Mapping Surface Flow Velocity of Glaciers at Regional Scale Using a Multiple Sensors Approach

We explore and compare the capabilities and limitations of different optical sensors (Sentinel-2/ESA, Landsat 7/8/USGS, Ven $\mathsf{\mu }$ s/CNES-ISA, Pléiades/AirbusD&S and Planet Labs images) for mapping the surface speeds of mountain glaciers on a regional scale. We present here our automated workflow designed to download data from institutional or commercial servers, prepare images, launch the feature tracking algorithm, calibrate glacier surface speeds, and our post-processing treatment to obtain filtered and time-averaged velocity maps. We applied our methodology to three regions: (1) the European alps; (2) the Peruvian Cordillera Blanca; and (3) the Southern Alps of New Zealand for years 2017 and 2018 and quantified ice velocity for every possible repeat cycle from few days up to 400 days. For these regions, we demonstrate the ability of our processing chain to derive precise time-averaged ice flow maps. The statistical analysis of the results provided by each individual repeat cycles shows that velocity mapping from Sentinel-2 is about twice more precise than that from Landsat 7/8. If Sentinel-2 captures more details than Landsat, some of the smallest glaciers (<250 m wide) remain challenging. Given the estimated precision for Sentinel-2, we also conclude that velocity fluctuations of the order of 10 m/yr can only be captured with repeat cycles longer than 60 days. Comparing Sentinel-2 with Pléiades, Planet and Ven $\mathsf{\mu }$ s imagery, we finally highlight the advantages of high-resolution sensors to map glacier surface speed with finer details in space and time.