Neutral atmospheric delay in interferometric synthetic aperture radar applications: Statistical description and mitigation

Abstract

Variations in the refractive index of the atmosphere cause variations in satellite‐based interferometric synthetic aperture radar (InSAR) observations. We can mitigate tropospheric effects by averaging N ‐independent interferograms. Because the neutral atmosphere is uncorrelated at timescales longer than 1 day, using this technique statistically reduces the variance, σ 2 , of the noise by a factor of N . Using zenith neutral atmospheric delays from Global Positioning System (GPS) data from the Southern California Integrated GPS Network, we find that the average variance depends on the distance between observations, L , and height difference, H , as σ = c L α + kH with estimated values for c , α, and k of about 2.5, 0.5, and 4.8, respectively, where σ is in mm and L and H are in km. We expect that the value of α is largely site‐independent but the value of c will depend on the water vapor variability of the area of interest. This model is valid over a range of L between approximately 10 and 800 km. Height differences between 0 and 3 km have been used in this analysis. For distances of 100 and 10 km with negligible height differences, σ is estimated to be approximately 25 and 8 mm, respectively. For a given orbit revisit time and image archive duration, we calculate the number and duration (assumed constant) of interferograms required to achieve a desired sensitivity to deformation rate at a given length scale. Assuming neutral atmosphere is the dominant source of noise, a 30° look angle, and an image revisit time of 7 days, detection of a deformation rate of 1 mm yr −1 over distances of 10 km requires about 2.2 years of continuous observations. Given our results, we suggest a data covariance structure to use when using InSAR data to constrain geophysical models.