Using a water index approach to mapping periodically inundated saltmarsh land-cover vegetation and eco-zonation using multi-temporal Landsat 8 imagery

Abstract

Saltmarsh land-cover (SLC) ecosystems, composed of unvegetated mudflats, saltmarshes, mangroves, and/or seagrass communities, are vulnerable to climate-induced impacts, such as sea level rise. Extracting a seamless and consistent waterline from satellite imagery is a major challenge because of environmental factors, such as turbidity, water depth and multiple types of underwater vegetation cover that introduce noise in the extraction of information. Hence, a water index, derived from multi-temporal Landsat 8 (OLI) data, acquired under different tides is proposed for mapping land-water across SLC wetlands by tracking waterlines. This provided inundation maps and defined eco-zones to specify south-eastern Bangladesh wetland composition. The NDWI_1 (McFeeters’s water index) applied to 42 OLI images and derived land-water difference maps generated inundation gradient maps with an overall classification accuracy of 87.8%. The simple intersection and union of region-of-interests extracted from the tide heights above the mean low-water springs enabled the mapping of four categories of wetland composition based on hydroperiods: a) irregularly inundated (II), regularly inundated (RI), irregularly exposed (IE; high floodplain), and subtidal (river bed and deep water sea). For all of the three study sites, mangrove, seagrass, non-mangrove and agriculture were all prominent on the IE eco-zone, while only saltmarsh was dominant on the II eco-zone. These maps of SLC wetland will enrich previous concepts of eco-zonation models that include salinity, erosion, accretion and rate of sea level rise as factors, suggesting that inundation extent and tidal phase complexities should be considered in the remote sensing of SLC composition for improved models of SLC vegetation response to climate change.