High‐Resolution Water Vapor Maps Obtained by Merging Interferometric Synthetic Aperture Radar and GPS Measurements

Abstract

AbstractPrevious research has demonstrated several successful meteorological applications of interferometric synthetic aperture radar (InSAR) by mapping high‐resolution precipitable water vapor (PWV) over large areas with satisfactory precision (e.g., 1–2 mm). InSAR, however, can only measure temporal changes in the PWV, that is, the differential PWV (ΔPWV) between two SAR acquisition epochs. The construction of absolute PWV maps using InSAR observations remains an intractable problem. We present here a new method for constructing high‐resolution PWV maps by fusing InSAR and global positioning system (GPS) measurements. Our method involves: (1) extracting temporal differences in the zenith wet delays (ΔZWD) from InSAR observations with the assistance of synchronous GPS measurements, and (2) using a constrained minimum variance estimator (CMVE) to construct the high‐resolution maps of the absolute ZWDs by fusing the InSAR‐ΔZWD and the GPS‐ZWD samples, and finally (3) transforming the high‐resolution ZWDs to PWVs using an elevation‐dependent proportionality. Only a single interferogram was needed in our method, so we did not use any temporal (i.e., time‐series) hypotheses, for example, we did not assume that the average value of the temporal turbulent ZWDs was equal to zero. We validated the new method over the Southern California region using four Sentinel‐1 interferograms related to eight SAR images acquired in different seasons and in different years. We also used synchronous GPS measurements from the Southern California integrated GPS network. Eight high‐resolution maps of PWVs related to the eight SAR acquisitions were generated, covering ∼250 km in area at a ∼160 m spatial resolution. Comparisons between the CMVE method and the other two methods (GPS‐only‐based interpolation and time‐series InSAR stacking) were conducted. Experimental results showed that the CMVE method performed significantly better than conventional methods for constructing high‐resolution PWV maps, which is of great interest to a wide community of geophysicists and meteorologists.