Zhang Yunjun , Heresh Fattahi, Xiaoqing Pi, Paul Rosen, Mark Simons, Piyush Agram, Yosuke Aoki

Time series analysis of synthetic aperture radar (SAR) and interferometric SAR generally starts with coregistration for the precise alignment of the stack of images. Here we introduce a model-adjusted geometrical image coregistration (MAGIC) algorithm for stack coregistration. This algorithm corrects for atmospheric propagation delays and known surface motions using existing models and ensures simplicity and computational efficiency in the data processing systems. We validate this approach by evaluating the impact of different geolocation errors on stacks of C-band Sentinel-1 and L-band ALOS-2 data, with a focus on the ionosphere. Our results show that the impact of the ionosphere dominates Sentinel-1 ascending (dusk-side) orbit and ALOS-2 data. After correcting for ionosphere using the JPL high resolution global ionospheric maps, with topside total electron content (TEC) estimated from GPS receivers onboard the Sentinel-1 platforms, solid Earth tides, and troposphere, the mis-registration RMSE reduces by over a factor of four from 0.20 m to 0.05 m for Sentinel-1 and from 2.66 m to 0.56 m for ALOS-2. The results demonstrate that for Sentinel-1, the MAGIC approach is accurate enough in the range direction for most applications including interferometry; while for the L-band SAR, it can be potentially accurate enough if topside TEC is available. Based on our current understanding of different error sources, we evaluate the expected range geolocation error budget for the upcoming NISAR mission with an upper bound of the relative geolocation error of 1.3 m and 0.2 m for its L- and S-band SAR, respectively.