Intertidal elevation dynamics and vegetation-geomorphology feedbacks revealed by ICESat-2 and Sentinel time-series deep learning

The data and code supporting this study are partially shared for peer review purposes. Full release is subject to data confidentiality agreements and intellectual property restrictions. Predicted DEM outputs (DEM_pred_2019.tif – DEM_pred_2023.tif): Annual intertidal digital elevation models for the Liao River Estuary (2019–2023), predicted by the trained Temporal Transformer model at 10 m resolution and referenced to the 1985 National Height Datum. All figures in the manuscript are derived from these GeoTIFF files. Calibration control points (矫正_RTK.xls): Coordinates and ellipsoidal heights of approximately 20 RTK-GNSS control points installed on stable artificial structures (e.g., embankments and sluice gates) within the study area. These points experience negligible local subsidence and were used exclusively for inter-annual datum alignment of ICESat-2 photon elevations to the 1985 national vertical datum. They were not used in model training or accuracy assessment. Independent validation data (RTK_2019.csv): A subset of field-measured RTK-GNSS points collected on natural intertidal surfaces in 2019, provided as a representative sample of the ~50,000 multi-year validation dataset used for independent model accuracy assessment. Training data sample (ATL03_with_features_2019_calibrated_normalized.csv): The processed and normalized training feature table for 2019, containing calibrated ICESat-2 photon elevations co-registered with multi-source remote-sensing features (inundation frequency, Sentinel-1 SAR texture, and Sentinel-2 vegetation indices). Data for 2020–2023 follow an identical structure but are withheld due to data confidentiality.code:https://github.com/ectdog/ICEsat2.git