Publications

Zhang, L; Dong, J; Zhang, L; Wang, Y; Tang, W; Liao, MS (2022). Adaptive Fusion of Multi-Source Tropospheric Delay Estimates for InSAR Deformation Measurements. FRONTIERS IN ENVIRONMENTAL SCIENCE, 10, 859363.

Abstract
Atmospheric propagation delay correction is the key to improving the accuracy of deformation measurement of satellite interferometric synthetic aperture radar (InSAR). The empirical phase-elevation models and external data-based models present uneven performances of atmospheric delay correction for InSAR deformation monitoring. In this study, based on our previous fusion of delays predicted by multiple weather models (FDWM), we propose a new approach of adaptive fusion of multi-source tropospheric delay (AFMTD) estimates derived from multiple models over wide areas, i.e., ERA5, GACOS, WRF, MERRA2, NARR, MODIS, Linear model, and Powerlaw model. The spatially varying scaling algorithm is employed to refine the tropospheric delays predicted by the weather models. Meanwhile, we adopt a multiple-window strategy to cope with the spatially lateral variation of tropospheric delays. The AFMTD not only improves the spatial heterogeneity of tropospheric delay, but also adaptively combines multiple models to achieve a more reliable delay estimation. This AFMTD method is incorporated into the StaMPS-SBAS procedure. We compared the AFMTD with other single models using ENVISAT ASAR and Sentinel-1 datasets over Los Angeles of Southern California. The result of ASAR first demonstrates the effectiveness and reliability of the AFMTD method by referring to the assumed ground truth of simultaneous MERIS observations. The results of Sentinel-1 data show that over 95% of unwrapped interferograms have the minimum root-mean-square values after AFMTD correction for both descending and ascending tracks. The validation against GPS observation presents that the RMSEs of InSAR displacement time series after AFMTD correction decreases at more than 90% of 125 GPS stations. The average reductions of RMSE are 35.79% and 36.28% for descending and ascending data, respectively, and the maximum improvement is more than 70%. Overall, the proposed AFMTD method outperforms any single model for InSAR tropospheric delay correction and provides an open framework to fuse multi-source tropospheric delay estimates.

DOI:
10.3389/fenvs.2022.859363

ISSN:
2296-665X