Time-Series Clustering of C-band SAR backscatter for classification of supraglacial winter lake behaviour

This repository contains the results and reproducible workflow of a study that applies time-series clustering of Sentinel-1 C-band SAR data to identify supraglacial winter lakes and classify their winter behaviour at two study sites across the Greenland Ice Sheet (Russell Glacier, NEGIS) and for two winters (2018, 2019). The provided csv files with the naming convention "S1_HV_TimeSeries_Clustering_studysite_Winter_year.csv" contain: Mean HV backscatter time-series for all individual supraglaclial lake basins (datetime columns "yyyymmdd") Clustering results from this study ("Cluster_number", "Cluster_label") Optical validation labels from melt seasons before and after winter ("Optical_label_autumn", "Optical_label_spring") Final winter lake classification lables ("Lake_classification_label") Misclassification labels ("Misclassification_source") The jupyter notebook implements the time-series clustering workflow used to classify supraglacial lake winter behaviour from Sentinel-1 C-band SAR backscatter data. The input to this notebook is background-corrected mean backscatter time series data from individual supraglacial lakes across the two study sites and two years, see csv files with the following naming convention "S1_HV_background_corrected_lake_TS_studysite_Winter_year.csv". Together, these files provide the data and code necessary to reproduce the winter lake classification results presented in the associated publication which is currently in review.