S0 S1 S2 S4 S8 S16 0.61 m -3.87 m (a) (b) (c) 0 6.28 15.76 rad -58.05 (d) (e) (f) FIgURE 8. (a) An a priori deformation model in the LOS direction with a negative value for displacement toward the satellite. (b) A scale image for the phase gradient estimation deduced from the a priori deformation model. S0 corresponding to the full-resolution single look complex image and Sn to the multilooking image after a complex average of n looks in range and 5n looks in azimuth. (c) The original differential interferogram. (d) A filtered interferogram by the multiscale phase gradient. (e) An unwrapped interferogram using the multiscale phase gradient by a least squares method. (f) A wrapped phase residual in the case of the 2005 Kashmir earthquake. (Figure used with permission from [33].) analysis of the redundancy and the complementarity between different measurements is of particular importance to provide useful information for the choice of the fusion strategy. Furthermore, the characterization of the displacement uncertainty is also essential for the choice of the appropriate fusion strategy. However, the uncertainty quantification investigation seems insufficient currently. In many studies, the detailed description of the displacement uncertainty is not available. 18 In the case of redundancy, if random uncertainty is present in the individual displacement measurement, all the measurements can be used jointly in linear inversion to maximize the reduction of the uncertainty associated with the fusion results, given enough computational capacity. For nonlinear inversion, the performance of this strategy depends on the data quality (i.e., noise level). This strategy can fail when it is difficult to adjust a model among a large number of noisy data. If systematic uncertainty is present ieee Geoscience and remote sensing magazine march 2016