Geomorphic approach to fill existing gaps in remote sensing surface water extent mapping

Optical-based remote sensing techniques can be an effective means to monitor surface water extent (SWE) in wetlands. However, these techniques have limitations which arise due to omission errors caused by cloud coverage, shadows, vegetation canopy above the water and the presence of water and vegetation within the pixel (mixed pixels). Moreover, regardless of the water index being used, images with small water bodies, or water bodies that have long perimeters compared to their area will have more mixed pixels and greater omission errors, leading to a failure to identify SWE. Geomorphic approaches have been recognised as a potential strategy to fill existing gaps in remote sensing applications and specifically, SWE mapping. These methods are generally based on digital elevation models (DEMs). DEM-based methods can identify of topographic depressions, which are areas that are geomorphologically prone to inundation. The aim of this study was to combine the use of optical remote sensing imagery and depression-based DEM data to develop operational SWE mapping which can deal with wetland complexity in the Murray-Darling Basin. SWE was mapped across a small watershed using Sentinel2 imagery. A stochastic depression analysis using a Monte Carlo approach was used to extract surface depressions from a DEM and generate a probability depression map. The domain cells of each SWE polygon were identified on the probability depression map and the 5th percentile probability of depressions was estimated. Finally, the SWE polygons were automatically extended with a region growing algorithm to areas where the probability depression map was greater or equal to the 5th percentile probability of depressions of that polygon. A case study in the Murray Darling Basin highlighted the improvement from the combined geomorphic approach over a remote sensing based approach.