Time-series surface velocity extraction of Petermann Glacier based on Sentinel-1 pixel offset-tracking and iterative SVD

Monitoring the Greenland glacier flow velocity is essential for the quantitative estimation of ice sheet material loss, the assessment of the impact of global climate change on ice sheet dynamics, and the evaluation of Greenland’s contribution to current sea-level rises. The offset-tracking technique is the main method for deriving glacier velocity by using the intensity information of SAR or optical images. Intensity offset tracking is less sensitive to decorrelation than the InSAR method and can be applied to images with long temporal intervals. However, glacier avalanche, ice avalanche, snowfall, and melting–freezing cycles on glaciers still cause changes in the scattering characteristics of the surface, resulting in changes of the SAR image intensity, leading to a loss of correlation in matching between images, especially in summer. To provide more accurate glacier flow velocity field, this research proposes a novel data processing strategy of processing Sentinel-1 SAR data and takes the famous Petermann outlet glacier in Greenland as an example to extract its glacier velocity based on image tracking. Noise and errors in tracking images formed by single pairs of Sentinel-1 images are removed through morphological opening operation, connectivity analysis, adaptive median filtering, etc. Meanwhile, annual and monthly Greenland ice flow velocity products are employed to select datum by taking its low-speed area as reference. We also introduce flow direction of the annual or seasonal glacier flow to filter out wrong matchings. Similar to the small-baseline analysis of the InSAR technique, redundant observation of tracking pairs with 6-, 12-, and 18-day intervals are then applied to the Singular Value Decomposition (SVD) method to solve the time series of glacier velocity and to avoid the possible rank deficit. SVD is iteratively performed to remove the observed coarse error that could not be eliminated in the previous processing by checking residuals of the observation after each iteration. We obtain the time-series glacier velocity for the Petermann Glacier from the year 2018 to 2020 with a temporal resolution of 6 days. Compared with the published glacier velocity products, our derived results are less noisy, more continuous, smoother, and cover more area than the CPOM product, which employs the same data source. Compared with the PROMICE product produced from multitrack SAR, data show that we share similar accuracy and effective data coverage, but the results of this research have higher resolution and are less noisy, especially in summer. We conclude that the proposed algorithm can effectively eliminate the anomalous matching of single offset-tracking pair for forming high spatial and temporal resolution glacier flow velocity time series with redundant matching pairs by using an iterative SVD method, which is essential for monitoring glacier flow velocity for the Greenland Ice Sheet with satellite SAR images. © 2024 Science Press. All rights reserved.