Site icon Conservation news

Low-tech challenges to high-tech forest monitoring: lessons from Ugandan parks

  • Remote sensing technology can provide useful intelligence to park managers but must be combined with an understanding of its limitations, as well as the tools and training needed for its use.
  • An assessment by park rangers in Uganda of satellite image-based deforestation alerts found that the alert locations at 30 m x 30 m resolution were sufficiently accurate to support reserve management.
  • The near real-time alerts of likely deforestation could make forest patrols more efficient and effective, but rangers must still have proper training, incentives, and resources to properly integrate alerts into their regular functions.

A pair of NGOs teamed up with rangers in four parks in Uganda to assess the potential for cutting-edge forest monitoring technology to support protected area management through early detection of deforestation.

The Jane Goodall Institute (JGI) and World Resources Institute (WRI) partnered with the Uganda Wildlife Authority (UWA) to train protected areas rangers in the Kibale Conservation Area in western Uganda on how to assess forest threats using a new system of near real-time alerts of deforestation events. The partners hope to determine how the information from these alerts can help slow deforestation within and surrounding protected areas (PAs). In an email exchange with Mongabay-Wildtech, they shared some of the surprisingly low-tech challenges and lessons they learned in rolling out new technologies in the field.

Forests of the Kibale region are home to Eastern chimpanzees and 12 other primate species. Photo credit: Jane Goodall Institute Kibale National Park also supports more than 375 bird species, including the emerald cuckoo Photo credit: George Powell


A JGI team trained over 50 protection staff in workshops conducted between November 2016 and May 2017 to use the online deforestation alerts, called GLAD alerts, produced by the University of Maryland and Global Forest Watch (GFW). The automated, weekly-updated GLAD alert system identifies likely sites of tree cover loss.

Bringing automated satellite image technology to the field

Kibale Conservation Area includes Kibale National Park, Toro Semliki Wildlife Reserve, Katonga Wildlife Reserve and Semuliki National Park. Kibale National Park alone supports 350 bird and 70 mammal species, including 13 primates, within more than 75,000 hectares of deciduous and evergreen forest.

It is one of Uganda’s most important forests for biodiversity, yet park rangers must constantly monitor the forest to address threats from poachers, loggers, and others illegally harvesting forest resources.

Deforestation along the boundary of Bwindi Impenetrable National Park, Uganda. Photo credit: Rhett A. Butler


Researchers have used satellite technology previously to detect forest loss on the outskirts of Kibale National Park and found that the growing human population around the park potentially threatened its integrity.

However, remote sensing technology that could help monitor the forest has not reached rangers in the Kibale protected areas (PAs).

Like managers of many of the world’s PAs, Kibale forest managers have lacked the capacity and timely intelligence to most effectively patrol and enforce the reserves’ boundaries.

During the training, the UWA ranger and monitoring teams tested the potential of the GLAD alerts to make the patrolling of large, often remote forested areas more efficient and effective. They first learned how to assess tree cover loss for their target areas using the GFW platforms, as well as interpret and manage the GLAD alerts and other types of forest loss GFW data.

The patrol and monitoring teams also learned how to collect evidence of forest change in the field to verify the alerts and the incorporate the alert and field validation data into an ecological monitoring program.

Challenges for deploying technology in the field

During the exercises designed to assess the technology’s utility for PA management, JGI and UWA identified several challenges to and conditions for successful use of the various forest monitoring tools in the field.

Training lead Lilian Pintea of Jane Goodall Institute explains to UWA staff how to access tree cover loss data in Global Forest Watch during a training workshop. Photo credit: Jane Goodall Institute

1: Limitations of technology

“One of the first questions that arises when applying remote sensing technology to support conservation action is, ‘Is it accurate enough to inform local action?’” Rachael Petersen, Impacts Manager for WRI’s Global Forest Watch, said.

To help assess this question with respect to the GLAD alerts, the UWA staff headed to the field to investigate 214 recent GLAD alerts indicating tree cover loss inside and just outside the boundaries of the four PAs. There, trainees learned how to use a Global Positioning System (GPS) unit to locate the sites of GLAD alerts and then to use the Open Data Kit (ODK) mobile app to collect and report field data to validate the forest clearing identified by the alerts.

Brenda Mirembe of JGI-Uganda instructs UWA rangers during a training in Kibale National Park. Photo credit: Jane Goodall Institute

GLAD alerts detect tree cover loss at the individual unit of a 30 x 30 meter satellite image pixel. Most of the alerts the Kibale rangers investigated had evidence of tree cover loss. However, the rangers found that many of the clearings reported in the alerts were not inside the exact forest loss pixel but in close proximity to the alert. Therefore, concluded Lilian Pintea, lead trainer and JGI’s Vice President for conservation science, “Real-time alerts like GLAD should be used as a possible indicator of illegal human activity, but rangers should search for evidence of new threats in the general area, instead of limiting themselves to the confines of the pixel.”

Moreover, he added, “some of the clearing identified by the alerts was not anthropogenic (human-caused) but caused by floods or elephants knocking down trees.”

Elephants eat and knock down trees, which in some cases may open up vegetation and resemble forest clearing in the small areas of satellite image pixels. Photo credit: Sue Palminteri

Regardless, said Petersen, the managers found the information in the alerts useful enough to encourage their rangers to investigate the alerts, even if some damage was not caused by humans.

The park managers in Uganda preferred to view all possible areas of forest change, including possible false alarms, before determining how to prioritize resources for field investigations and patrols. This strategy may be practical in large, poorly studied parks. The alerts detected and located two new clearings in Kibale National Park where the UWA rangers were previously unaware of illegal activities, and the rangers used the location data to find these areas and prevent additional deforestation.

One of two sites of illegal logging and encroachment in Kibale National Park located with the help of GLAD alerts. The DigitalGlobe 2007 and 2017 satellite images stored in the Open Data Kit data set confirm the location and extent of forest loss from logging and encroachment indicated by the alerts. Image credit: Jane Goodall Institute/World Resources Institute

2: Remote study sites with limited resources

Remaining challenges were strikingly low-tech and could apply to introducing field teams to any new monitoring technology. For example, Petersen said that some ranger outposts lacked functioning solar batteries to charge the mobile devices they would need to view the weekly forest loss alerts.

Not surprisingly, these outposts, like those in many PAs, had very limited internet access, which made it difficult to load and take full advantage of the GFW website where the rangers first access the alerts. Once they subscribe, field teams also need email access to receive each new alert.

Lastly, some of the sites identified by the alerts occurred in very remote parts of the park that rangers could not access. An inability to get around hinders the effective use of any sensor or other advance warning technology, especially in zones of steep topography or violent conflict.

3: Human capacity and interest

Effectively integrating the intelligence gained from alerts and other remote sensing technology into standard operating procedures requires that rangers be motivated to try the new tools and receive the training and resources to do so.

Uganda Wildlife Authority rangers learning to validate forest loss on the ground using GPS and Open Data Kit technologies. Photo credit: Jane Goodall Institute

For example, in Kibale, Petersen said, “Trained rangers were often transferred to other parks, requiring a continuous refresher training for their replacements. Some rangers had little interest or incentive to visit alert areas. Others struggled to navigate in the field using GPS devices.”

Pintea added, “Some park managers were skeptical of the need for [near real-time alerts] in areas with relatively few threats, [whereas] others felt that habitat loss at any scale must be addressed and that the deforestation alerts could be an important tool to help them do so. As one ranger said, ‘Even if you do not have fires often in your house, you should not throw out your fire-fighting equipment.’”

By the end of the training, he continued, the UWA managers were interested in both incorporating the GLAD system into their operations and scaling it up across the country’s protected area network.

The JGI-UWA collaboration in the four Kibale protected areas suggests that ensuring adoption requires time to build trust in the accuracy and utility of the alerts, as well as increasing the capacity and resources to use them.

A L’hoest’s monkey in Uganda appreciating additional forest protection support. Photo credit: George Powell