A Collection of Novel Algorithms for Wetland Classification with SAR and Optical Data

• Main Author: Salehi, Bahram
• Other Authors: Brisco, Brian, Mahdavi, Sahel, Granger, Jean, M.Manesh, Fariba, Amani, Meisam, Mahdianpari, Masoud
• Format: Article in Journal/Newspaper
• Language: English
• Published: 2022
• Subjects: Science / Environmental Science; bisacsh:SCI026000; Newfoundland
• Online Access: https://openresearchlibrary.org/viewer/60339f98-4e42-481d-b5a4-092d3ffc3df1; https://openresearchlibrary.org/ext/api/media/60339f98-4e42-481d-b5a4-092d3ffc3df1/assets/external_content.pdf

Wetlands are valuable natural resources that provide many benefits to the environment, and thus, mapping wetlands is crucially important. We have developed land cover and wetland classification algorithms that have general applicability to different geographical locations. We also want a high level of classification accuracy (i.e., more than 90%). Over that past 2 years, we have been developing an operational wetland classification approach aimed at a Newfoundland/Labrador province-wide wetland inventory. We have developed and published several algorithms to classify wetlands using multi-source data (i.e., polarimetric SAR and multi-spectral optical imagery), object-based image analysis, and advanced machine-learning tools. The algorithms have been tested and verified on many large pilot sites across the province and provided overall and class-based accuracies of about 90%. The developed methods have general applicability to other Canadian provinces (with field validation data) allowing the creation of a nation-wide wetland inventory system.