Modeling interferograrn stacks

Synthetic aperture radar interferometry is limited by temporal and geometrical decorrelation. Permanent scatterers (PSs) are helpful to overcome these problems, but their density in agricultural and out-of-town areas is not always sufficient. The forthcoming availability of satellite platforms with thinner orbital tubes and shorter revisit times will enhance the use of interfero-gram stacks, which are usable for distributed and progressively decorrelating targets, like those found in agricultural areas. To estimate the possibilities of the interferogram stack technique, a Markovian model for the temporal decorrelation is considered. ERS-1 data measured in C-band over Rome with a three-day repeat cycle are used to identify the parameters for this model, namely, the decorrelation time (estimated as 40 days) and the short-term coherence (estimated as 0.6). In the hypothesis of small deviations from a model of the motion, the optimal weights to be used to combine a sequence of interferograms taken at intervals that are shorter than the decorrelation time are calculated in the cases of progressive and sinusoidal ground motion. The dispersion of the optimal estimate of the motion is then determined. This model is extended to frequencies other than C-band. These evaluations are compared with the known results obtained for PSs. As an example, the case of a time interval between the takes of T = 12 days is considered. With N consecutive images, interferograrn stack results are equivalent to PSs if the pixel count in the window used to smooth the interferograms grows with N-2.