Copy of DSC00026.jpg

Abundance Maps of Non-Timber Forest Products

Non-timber forest products (NTFPs), such as Brazil Nuts are the bedrock of food security and economic independence of traditional communities across the Amazon Rainforest. While this method of agroforestry promotes productive standing forests, it is difficult to scale, improve productivity and coordinate logistics. However the global demand for natural products is increasing, and the Amazonian people see an opportunity for strengthening their economic resilience sustainably.

Luciano Quilombola collecting a Brazil Nut seed pod - Cachoeira Porteira - PA - Brazil,

Bioverse, with the support of UNICEF’s Innovation Fund have built and field-tested technologies that help vulnerable communities protect their natural ecosystems and contribute towards improving their livelihoods. In 2020 and 2021 the Bioverse team has worked alongside the Kayapo and Quilombola people, who are inhabitants of the Amazon rainforest, to collect ground truth data for important NTFPs.

 

The territory of these people is under threat from logging, mining and ranching. The Forest Abundance Maps project helped them draw more economic value from their forests in a sustainable way. This built greater economic resilience in the participating communities and also created new economic incentives for keeping the rainforest standing.

The engine of value at the center of the project is the Brazil nut tree (Bertholletia excelsa), a native tree species that can yield $1500 USD of value per year (based on market price in the USA) from the nutritious nuts that it produces. Bioverse started with a hypothesis that artificial intelligence (and specifically machine vision) could locate Brazil nut trees that the local harvesting community was not previously accessing. Locating these trees directly increases the per hectare real economic value of the standing forest to the community, making it easier for them to resist pressures from logging, mining or ranching. 

 

The target area for the pilot was 400 Sq kilometers and the community identified 680 trees during our baseline GPS-powered survey. Bioverse’s work with UNICEF Innovation Venture Fund successfully identified 9240 additional trees—an increase of 1800%.

Figure 2 - Satellite image and Brazil Nut spatial distribution from unsupervized ML based classification.

The project relied upon three different layers of technology. First, Bioverse used multispectral high resolution satellite imagery for a detailed picture of the forest canopy (figure 1). Second, was higher resolution canopy images taken from drones (figure 2). This drone layer was necessary to validate training inputs for the classification algorithm and will not be required for further detection of Brazil nut trees. Third, was handheld mobile technologies, initially to provide GPS information about known tree locations and subsequently to provide harvesters with in-hand Forest Abundance Maps to help them locate trees that they had not previously harvested from.

 

For the community, these app-based Abundance Maps can be a true game changer. We work alongside the formal cooperatives of the participating communities to ensure that sensitive information (like tree locations) is only shared when, how and with whom the community decides. As the harvesting community grows accustomed to using these mobile tools, we at Bioverse can add to the power and the capacity of the app.

One-by-one, machine vision trained alongside beneficiary communities can help locate resources that will bring prosperity to the community. The mission of Bioverse is to protect biodiversity around the world. A core value of the company is to use our technology to identify only those resources that are valuable when they are left standing in a healthy web of life and to only locate those resources for and with the benefit of the communities who take responsibility for protecting the rainforest.

While these tools and services can help to improve the livelihoods of the harvesters in question, it is not yet viable for Bioverse to support its operating costs (or its R&D) from user fees. Therefore  Bioverse is seeking to partner with established companies that are already part of the non-timber related forest products supply chain. 

Figure 2 - UAS borne image, Brazil Nut ground truth routine. Cachoeira Porteira - PA Brazil 

Our team also recognizes the sensitivity of data obtained via the operation of the Abundance map projects which preclude building business models around sale of data. So for the next couple of years, Bioverse hopes that conservation-focused philanthropists, impact investors and the emerging natural products industry will fund product development and scale-up. When the Abundance Maps support multiple species and languages and once the communities of the Amazon are in a less economically precarious state, the possibility of opening the solution for paid users in private property may become a possibility.

Bioverse is able to support this work and technology through funding gaps because of the profitable work that it undertakes in the ag-tech space, specifically on biodiversity monitoring for pest and diseases control. In the near term, Bioverse will seek funding to develop algorithms that identify additional productive tree species like Cumaru (Dipteryx odorata), Andiroba (Carapa guianensis), and Copaiba (Copaifera langsdorffii), which are growing in the same areas. Bioverse also seeks funding to scale up the creation of their maps for other indigenous communities who draw their livelihoods from the forest. This approach is not limited to the Amazon Rainforest, as we are in conversations with INGOs in the Democratic Republic of Congo, who are committed to addressing deforestation in Central Africa while bringing greater prosperity to forest-dwelling or forest-adjacent communities.

 

In line with the digital development principles, Bioverse builds these products in close collaboration with the communities who use them, while helping them to protect their data and privacy. Bioverse is also helping to build the commons by making the machine learning aspects of the technology open source, which can benefit researchers, governments and other civil society organizations working with forests.