Using ICESat-2 to characterize coastal ecosystems

Abstract

Coastal seascapes (seagrasses, mangroves, coral reefs, tidal flats) support the livelihoods of local communities, offer protection from extreme weather events, provide 25% of the oceanic carbon pool and support 25% of global biodiversity. Characterizing important ecosystems within this environment is an initial step to understanding their distribution and how they may alter within a rapidly changing world. We used ICESat-2 data to successfully characterize both aboveground and sub-aquatic ecosystem structure, namely within mangrove forests and seagrass/coral systems. We compared ICESat-2 ATL08 data with TanDEM-X data to accurately characterize mangrove forest canopy height (r2: 0.70, MAE:-1.5 m) demonstrating its ability to model height where wall-to-wall high-resolution DEM data may not be available. Furthermore, we implemented machine learning (e.g. K-NN and Isolated Forest) algorithms to successfully filter noisy ICESat-2 ATL03 photon data to extract water surface and benthic surface heights. Sub-aquatic surface heights were located between approximately 0-30 m below the water surface and were compared against locally sourced bathymetric data, demonstrating that benthic surface height can be accurately estimated and applied to regions where high-resolution bathymetric data is unavailable. This approach is readily scalable to large datasets such as ICESat-2 which contains millions of individual photons, via the use of well developed, powerful open source software. This work has important implications for characterizing tropical coastal ecosystems, particularly sub-aquatic habitats which are not currently readily mapped with existing remotely sensed data.