Tm. Vandam et al., ATMOSPHERIC-PRESSURE LOADING EFFECTS ON GLOBAL POSITIONING SYSTEM COORDINATE DETERMINATIONS, J GEO R-SOL, 99(B12), 1994, pp. 23939-23950
Earth deformation signals caused by atmospheric pressure loading are d
etected in vertical position estimates at Global Positioning System (G
PS) stations, Surface displacements due to changes in atmospheric pres
sure account for up to 24% of the total variance in the GPS height est
imates. The detected loading signals are larger at higher latitudes wh
ere pressure variations are greatest; the largest effect is observed a
t Fairbanks, Alaska (latitude 65 degrees), with a signal RMS of 5 mm.
Out of 19 continuously operating GPS sites (with a mean of 281 daily s
olutions per site), 18 show a positive correlation between the GPS ver
tical estimates and the modeled loading displacements, Accounting for
loading reduces the variance of the vertical station positions on 12 o
f the 19 sites investigated. Removing the modeled pressure loading fro
m GPS determinations of baseline length for baselines longer than 6000
km reduces the variance on 73 of the 117 baselines investigated. The
slight increase in variance for some of the sites and baselines is con
sistent with expected statistical fluctuations. The results from most
stations are consistent with similar to 65% of the modeled pressure lo
ad being found in the GPS vertical position measurements. Removing an
annual signal from both the measured heights and the modeled load time
series leaves this value unchanged. The source of the remaining discr
epancy between the modeled and observed loading signal may be the resu
lt of (1) anisotropic effects in the Earth's loading response, (2) err
ors in GPS estimates of tropospheric delay, (3) errors in the surface
pressure data, or (4) annual signals in the time series of loading and
station heights. In addition, we find that using site dependent coeff
icients, determined by fitting local pressure to the modeled radial di
splacements, reduces the variance of the measured station heights as w
ell as of better than using the global convolution sum.