Improved surface displacement estimation through stacking cross-correlation spectra from multi-channel imagery

Studying sporadic and complex geophysical surface flows, like earthquakes or sea surface circulation, are challenging cases. If a satellite is able to image an event, it becomes essential to pull out as much information as possible. In this contribution we demonstrate a method to increase the coverage and signal-to-noise ratio for displacement estimation, making such surface flow estimates more complete. We leverage upon the redundant offset information acquired by multi-channel push-broom imagery. The individual cross-correlation spectra (cross power spectral density; Fourier transform of the cross-correlation function) of different spectral bands are averaged in the frequency domain before sub-pixel offset-estimation by phase-plane fitting. The method is demonstrated near Kaikōura, where in 2016 a surface rupture occurred. RapidEye data from two different dates were used to reconstruct the displacement. In addition, the circulation along the coast is estimated from data from a single date where multiple spectral bands were acquired within seconds which made stacking of cross-correlation spectra possible. The demonstrated methodology is applied to data from the already decommissioned RapidEye constellation, but can be adopted to other pushbroom systems, such as the Landsat legacy or Sentinel-2.