Blog: Improved Habitat Monitoring in Alberta Using Lidar and Machine Learning in Wellsite Identification

Scientists at the Alberta Biodiversity Monitoring Institute’s (ABMI) Imaging Centre are leveraging the power of an impressive partnership: remote sensing (lidar1) and machine learning. By combining these advanced technologies, imaging experts can now automatically and accurately delineate the outlines of wellsites in the boreal forest of Alberta; proving to be a boost to the science of land mapping as well increasing the level of detail of existing land-use maps. Manually delineating the accurate boundaries of wellsites across the province would be by every calculation a mammoth task. Undeterred, the Imaging Centre’s team devised a strategic solution that began with taking advantage of the ABMI’s Human Footprint Inventory (HFI) as reference data. The HFI dataset already presented a large amount of reliable information on wellsites and was created and digitized using three different methods: 1) Data buffered by the Government of Alberta using intelligent buffering based on the type of wellbore. 2) Use of the wellsite’s existing disposition boundary or defined area for oil and gas activities and 3) Manually digitizing wellsite locations using medium-high resolution satellite or SPOT imagery. Later, wellsites in the HFI are verified and attributed using AER wellbore data or disposition datasets. The team combined this HFI data and relevant lidar-derived products with machine learning to generate accurate boundary boxes around wellsites. They developed a two-step process: first, detecting wellsites, and second, delineating their accurate boundaries.