CanopyWatch – Enhancing Deforestation Monitoring and Conservation in the Congo Basin using Machine Learning

Challenge Started!

Enhancing deforestation monitoring and conservation in the Congo Basin rainforest through improved detection algorithms, expanded deforestation types, and increased frequency of satellite image analysis. In this 8-week challenge, you will join a collaborative team of 50 AI engineers from all around the world.

The problem

The Congo Basin is home to the world’s second-largest tropical rainforest, spanning 2.5 million square kilometers over six countries. It is also the world’s last tropical carbon sink, as well as the home of over 1,000 threatened species. Environmental efforts receive up to €300 million per year, but deforestation, biodiversity loss, and carbon emissions continue unabated. At the same time, the Congo Basin provides food, medicine, water, materials, and shelter for over 75 million people.

Deforestation poses a severe threat to the Congo Basin rainforest. This Omdena-Project Canopy challenge aims to address the critical problem of tree cover loss in the Congo Basin rainforest and provide actionable insights for local communities, civil society, and other stakeholders and policy-makers.

Unlike other tropical regions, deforestation in the Congo is complex and multi-causal. It is more akin to ‘death by a thousand cuts’ than the outright land-grabbing and clear-cutting seen in the Amazon and Southeast Asia. Slash-and-burn agriculture, logging, mining, and industrial agriculture – both legal and illegal – all contribute to the ongoing loss of forest cover. Moreover, these causes reinforce one another in complex ways. For example, new logging roads open up access to previously untouched primary forest, which is then exploited for informal agriculture, mining, poaching, and other activities.

Platforms that use remote sensing to detect deforestation already exist, but they do not detect deforestation by type. This is needed to understand the complexity of the threats facing the Congo Basin rainforest. Decision-makers cannot set effective policies if they do not understand the causes of deforestation or the extent to which different drivers conspire to compound the losses.

At the same time, millions of people depend on the rainforest to provide them with food, medicine, and shelter. In many cases, they hold legal title to their land and are deeply invested in maintaining the integrity of their rainforest homes. However, they currently lack the ability to know where – and why – deforestation is happening within their boundaries. CanopyWatch is designed to bridge this gap, by providing forest communities – and the local NGOs that serve them – the ability to be notified of new deforestation activity, in as near-real time as possible, so that they may intervene effectively.

Beyond addressing the immediate and ongoing causes of deforestation, CanopyWatch will provide a valuable foundation for generating many additional actionable insights:

Carbon Emissions and Climate Change: The Congo Basin rainforest plays a crucial role in mitigating climate change as it acts as a carbon sink, absorbing significant amounts of carbon dioxide from the atmosphere. However, deforestation and forest degradation release carbon emissions, contributing to global warming. When combined with carbon stock estimation, the insights provided can assist decision-makers in identifying which parts of the forest are both the most carbon-sequestering and under the greatest threat.
Biodiversity Loss: The Congo Basin rainforest is known for its exceptional biodiversity, hosting numerous plant and animal species, many of which are found nowhere else on Earth. However, the region is experiencing significant biodiversity loss due to various factors such as habitat destruction, poaching, and illegal wildlife trade. When combined with biodiversity maps, the risk to threatened and endemic species can be better understood, providing insights and recommendations to decision-makers on how to protect and preserve the diverse ecosystems and species in the Congo Basin.

The impacts of the project can be significant:

Empowering Local Communities: The project can assist local communities in addressing illegal logging and mining within their forests. When communities have accurate and timely data, they can be much more self-sufficient. In turn, this self-sufficiency contributes to community stability and resilience.
Preservation of Biodiversity: With actionable insights, decision-makers can implement conservation measures that help protect the unique and diverse species in the Congo Basin. Preserving biodiversity has ecological, economic, and cultural benefits and contributes to the overall health and resilience of the rainforest ecosystem.
Climate Change Mitigation: By addressing deforestation, the project can help reduce carbon emissions from the Congo Basin rainforest. This contributes to global efforts in mitigating climate change and meeting targets set under international agreements like the Paris Agreement. Preserving the rainforest’s role as a carbon sink is crucial for maintaining a stable climate and minimizing the impacts of climate change on a global scale.

In summary, this project aims to address deforestation, biodiversity loss, and carbon emissions in the Congo Basin rainforest. By providing actionable insights, decision-makers can implement measures that promote local communities, preserve biodiversity, and contribute to climate change mitigation.

The project goals

The main goal of the project is to create the next iteration of CanopyWatch, an application that combines satellite imagery and machine learning to detect deforestation in the Congo Basin rainforest, by type of deforestation (eg, logging, slash-and-burn, industrial agriculture).

The objectives of this Omdena-Project Canopy Challenge are:

Improve precision & recall of the existing, core ML algorithms used to detect logging and slash-and-burn deforestation.
Expand the types of deforestation to include, in preferential order, industrial agriculture, mining, and/or habitations (>80% precision and recall).
Evaluate and possibly improve the algorithms used to secure cloud-free optical band imagery from Sentinel-2.
- Evaluate and possibly improve the process by which cloud-free maps are assembled.
Increase the frequency of complete Area of Interest image pulls.
- Preference is annual pulls, from the beginning of Sentinel-2 service (2016).
- Stretch goal: Create a pipeline that will automate the process of regular imagery pulls and assembly of cloud-free images.
Recommend a new front-end display service.
Increase final accuracy with post-processing workflows.
- Elimination of false positives by omitting ‘orphan’ chips that have no contiguous/near neighbors.
- Use of Open Street Map metadata to distinguish commercial roads from logging roads and exclude the former from prediction results.
- Stretch goal: Create a pipeline that will automate the inference and display of future imagery pulls.
Stretch goal: Integrate SAR (radar) imagery from Sentinel-1 to augment the array of the optical band (RGB, NIR, NDVI) imagery from Sentinel-2.

Why join? The uniqueness of Omdena AI Innovation Challenges

A collaborative experience you never had in your working life! For the next eight weeks, you will not only build AI solutions to make a real-world impact but also go through an entire data science project lifecycle. This covers problem scoping, data collection, and preparation, as well as modeling for deployment.

And the best part is that you will join a global and collaborative team of changemakers. Omdena AI Challenges are not a competition or hackathon but a real-world project that will take your experience of what is possible through collaboration to a new level.