Monitoring Mangrove Biodiversity with COBEC

It’s an exciting time to be involved in biodiversity monitoring, especially in the context of marine restoration! Over the past year, we’ve been working with COBEC, a Kenya-based nonprofit, to restore mangrove forests in Marereni as part of one of the first active marine biodiversity credit projects in the world. Led by Seatrees, this project issues Biodiversity Blocks—a new type of biodiversity credit—and monitors the broader ecosystem to detect changes in biodiversity that result from mangrove restoration.

What’s truly special about this project is how it ties community-led restoration efforts directly to rigorous biodiversity monitoring. Over 600 community members have already planted more than 300,000 mangrove trees. These individuals are not only restoring degraded forests; they are also growing seedlings in local nurseries to generate income. Alongside the planting, the team is monitoring biodiversity metrics to track how restoration is impacting the broader ecosystem, with a strong focus on building the scientific capacity of local researchers and communities.

There are also several complementary benefits built into the project. Local community scouts patrol mangrove areas alongside government rangers to help protect the broader forest from illegal logging. The project is also providing 300 scholarships over 10 years, and supporting woodlots for non-mangrove tree species—providing income alternatives to reduce pressure on mangrove forests.

Tech-Enabled Biodiversity Monitoring: The Promises and Pitfalls

We’re living through a period of incredible innovation in biodiversity monitoring. New tools like drones, eDNA analysis, photogrammetry, and bioacoustics are opening up new data collection methods that would have been unimaginable a decade ago. If that wasn’t enough, a generation of new AI tools are streamlining analysis workflows and helping us process complex data in real time.

But these technologies come with their own set of challenges. Hardware can be expensive. Data storage is often limited in remote settings. And while AI offers lots of promise, it also has limitations—especially when it comes to precision. Many AI models still struggle with confidence estimation, generalize too broadly, or simply haven’t been trained on region-specific datasets. To complicate things further, the training and technical support needed to deploy these tools effectively is often overlooked, particularly in under-resourced environments.

Those are just a couple reasons why there’s still enormous value in combining modern tools with traditional approaches. Going into the field, making in-situ observations, using field guides, and leveraging local knowledge all remain essential to building a clear and contextual understanding of biodiversity.

Learning from the Field: Identifying Invertebrates in Kenya’s Mangrove Forests

One of the biodiversity metrics we're closely monitoring in the Marereni mangrove forests is invertebrate species diversity. Mangrove invertebrates – think crabs, snails, barnacles, and oysters – play a huge role in making these ecosystems function. They’re critical herbivores, they help decompose organic material, and many are a vital food source for other animals in the forest. Fiddler crabs alter the sediment through burrowing, enhancing nutrient cycling. Oysters growing on mangrove roots filter toxins out of the water. Mud crabs form the foundation of a fishery that supports local livelihoods. 

In other words, these little creatures punch well above their weight!

Invertebrates can also provide a window into ecosystem health. We're trying to understand whether certain species are indicators of forest condition—found only in healthy mangroves, for instance—and how species composition shifts in response to restoration. But to do this properly, we need accurate species identification.

Local scientists at COBEC are our eyes and ears in the mangrove forest. Led by Program Coordinator Julius Sila, COBEC’s team of researchers regularly ventures into the field to count and identify all observed invertebrates within predefined quadrats. 

COBEC’s initial approach was to use Google Lens and ChatGPT for species identification based on these photos. While good for an initial first pass, this approach came with big limitations. These AI tools gave no clear measure of confidence for the identifications, and often mistook local species for visually similar ones found elsewhere in the world.

This isn't surprising—many species are incredibly difficult to distinguish without looking at one or two particular traits (which is why field guides often direct your attention to specific features, rather than overall size, color, or general appearance). And unless an AI tool has been specifically trained on these distinguishing traits, it’s likely to get it wrong.

So we didn’t throw in the towel on AI; we just got smarter about which tools we used. That’s where iNaturalist came in.

With iNaturalist, we gained multiple advantages: it provides a percent confidence estimate for each identification, allows us to constrain results geographically, and opens up access to a massive global community of expert citizen scientists who can weigh in on IDs. In many ways, it's like crowdsourcing science—and it's been a game changer for us.

Another bonus: the photos taken by COBEC staff can be revisited later, not just for validation but for training and learning. It’s an ongoing feedback loop that enhances local capacity and helps ensure the data we collect is as rigorous and transparent as possible.

We plan to continue integrating iNaturalist into our biodiversity monitoring framework. Beyond improving species ID accuracy, it fosters a stronger connection between local scientists and the global research community. 

As we think about adding more technologies—like drone-based photogrammetry or passive acoustic recorders—we’re also committed to pairing these tools with human insight. Whether that’s from community members with lived experience, local ecologists, or citizen scientists, their input helps us interpret what the data really means.

Lessons Learned

A few key takeaways have emerged from this experience:

1. Don’t ditch the basics. New technology is powerful, but it’s not infallible. Fieldwork, local knowledge, and critical thinking are still foundational. If an AI tool gives you a result, you should always ask: Does this result make sense, based on my understanding of this ecosystem’s local context?

2. Conservation is a process. Introducing new tech takes time. Committing time and resources to training and onboarding are essential—especially in remote or resource-constrained areas. You can’t just hand someone a drone and say, "Go." You need to build systems of learning and support.

3. Accessibility is the future. Making technology easier to access pays huge dividends. Whether it’s species ID with iNaturalist or low-cost hardware for collecting data in the field, democratizing tools empowers a wider range of people to participate in science.

A related example worth highlighting is our partnership with Samsung and the Scripps Institution of Oceanography for monitoring coral reef restoration. To create an alternative to expensive DSLR cameras (commonly used by academic researchers), we’re using Samsung phones to photograph coral reefs and generate high-res 3D models. It’s pushing the limits of photogrammetry while simultaneously making the technology more accessible. Our goal is to make it so any group with a smartphone can document and track reef recovery over time.

That’s accessibility in action—and it's exactly the kind of innovation we need more of in the conservation space.

Working with COBEC on species ID has been a rich and rewarding experience. It’s reminded us that while tools and tech are important, it’s people—their insights, experiences, and curiosity—who make the work meaningful. And when we combine local leadership with global connectivity, powerful things can happen for both ecosystems and communities.