Global Delineation of the Earth's Coastal Critical Zone Based on Functional Identification and Machine Learning (2022 Baseline)

This dataset establishes the inaugural global physical benchmark for the Earth’s Coastal Critical Zone (ECCZ), anchored to a 2022 baseline. Unlike traditional geometric delineations, this dataset synthesises a high-dimensional feature space comprising 18 key indicators spanning three core domains: the physical environment (soil properties, hydrology, geomorphology), anthropogenic intensity (night-time light, population density), and ecosystem structure. The spatial boundaries were delineated using a spatially stratified Support Vector Machine (SVM) classification architecture, designed to resolve the complex non-linear coupling between natural and human processes. Following rigorous post-processing (including spatial filtering and topological correction), the final output is provided in Vector Shapefile (.shp) format to facilitate seamless integration into Geographic Information Systems (GIS) for spatial analysis and regional planning. Technical Specifications: Format: ESRI Shapefile (Vector Polygon) Coordinate Reference System (CRS): WGS 84 (EPSG:4326) Baseline Year: 2022 Version Update: The previous version did not adequately account for spatial autocorrelation during machine-learning sample selection and model evaluation. Because geographically adjacent samples may share similar environmental characteristics, the use of spatially dependent training and validation samples may have resulted in optimistically biased estimates of model performance. In the current version, this issue has been addressed by introducing a spatially blocked sampling framework that improves the spatial independence of the samples used for model development and evaluation. The revised strategy provides a more rigorous and realistic assessment of model generalisation across geographically distinct regions. The ECCZ boundaries have consequently been regenerated using the corrected modelling framework.