Piloting Next-Generation Wetland Mapping in Alberta with Advanced Artificial Intelligence Methodologies
Held Wednesday, May 7, 2025 - 3:00 p.m.-4:30 p.m. EDT
INTRODUCTION [Presentation - PDF]
Ian Grosfelt, National Association of Wetland Managers
PRESENTERS [Presentation - PDF]
- Rebecca Edwards, Ducks Unlimited Canada
- Lyle Boychuk, Ducks Unlimited Canada
ABSTRACT
This presentation provided an overview of a collaborative project, delivered by the Government of Alberta, Ducks Unlimited Canada, and the Alberta Biodiversity Monitoring Institute, aimed at piloting and refining state-of-the-art wetland mapping methodologies specific to Alberta’s boreal, parkland, and grassland region. Wetlands across Alberta, which act as nature’s water filters, biodiversity hot spots and carbon storehouses, are facing significant changes due to climate change and human disturbance, therefore, spatial inventorying of wetlands is critical and required to support legislation and policy development surrounding the management and monitoring of these sensitive ecosystems.
In 2014, the Government of Alberta (GOA) released a provincial wetland inventory, the Alberta Merged Wetland Inventory (AMWI). However, the AMWI is derived from various image sources, across a range of dates, while using different methodologies, resulting in variations in quality and accuracy. In 2020, the Government of Alberta published standards for addressing wetland inventory updates across the province, including requirements for minimum mapping units, classification accuracy, and classification detail; thus, an updated wetland inventory is imperative to ensuring a consistent and accurate wetland inventory covering Alberta. Recent advancements in remote sensing methods and Earth observation data present encouraging avenues for meeting these standards. In response to this, the project team developed, advanced, and evaluated several methodologies across four pilot areas targeted at achieving the provincial benchmarks. These methodologies leveraged multiple artificial intelligence (AI) technologies (e.g., machine and deep learning), Earth observation datasets, recent airborne LiDAR acquisitions, and airborne and field verified wetland surveys. The resulting inventories successfully tested and advanced AI methodology. Three of the pilot areas (excluding parkland) met the provincial mapping standards (greater than 70% accuracy) at various thematic levels, ranging from general wetland detection to wetland class and form. Machine learning outperformed deep learning in the form classification due to its ability to classify minority classes where, with more training data, deep learning will continue to get more accurate results. Lastly, LiDAR was critical for identifying isolated depressions in the Prairies and had significant impact at the form level in the Boreal. These pilot inventories offer crucial insights into scaling these methods to a provincial level and provide valuable guidance for advancing wetland mapping and monitoring efforts across Alberta.
BIOS
Rebecca (Becky) Edwards is the lead remote sensing specialist for DUC’s National Boreal Program. She completed her MSc in Geography specializing in remote sensing from Queen’s University focusing on vegetation change in the Arctic. Becky has been working with DUC for over 7 years with experience in wetland remote sensing, time series analysis, northern landscapes, and wetland ecology. She has led and supported many large-scale wetland inventory products across the Boreal to support the Canadian Wetland Inventory (CWI) and continues to advance novel wetland classification methodologies.
Lyle Boychuk is the Manager of GIS and Inventory Programs; Prairie Region for Ducks Unlimited Canada Lyle has been working in geomatics for 27 years and is currently in his 23rd year of service with DUC. He manages a team of specialists responsible for building GIS applications in support of conservation programs and managing the delivery Canadian Wetland Inventory (CWI) projects across Prairie Canada. His professional interests include the use of remote sensing to characterize Prairie Wetlands, the application of terrain data for quantifying wetland hydrology; and the use of UAV technology to support conservation program delivery.