Over the last decades, more than half of Indonesia’s tropical rainforests have vanished, mainly due to palm oil cultivation, gold mining, logging and creation of arable land. To preempt further deforestation, WWF, Deloitte and a consortium of public an private entities have developed a tool that predicts deforestation through automated assessment of spectral satellite imagery and other geospatial data. The first phase of this pilot, part of the WWF Forest Foresight programme, ran from 2020 onwards in Kalimantan (part of the island of Borneo).
Because of the vast amounts of data our AI-model needed to process, we migrated the initial prototype to a cloud-native, scalable configuration. This allowed us to cut data processing time from twelve to four hours and significantly speed-up prediction generation for the entire island of Borneo (which harbours 425,124 square kilometers of rainforest).