Abstract

Satellite data has never been more numerous and accessible. High resolution images of the Earth are being collected daily, and conservationists are exploring ways to use these big datasets to save species and protect natural habitats. We have been developing approaches to use this data and cloud computing technologies to address one of the most significant limitations to biodiversity conservation - insufficient monitoring and enforcement of laws and regulations that protect species habitats. We have adapted and developed two generalized methods that automatically detect landscape changes in a variety of habitat types using free and publicly available data and tools. After testing the algorithms at over 50 sites of known change in the United States, we found these approaches were effective (AUC > 0.90) at distinguishing between areas of habitat loss and areas of no change. We then applied these methods to detect and measure ongoing threats to imperiled species habitat: oil and gas development in the range of the Lesser Prairie Chicken, and sand mining operations in the range of the dunes sagebrush lizard. The habitat loss we detected resulted in petitions to place both of these species on the Endangered Species List, illustrating how these algorithms can be used to help close the implementation gap of monitoring and enforcement in biodiversity conservation. The explosion of available imagery has also coincided with the growth of computer vision and deep learning models, and there is growing interest in applying these deep learning approaches to train computers to find objects in satellite images. We are using computer vision to help automatically map landscape features that would take months to manually delineate, in a case study where knowledge of their location can guide on the ground conservation action.