On the Consistency of Tropospheric Delays Over Mountainous Terrain Retrieved From Persistent Scatterer Interferometry, GNSS, and Numerical Weather Prediction Models

The tropospheric refraction along the signal path is the same for global navigation satellite systems (GNSS) and radar interferometry. However, different observation geometries, spatiotemporal sampling, signal processing methods, as well as signal wavelengths yield rather complementary measurements. The origin of this research is the question whether tropospheric delays retrieved at GNSS permanent stations can support persistent scatterer interferometry (PSI) processing for the retrieval of surface displacement in mountainous terrain, which is challenging because of spatial gaps due to radar layovers, shadowing, and temporal decorrelation in combination with strong variations of water vapor. We analyze maps of tropospheric path delays obtained by collocation of GNSS-estimated delays and PSI processing of an interferometric stack of Cosmo SkyMed X-band synthetic aperture radar (SAR) data in a mountainous region in Valais, Switzerland. We aim to assess the consistency and differences among the datasets to better understand their ability to sense small-scale structures in the lower atmosphere. In addition, we compare them with maps of tropospheric path delays derived from Consortium for Small-Scale Modelling (COSMO-2) numerical weather model (NWM) data. We investigate several factors affecting the interpolation of the GNSS zenith delays to the locations of the persistent scatterers, such as assumptions in the collocation, network size, and resolution. We assessed the meteorological parameters of the NWM to find potential correlations between specific meteorological conditions and different levels of (dis)agreement of delay maps; a clear correlation was not found. We found that the delays estimated from collocated GNSS measurements and PSI tend to have a different dependency on the terrain altitude. The PSI-derived path delays obtained from the X-band SAR data stack capture small-scale spatial variations also visible in NWM delay maps; whereas, at a larger scale, mismatches are found. It appears that the current GNSS network in the mountainous area of the Valais is not dense enough to capture strongly varying tropospheric refraction. We can conclude that denser networks (with a resolution of 5–10 km) in the interferometric SAR (InSAR) footprint region and a careful choice of the assumptions in our interpolation method would make GNSS more suitable for helping PSI processing.