Approximately 44,500 Sentinel-1 SAR acquisitions were processed to generate around 341,400 interferograms at ~100 m resolution using the automated COMET-LiCSAR system. The Tibetan Plateau was divided into 127 ascending and 114 descending frames (each approximately 250 km wide), forming short-baseline interferometric networks with additional 6- and 12-month pairs to reduce phase bias.

Using LiCSBAS, line-of-sight displacement time series and average velocities at ~1 km resolution were estimated. Corrections were applied for tropospheric, ionospheric, and solid Earth tidal effects. Velocity uncertainties were flattened via semi-variogram analysis, and Eurasian plate motion was subtracted from all frames.

A total of 18,203 GNSS velocities and 6,607 levelling rates were compiled from 131 studies published since 2013. The dataset was refined by removing outliers, aligning studies via Euler poles, eliminating outdated or duplicate entries, and merging collocated measurements.

To construct a unified coarse velocity model integrating GNSS, levelling, and InSAR data, the velmap approach was followed. This involved inverting for 3D velocities at each node of a triangular mesh spaced by ~0.2° in longitude and latitude, along with frame-specific reference frame adjustment parameters and linear-with-height atmospheric correction terms. The reference frame adjustment parameters consisted of a second-order polynomial surface, and regularization was applied using Laplacian smoothing.

East-west and vertical velocities were derived from georeferenced mosaics of ascending and descending line-of-sight velocities, using coarse north-south velocities as constraints. Horizontal strain and rotation rates were calculated from velocity gradients, with spherical corrections applied. A 150 km median filter was applied to east-west velocities to balance resolution and noise.

Further details are available in Wright et al. (2025, Science).