In the fields of Geodesy and Geophysics, satellite Interferometric Synthetic-Aperture Radar (InSAR) is used to map deformation on Earth's surface. However, when InSAR satellites are used alone, their spatial resolution is limited due to atmospheric noise and the effects of General relativity on the GPS satellites. Medium-to-long wavelength (MLW) noise in areas where GPS stations are not uniformly distributed due to GPS clock drift are corrected using a machine learning de-noising algorithm.

So far, a systematic "rainbow" pattern characteristic of MLW noise has been removed from each image. In its place, we are left with a higher effective resolution.   Further analysis of the resolution increases can be checked by comparing the InSAR subsidence velocity with a velocity map produced with seismic analysis (REF). This work is a qualitative preliminary demonstration of the use of machine learning for GPS synchronization corrections, so future work may look at the quantitative gains produced.