An analysis of the uncertainty in InSAR deformation measurements for groundwater applications in the San Luis Valley, Colorado

Interferometric synthetic aperture radar (InSAR) data provide spatially dense maps of the deformation of Earth's surface, where one pixel represents 50 m by 50 m. At a number of field sites it has been shown that the deformation measured by InSAR is related to changes in hydraulic head in underlying confined aquifer systems. In an agricultural area in the San Luis Valley (SLV), Colorado, we have shown that InSAR deformation measurements exhibit seasonal trends similar to hydraulic head measurements made in wells. However, when we attempted to estimate hydraulic head from the InSAR deformation measurements we found that the estimate of the uncertainty in these measurements was not accurate enough to assess the agreement between the two datasets. Here we more accurately estimate the uncertainty in the InSAR deformation measurements and present a methodology that uses this uncertainty to optimally process the data for groundwater applications. In this study we used data from the ERS-1 and ERS-2 satellites, which have 30 acquisitions archived from 1992 - 2001. Small Baseline Subset (SBAS) analysis was used to produce a time-series of the deformation for all pixels with data quality above a selected threshold. The deformation is derived from a change in the electromagnetic phase between two different acquisition times. Four major components contribute to uncertainty in the measurement of the phase ( ): integer phase ambiguities, incorrect orbital parameters, atmospheric phase effects and decorrelation of radar signals. We assumed that uncertainty due to integer phase ambiguities and incorrect orbital parameters is small in the SLV. We proceeded to address the two other components of uncertainty, atmospheric phase effects and decorrelation of radar signals. We first used the trends in the hydraulic head to help us identify acquisitions that may have been corrupted by atmospheric phase effects. We found that our technique works well with synthetic data. However, with real data from the SLV we observed an inherent trade off between the number of data samples and the amount of uncertainty due to atmospheric phase effects. Next we quantified the uncertainty in the phase due to decorrelation of radar signals. We then propagated this uncertainty through the SBAS processing chain to produce the variance of the deformation estimates at each acquisition time. Having developed a methodology, we then used synthetic data to better understand how to select pixels for SBAS processing. It is common for the selection of pixels to be based on how the radar signals decorrelate over time. However, having shown that the decorrelation is related to the uncertainty in the deformation estimate, we were able to show how one could select pixels based on the desired uncertainty in the final deformation estimate. The ultimate goal of our research is to investigate how we can use InSAR deformation measurements to estimate hydraulic head in confined aquifer systems. Quantifying and understanding the uncertainty in the deformation measurement is an important first step in this process.