Assessing land use following tropical deforestation: combining remote sensing and deep learning
This PhD research explored the use of deep learning methods to gain an understanding of the proximate drivers of deforestation in the tropics. The use of satellite imagery and deep learning to identify land-use offers promising improvements over traditional methods, enabling the generation of continental scale, up-to-date, maps. This work provides strategies to facilitate automated land-use mapping and shows through case studies how to achieve that goal. Specifically, the deep learning methods were developed to classify land-use following deforestation and test the application of the method over national and continental scales mapping. This research contributes to large-scale forest and land-use change monitoring in the context of climate change, and the Sustainable Development Goals. It is also useful for companies that want to measure their commitment to forest conservation initiatives to meet the climate action and net zero goals initiatives.