AI-driven mapping helps advance wetland inventory across Alberta

“Our teams also conducted helicopter and ground-based wetland surveys to test our machine learning models,” says Rebecca Edwards,
advanced remote sensing analyst, DUC boreal operations. “Field validation is essential to ensure our classifications meet Alberta’s mapping standards.”

How AI is improving wetland mapping

The study tested several artificial intelligence techniques, combining airborne LiDAR data with multi-season satellite imagery to detect wetland features and classify wetland types. These include shallow open water (bare and aquatic bed), marsh (graminoid), fen (wooded coniferous, shrubby, graminoid), bog (wooded coniferous, shrubby, graminoid) and swamp (wooded, coniferous, deciduous, mixedwood and shrubby).

Encouraging results

The study’s wetland classifications met or exceeded Alberta’s provincial mapping standards across the four pilot areas. In most use cases, the models achieved high levels of accuracy in distinguishing wetlands from surrounding uplands.

The experts found that:

• Deep learning models delivered the highest overall mapping accuracy, producing precise boundaries for wetland areas at the wetland class level.
• Machine learning models captured finer details, particularly improving the detection of rare or less common wetland form level.
• Combining multiple data sources like LiDAR and satellite imagery significantly improved model performance.

In some cases, integrating LiDAR-derived elevation data increased classification accuracy by more than 10 per cent. 

Together, these approaches demonstrate that artificial intelligence can help modernize large-scale landscape monitoring and support the founding of wetland inventories.  

Why this research matters

Accurate wetland mapping plays a critical role in conservation planning, environmental monitoring and land-use management.

The study highlights the consistency of how artificial intelligence or AI-driven tools help scale wetland inventories across Alberta and potentially across other parts of Canada. By automating portions of the mapping process, experts can process large geographic areas more efficiently while maintaining high accuracy.

Reason’s the work is important:

When combined with standardized validation processes to has the potential to produce authoritative wetland inventory that meets national and provincial standards.

DUC has continued to refine and adapt the methodology to other parts of Prairie Canada, validating that the methods are robust, and adaptable. This work is the foundation for DUC wetland mapping program which continues to invest into these methods.

Meet the researchers behind the report

The study brought together representatives from the Government of Alberta’s Environment and Protected Areas (EPA) program, Alberta Biodiversity Monitoring Institute (ABMI), scientists and remote sensing experts from multiple organizations working to improve wetland monitoring across Canada. The Alberta government contracted DUC and the ABMI to collaboratively develop methods aimed at meeting minimum mapping standards. The project was led by an EPA technical committee and funded the wetlands replacement fund.

Click to learn more about innovative approaches to mapping.

DUC experts involved in the study include:

• Joshua Evans, Advanced Remote Sensing Analyst, Prairie Regional GIS
• Rebecca Edwards,
Advanced Remote Sensing Analyst, Boreal Operations
• Lyle Boychuk,
Manager of GIS/Inventory, Prairie Regional GIS

Together, the team evaluated how different AI models, including Random Forest, XGBoost and deep learning convolutional neural networks, performed when mapping wetlands using integrated remote sensing datasets. 

Their work contributes to a growing body of research exploring how artificial intelligence can support wetland science, biodiversity monitoring and large-scale environmental mapping. 

Supporting the future of wetland science

As Canada works to better understand and manage its wetlands, advances in data science and remote sensing are opening new possibilities.

AI-driven mapping approaches can help experts to identify wetland boundaries, classify wetland types and monitor landscape change more efficiently than traditional methods alone.

For Ducks Unlimited Canada, this type of research supports ongoing efforts to improve wetland science and provide decision-makers with the tools needed to manage landscapes sustainably. 

By combining cutting-edge technology with field expertise, scientists are helping ensure that wetlands and the wildlife and communities that depend on them, remain a vital part of Canada’s natural heritage.

Learn about DUC’s Canadian Wetland Inventory Initiative

Read the Government of Alberta project returns for wetland inventory pilots