“Mauritius has the potential to become a regional reference for AI-driven fisheries management”

Dr Hidekazu Kasahara, Director of Ocean Eyes, Japan 

Mauritius is positioning itself at the intersection of fisheries, ocean intelligence and artificial intelligence through a collaboration between local startup Ocean Economy AI Lab and Japanese technology company OceanEyes. Their objective is to transform satellite imagery, oceanographic modelling and fishermen’s practical knowledge into predictive tools capable of identifying potential fishing grounds with greater precision. In this interview with Bizweek, Dr Hidekazu Kasahara, consultant to Ocean Economy AI Lab and representative of OceanEyes, explains why Mauritius already possesses important foundations for AI-driven fisheries, how predictive models are developed, and why human judgement will remain central despite advances in technology. He also discusses the economic implications for fishermen, particularly in reducing fuel costs and operational uncertainty, while outlining how Mauritius could emerge as a regional hub for ocean-data innovation in the Indian Ocean.

R.V.

Mauritius is often viewed as a small fisheries market. Yet, you suggest that the country already possesses important foundations for AI-driven fisheries. What strengths do you see here, and what remains to be developed?

Before travelling to Mauritius, I reviewed several documents prepared by the Mauritian government and the European Union. What stood out immediately was that Mauritius has already begun building an ecosystem around ocean observation data and digital applications intended for fishermen.

This is important because it means the country is not starting from the beginning. The authorities already recognise the strategic value of oceanic data, and that awareness is the first step towards more advanced AI-based fisheries systems.

 

“A Potential Fishing Ground model successfully developed in Mauritius could also serve countries across the region”

 

Across Japan, parts of Europe and a limited number of other countries, fishermen are already using satellite and oceanographic information to identify productive fishing grounds. Even so, so-called “smart fishing” remains relatively new at a global level.

Mauritius therefore already possesses a significant advantage: an emerging culture of ocean-data literacy. That is a strong platform on which more sophisticated technologies can be developed.

So, you believe Mauritius already has a sufficient level of ocean-data literacy to move towards AI applications?

Yes. The groundwork has clearly been laid through government initiatives and the introduction of ocean-data applications.

The next stage requires leadership and coordination to encourage adoption within the fishing sector. In many countries, including Japan, fishermen tend to be cautious about new technologies. Adoption can therefore be slow.

“Across Japan and parts of Europe, fishermen are using satellite and oceanographic information to identify productive fishing grounds”

What is interesting in Mauritius is that organisations such as Ocean Economy AI Lab already understand the long-term potential of these solutions and can help drive that transition. That leadership is extremely valuable.

In Japan, progress has largely depended on the involvement of the central government and local authorities. Even today, many fishing communities remain conservative, partly because of ageing demographics within the industry.

Is that conservatism still visible in Japan’s fisheries sector today?

It varies from one community to another. In more industrialised fishing regions such as Kochi or Yaizu, particularly among skipjack fleets, data-driven fishing methods are already widely used, and some operators have integrated AI tools into their activities. In southern Japan, including areas such as Fukuoka, local governments have also encouraged smaller and medium-sized fishing operators to adopt these technologies.

How long has this transition been underway?

For more than a decade.

Over that period, have these technologies significantly increased fish catches?

Not necessarily in terms of total volume. Japan’s fishing population has been declining for demographic reasons. However, the technology has substantially improved operational efficiency. Fishermen spend less time searching for fish, which directly reduces fuel consumption and operating costs. In practical terms, they can identify promising fishing grounds much faster than before.

Ocean Economy AI Lab aims to convert highly technical ocean data into practical tools for fishermen. What is the greatest challenge in making that data operationally useful?

The first and perhaps most difficult challenge is obtaining sufficient high-quality data. Without reliable datasets, even the most advanced AI models cannot function effectively. Gathering this information requires coordination, management and institutional support. In Japan, local governments often facilitate access to fisheries and oceanographic data, and we collaborate closely with authorities in regions such as Tokyo, Chiba and Shizuoka. In countries such as Indonesia, however, the process becomes far more complicated because the country is geographically fragmented and fishing practices differ significantly from one island to another.

Once sufficient data has been collected, we can train deep-learning systems. But even then, localisation remains essential. Fishermen in different regions require different forms of information. Japanese fishermen, for example, often want much more than a simple Potential Fishing Ground map. They also want access to temperature profiles at different depths, current movements and other oceanographic indicators. They trust their own experience and use our systems as decision-support tools rather than as absolute answers. 

In some Southeast Asian countries, fishermen may rely more directly on the probability maps themselves because ocean-data literacy is less developed. Even so, weather conditions and sea-state information remain essential. A forecast showing a productive fishing area is of little value if conditions are unsafe. This is why safety information must always accompany predictive fishing tools.

AI systems depend heavily on data quality. What types of local Mauritian data will be most critical for building a reliable forecasting model?

Our datasets are generally built around four core variables: time and date, geographical coordinates, fish quantity and fish species. Beyond that, depth information is extremely important because many species occupy different water layers depending on temperature and environmental conditions. We also need information on fishing methods and equipment because these factors influence catch data. In addition, local ecological knowledge is invaluable. Fishermen’s observations, behavioural patterns of species and accumulated field experience all contribute to improving forecasting accuracy. For example, if a species is known to inhabit waters below a specific depth, that information becomes highly relevant when training the model.

Does Mauritius already possess enough data to begin building such systems?

As a starting point, we are already using public datasets provided by the Indian Ocean Tuna Commission.

Can these datasets be used commercially?

Yes. We are currently using them in relation to large tuna species.

How exactly does Ocean Eyes combine satellite imagery, ocean models and fishermen’s expertise to predict fishing grounds?

The system relies on three principal technological components. The first is an ocean numerical model capable of calculating sea conditions such as temperature, salinity and currents at different depths. These calculations require significant computing power and are generally carried out using supercomputers. Historically, this capability has been concentrated within navies, research institutions, universities and meteorological agencies. Only a very small number of private companies worldwide possess this expertise. 

The second component is artificial intelligence itself, particularly deep-learning systems capable of estimating the probability of fish presence. The third component involves satellite-data analysis. We combine ocean-condition forecasts with satellite observations – including chlorophyll concentration levels – and integrate them into our AI models. Local fishing knowledge is then incorporated into the training process. By combining these three layers, we can generate Potential Fishing Ground forecasts.

It is an extremely sophisticated process.

Yes, it is.

But the methodology is already used in Japan?

Yes. The system is already functioning.

Which species have you worked on so far?

We have developed models for between five and seven species. Tuna and skipjack are among the most important. We have also worked on squid, which is highly valued in Japan, as well as Alaska pollock, mackerel and several local Japanese species. We are now exploring lagoon species, although forecasting becomes much more difficult in shallow coastal waters.

So, the technology performs best in deep seas.

Yes. It is particularly effective for species such as tuna and skipjack that operate in deeper waters. We already have extensive operational experience in Japan and Indonesia.

Your first commercial product is a Potential Fishing Ground mapping tool. In simple terms, what will fishermen see?

The map displays sea-surface temperature as a background layer. Areas where temperature changes rapidly are highlighted because fish often concentrate in those zones. The system then identifies areas according to probability levels. Red zones indicate the highest probability of productive fishing grounds, while blue zones represent lower probabilities. For Japanese fishermen operating hundreds of kilometres offshore, this information is extremely valuable because it reduces search time and fuel consumption.

AI can assist decision-making, but fishermen still retain ultimate responsibility at sea. How do you view the relationship between technology and human judgement?

The system provides probabilities, not certainties. The final decision always rests with the captain because many factors must still be considered, including weather conditions, tides, distance, engine reliability and overall safety. Technology can reduce uncertainty, but it cannot replace professional judgement and experience.

Among the benefits you mentioned – reduced uncertainty, lower fuel costs, time savings and improved safety – which is currently the most urgent for fishermen? 

Fuel costs are probably among the most urgent challenges globally. Reducing unnecessary navigation time translates directly into lower operational costs. For vessels operating far from shore, this can make a major economic difference.

The first phase of the project focuses on tuna species such as bigeye, yellowfin, albacore, skipjack and bluefin. Why are tuna species the logical starting point?

There are several reasons. First, we already possess substantial operational experience with these species in Japan and Indonesia. Second, tuna fisheries represent high-value commercial activities. The additional revenue generated through improved efficiency can justify investment in these technologies.

There is also a technical reason. Forecasting accuracy tends to be higher offshore than near coastlines, because satellite observations close to land are more affected by clouds, rainfall and atmospheric interference. Since tuna species generally inhabit deep waters, they are particularly well suited to this type of forecasting model.

Looking ahead, how could Mauritius position itself as a regional centre for ocean-data and AI-driven fisheries innovation?

The Indian Ocean is shared by many countries, but it is ultimately one interconnected oceanic system. If OceanEyes and Ocean Economy AI Lab successfully develop and deploy a strong Potential Fishing Ground model in Mauritius, the technology could eventually serve not only Mauritius, but also countries across the region. Mauritius therefore has the potential to become a regional reference point for ocean-data applications and AI-driven fisheries management.