A Review Of Evolving Remote Sensing And Automated Techniques In Rock Glacier Mapping
https://doi.org/10.1016/j.earscirev.2026.105473 <-- shared paper
--
H/T Sunil Tamang
“HIGHLIGHTS:
Mapping mainly relies on manual digitisation based on geomorphic features.
Machine Learning shows potential but remains underutilised in automated mapping.
Current approaches do not adequately validate rock glacier extent mapping.
Open data and code sharing will enhance validation and reproducibility.
ABSTRACT: Accurate and consistent mapping of rock glaciers is necessary for various scientific and applied studies, including understanding geomorphic processes and evolution, hydrological dynamics, geohazard assessment, mountain biodiversity, and permafrost studies. Although rock glacier mapping practices and methods have evolved alongside progress in geospatial technologies, none of the existing literature has systematically focused on the full range of methods used for rock glacier mapping. This review provides the first structured comparison of manual, interferometry-aided, and machine learning (ML)-based mapping approaches specific to rock glaciers, with an emphasis on spatial delineation, classification, validation and uncertainty. Advantages and limitations of each method are examined thoroughly. Despite rapid progress in remote sensing, digitisation based on visual interpretation of geomorphological features is still widely used. In contrast, the application of automated techniques such as ML models in rock glacier extent mapping is underexplored. Discrepancies have been observed between ML-mapped polygons and manually digitised ones, raising concerns about the accuracy of mapped extents of rock glaciers. Addressing uncertainties in rock glacier mapping is crucial, not only to ensure the reliability of mapping methods but also because these uncertainties can propagate into subsequent studies. Current validation approaches primarily assess the presence of rock glaciers, but do not adequately validate their mapped extent. To address this gap, [the authors] recommend integrating field-based mapping, uncertainty quantification, the use of high-resolution datasets, and, where applicable, local indigenous knowledge. Lastly,[they] advocate for the sharing of mapped rock glacier data and ML code to promote open, reproducible, and collaborative research…”
#GIS #spatial #mapping #rockglacier #glaciers #permafrost #remotesensing #interferometry #earthobservation #spatialanalysis #machinelearning #AI #machinelearning #ML #deeplearning #CNN #metrics #inventory #earthobservation #GeoAI #geostatistics #InSAR #LiDAR #radar #satellite #review #literaturereview #geomorphology #geomorphometry #hydrology #geohazard #risk #hazard #engineeringgeology #biodiversity #permafrost #cryosphere #ice #snow #geology #assessment #survey #research
#GeospatialResearchInstituteToiHangarau | #UniversityofCanterbury