Advances in Analyzing InSAR Persistent Scatterers in Non-urban Environments
Andrew Hooper(1) , Howard Zebker(1)
, and Paul Segall(1)
In Hooper et al. (GRL, 2004) we presented a new InSAR persistent scatterer (PS) method for analyzing episodic crustal deformation in non-urban environments. We have since made significant improvements to the method as well as successfully applying it in various settings, including those where conventional interferometry has proven difficult.
Our algorithm for detecting PS depends on spatial correlation with other PS. In our original algorithm we assumed a fixed length-scale for the spatial correlation over the entire region of interest. However, the actual distance over which PS are correlated, which is primarily dependent on atmospheric and deformation gradients, will likely vary over the region. We have modified the algorithm to allow for a spatially-variable length-scale that is determined from the data itself using a modified adaptive phase filter. We validated the method in the Lost Hills region of California where the deformation gradient due to oilfield-related subsidence varies greatly across the region.
We have also developed three-dimensional unwrapping algorithms to incorporate the time dimension characteristic of PS datasets, in addition to the two usual spatial dimensions. These algorithms enable us to unwrap the PS more reliably than using existing two-dimensional algorithms, as demonstrated on our Long Valley dataset.
The PS method provides a means to remove atmospheric noise and produces time series that are convenient for deformation modeling. A further potential advantage comes in being able to tease out a signal in vegetated areas where conventional InSAR does not work well. One such area is Mt St Helens, which began erupting in September 2004 after a hiatus of 13 years. However, it is not known when the lava currently erupting entered shallow depths. One potential time period is during the earthquake swarms in the first half of 1998. GPS data covering this period suggest little deformation, but are noisy. Using our PS method we analyzed SAR data acquired over this time period and also concluded that no significant deformation occurred.