<p class="western" style="line-height: 150%; margin-bottom: 0in;">Recent atmospheric warming has led to widespread permafrost degradation and subsidence induced by ground ice thaw across the pan-Arctic permafrost domain of North America and Eurasia. Reported subsidence rates reach up to 2 cm/year in areas with low ice content and exceed 3 cm/year in ice-rich regions. Quantifying long-term subsidence trends over extensive areas is critically important to detect and assess permafrost changes and impacts. Unlike the development of localized thermokarst features, the isotropic thaw subsidence is relatively slow and spatially more uniform across broad spatial scales. Consequently, long-term monitoring at numerous sites and over large areas is essential for understanding the impacts of permafrost thaw and for improving numerical simulations of permafrost degradation and ground surface subsidence. Here, we show that Sentinel-1 SAR interferometry can be successfully used to monitor long-term surface subsidence rates in the Lena Delta in northeastern Siberia. The observed ground surface subsidence rates between 2016 and 2021 vary significantly among the three geomorphological units of the Lena Delta and turn out to be closely linked to ground ice content. The largest subsidence rate, at 16.0 ± 7.2 mm/year, occurs in the Yedoma, the most ground-ice rich geomorphological unit. In contrast, the most stable geomorphological unit, with subsidence rates of 1.0 ± 4.4 mm/year, is associated with a predominantly sand-dominated lithology having low ground-ice content and dry vegetation cover. The youngest and most active part of the delta, consisting of Holocene river terraces with polygonal tundra and the present-day floodplains, is characterized by heterogeneous subsidence rates of 7.0 ± 5.3 mm/year. For continuous monitoring of long-term ground surface subsidence over larger Arctic regions using Sentinel-1 data, a long series of satellite acquisitions with preserved long-term coherence (e.g., over one year of summer acquisitions) is required. While Sentinel-1 C-band data suffer from poor coherence over one-year intervals, SAR systems with longer wavelengths (e.g., L-band) can achieve better performance for this purpose.