Extraction of Velocity Time Series With an Optimal Temporal Sampling From Displacement Observation Networks

Fusion of Multi-Temporal and Multi-Sensor Ice Velocity Observations

“…The proposed method is an extension of (Charrier et al, 2021a) and (Charrier et al, 2021b). In (Charrier et al, 2021b), the mono-sensor dataset had enough redundancy to obtain directly a regular time series at user-defined temporal sampling using an improved temporal closure formulation. However, in the case of a multi-sensor data set with numerous gaps, the problem is more complex.…”

“…The SBAS approach uses interferograms produced by image pairs spanning small temporal and geometrical baselines to minimize temporal and geometrical decorrelation. Later, this approach has been applied to offset-tracking displacement measurement of SAR images (Casu et al, 2011, Euillades et al, 2016, Guo et al, 2020, Charrier et al, 2021a, Charrier et al, 2021b and optical images (Bontemps et al, 2018, Lacroix et al, 2019, Ding et al, 2020.…”

“…However, most of the authors who use the temporal closure approach consider only one sensor (Altena et al, 2019, López-Quiroz et al, 2009, Casu et al, 2011, Berardino et al, 2002, López-Quiroz et al, 2009, Euillades et al, 2016, Charrier et al, 2021a, Charrier et al, 2021b or invert a time series from each sensor apart (Kim et al, 2015, Ding et al, 2020. Only a few studies choose to include datasets from different sensors in one inversion.…”

“…Indeed, (Bontemps et al, 2018) are interested in cumulative displacement (CD) time series, i.e. displacements between each date and a reference date, to study slow moving landslide whereas velocities between consecutive dates, named Leap Frog (LF) velocity time series in (Hadhri et al, 2019, Charrier et al, 2021b, are more appropriate to analyze intra-annual variations of fast moving targets such as glaciers (Greene et al, 2020, Derkacheva et al, 2020. The analysis of an irregular LF time series can raise issues because: 1) The resulted velocities does not have the same temporal sampling.…”

“…2) displacements with very short temporal sampling are inverted if the images from the different sensors are acquired a few days apart (for example up to 1 day apart between Venµs and Sentinel-2 satellites). However, very short temporal sampling velocities are likely to be more noisy (Casu et al, 2011, Millan et al, 2019, Charrier et al, 2021b. 3) the dates between which the displacements are inverted can be different from one pixel to another because of different data quality since outliers are rejected in most of the available datasets (Millan et al, 2019, Gardner et al, 2018, Fahnestock et al, 2016.…”