Since 2015, World Resources Institute (WRI) and Orbital Insight have worked together with a grant from the Generation Foundation to find new applications of computer vision and deep learning that can improve the Global Forest Watch (GFW) project in examining the world’s forests.

Forests worldwide are disappearing at an ever-increasing rate in parallel to growing agriculture and commodity production. Industrial scale plantations that are needed for palm oil production, for example, have been specially harmful to the environment and have led large-scale deforestation. Through the WRI’s expertise in forest monitoring and with Orbital Inisght’s high resolution satellite data analysis, the project has been able to better assess the condition of forests and map palm oil plantations in Malaysia, Indonesia, Cambodia and Colombia, with plans to do the same for Papua New Guinea, Peru, Liberia, Guatemala and Honduras soon.

According to the non-profit organization (WRI), unlike traditional methods for mapping forest cover that rely on how “green” certain pixels are in a satellite image, GFW’s deep learning model looks at the broader context of an image and can differentiate plantations based on their color, size, shape and pattern. Over 600,000 high resolution satellite images are used to teach the system to identify patterns of trees and roads more indicative of plantations. Additionally, to create a model that identifies industrial oil palm plantations, the GFW has been using used supervised machine learning. Thanks to satellite imagery of oil palm plantations alongside non-plantation areas such as cities, bodies of water and natural forest, the model has been taught to determine what a plantation looks like. The WRI team manually labeled over 3,000 satellite images that covered a diverse set of plantations and geographies in order to provide the model with sufficient training data.

The GFW also uses a technique called convolutional neural network (CNN) to train their algorithm to accurately distinguish different plantations. Using the training dataset annotated by manual labour, the algorithm is able to identify industrial oil palm plantations based on clues such as texture and pattern of trees and roads. Finally, the WRI team evaluated the performance of the artificial intelligence models by comparing the predictions to ground truth markings.

The GFW project allows for comprehensive monitoring of deforestation with a specific focus on changes of landscape triggered by human commercial activity. Any government, farmer or member of civil society can monitor a selected area of the world for free thanks to the GFW.  As such, the project contributes to the SDG 12 (Responsible Consumption and Production) as it casts a spotlight on the overextending activity of palm oil plantations, among other, the SDG 13 (Climate Action) and the SDG 15 (Life on Land).