Reliability of Sentinel-1 InSAR distributed scatterer (DS) time series to estimate the temporal vertical movement of ombrotrophic bog surface

Abstract

<p>A better understanding of the short term and seasonal peat surface vertical displacements (bog breathing) (Roulet 1991) initiated by changes in the water table is needed to improve spatial models of greenhouse gas emissions (Dise 2009). Synthetic Aperture Radar Interferometry (InSAR) is a promising tool for the task but accounting for relatively large peat surface displacements (Fritz 2006, Howie & Hebda 2018) may cause propagation of ambiguity errors and unreliability (Alshammari <em>et al</em>. 2018, Heuff & Hanssen 2021). This is usually overlooked and the absence of ground levelling data for validation is characteristic of InSAR research in peatlands (Cigna & Sowter 2017, Alshammari <em>et al</em>. 2018).</p><p>We calculated distributed scatterer (DS) time series over 2014–2020 for Sentinel-1 relative orbits number (RON) 80 (descending) and 160 (ascending). The high frequency continuous in situ ground levelling measurements cover the snow and ice-free period of 2016 (April–October). Limited by the availability of Sentinel-1 data, 13 images from both stacks were evaluated against the levelling of a hummock plot. DS points used in the comparison were located around the plot at 125–315 m. The bog points were referenced to the stable DS points from a nearby village (4 km away) to account for atmospheric effects. InSAR line of sight deformation results were projected to vertical dimension (u<sub>LOS</sub>).</p><p>Concerning only the dates when we had SAR acquisitions, the largest change relative to the maximum surface level of the period is -6.6 cm and median change -2.4 cm for RON 80, and -7.5 cm and -2.4 cm for RON 160. The maximum deviation between the u<sub>LOS </sub>and the levelling is 5.6 cm and median 2.11 cm for RON 80. For RON 160, the maximum deviation is 5.85 cm and median 2.81 cm. The Spearman correlation coefficient (r<sub>s</sub>) between the u<sub>LOS </sub>and the levelling is 0.84 for RON 80 and r<sub>s </sub>= 0.81 for RON 160 (<em>p</em>-value <em>< </em>0.001 in both cases).</p><p>To reduce the need for ambiguity resolution in the DS time series, we used relative changes between two consecutive acquisitions (baseline of 12 or 6 days) instead of accounting for the absolute change. The in situ relative surface changes between the consecutive acquisition dates of RON 80 are -2.55...2.1 cm (median -0.08 cm) and the deviation of the DS from the levelling is -1.17...1.28 cm (median 0.38 cm). For RON 160, levelling values are -0.9...3.3 cm (median -0.3 cm) and the deviation -3.06…0.81 cm (the former is -0.45 if the 12-day image pair corresponding to the change larger than the u<sub>LOS </sub>height of ambiguity is removed), median 0.23 cm. Between the levelling and DS data r<sub>s </sub>= 0.67 (<em>p</em>-value 0.035) and 0.77 (<em>p</em>-value 0.005), respectively for RON 80 and 160. Based on the in situ levelling, we demonstrated that 1) Sentinel-1 DS time series severely underestimate real surface changes over the bog and 2) despite a serious ambiguity problem, DS time series contain the useful signal because 6-day surface changes are relatively small and usually do not need ambiguity resolution.</p>

Bridging Loss-of-Lock in InSAR Time Series of Distributed Scatterers

“…Due to the loss-of-lock phenomenon, attempting to interpret the entire ESM time series of phases at once is not possible and will result in several types of error, such as interpreting a noisedominated signal as real deformation, or phase unwrapping errors when transitioning from incoherent to coherent interferograms [6], [8], [9], [19], [20]. Thus, a different approach is required.…”

“…The typical approach to DS InSAR processing involves applying a minimum cost flow spatial unwrapping algorithm to the data, such as the well-known SNAPHU algorithm [26]. However, this approach is not well-suited to peatland observations due to rapid soil movements and the high degree of multilooking required [8], [20], [22]. Heterogeneity in both the type and depth of the soft soil layer of the Holocene will result in different responses to the seasonal weather conditions which the ground is exposed to, leading to spatial differences in the seasonal amplitude of the reversible displacement.…”

Spatio-Temporal Characteristics and Driving Mechanism of Alpine Peatland InSAR Surface Deformation—A Case Study of Maduo County, China

“…Umarhadi et al [17] assessed the relationship between subsidence rates and LULC in tropical peatlands using the SBAS-InSAR method. Tampuu et al [18] used the SBAS-InSAR technique and seasonal-annual search approach to reveal a consistent rate of peatland subsidence and validated the InSAR results through field surveys. Samuel et al [19] combined InSAR data, processed using the ASPIS algorithm to monitor ground motion between 2017 and 2021, with optical and LiDAR data to investigate the rate of subsidence across palsa peatlands in northern Sweden.…”