Automated grounding line delineation using deep learning and phase gradient-based approaches on COSMO-SkyMed DInSAR data

<p>This repository contains the file 'Natalya_Maslennikova_data_2024.zip,' which includes 171 DInSAR interferograms (both phase and coherence of a DInSAR signal) and three versions of the grounding lines: manually mapped ('gl_manual_mapping.shp'), mapped by the neural network ('gl_neural_network.shp'), and mapped using the phase gradient-based approach ('gl_phase_gradient.shp'). The X-band DInSAR data were acquired by the ASI&rsquo;s COSMO-SkyMed mission between 2020 and 2022 over five East Antarctica glaciers: Stancomb-Wills (STA: 32 interferograms), Veststraumen (VES: 35 interferograms), Jutulstraumen (JUT: 42 interferograms), Moscow University (MOS: 27 interferograms), and Rennick (REN: 35 interferograms).</p> <p>The 'gl_neural_network.shp' and 'gl_phase_gradient.shp' shapefiles contain the following fields: the name of the glacier (the 'Glacier' column), four dates in YYYYMMDD format when the corresponding DInSAR interferogram was acquired (the 'Primary1', 'Secondary1', 'Primary2', 'Secondary2' columns), and the double difference name in 'Primary1_Secondary1-Primary2_Secondary2' format (the 'DD' column). The 'gl_manual_mapping.shp' shapefile contains 'Glacier', 'Primary1', 'Secondary1', 'Primary2', 'Secondary2', and 'DD' columns as well, along with the 'Revisit' column (difference in days between the 'Primary1' and 'Primary2' images were acquired), the 'Time' column (acquisition time in 'HHMMSS' format), the 'Coherence' column (average coherence of the corresponding DInSAR interferogram), and the 'Train_1', 'Train_2', 'Train_3' columns, which contain indicators 'Y' for 'yes' and 'N' for 'no', showing whether or not the DInSAR interferogram was used in the corresponding training session.</p>