A Spectral Model for Multilook InSAR Phase Noise Due to Geometric Decorrelation

We find that the multilook phase noise associated with geometric decorrelation in a synthetic aperture radar (SAR) interferogram does not follow the spectral shape of the signal in the range dimension because the phase noise arises from the uncorrelated ends of the frequency bands of the single-channel radar data contributing to the interferogram. The phase noise due to geometric decorrelation therefore diminishes much more quickly than the square root of the number of looks in range when the number of looks is large, and it is thus not well characterized by the common practice of computing the multilook phase noise from the Cramer-Rao bound as a function only of the total coherence and the total number of looks. Large discrepancies may result in cases for which the phase noise due to geometric decorrelation dominates, such as when simulating data with no receiver thermal noise. We present here a more accurate model of the multilook phase noise due to geometric decorrelation in addition to thermal noise. Our model depends on the spectra of the individual channels and on the spectrum of the multilook averaging window in range. Use of this model may have bearing on decisions of whether common-band filtering is necessary for particular applications.