Glacier Image Velocimetry: an open-source toolbox for easy and rapid calculation of high-resolution glacier velocity fields

Abstract

Abstract. We present Glacier Image Velocimetry (GIV), an open-source and easy-to-use software toolkit for rapidly calculating high-spatial-resolution glacier velocity fields. Glacier ice velocity fields reveal flow dynamics, ice-flux changes, and (with additional data and modelling) ice thickness. Obtaining glacier velocity measurements over wide areas with field techniques is labour intensive and often associated with safety risks. The recent increased availability of high-resolution, short-repeat-time optical imagery allows us to obtain ice displacement fields using “feature tracking” based on matching persistent irregularities on the ice surface between images and hence, surface velocity over time. GIV is fully parallelized and automatically detects, filters, and extracts velocities from large datasets of images. Through this coupled toolchain and an easy-to-use GUI, GIV can rapidly analyse hundreds to thousands of image pairs on a laptop or desktop computer. We present four example applications of the GIV toolkit in which we complement a glaciology field campaign (Glaciar Perito Moreno, Argentina) and calculate the velocity fields of small mid-latitude (Glacier d'Argentière, France) and tropical glaciers (Volcán Chimborazo, Ecuador), as well as very large glaciers (Vavilov Ice Cap, Russia). Fully commented MATLAB code and a stand-alone app for GIV are available from GitHub and Zenodo (see https://doi.org/10.5281/zenodo.4624831, Van Wyk de Vries, 2021a).

GLAcier Feature Tracking testkit (GLAFT): A statistically- and physically-based framework for evaluating glacier velocity products derived from satellite image feature tracking

“…For example, when the satellite images are high-pass filtered before computing the cross-correlation surface, the resulting velocity maps often display improved quality as represented by a low δ u value (Figure 4a). This observation aligns with several past studies (Dehecq et al, 2015;Fahnestock et al, 2016;Van Wyk de Vries and Wickert, 2021). The δ u values also decrease with increasing matching template size, a classic trade-off between spatial smoothing and noise (Ahn and Howat, 2011, Figure 4b).…”

“…Since SNR correlates to the matching correctness and provides a good pixel-based quality assessment, many feature tracking tools generate a SNR map as part of the standard output along with the velocity grid, such as CARST (Zheng et al, 2019(Zheng et al, , 2021 and GIV (Van Wyk de Vries and Wickert, 2021). Ideally, we want to exclude incorrect matches from propagating to the derived velocity map and calculate uncertainties for correct matches.…”

Pre-collapse motion of the February 2021 Chamoli rock-ice avalanche, Indian Himalaya

“…Leprince et al, 2007), measuring glacier flow velocities (e.g. Bindschadler and Scambos, 1991;Heid and Kääb, 2012;Millan et al, 2019;Van Wyk de Vries and Wickert, 2021), and measuring landslide displacements (e.g. Behling et al, 2014;Lucieer et al, 2014;Manconi et al, 2018;Dai et al, 2020a;Dille et al, 2021).…”

“…We used two different feature-tracking toolboxes: GIV (Van Wyk de Vries and Wickert, 2021) and AutoRIFT (Lei et al, 2021). Both GIV and AutoRIFT are based on three core components: a pre-processing module which applies one or more filters to images to enhance distinct surface features for tracking, a multipass 2D image correlator, and a post-processing module to identify and filter erroneous displacement values (Van Wyk de Vries and Wickert, 2021;Lei et al, 2021). The GIV toolbox is written in MATLAB and performs image cross correlation in the frequency domain, while AutoRIFT is written in python/C++ and performs the cross correlation in the spatial domain.…”

“…We processed these image pairs using GIV. We also resampled the velocity timeseries to monthly resolution (see Fig 4 c-g) using a weighted averaging scheme described in Van Wyk de Vries and Wickert (2021).…”