Premium
Comparison of Kalman filter estimates of zenith atmospheric path delays using the Global Positioning System and very long baseline interferometry
Author(s) -
Tralli David M.,
Lichten Stephen M.,
Herring Thomas A.
Publication year - 1992
Publication title -
radio science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.371
H-Index - 84
eISSN - 1944-799X
pISSN - 0048-6604
DOI - 10.1029/92rs01779
Subject(s) - zenith , very long baseline interferometry , global positioning system , geodesy , remote sensing , geodetic datum , kalman filter , environmental science , standard deviation , root mean square , meteorology , geology , mathematics , physics , computer science , statistics , telecommunications , quantum mechanics
Kalman filter estimates of zenith nondispersive atmospheric path delays at Westford, Massachusetts, Fort Davis, Texas, and Mojave, California, were obtained from independent analyses of data collected during January and February 1988 using the Global Positioning System (GPS) and very long baseline interferometry (VLBI). The apparent accuracy of the path delays is inferred by examining the estimates and covariances from both sets of data. The ability of the geodetic data to resolve zenith path delay fluctuations is determined by comparing further the GPS Kalman filter estimates with corresponding wet path delays derived from water vapor radiometric (WVR) data available at Mojave over two 8‐hour data spans within the comparison period. GPS and VLBI zenith path delay estimates agree well within one standard deviation formal uncertainties (from 10–20 mm for GPS and 3–15 mm for VLBI) in four out of the five possible comparisons, with maximum differences of 5 and 21 mm over 8‐ to 12‐hour data spans. For one comparison, the maximum difference between GPS and VLBI is 50 ± 20 mm and clearly shows an unexplained systematic difference which is probably related to poor elevation angle coverage in the VLBI data at the time, GPS fiducial network sensitivity, and poor azimuthal coverage. The root‐mean‐square differences between GPS estimates of the total path delays and WVR measurements added to the hydrostatic delay component determined from surface barometric pressure data are between 9 and 15 mm, however, with biases of 10–15 mm.