. And then you have a payload option that basically ties any Bristlemouth sensor into.

So you can measure currents, you can measure acoustics. It basically becomes a similar platform.

Chuck: And so as you do this, what kind of sensors or how many sensors are on a particular spotter? Is it 10, 20, 50? Like, how many sensors are there?

Tim: It’s really dependent on the configuration. The basic system measures everything surface weather-related, and then you can add basically theoretically as many sensors as you’d like, limited only by the power that the Spotter can produce. And so this also goes up when you think about the Scout, which has a much bigger solar acquisition system and, because it has more surface area, we can add more sensors there and still maintain a very high duty cycle.

So it’s always a trade-off that if you add more sensors, depending on your power system, you’re gonna have to limit your duty cycle. So how often can you sense? If you want monitoring, you’re gonna have to be sensible about how many sensors you put on a certain system. And this goes back into, I think, what’s pretty important that we sometimes forget is that power is everything in the end, right?

So if we lose power, everything stops. When you want small, scalable systems, low power sensing capabilities is absolutely essential. And so we make active investments in helping other companies actually reduce the power consumption of sensors because that will unlock the scalability of sensing.

So, it’s something that’s very critical, and obviously, a bigger platform like our Spotter Scout can mitigate some of that, and you can put more sensors on it. But ultimately, we wanna bring everything back down. The lower power every individual node can be, the more nodes we can have or the more duty cycle we can have.

Chuck: And it’s all transmitted through satellite from a communication standpoint. Is that right?

Tim: It’s flexible from a telemetry perspective. We have across the ecosystem now satellite cellular, which operates almost globally. So basically, any coastal region where cellular is available, it automatically switches. We also have Starlink on the Spotter Scout.

And so basically you have different options from a communication perspective, depending on like, where you operate and what kind of data density you need. And that flexibility will only increase. We also have line of sight communications like LoRa on some of our systems, and we’re agnostic to that sort of pipeline through which you want to push the data out because different applications might have different needs there.

And again, you’re only limited by the amount of real estate you have to put antennas on and the amount of power you have to communicate data away from the system.

Chuck: And what’s the magnitude of cost decline that has happened in the 10 years or so, you’ve been doing this work? Is it 20X, 50X, 500X? 

Tim: Yeah, it depends on where you look, like from our platform, like where it was, say five years ago, to like where it is today, it’s a little bit cheaper, but not huge. I think it’s more like the development of new technologies based on advances elsewhere.

So, for instance, like the reason why Spotter exists, the reason why Spotter Scout exists, the reason why a lot of these other platforms exist are through advances in IoT, like just consumer electronics, right? And then, you’re combining that with like new developments in material sciences, you can build things that are much more repeatable, scalable, smaller scale, and you can get things that you couldn’t do before from small platforms, now.

That, if you look at maybe the last 30 to 40 years, it’s several orders of magnitude. Like basically things that you had to do with a massive piece of equipment with a boat, but you can do now with an airplane flying over and dropping buoys out. That is several orders of magnitude.

Similarly applies to unmanned surface vehicles, unmanned underwater vehicles. So it is rapid in terms of what sort of like new hardware development can do in terms of step function drops in terms of cost. But what is really important that we shouldn’t forget, while we drop under a certain threshold level of cost, at the same time, we need to really work hard to create value with that information, because the delta between those two is what drives growth.

It’s not the cost or the value, it’s the difference between the two.

Chuck: So in the past, you’ve talked a little bit about the small pie idea that the data that comes from this has been bespoke and smaller, hard to scale up. But with AI now and some of the ability to help bring all these different pieces of data together, you’re unlocking opportunities that weren’t there before.

Tim: Yeah. If you think about the challenge of building maybe any business, I don’t know, but certainly a business in the ocean space, there are three buckets of challenges. One is getting information. The other one is models. So effectively, you could equate that to expertise, like how do you translate the observation to something that’s actually useful to somebody?

And then the last bit is distribution, like how do you get this in the hands of somebody that can extract value from it? So that last part in the past was actually quite expensive and valuable, and you can think of it as this is both like your software interface APIs or otherwise.

So something that delivers basically the last mile, like delivery of your model output to a customer in a way that actually delivers value. And there is the go-to-market part to it as well. Now I think the go-to-market part obviously has some benefits from having AI, ML tooling.

But I think primarily the delivery systems, the cost of developing those has significantly decreased. Now, AI, ML will impact all of this, like all three buckets. I think on the data acquisition part, there will be some benefits, but it won’t be revolutionary overnight. I think on the modeling side, for sure, like data science in some aspects has gotten a lot easier.

But again, you still need expertise and hyper expertise in many places to really benefit there. But I think the delivery, the last mile delivery, has gotten a lot easier. So if you think about the ocean space as like a lot of smaller TAMs with distinct users and distinct pain points, but using mostly similar data and models, then you can see that basically you could actually unlock a much bigger TAM by building different delivery systems quite quickly and cheaply, which is something that wasn’t as easy as before.

Was it possible? Yes. You would just require more investment, and you’d have to go slower, probably. I think the availability of modern AI/ML tooling, particularly in the development of dashboards, APIs, and coding has significantly reduced the cost there, and yep, removed that as a barrier to entry, so to say.

Chuck: Yeah, I thought that was interesting when we were prepping for the call that you were talking about, that it doesn’t lower the cost of the data acquisition, but it actually makes the data more valuable because you can service more niches or more opportunities than you were able to before, specifically because you can actually package it much more efficiently and be able to meet the needs of the specific customers along the way. How did you discover that? Or what was the insight that got you there on that?

‘Cause that’s really interesting, ​’cause there are lots of companies trying to figure out what’s the value proposition for AI for them, and efficiency is the number one, quote-unquote, ​“obvious thing” here, but you’ve actually figured out how to make it actually more valuable rather than more efficient.

Tim: Yeah, I think the efficiency is a nice add, but everybody gets that, right? So I think if you think about it as like– Maybe one way to think about it is like you can see AI/ML as the development of the combustion engine. It gave us the technology to allow us to move things around, but as a result, we have to go dig for oil because it requires that to run.

And I think in many ways AI/ML provides an accelerant. It can go faster. We can do things much faster and more efficiently. But ultimately, we need a lot of information to make those systems work. And I think that it’s a supply-demand problem where as soon as you create a system that can more efficiently use information to create value, then there’s gonna be a much larger demand for information.

So effectively, the price of information has gone up tremendously versus some of the other parts of the systems, the cost of that has come down significantly. So I think this is a very data-centric era. And so in this sort of paradigm, information is much more valuable but also much more needed than ever before.

Now, how does it make it easier to collect data? Collection of data still is sensing. Like something has to physically interface with the world in some form or shape to make a measurement. And that could be an indirect measurement, like a proxy measurement, but it still has to be some sort of interaction.

I think the benefits there of AI/ML are not nonexistent, but they’re certainly less direct than it is, say, in software development or data science, or model development. So I think bringing it back to what we talked about earlier, like you need observations to understand the world, and from those observations, you build models of the world.

I think the models have certainly benefited a lot and continue to benefit, and have maybe in some places radically changed how they operate with AI/ML. The data itself is not that different in many ways. We still need physical observation somehow. A sensor needs to interface with the real world to collect information that can feed those systems.

And as a result, I think just because of a balancing relationship where basically one just got a lot cheaper and better and faster, and the other one basically is now more needed, so you’re gonna increase the value on that side. And I think it’s obvious, and I think everybody is seeing that.

And it also automatically becomes the most defensible mode for any company, its unique and proprietary information.

Chuck: Yeah, it’s pretty amazing how these models are so powerful, but they are so just ignorant to some degree, until you start to really put that data in there and be able to help it understand and get the context of your particular instance. And I think that offers a tremendous moat and defensibility as you continue to grow.

And every time one of these spotters are out there sending you back data, that just continues to build the moat that you’ve got.

Tim: Of course. LLMs basically wouldn’t exist without the internet, right? So the access of large amounts of information makes this possible, and without it, it will not be or will be very limited. And I think that is maybe one of the more exciting things, and like changes in what we will see in the ocean space.

The radical increase in the data density will unlock capabilities we’ve never seen before. But it will ultimately rely on information. Has always relied on information, by the way. Nothing has changed there. It’s just that basically the acceleration is much faster. It is much more clear to everyone that ultimately all of our understanding of the world is based on observation.

Again, nothing new, but it is more obvious.

Chuck: And as we see a changing climate, these observations, are they more valuable, or do we get concerned that because of changes happening so rapidly, I’m sure in some of these climates that it’s not as linear, the data’s not as informative as it could be? Or how do you think about that?

It’s gotta be a really challenging model, ​’cause if the next 10 years are gonna be different than the last 10 years, like how do you then use that data effectively? How does that work?

Tim: Obviously, in any system, you have different timescales. And if you see different dynamics on different timescales, you still ultimately have to go back to how good did you understand the system in the first place? So I would argue that in a system that has zero changes at any timescale, you don’t need to measure all that much because once you’ve captured the system, you basically learn more and more about the same system.

In a changing system, the only way to resolve those changes is to actually have much more data at any given instance in time. So again, if we assume that we’re in a rapidly changing system, I think that will only up the ante on like how much data do we collect at any instant in time, because in a number of instances of time, we’re gonna be in a different system.

So this is our only chance to measure the system in its current state because it will change.

. So you have to increase the temporal resolution of your observations as well.

Chuck: As you zoom out from where we are today, is there like a holy grail of ocean data, and if so, what is it? And if not, how do you see the next five to 10 years?

What’s gonna be the frontier of this ocean intelligence that’s out there?

Tim: A good question. I think I probably have difficulty looking at the future beyond two years at this point because things are changing rapidly. But I think, look, if you just step back regardless of the timeframe, if you look at what we want to achieve is that ocean intelligence is no different from the rest of environmental intelligence, atmospheric intelligence.

It shouldn’t be a thing anymore. Right now, it’s a completely standalone thing. It’s operating in its own space, and this is similar to what we’ve done in model development for forecasting the weather. Obviously, the ocean and the atmosphere are one system, and in models, back in the ​’50s and ​’60s, because of computational limitations, we broke it up.

Let’s model only the atmosphere, model only the ocean, model only the surface waves in the middle. And then, over the last decades, we’ve started to pull these models together. We started calling them coupled models. They shouldn’t have been decoupled in the first place, obviously, but we did that for practical reasons.

Also partially for social reasons because scientists were working together in either atmospheric scientist groups or ocean scientist groups. And as a result, basically, we have these models that we now have to couple back together. I think the holy grail would be that we have sufficient information to look at ocean intelligence similarly as we do on like atmospheric intelligence or land-based intelligence, that we have an abundance of information.

The sort of like trust levels of what the intelligence means is similar in terms of this is now within operational ranges. I can make decisions based on this. And we have the frameworks both for the data acquisition and the analysis of that data and the delivery of those insights in place to operate in the ocean and under the surface of the ocean, as we would in, say, the lower atmosphere.

And that I think would be a great state to shoot for, and we’re a long way away from that. And I think, again, the challenge is not the models. We have, I think, very similar models for underwater as we do over the water. It’s really data acquisition. The scalability of information, the sensing platforms, the scale of everything.

Basically, economies of scale are gonna matter to be able to accelerate fast. And if we set as a North Star that ocean intelligence is no longer something that we talk about in the sense that it is basically just like the intelligence of our environment that would be that would be a wonderful outcome.

Chuck: Yeah, that’s interesting. And that’s gonna come from a lot of companies working together too, right? Is there a consortium happening around that, and what does that look like at this point? And how’s that going?

Tim: Look, as anywhere else, I think there is an incredible ecosystem of startups in the ocean. I think that has been accelerating over the last, I don’t know, 5 to 10 years, which is, I think, incredible, and very important and valuable. I think, as startups do, we initially focus on a specific subtopic and don’t do anything else because that would just risk the success of the overall outcome.

And as you reach a certain scale, and I think that is still coming in this space in terms of okay, what are gonna be like the big players to actually scale things up and how do we partner up across those? We will start to see how that sort of commercially evolves.

Like I’m a scientist by training, but I firmly believe that it is the commercial drivers that will lead us to success here. And so we have to find value propositions, and all of these startups are working on that.

And I think ultimately the partnerships and the developments, like how these things all fit together, will be quite organic with a little bit of help here and there, obviously. But yeah, I think it’s a super exciting time. 

Chuck: So what is one of the coolest things you’ve discovered or have realized with your data about the ocean? You’ve had to have seen some really interesting eureka moments or aha moments throughout time.

Tim: Yeah, we’ve seen a lot of eureka moments. I don’t know if a lot of those eureka moments were more into technology, ​“Oh, shoot, we can do that. We can measure that. We can make a cheaper measurement of that,” versus the measurement itself. Because we mostly look at aggregated data.

We don’t look at event-based data necessarily. And so there are interesting things around what is the occurrence of very large waves in the ocean? Sort of one of those questions that everybody’s always dreaming about is are these freak waves real or not?

So that we have actual observations now, so we can say ​“Yeah, they’re real.” But it’s mostly about every time we find that we can measure something cheaply, like we can take something else and interpret it in a different way, so that we can have an indirect measurement.

I think those are the biggest eureka moments, because every time we do that, we unlock more information that becomes scalable. That is, I think the biggest thing that we should push for is if something is very expensive to measure, then either find an indirect way of measuring or make the way of measuring it cheaper.

It’s a mundane effort. It’s not super exciting or sexy, but I think it is the most important thing to work on right now in the ocean. Like just figuring out ways of getting data to be lower cost, so less costly per data point.

Chuck: Tim, awesome. Really enjoyed the conversation today. Love what you guys are doing. And obviously, you’ve chosen an area that is just really difficult to operate in, let alone succeed in.

Kudos to you and the work you and your team are doing. Look forward to watching you in the future and seeing where else you go.

Tim: Thanks, Chuck. Appreciate it.

Tonya: Thank you for tuning into the S2G podcast. If you enjoyed this episode, please subscribe, leave us a review, or share it with a friend. For more information or to connect with us, check out s2gin​vest​ments​.com.