New Tool Revolutionizes Coastal Shoreline Mapping with Decades of Satellite Data
Coastal scientists and resource managers have long sought a reliable and efficient way to monitor and manage the world’s shorelines. Thanks to extensive collections of satellite data spanning more than four decades, this endeavor is now more feasible.
However, it’s the recent advancements in image processing, web-native geospatial data rendering, machine learning, and deep learning that have truly transformed the enterprise of automated measurements from remotely sensed imagery, enabling the reliable extraction of shorelines—known as satellite-derived shorelines (SDS).
Researchers from USGS and New South Wales Department of Planning and Environment have introduced CoastSeg, an innovative, interactive browser-based program that aims to promote and democratize the adoption of SDS detection workflows among coastal scientists and resource management practitioners. SDS, a burgeoning sub-field of coastal sciences, focuses on detecting and post-processing a time-series of shoreline locations from publicly available satellite imagery.
Streamlined Functionality for Coastal Scientists
Shoreline mapping comprises two components; a) detection of the waterline in each satellite image, and b) estimating the shoreline location from the instantaneous waterline, by applying a horizontal correction that accounts for tide height. The primary aim of CoastSeg is to serve as a comprehensive toolkit that encompasses core SDS workflow functionalities. This includes efficiently handling essential tasks such as file input/output, image downloading, geospatial conversion, mapping 2D shorelines to 1D transect-based measurements, and applying tide corrections to shoreline measurements. These functions are crucial to a basic SDS workflow, regardless of the specific waterline estimation methodology used. CoastSeg is also a convenient way to access all Landsat and Sentinel satellite imagery for any coastal region in the world, for purposes of shoreline mapping as well as other monitoring and research purposes.
CoastSeg’s development team has re-implemented and enhanced the algorithms and workflows behind a popular existing toolbox, CoastSat, within the new platform. This improvement allows users to perform the well-established CoastSat SDS workflow, which has been well-tested and reported in the literature, but in a more accessible and convenient browser-based environment, underpinned by access to an underlying database of coastal reference shorelines, transects, and other auxiliary data. Tidal prediction has been baked into the CoastSeg program, allowing users to apply tidal correction without third-party software, facilitated through the pyTMD project, which exposes a number of astronomical tide prediction suites. This transition not only streamlines the process but also broadens the potential user base by lowering technical barriers.
In addition to the CoastSat workflow, which uses a neural network classifier trained on limited data to estimate shoreline position, CoastSeg provides an alternative workflow based on a more powerful deep neural network model, which has been trained on data from significantly more locations. This workflow, as well as the CoastSat workflow, both depend on a core set of common functions within an API, and yield equivalent outputs, so direct comparisons can be made between methods. Users can use either or both methods from within the same platform to carry out SDS mapping. Users are also encouraged to contribute alternative workflows that use the CoastSeg API, and the CoastSeg team continues to explore cutting-edge algorithms for enhanced shoreline detection. A comprehensive validation of CoastSeg’s deep-learning-based SDS workflow is now underway at numerous global sites with field survey data.
Collaboration and Reproducibility at the Forefront
A standout feature of CoastSeg is its focus on facilitating experimental and collaborative contexts. The platform introduces ‘sessions’, a novel mechanism for saving the current state of the application into a session’s folder. This capability allows users to share their sessions with peers, enabling the replication of experiments, building upon previous work, quality assurance, or accessing data downloaded by others. Such functionality is critical for oversight, reproducibility, and practical needs based on division of labor.
Ongoing and future work includes further refinement of the CoastSeg platform within the Typhoon Merbok Disaster Emergency Recovery Efforts Project, where it is being used to map the evolving shorelines of western Alaska over the past four decades. The Merbok project also provides a challenging environment to develop more advanced waterline and shoreline mapping workflows that will be eventually incorporated into CoastSeg, owing to the relatively poor visibility, presence of sea ice, geodetic imprecision, and other related challenges in the region.
One crucial aspect of this effort is the integration of additional satellite imagery sources, such as higher-resolution PlanetScope imagery from Planet Labs, made available to federal researchers through the Commercial Smallsat Data Acquisition program. CoastSeg also may serve as an ideal platform to implement more advanced image preprocessing algorithms that may benefit shoreline mapping, such as image super-resolution, where machine learning is used to provide more finely resolved pixels, and other potential image processing related to maximizing signal-to-noise in input imagery (summarized by Vitousek et al., 2023).
CoastSeg’s innovative approach aims to significantly enhance the efficiency and accuracy of shoreline mapping. By leveraging decades of satellite data and cutting-edge technological advancements, CoastSeg can be a valuable tool for coastal scientists and resource managers worldwide, paving the way for more informed and effective coastal management practices.
Related
Typhoon Merbok Disaster Emergency Recovery Efforts
Remote Sensing Coastal Change
Related
Typhoon Merbok Disaster Emergency Recovery Efforts