| Abstract | GPS and InSAR are the primary geodetic tools for measuring plate boundary crustal deformation. The existing EarthScope PBO project maintains over 572 GPS stations spanning the 2004 - 2008 epoch, and satellites such as ERS, ENVISAT and ALOS provide an enormous InSAR dataset spanning 1992 - 2008 through the WInSAR archive. However, the optimal way to combine information from these two sources is not clear. GPS data, with an average spacing of ~10 km in California, provide better constraints on long wavelength deformation signals (~40 km), while InSAR, with 100 m spatial sampling, provides better constraints on short wavelength signals. Interseismic deformation must be accurately measured (error < 2 mm/yr) over this entire range of length scales to be able to estimate the depth and extent of locked fault patches where stress is accumulating. We are developing a remove/filter/restore procedure to optimally combine GPS and InSAR observations. The first step is to construct a wide-area 3-D dislocation model based only on the GPS data. Then the dislocation model is removed from each LOS interferogram. Next, the residues are high-pass filtered using a Gaussian filter with 0.5 gain at 40 km based on the spacing of GPS stations in California. Because the residual interseismic signal and noise have different scale dependencies, filtering an interferogram can increase the signal-to-noise ratio by as much as 20%. Multiple interferograms are stacked to further reduce the atmospheric error. Finally the complete deformation field is constructed by adding the GPS based dislocation model to the stack of filtered interferograms. Applying this procedure to a large stack of ERS interferograms in Southern California demonstrates better than 2 mm/yr accuracy over a wide range of length scales (100 m to 500 km). Extending the method to Northern California, where InSAR coherence is poor at C-band, will require using the longer wavelength L-band data from ALOS, which is shown in the data we processed. Our analysis shows that ALOS data will be able to make major contributions toward measuring interseismic deformation in about 3 years. |
|
GPS spacing in California. (a) GPS distribution in California and the San Andreas fault. The solid line is the main trace of the SAF and the dots are GPS stations used in this study. (b) Histogram of distance between GPS stations. The bin size is 5 km and the red line is the best fitting Poisson probability density function with a lamda of 16.7 km. (c1-4) Histogram of distance from GPS stations to the SAF. The cumulative histogram is normalized in the way that dividing number of stations within a certain distance from the fault by the number of segments the SAF can be divided with this distance. (d1-3) Normalized cumulative histograms for three groups: 1. Northern California (y axis 950-1300 km); 2. Central California (y axis 700-950 km); 3. Southern California (y aixs 400-700 km). |
|
| Filter effect on interseismic signal observed by ALOS ascending interferograms. (a) A synthetic ALOS ascending interferogram based on GPS data. (b1-4) filtered interferograms with different Gaussian filters. The number on the top right corner is the 0.5 gain wavelength of the Gaussian filter. (c1) Maximum signal for four different areas changes with different filters. (c2) Signal-to-noise ratio changes with different filters. As the data shows, filtering can increase the signal-to-noise ratio by as much as 20% compared to no filtering. |
|
