Abstract
In response to the necessity for enhanced wetland inventories, this study aimed to assess effectiveness of employing high spatial resolution SPOT imagery (1.5m resolution by Planet Lab) for classifying and detecting wetlands in Atlin, British Columbia, Canada. Utilizing the Random Forest Classifier, featured for its capability in handling high-dimensional spatial data, the research aims to contribute to the local understanding of wetland status through advanced raster analysis. The application of the Random Forest Classifier yielded an overall classification accuracy of 86%, underscoring the method’s applicability for wetland delineation in Atlin. The generated wetland map, featuring a 10m spatial resolution, integrates topographic, vegetative, and textural indices, presenting a valuable tool for assessing the variable importance in wetland classification. Despite its high accuracy, the study acknowledges the irreplaceable value of field assessments for comprehensive wetland evaluation by ecological uniqueness of wetlands. This research not only demonstrates the potential of high-resolution SPOT imagery in environmental monitoring but also encourages further application of machine learning techniques in the preservation and management of critical wetland ecosystems.
MGEM Student: Ashley Yang
Key words: Wetland Mapping, SPOT Imagery, Supervised Classification, Remote Sensing, Machine Learning, Random Forest
Explore this project
Download the final report and data
Cite this project
Yang, Ashley, 2024, “Wetland Mapping: Application of Supervised Classification Using Random Forest in Wetland Prediction“, https://doi.org/10.5683/SP3/PGOUF2, Borealis, V1