Hello and

welcome to this special episode of the CORDIScovery Podcast.

In today's episode, we are going to be looking at two key

challenges that are deeply connected

to our future: food security and the health of our oceans.

Researchers are developing innovative solutions to make food systems

more resilient and to safeguard marine ecosystems.

I'm joined by representatives of two projects

that have received funding from the EU Horizon Europe program.

And these projects illustrate how science and innovation can help secure

a healthier, more sustainable future, both for people and for the planet.

So Wibi, Jérôme, hello. Hello.

Nice to meet you.

Thanks for joining us.

Wibi is a geo visualization

data scientist based at the OpenGeoHub Foundation in the Netherlands

and is part of the consortium of the AI for Soil Health project.

Yeah, exactly.

Jérôme is an Earth Observation expert and the Product Manager

of the European Digital Twin of the Ocean, the EDITO project.

So Wibi, if I could start with you, your project aims

to use artificial intelligence to improve soil health.

And you're working,

if I understand correctly, on the development of a soil digital twin.

How does that relate to the EU soil mission,

which aims to transition to more healthy soils by 2030?

That's correct.

But when we’re talking about soil digital twins,

we try

to adopt the concept of digital twins, but right now we're not exactly in the

digital twins for now.

You're sort of working towards.

Yeah working towards it.

I know it's more of like a decision support tools, if I may say so.

We created the data sets in the European scale 30m, which then can be used

for decision support information for general people, farmers, researchers.

Because we created this data, also it’s fully open source.

Everybody can access it. Everybody can use it.

No licensing, just CC BY 4.0, everybody can use it basically.

And So, the data was already there or?

The data is already there.

It's already been published.

Then as we know, soil is pretty important stuff in this world.

So especially for farmers and stuff like that, by them

knowing what kind of soil properties they have.

Sorry, I should have explained that from the beginning.

So, we have several soil properties.

The data set is from the Soil Scientist.

So, we have a legacy data from all the different countries,

and we collected them, we filtered them and we cleaned them.

We also measured new samples

in a pilot site with the help of the soil missions

community and the living lab, basically.

And we use those points

to create the map with machine learning algorithms, basically.

And combine it with satellite data.

So, this kind of like if I understand correctly, a citizen science

dimension to your project in that you're getting people

to help you collecting the data that you can then feed into?

That was the next data variables basically.

But right now, it's collected by a proper soil scientist.

So, it's like proper data of measurement of soil properties, right?

Carbon acidity and stuff like that.

And then we map them.

We use one method of AI, which is machine learning, to find a pattern.

And they will predict the basically like the value around the points.

I will get like a map of the whole of Europe.

Right. And what is the data showing you?

I mean, how concerned should we be about the states of Europe soils?

As for the state of Europe soils, I think it really depends on the places.

But in general, we can see a decline

if you visit our website, edito.eu, and check one of the soil properties,

because our data is from 2000 to 2022, So, it's temporal,

everybody can see the changes.

And you can notice, like most parts of Europe have a decline

in soil properties, which is of course not good.

Like decreasing soil carbons, but mostly this decreasing

soil carbon happens in the city because of the impervious surface

around the coastal area and stuff like that.

But of course, the fields are also affected because of the way farmers

hand their fields with the chemical they use, and everything.

This also affects the carbon content, for example, in their area.

So, this is also useful for them to check later temporal factors

like we can get like a statistic basically like a line of years.

And they can check “oh I have like a pretty good

peak increase in certain years.

What did I do at the time?

Why do I have like a big decline in the next year?” So that's why I say

that more like this is a support tool, because they can check by themselves

and they can see and, go back to, how they handle their farms.

And they can review their methods further.

Yeah. Yeah.

And what actual practical tools are you putting in the hands of farmers?

I believe you're working on a smartphone app.

Is that already available?

The smartphone will be available next month, the first version.

So the way we use AI is two things.

One is machine learning that I just told you about,

and the other one is large language models.

This one is pretty common, like chat bots and stuff like that.

We make a prompt and ask, and they give the answers back.

Basically, if we are not a real soil scientists, we wouldn't be able to tell

what does the soil properties mean if it's certain levels, what does it mean?

We want to know basically.

So, because of that we created the mobile phone app too.

So, this will include the chat bot.

This chat bot will try to interpret the soil properties

and tell the user what does it mean for them.

For example, the farmers can tell “Can you see my location right now

through the geo location?

Tell me what is the pH of my farm, and what does it mean for me.”

Stuff like that. So

we tune it, or we train it in a simple terms,

to make a simple lines or simple sentences that is easily understood

by the general public because we already have, like

several other platforms, that of course, the researcher can use easily

and it's more complex, but that's not ideal for the other users.

Yeah, basically that.

Thanks very much.

And I think we can come back to some of those issues,

perhaps, in a more general discussion in just a moment.

But first of all, let me turn to Jérôme and bring you in.

So, one of the main outputs

of your project or what you're working on is the digital twin of the ocean.

Can you just explain to us briefly what exactly that is?

Yeah. Okay.

For us, perhaps, a digital twin of the ocean

is like a replica of the ocean, a digital replica of the ocean.

So, you can see it like a globe

where you can see all the physics of the ocean,

which means, for instance, the ocean, it's like, you see the water

temperature, the salinity and the winds occurrence, everything related to ocean.

And it's interactive.

You can play with this globe.

So, it's like Google Earth, you see the Earth, and you can play with it.

And here we are talking about science and the ocean.

It's really about replicating

the physics of the ocean, the physics and the biology of the ocean.

Meaning, for instance, that you can see

past conditions of the ocean, but also, future conditions and forecasts.

And, here at Mercato, I work at Mercator

Ocean International, we are in charge of the Copernicus Marine Service.

And basically, Copernicus Marine

Service is about to giving forecast about the condition of the ocean.

So, on a daily basis, you have a ten days forecast of all of the ocean,

will change the temperature, the salinity, the current, and so on.

And the digital twin of the ocean is about taking this data and also other

data, from

in-situ data,

satellite data and so on, and to put in place computing resources

on top of that, AI, so we can give and develop applications what we call,

what-if application, that allows decision maker from the public

to ask questions about what if I'm doing this, what is the reason?

What difference does it make, figure out variables and different outcomes.

And I think, I mean, Wibi you mentioned that you have data

covering a 2- or 3-year period, 2020 to 2022.

I think you have much longer time series data for the digital twin of the ocean,

right Jérôme?

Yeah, exactly.

For the physics of the ocean,

we have models from the 90s,

so 40 years of forecast.

It's the forecast we call that hindcast,

so 40 years of reanalysis of these forecasts originally.

So basically, we have 40 years off sea temperature, current salinity and so on.

So it's very important because this way we can see the trends and we can

see change, or instance, due to climate change, through all these parameters.

And this is very important in particular for AI,

for instance, because you know that AI you need to train it with data.

And the more the data you have, the better it is.

The older they are, the more the data and the more accurate

it is if you want to predict change.

Yeah. Oh, sorry.

Wibi, please to come in.

No, I just I'm agreeing with it.

The data is very important for, for example, for our case

soil properties like various. Right.

They have chemicals, they have physical, they have biological attributes as well.

For us, the biological itself, we didn't map it because the model is not added up

so good because of the lack of data and the noise inside of the data.

So, in this kind of process where you use AI and the type

of machine learning it is very important to have good data.

So clean data and enough data for like analysis especially for him, it’s

like a whole globe.

Yeah, that is like proper data with distribution as well.

It’s a good point.

Yeah.

And can I ask Jérôme, I mean what is different

between this digital twin approach and you know, computer modeling

which has been around for quite some time, is it the AI dimension or what?

What's new?

Yeah.

In fact, of course, Digital Twin is using a computing model of course.

But you can see it as the computing models are a bit

static.

You have just forecast data that is a bit static. It's very science related.

And the digital twin is about dynamic stuff.

It's interactive.

You use this model, but you can play with it.

And it's easier for non-expert people to understand them.

So, it's more so in the ways, there's a dynamic of the stuff.

And AI and of course, on top of it, can bring some new stuff about this

new way to extract information and to ease the understanding of the ocean.

But really, it's really about modeling, of course,

because we can use a lot of modeling, AI is training with modeling and so on.

So, it's not at all the end of old school modeling, I would say not at all.

But it's like taking a snapshot with some modeling versus

having an interactive way to play with it.

So that will be the main difference with a digital twin of the ocean.

And how can you be sure or check that, I mean the

ocean environment

is a very complex and multi dimensional ecosystem.

How can you be sure that the model correctly reflects

the interplay between the different factors?

Yeah.

As you say, the ocean is an incredibly complex medium.

I mean, it interacts with everything and of course there is no one model,

a fantastic model that will magically explain everything.

So, we have basically a particular model for each of the parameters.

And you can see the digital twin of the ocean as

multiple layers, like for instance, you have the layers

where you have the ship routing, you have the physical information,

you have fishing efforts, you have pollution, any kind of information.

You can mix everything.

And we have then you have a model

to track every of these aspects.

And with the digital twin, you are able to mix everything

to try to extract information from them to do this.

So of course, in this area, AI is very useful because it helps you

to manage all this data because on the human level it's

very difficult to manage when you have a lot of databases and AI can help you.

AI can help you find trends, for instance.

Subtle changes are extracted from this from

this tool.

It sounds like for the digital twin of the ocean,

you have more years of data that you can, you can use perhaps more diverse data.

In your journey towards a digital twin of the soil.

Do you think you need more data?

You mean for the other years?

Other years, I don't know, perhaps other factors.

Yeah. We do need more data for that.

So, for the current soil properties, the layers, the map that is out, it’s

the one with actual clean

and enough data for us to model and map it.

Of course,

from the legacy point itself, the data that's collected by soil scientists

in Europe; it also contains data

sets starting from 1960s, 1970s.

But okay, the distribution of this data itself

is not comparable to the one in 2000 to 2022.

So, if we can model it, we can map it, but the quality will not

be the same basically. So that's why we didn't do that.

It's the same reason for the biological properties of the soil.

And I think,

for the digital twin, do you have a what-if scenario, right?

This way I'm saying that we are more like a decision support tool, because I know

there's a clear line between digital shadow and digital twins.

And what part of it is the scenario-based situations,

which is the what-if that they had, for us we are not there yet.

So, this is very interesting.

I love what-if scenarios.

I really love your setup too.

Thank you.

Can I ask you Jérôme, I mean, would these projects be working

amongst others with farmers to provide them with tools they can use?

I mean, you potentially have a lot of uses, for the digital twin of the ocean,

you know, everything from politicians and policymakers

to the shipping industry, fishermen, the fishing industry.

You know,

what kind of use are they making of the tool?

Do you do anything, in terms of outreach to those different communities

to explain to them what's available and encourage them to use it?

Yeah.

And, so the digital twin, the European Digital Twin of the Ocean,

the EDITO project, as we call it, is pretty new.

It started three years ago, but it's based on, as I explained,

it's based on data that has been in for a long time.

And in particular, it's based on the Copernicus Marine Service.

So, the Copernicus Marine Service, that's a European service.

So, perhaps I can introduce the Copernicus project.

It's a flagship project from the European Commission.

So basically, the Copernicus project is a

monitoring project on the Earth, and,

so it means that we have satellites,

we have in-situ data and satellite data to monitor all aspects of the Earth.

And there are six services.

So, marine service, land service, emergency service,

weather service, climate change service and so on.

And so, the marine service is about giving forecast data on the ocean.

And this is, of course, used by a lot of people from scientists

to industry and to private companies that make business of it.

There is something like more than 100.000 registered users,

millions of downloads per month.

And so, this is a basic data that is used.

So the European Digital Twin of the Ocean is built on top of that and built on top

of also another network which is called EMODnet, which is about in-situ data.

It is here to fill the gap

between just giving the data and help people to make decision and to

have information extraction of data.

So, we put in place the infrastructure

and the tools, so computing resources and tools to do that.

And so to more directly answer your question, for the moment,

for the first three years of the project, it was basically more a research project,

meaning that we were also discussing a lot with researchers

and scientists that developed models on top of it.

And we developed some what-if scenarios just to showcase

how it can be used not only for scientists,

but also for decision makers and the general public.

So for instance, for fishery,

we developed a what-if scenario on Sargassum algae.

Sargassum is an algae

that goes to the Caribbean, and when it goes to the beach,

it's a bit of a problem because there are cases

that are linked to this

sargassum and it's bad for tourism.

And we have a model that detects sargassum

from satellites and then a modeling of the drifting of the sargassum.

And then you know when it will arrive in your

economic zone.

And then you can go with your boat to fish for it.

To fish for it,

because it can be also a resource, because if you take it, you can use it

to make bricks, for instance, to build houses.

So, this is an example of something that is a problem.

Sargassum on the beach can be also an opportunity and resource.

But to know that, you need to know as a fisherman

when the sargassum will come, is there a sufficient sargassum

so it's interesting economically to take the boats, and so on.

So, this is the kind of example we have.

And the idea now with the next phase of the EDITO project

is to onboard this kind of, I would say public

or private, user to use the platform and to make the best of it.

Okay.

And Wibi, is there an equivalent example that you could mention

either that has happened or that you could see happening,

based on the data you're making available on soil health?

So, is there a kind of use case or scenario that you could mention?

Not a scenario-based situation.

But if you want me to give an example of the usage of this data.

I can tell you that because we are also like, within the soil missions.

So, we also have collaboration with the Living Lab and stuff like that.

In fact, yesterday we have like, course in Spain at the Living Lab as well.

So, in this course we also use this platform

that we currently have published, the Eco Data Cube,

and we educate farmers basically on how can they use this.

And I think yesterday they were testing it on the scale of their

farm, for example, one person's farm and they checked in the temporal factors

by the statistics by time, from 2000 to 2022.

And they do exactly what I told you the first time, like

oh, there is like a step increase and step decline in these certain years.

And they go back to their farming methods and they check what kind of chemicals

they use, what kind of stuff they used to take care of their farms.

And this actually makes sense and correlated a lot with the data sets

that we have, the layers that we have, and they can actually tell

which part of their farming methods that caused this.

And then they can plan for the future, what kind of method

they should have done for certain kind of plants, for example.

Because our data is also based on depth as well.

So, depending on what type of plant they were planting,

they need to see which depth does it correlate to.

So of course, each step has different pattern.

I see.

Perhaps, just a last question for both of you.

I mean things are obviously evolving very quickly with artificial intelligence

and so on.

If you could look into your crystal balls, into the future, perhaps

five years ahead, where would you see, ideally, your project being?

What further breakthroughs do you think you may be able to make?

And what difference could it bring?

Perhaps Jérôme, you would like to start off?

Yeah.

I think, ideally, it would be that EDITO

is the reference for oceanography.

Like, always I take this similarity, it's not a good one,

but if you think of video,

you think YouTube, if you think picture, you think Instagram.

Let’s say

you think I need oceanography information and data; you go to the EDITO platform.

And, with the artificial intelligence coming in.

Ideally, what would happen I think, is that now

we start with chatbots that when you ask a question,

you say “I would like to see the temperature in

the Gulf of Naples” for instance, “next week”,

the chatbot will tell you how to get the results.

So, the next phase will be

the assistant, the AI

will give you the result, meaning it is able to launch processes, right?

To use all this data, all these models, to combine them,

to give you the result, and to really give you information.

And this could be, I think, a game changer to move

this amount of data, which is enormous, that is really embedded

by a kind of ocean assistant, an ocean intelligence.

That gives you an answer directly.

Because a very important thing

is, I think, if there is only one thing to remember from this

is that the healthier the oceans are, the healthier our future in fact.

And it's important that we make better choices.

And the digital twin is not just a scientific tool,

but it's like a window to open to people to see the ocean, to understand it.

Yeah.

And better knowledge means better choices and better choices, it

means a better planet.

So better knowledge means better choices.

Do you share a similar vision for your project Wibi?

Absolutely.

We're also very excited with the mobile app including the chatbot,

because when we did the first feedback round

with the farmers for example, they were clearly divided into two parts.

The one that likes visualization

and the other one that doesn't really care about it at all.

They just want to know what is happening.

And chatbot for sure, as he explained, will help a lot because nobody knows what

pH means, nobody knows what carbon content in the soil means.

What does it do with my farm?

With the chatbot they can simplify the terms,

and they can give like a simple explanation of how it works.

And maybe they also can suggest what to do based on how we trained large language

models, as long as we feed it with a lot of information, that will work.

I'm very excited for this one.

And we also will have the first launch, next

month, actually, in November for the mobile app.

November 2025.

I can already see that the chatbot is working quite well in the backend.

So, I'm very excited for this one too.

And for another point,

because our company is like a nonprofit

organization, we try to give open-source data,

especially for researchers, and that we already kind of achieve.

So, we have already published European data freely.

Everybody can access, everybody can take a look at it.

Especially the researchers, they can use it.

Aside from the European scale, we also created the global scale

so we have another platform for the global soil data.

That is also already published; the paper is out, the data is out.

Everybody can use it.

So, we also achieved some things out of this project specifically.

Shout out to my boss, that's their ideology which is very nice.

They are researchers.

They know how hard

researchers find the data sets, and they create that new data set

that is completely open.

You don't have to pay.

It’s a very nice mission, very nice vision for them.

Before I do a short closing,

is there anything that you wanted to mention that you haven't had a chance to?

No, I agree; I learned a lot about Jérôme's projects as well.

It's very interesting for me personally.

It's very nice talk.

Do you have an Apple application or it will be on Android?

Oh, it will not be an Android.

It will be on the web, basically a website.

The way we created the architecture will mimic the device amazingly.

We tried, we were thinking of publishing it

on the App Store, but I don't think it's an option for now.

Of course, because of registration, authentication, and stuff like that.

It's easier on the website,

but when you use it on the phone, it will look like a mobile app for you.

So, thanks very much Wibi and Jérôme for joining us.

Thanks also to our audience for tuning in to this episode.

You can follow us on Spotify and Apple Podcasts

and do check out the podcast homepage on the Cordis website.

Subscribe to make sure that the hottest research in EU funded

science isn't passing you by.

You can also find more information and project examples on the Cordis website.

On the European Commission's Research

and Innovation website and in our online magazine, Horizon.