Road networks stretching into the forests of the Congo Basin: satellites and AI help monitor

The Congo Basin rainforests are the second largest tropical forests in the world, store large amounts of carbon and host high levels of biodiversity. Historically these forests have remained largely intact, but more recently road development has posed a significant threat.

A new study uses satellite imagery and artificial intelligence to map forest roads in the Congo Basin in greater detail than ever before, providing vital new information for forest conservation and management.

Most of the Congo Basin’s road network has been built to facilitate industrial selective logging, which harvests mature, valuable timber while leaving most of the forest intact – a relatively sustainable practice that brings significant benefits to the local economy.

However, building temporary roads to facilitate access to forests leads to large amounts of carbon emissions and has long-term impacts on forest ecosystems. Another negative impact is that roads open up remote forest areas, which can lead to illegal hunting, mining and agricultural development.

Advanced detection methods to visualize any road

Timely and transparent information on where and when road networks are being constructed is crucial for sustainable logging practices and forest and biodiversity conservation. Current forest road maps rely on labor-intensive manual digitization of satellite imagery and are therefore highly incomplete or outdated.

Researchers have developed a method to automatically detect forest roads from satellite imagery. The study was recently published in the journal Remote Sensing of Environment. The study demonstrates an advanced detection method based on a combination of deep learning techniques and high-resolution optical and radar satellite imagery.

“Optical sensors provide detailed images in clear weather, while radar sensors are able to ‘see through’ clouds during long rainy periods in the tropics. This way, even the narrowest, busiest road sections can be pinpointed with precision,” says Bert Slagger, a postdoctoral researcher at Wageningen University and lead author of the study.

The algorithm processed and analyzed hundreds of thousands of satellite images on Google Earth Engine, a cloud-based processing platform.

“We’ve mapped about 50,000 km of road development in the Congo Basin over the past five years,” Slagter explains, “and this inventory will be continually updated for years to come. We also aim to apply this methodology to the Amazon and Southeast Asian rainforests.”

The Congo Basin forest roads map reveals some striking trends. The western countries of the region (Cameroon, Equatorial Guinea, Gabon, and the Republic of Congo) have undergone extensive road development in recent years as a result of the extensive logging activities permitted in these countries. Moreover, almost a quarter of the Congo Basin forest roads are clearings in pristine forests with no signs of previous human impact.

Towards tree harvesting that minimizes damage to the environment

Further explaining the importance of the study, senior co-author Kurt Fesenmeyer, a forest spatial data scientist at The Nature Conservancy, explained, “Integrating this satellite monitoring system into forest management operations will represent an important step in balancing profitable logging operations with minimizing environmental damage.”

“This could include promoting reduced climate impact logging (RILC) practices, such as improving road design, preventing road expansion into protected areas, and properly closing abandoned roads to avoid further human impacts on remote forests.”

Moreover, because roads are a strong indicator of human impact on natural systems, a continuously updated roadmap is crucial for developing large-scale forest conservation strategies. A detailed, up-to-date roadmap would significantly improve common conservation indicators such as “human footprint” and “intact forest landscapes,” enabling governments and NGOs to better protect critical areas from deforestation and forest degradation.