Wetland mapping with multi-temporal sentinel-1 & -2 imagery (2017 – 2020) and LiDAR data in the grassland natural region of alberta

ABSTARCTIn the Grassland Natural Region (GNR) of southern Alberta, wetlands are relatively small-sized disconnected prairie pothole marshes, swamps, and shallow open water habitats often surrounded by grasslands, parkland forests, agricultural lands, urban areas. These susceptible to climatic variability, resulting in temporally spatially dynamic that difficult map accurately. This study hypothesizes seasonal synthetic aperture radar (SAR) optical imagery will capture temporal variations spring/summer fall months 2017, 2018, 2019, 2020. We propose these data combined with topographic variability offered LiDAR-derived wetness index (TWI) shall result accurate delineation wetlands. Using a combination open-access government databases, we generated ground training develop classification models perform accuracy assessments. The wetland products’ overall results ranged from 63.2% 75.7%. pixel-based random forest (RF) classified dataset (Dataset 5 – multi-temporal (2017–2020) S1 SAR (VH) S2 (B8 B11) bands fused TWI) had highest (75.6%). RF significantly outperformed similar CART (Classification Regression Trees) SVM (Support Vector Machine) classifications, which accuracies 67.4% 63.2%, respectively. In addition, optimal product best F-score values for upland classes: 0.61 (marsh), 0.82 (open water), 0.75 (swamp), 0.8 (uplands). Overall, methodology adopted this is promising mapping spatial distribution across seasonally GNR Alberta.