Validation of point scatterer phase statistics in multi-pass InSAR

Abstract

In radar images, point scatterers provide reflections that can often be attributed to a single physical object, usually smaller than the resolution cell size. Stable point scatterers with high backscatter coefficients through time (mainly 'man-made features') are potential targets for InSAR deformation analysis. A-priori knowledge about the observation statistics of InSAR point scatterer measurements is necessary to give an independent quality description of the estimated (deformation) parameters. Using the Signal-to-Noise Ratio (SCR) from SAR images an a-priori estimate of the phase variance can be obtained. A controlled corner reflector experiment has been set up with leveling as an independent validation technique. In the period from March 2003 to June 2004 the movements of five corner reflectors in the area near Delft University of Technology have been monitored using leveling and repeat-pass InSAR (ERS-2 and Envisat). Double-difference observations, the first observations that bear interpretable information, are used for comparing the InSAR and leveling measurements. The stochastic model is set up as a variance-covariance matrix, estimating a-posteriori variance factors for the observation groups leveling, ERS-2 and Envisat. From the results it can be concluded that the a-priori phase variance based on SCR for the corner reflectors (σ ∼0.4mm) is significantly overestimated. The estimated a-posteriori InSAR phase standard deviation is 3.1 mm for Envisat and 3.9 mm for ERS-2 (mm, vertical).