G21B-01 INVITED 08:00h
Coordinate Time Series Generation
Analyses of global positioning system (GPS) and other spaced based geodetic data generate results which have rank deficiencies and poorly resolved components that need to be resolved through the realization of a reference frame. A true rank deficiency exists for very long baseline interferometry (VLBI) measurements to extragalatic radio sources because these measurements are effectively insensitive to translation of the terrestrial coordinate system. Systems that use earth orbiting satellites have sensitivity to the center of mass of the Earth but the stability of this origin can be compromised by mismodeled orbital dynamics. Rotational rank deficiencies are common to all space based geodetic systems primarily because the transformation between terrestrial and celestrial frames is through the rotation axis of the Earth which moves by many meters with respect to the body of the Earth. Scale might not be considered a rank deficiency in most systems, but uncertain knowledge of the antenna characteristics on systems such as GPS can lead to scale uncertainties. In this paper, we examine methods for reference frame realizations on spatial scales of kilometers to global. The method we use is based on Bayesean estimation with conditions applied through the covariance matrix of estimated positions. Examples will be shown for times series generated on different temporal and spatial scales with emphasis on potential systematic differences in the time series that can develop with various frame realization assumptions.
G21B-02 INVITED 08:20h
Spatial and temporal filtering for high-rate GPS regional networks: Implications for reference frame definition
GPS data from regional networks such as those established in the Western U.S. over the last decade (BARD, BARGEN, EBRY, PANGA, SCIGN) are routinely processed with respect to ITRF in 24-hour segments, based on 30 s sampled data (often decimated to even lower rates). (The Plate Boundary Observatory's strategy appears to be the same, although data will be collected routinely at a 15 s sampling rate.) The result is a time series of daily site positions, which are spatially filtered to remove a common mode signal. This spatial filtering process is performed simultaneously with a parameter fit to the time series that includes rates (velocities), offsets (coseismic and/or instrumental), seasonal terms (annual and semiannual), and perhaps postseismic terms (exponential or logarithmic). There are several variations to this approach, as well as several complications, such as when the network in question becomes larger. Nevertheless, the spatial filtering results in a significant reduction in the rms for the resulting daily position time series, about 1 mm in the horizontal components, and 3-4 mm in the vertical. There is an ongoing effort in southern California to upgrade SCIGN stations to high-rate sampling (1 Hz or higher), and real-time ($<$1 s latency) streaming and instantaneous positioning. More than 50 stations have already been completed. I demonstrate with upgraded SCIGN stations that routine instantaneous positioning of ultra-high-rate data (10-20 Hz) is possible and provides several advantages over the traditional approach mentioned above that can be exploited to increase the sensitivity of CGPS data for the detection of transient deformation over a range of temporal scales. (One of the primary goals of the PBO project is to observe and quantify transient deformation.) In this paper I discuss ways to maintain a consistent regional reference frame for this observational/analysis approach that takes into account various scientific and civil objectives.
G21B-03 08:40h
Improving reference frame stability by modelling common modes of surface displacements using Empirical Orthogonal Functions
For a number of scientific and non-scientific applications, monitoring of point motion with respect to a regional or global reference frame is fundamental. Applications include but are not limited to tectonic plate motion, displacements of the Earth surface due to atmospheric, hydrological and cryospheric loading, sea level variations, natural and man-made land subsidence, and stability of large infrastructure. The geodetic tool most widely used today for this is GPS, either in campaign mode where points are re-visited episodically, or with continuous GPS (CGPS) stations operated over a long time. Time series of point coordinates derived from GPS observations are affected by a number of perturbations due to reference frame instabilities, atmospheric and ionospheric mismodelling, orbital errors, unaccounted point displacements due to local to regional processes, and local processes at the GPS site. These perturbation result in actual and apparent point motions on widely varying spatial and temporal scales, which may mask the signals actually sought for in a particular application. The spatially coherent common modes in the 3-D displacement fields observed by regional CGPS networks (with examples from Europe and the Basin and Range Province, USA) are determined with the help of a Principal Component Analysis (PCA) leading to Empirical Orthogonal Functions (EOF). The main common modes are used to create a regional and local displacement field, respectively, which can be used to decontaminate GPS determined point coordinates for these larger scale displacement. This new methodology is compared to standard approaches of spatial filtering that use simple averaging [Wdowinski et al., 1997]. The common modes also allow to identify the contribution of different forcing factors to the observed displacements of the solid Earth's surface in order to separate unaccounted point displacement from other effect. As an example, the common modes due to atmospheric pressure loading are identified and quantitatively assessed.
G21B-04 08:55h
Mapping of GPS Networks Into ITRF: Regional Versus Global Approach
To align daily (or combined) GPS solutions with any realization of the International Referential Reference System (currently, ITRF2000), the usual first requirement is to process the network of interest together with a subset of stations with well-known position and velocity estimates in ITRF2000. These reference stations are then used to compute the transformation parameters to project the daily (or combined) "unknown" reference frame onto ITRF2000. The selection of the reference stations is a fundamental step of the mapping process. Although ITRF2000 is a large improvement (quality, quantity) in comparison with the previous ITRF versions, the amount and distribution of GPS stations (typically IGS stations) with well-determined ITRF2000 position/velocity representation is still not optimal. The problem is significant for many regions since more than two-thirds of the reliable IGS stations are located in Europe and North America). Regional networks of GPS stations (located on the North Atlantic and European areas) are used to discussed several issues related with the selection of the best set of reference stations for the mapping of episodic and permanent networks. Special attention is dedicated to the choice between a regional (continental size) and a global mapping approach. Based on several tests carried out, the selection of a global reference network is favored over a regional set of mapping stations. The tests show that a regional approach does not provide a significant improvement for the specific network locally. Moreover, the global approach is more consistent with ITRF2000 on a planetary scale (although the scope is restricted to a North Atlantic - European scale). For the analyzed permanent networks, the small number and uneven distribution of reliable IGS stations, within and around them, also favor the use of a global network to estimate the mapping parameters. The possibility of extrapolating existing errors in the transformation parameters also supports this option.
G21B-05 09:10h
Rate Difference Between VLBI and GPS Reference Frame Scales
For analysis of VLBI data, we apply different reference frame procedures depending on the problem being investigated. One class of solutions is for determination of the secular variation of site positions. Here we estimate the positions and velocities of sites over their entire histories of observations. To place the solution in a particular reference frame, we apply no-net translation and rotation constraint conditions. In comparisons between VLBI and GPS (JPL) site velocities at 22 colocated sites, both given in their own realizations of the ITRF2000 frame, we have found that GPS vertical rates are an average of 1.5 mm/yr (or about 0.2 ppb/yr) greater than the VLBI vertical rates. GPS baseline length rates of change are also on average about 0.2 ppb/yr greater than the corresponding VLBI baseline length rates. We investigate whether this difference can be explained by reference frame and modeling differences between the two techniques. The discrepancy is examined in the context of choosing the appropriate reference frame for studies at specific sites. For example, we recently investigated whether observed geodetic rates at Ny-Alesund, Svalbard in the Arctic can be explained by effects including post-glacial rebound and present-day ice melting, where the discrepancy between the vertical rates from the two techniques is significant compared to these effects. We also compare VLBI velocities with collocated GPS site velocities from several different GPS analysis centers in order to help resolve the source of the apparent scale rate difference.
G21B-06 09:25h
Towards improving the accuracy of GPS-based global vertical velocity estimates
The Global Positioning System (GPS) has demonstrated the ability to measure horizontal motions with sub-mm/yr accuracy. Unfortunately, the accuracy of vertical velocity estimates, particularly global estimates, is significantly poorer, and the effect of some of the GPS error sources on these estimates is not fully understood yet. An improved accuracy of globally-referenced vertical velocity estimates would serve to enhance investigations of a variety of geophysical phenomena of broad scientific interest such as global sea-level and climate change, coastal hazards, surface loading, tectonic deformation, ice mass and geoid variability. In this talk, we will present recent studies leading towards improving the accuracy of global vertical rates with GPS, with especial emphasis on network geometry, reference frame design, global scale factor as well as geophysical modeling of the ionosphere and global isostatic adjustment.
G21B-07 09:40h
Towards a Refined Realisation of the Terrestrial Reference System
Global reference frames provide the framework for scientific investigations of the Earth's system (e.g. plate tectonics, sea level change, seasonal and secular loading signals, atmosphere dynamics, Earth orientation excitation), as well as for many practical applications (e.g. surveying and navigation). Today, space geodetic techniques allow to determine geodetic parameters (e.g. station positions, Earth rotation) with a precision of a few millimeters (or even better). However, this high accuracy is not reflected by current realisations of the terrestrial reference system. To fully exploit the potential of the space geodetic observations for investigations of various global and regional, short-term, seasonal and secular phenomena of the Earth's system, the reference frame must be realised with the highest accuracy, spatial and temporal consistency and stability over decades. Furthermore, future progress in Earth sciences will fundamentally depend on understanding the Earth as a system, into which the three areas of geodetic research (geometry/deformation, Earth rotation, gravity) are to be integrated. The presentation focusses on various aspects that must be considered for a refined realisation of the terrestrial reference system, such as the development of suitable methods for the combination of the contributing space geodetic observations, the realisation of the TRF datum and the parameterisation of site motions. For this purpose we investigated time series of station positions and datum parameters obtained from VLBI, SLR, GPS and DORIS solutions, and compared the results at co-location sites and with other studies. Furthermore, we present results obtained from a TRS realisation based on epoch (weekly/daily) normal equations with station positions and daily Earth Orientation Parameters (EOP) using five years (1999-2004) of VLBI, SLR, GPS and DORIS data. This refined approach has major advantages compared to past TRF realisations based on multi-year solutions with station positions at a given epoch and constant velocities, as for instance non-linear effects of site positions and datum parameters can be considered, and consistency between TRF and EOPs can be achieved. First results of this new approach are promising.