Ice front positions for Greenland glaciers (2002–2021): a spatially extensive seasonal record and benchmark dataset for algorithm validation

Abstract. Glacier terminus (ice-front) positions are key indicators of glacier dynamic variability and ice–ocean–atmosphere interactions and provide essential time-varying boundary conditions for ice-sheet modelling. High-precision, spatially extensive records are therefore critical for quantifying terminus recession, improving estimates of dynamic mass loss, and supporting the development and validation of automated front-detection algorithms. However, existing manually delineated datasets are largely restricted to marine-terminating glaciers and exhibit heterogeneous spatial coverage, temporal sampling, and delineation criteria, which limits ice-sheet-scale representativeness and propagates into consolidated and automated products that depend on them for training and evaluation. Here we present Greenland Terminus Position Dataset (GrTPD), an independent and internally consistent dataset of ice front positions in Greenland, providing spatially extensive, seasonally targeted coverage (at least twice per year) across marine-, land-, and lake-terminating glaciers. The dataset comprises 19 171 terminus delineations for 465 glaciers spanning 2002–2021, derived from multi-source satellite imagery (Landsat, Sentinel-1/2, MODIS, ENVISAT, ASTER, and ERS). Delineations were produced using standardized workflows implemented in Google Earth Engine (GEE), ENVI, ArcGIS and QGIS. Each record is accompanied by metadata describing acquisition date, satellite, method, glacier identifier, official name, glacier type by terminus and origin, vertex count, and ice front length. Positional accuracy was evaluated using average minimum distance (AMD) comparisons against the integrated manually delineated TermPicks dataset and the automated AutoTerm product for overlapping glaciers. Mean AMD of 86 m relative to TermPicks and 115 m relative to AutoTerm indicates high geometric fidelity and positional consistency. Across spatial aggregates and time series, GrTPD shows closer agreement with TermPicks than with AutoTerm, consistent with the greater sensitivity of automated delineations to image quality, low-contrast ice–ocean mélange conditions, and heterogeneous terminus geometries. By extending coverage beyond marine-terminating glaciers, GrTPD enables a more comprehensive assessment of Greenland glacier terminus variability and provides a high-quality benchmark for validating and intercomparing automated delineation products. The dataset is publicly available from Zenodo: https://doi.org/10.5281/zenodo.19181770 (Xi et al., 2026).