G51D-01 08:00h
Detection of time-varying deformation signals in continuous GPS data
With increasingly large networks of continuous GPS receivers like the Plate Boundary Observatory, the need is clear for an automated way to screen the data for evidence of temporally varying deformation. An automated method will not only reduce the burden on network operators but also introduce an objective way of analyzing time series in order to detect a subtle signal that may not be discernable by eye. Several factors complicate the detection of such signals. (1) The potential source geometry is unknown. (2) The spatial distribution of stations is irregular. (3) The data are contaminated by time-varying noise sources that may mimic transient deformation. Methods have previously been developed to extract and model transient deformation signals in the context of a known deformation source by exploiting the spatial coherence of a source signal and the spatial incoherence of local time-varying noise. In contrast, this work is focused on development of an analysis tool that will exploit the spatial coherence of deformation signals independent of any assumed model geometry. Several approaches have been explored including the use of splines, principal mode analysis, rate-change detection in time series, and weighted averages of neighboring station displacements. The varying geometry and station density of existing networks, and the expected range in the source complexity and temporal character of transient signals suggest using a hierarchy of detection methods. In such an approach, initial analyses will flag areas for further examination. Ultimately, once the geographic location of a potential signal has been refined, source-specific modeling can be pursued. Application of these approaches using time series from SCIGN, PANGA, and GEONET and a range of input synthetic transient signals illustrates the strengths and limitations of these methods in detecting and characterizing aseismic transients.
G51D-02 INVITED 08:15h
Lower Crustal Transients and Surface Mass Transport: Time Series Filtering for Signal Detection at Two Spatial Extremes.
Two examples are presented of GPS time series filtering at two spatial extremes: (1) the recent discovery of a lower-crustal transient event by filtering out "common mode" errors and seasonal signals [Smith et al., Science, 305, 2004]; and (2) the detection of global-scale surface mass transport by use of spherical harmonics to smooth through local- to regional-scale noise [Gross et al., GRL, 31, 2004]. These examples illustrate the paradigm that one geodesist's noise is another's signal. Specific attention is given to the actual methodology applied to extract the signals of interest at these two spatial extremes, starting at the level of the actual GPS data processing, through to time series analysis, with a discussion on relevant reference frame issues.
G51D-03 08:30h
Strong Ionospheric Disturbances Observed by a Dense GPS Array After Large Earthquakes: Case Study of the 2003 Tokachi-oki Earthquake and its Geophysical Mechanism
Ionospheric disturbances have been detected after, e.g. Northridge (Calais and Minster, 1995) and Denali (Ducic et al., 2003) earthquakes. Similar signals observed after the 2003 Tokachi-Oki Earthquake, the largest earthquake in Japan after the completion of GEONET, a nationwide array composed of over 1000 CGPS stations. We followed a standard procedure: applying a band-pass filter for the ionospheric combination of the L1 and L2 phase signals and calculating subioonospheric points (SIP) assuming thin ionosphere at the height of 350 km. Owing to the high density of SIP, many interesting features are observed and several important parameters were constrained, e.g. (1) apparent propagation speed, (2) directivity of disturbance signals, (3) decay during propagation, etc. As for (1), the observed speed of about 1 km/sec is significantly smaller than the Rayleigh Wave velocity, significantly faster than Travelling Ionospheric Disturbances (TID), but is consistent with the sound velocity at the ionospheric heights. The acoustic wave generated by sudden vertical movement of the Earth's surface first propagate upward. Then it will be refracted by height-dependent velocity structure resulting in horizontally propagating wave through the ionosphere. The observed TEC variation, with a wavelength of a few hundred km, may reflect electron density oscillation caused by the passage of such an acoustic wave. Regarding (2), there was a clear indication that the wave does not propagate northward. As first suggested by Calais et al. (1998), such a blocking is considered to be due to interaction between the geomagnetic field and the movement of charged particles comprising the ionosphere associated with the acoustic wave propagation. The model predicts that there will be no southward propagation of ionospheric disturbances caused by earthquakes in southern hemisphere mid-latitudes, which needs be confirmed by future earthquakes. The point (3) enabled the authors to define the empirical equation to calculate "Ionospheric disturbance magnitude" using the focal distance and disturbance amplitudes. Because the ionospheric disturbance monitoring does not require precise orbit information, such magnitudes could be determined near real time. This may help us, e.g. issue early warning message of Tsunami.
G51D-04 INVITED 08:45h
The effect of the 2nd order ionosphere correction on GPS orbits and station positions
The Global Positioning System (GPS) transmits two frequencies, allowing users to correct for the first-order ionospheric signal group delay (or phase advance) that can be as large as 100 meters on L1 frequency. The second-order ionospheric term, caused by the Faraday rotation effect induced by the Earth magnetic field, is about 1000 times smaller and usually ignored. Implementation of the 2nd order correction suggested by Bassiri and Hajj [1993] was shown by Kedar et al. [2003] to affect station positions from sub-daily to annual time-scale. The correction causes a latitude-dependent 0.1-0.5 cm southward shift in the position, which is roughly proportional to the integrated electron density above the receiver, and has strong diurnal, seasonal and decadal signatures. We will present new results incorporating the 2nd order correction in routine sub-daily and daily data processing, and discuss its effect on station positions as will as on GPS orbit estimates.
G51D-05 INVITED 09:00h
Testing monument stability with short-baseline GPS measurements
The two permanent GPS sites at Pinon Flat Observatory (PFO)- called, naturally, PIN1 and PIN2 -are among the oldest permanent GPS sites in southern California. These sites were built in 1989 as prototypes for a continuous GPS network, an idea only just being implemented in a few places. The monumentation (how the GPS antenna was attached to the Earth) used at these sites was based on long experience operating several laser strainmeters at PFO, and is the basis for the drilled-braced monument design currently used by networks such as SCIGN and PBO. The record of GPS measurements between these sites provides long-term data between two closely spaced and well-anchored monuments, and so gives us a unique perspective on potential sources of noise such as monument stability, near field environmental effects and equipment effects. Because the two sites are only 50 meters apart, many sources of noise (and signal) are common to both, and cancels when we estimate the baseline vector between them. The day-to-day scatter in the baseline series (1-sigma 0.07 mm in the horizontal) is almost a factor of 10 smaller than is typical for a globally-referenced position series, and significantly less than the scatter from differencing the global series between these sites (0.35 mm). We present the results of a Maximum Likelihood Estimation (MLE) of the noise content which we assess in the context of monument stability. We also discuss an unexpected annual signal in the north-south component, a change in the white noise content attributable to an equipment change and outliers and offsets due to known local environment changes.
G51D-06 09:15h
Consistent Long-Time Series of GPS Satellite Antenna Phase Center Corrections
The current IGS processing strategy disregards satellite antenna phase center variations (pcvs) depending on the nadir angle and applies block-specific phase center offsets only. However, the transition from relative to absolute receiver antenna corrections presently under discussion necessitates the consideration of satellite antenna pcvs. Moreover, studies of several groups have shown that the offsets are not homogeneous within a satellite block. Manufacturer specifications seem to confirm this assumption. In order to get best possible antenna corrections, consistent ten-year time series (1994-2004) of satellite-specific pcvs and offsets were generated. This challenging effort became possible as part of the reprocessing of a global GPS network currently performed by the Technical Universities of Munich and Dresden. The data of about 160 stations since the official start of the IGS in 1994 have been reprocessed, as today's GPS time series are mostly inhomogeneous and inconsistent due to continuous improvements in the processing strategies and modeling of global GPS solutions. An analysis of the signals contained in the time series of the phase center offsets demonstrates amplitudes on the decimeter level, at least one order of magnitude worse than the desired accuracy. The periods partly arise from the GPS orbit configuration, as the orientation of the orbit planes with regard to the inertial system repeats after about 350 days due to the rotation of the ascending nodes. In addition, the rms values of the X- and Y-offsets show a high correlation with the angle between the orbit plane and the direction to the sun. The time series of the pcvs mainly point at the correlation with the global terrestrial scale. Solutions with relative and absolute phase center corrections, with block- and satellite-specific satellite antenna corrections demonstrate the effect of this parameter group on other global GPS parameters such as the terrestrial scale, station velocities, the geocenter position or the tropospheric delays. Thus, deeper insight into the so-called `Bermuda triangle' of several highly correlated parameters is given.
G51D-07 09:30h
An Atmospheric Turbulence-Based GPS Stochastic Model
The precision of GPS position estimates depends on several error sources. One of the limiting GPS error sources is due to fluctuations of water vapor content in the atmosphere, which causes variations in slant-path propagation delays. In GPS data processing, atmospheric propagation delays are typically modeled using a simplified parametric model that does not fully account for water vapor fluctuations. We have developed a stochastic model for delay variations in GPS that is based on the fluctuations of the atmospheric refractive index. We have implemented this model in a least-squares method and, using GPS data, have thus the ability to estimate a site-dependent, time-varying turbulence strength parameter. Preliminary results are very promising and suggest that turbulence is an important error source in GPS. We will describe this variance-covariance model and the scaling turbulence variance estimates. These results may have implications for the interpretation of signals, and its separation from noise, in analysis of GPS time series, and may benefit both geodesists and meteorologists.
G51D-08 09:45h
Can a high resolution Numerical Weather Model (NWM) improve site positions?
In recent years, NWM-based mapping functions have utilized global gridded NWMs with horizontal resolutions of 50km to 200km. These have resulted in significant improvement in estimates of the hydrostatic component of the delay and estimated site positions, while little or no improvement has been realized in estimates of the wet component. This lack of improvement has been attributed to the much smaller horizontal scale for variation in the water vapor distribution ($<$1-5 km) than is represented in the global NWM (50-200 km). We have begun an investigation to 1) see if mapping functions derived from a high-resolution numerical weather model can improve the wet delay component, and 2) evaluate site position errors as a function of NWM resolution. Eventually we will look at the effects of time-varying azimuthal asymmetry, since NWMs can provide a full 4-dimensional (three spatial dimensions, plus time) picture of the distribution of mass in the atmosphere. As a first step, however, we report preliminary results using only symmetric mapping functions. Using radiosonde profiles as "truth" for calculating the mapping functions, we will evaluate site position errors (bias and RMS) for mapping functions calculated from MM5 runs at four horizontal resolutions, from 81 km down to 3 km. For our evaluation, we use the eight globally distributed sites of the CONT02 VLBI observation sessions which covered fifteen successive days in 2002 October. The effective site position errors will be calculated for the GPS constellation specific to each site. We anticipate that the results will provide guidance on the minimum resolution required for NWM-derived azimuthally symmetric mapping functions.