Fecha de vencimiento y amortización de los valores

In Chapter 4 the GPS processing stages using GAS and the procedures for automated processing were introduced. Using these, a 13 CGPS station data set was processed and analysed on a daily basis, using data up to September 2001. During the re–analysis of the data set archived in the processing data archive procarch (§4.3.1), this 13 CGPS station network was extended to a total of 25 CGPS stations, including additional data for several IGS stations for the period of 1997 to 1998. Both analyses differ slightly in the applied processing strategies, which have been denoted as Strategy 1 and Strategy 2 in this thesis.

For Strategy 1, the data for nine UK CGPS stations were processed along with data from four IGS stations in Europe, used as fiducial stations. The nine UK CGPS stations used were: Aberdeen (ABER), Camborne (CAMB), Hemsby (HEMS), IESSG (IESG), Lerwick (LERW), Newlyn (NEWL), Liverpool (LIVE), Lowestoft (LOWE) and Sheerness (SHEE) (see Tables 5.1 and 5.2 for information on station categories and data availabil- ity). The IGS stations used for this analysis were Kootwijk (KOSG), over the complete

Chapter 6. Analysis of the Continuous GPS Network 140

timespan, Onsala (ONSA), Wettzell (WTZR), and Villafranca (VILL), for a limited time span from 13 September 1998 onwards.

As a result of the daily GPS processing, a series of loosely constrained daily GPS network solutions were obtained for the period up to 25 September 2001. In these solutions the final IGS precise ephemeris was held fixed and the coordinates for KOSG were tightly constrained. The coordinates of KOSG were taken from the same ITRF realization as the IGS precise ephemeris. In each case, the coordinates of KOSG were motioned to the observation epoch using the station velocity given in the appropriate ITRF realization. All of these loosely constrained daily GPS network solutions were computed using the ionospherically free double difference observable with integer ambiguities free and systematic error models for solid Earth tides (§3.4.4_{), ocean tide loading (§}3.4.4) and antenna phase centre variations (§3.4.3). For the modelling of the hydrostatic zenith tropospheric delay, the Saastamoinen model (Saastamoinen, 1973) with the Niell dry mapping function (Niell, 1996) was used. The wet zenith tropospheric delay was then modelled as a random walk process (§3.4.2) with a process noise of 0.4 cm/√hr, and mapped to the appropriate station to satellite elevation angle using the Niell wet mapping function, as previously described in Teferle et al. (2002a). Due to the application of the random walk process, it was not possible to fix the ambiguities to integers within GAS (Penna,1997). Not fixing ambiguities to integer values has been reported to increase the formal errors by between 20 to 30% (Johansson et al.,2002). However, with a focus on the height component it was decided to accept the possible worsening of the solution caused by not fixing ambiguities in favour of the improvement in the height obtained from the refined tropospheric modelling.

In order to form coordinate time series for each station, the baseline vectors and associated variance-covariance information, obtained from each loosely constrained daily GPS network solution, are transformed to a common reference frame. To do this, in 1997 and partly 1998, the IGS station at KOSG was tightly constrained to its ITRF97 coordinates, motioned to the observation epoch. However, from 13 September 1998 onwards, a network of four European IGS stations, consisting of ONSA, KOSG, WTZR and VILL, was used in this transformation, with each station tightly constrained to its ITRF97 coordinates, motioned to the observation epoch. Hence, for each of the nine

Chapter 6. Analysis of the Continuous GPS Network 141

UK CGPS stations, continuous daily coordinate time series were obtained in the ITRS97 (Teferle,2000).

Panafidina and Malkin(2001) andMalkin and Voinov(2001) reported that the official EUREF weekly coordinate time series contain jumps and systematic seasonal errors, especially in the height component. It was reported that these systematic errors may have been caused by distortions of the reference frame caused by errors in the modelling of the movement of fiducial stations, if more than one fiducial station is used with tight constraints. Errors in the velocities of fiducial stations, e.g. peculiar station motion, local displacements or equipment changes, would result in errors distributed over the whole processed network. Furthermore, these distortions increase towards the edges of the network, especially if fiducial stations are concentrated near the centre of the network (Panafidina and Malkin, 2001). Their results from a re–analysis of selected EUREF stations showed that solutions based on a non–fiducial strategy (Blewitt et al.,1992;Heflin et al., 1992; Rius et al., 1995; Park et al., 2002) were most likely to be free of seasonal signals. In addition it was noted that, this non–fiducial strategy would also enable the EUREF weekly solutions to be transformed to any reference frame and re–transformed to another one using a much simpler procedure than that needed for removing constraints (Mareyen and Becker,2000).

With the re–analysis of the 25 CGPS station network, including stations Aberystwyth (ABYW), Barking Barrier (BARK), Brest (BRST), Dunkeswell (DUNK), Herstmonceaux (HERS), Hurn (HURN), Morpeth (MORP), National Physical Laboratory (NPLD), North Shields (NSTG), Pershore (PERS), Portsmouth (PMTG), and Sunbury (SUNB), this non– fiducial strategy, was introduced such that no IGS stations were constrained at the network processing stage. This is referred to as Strategy 2 in this thesis.

The second major difference between Strategies 1 and 2, was that additional data for the IGS stations ONSA, WTZR and VILL for the period prior to 13 September 1998, were obtained in order not to change the processing configuration throughout the total observation period, thus reducing the possibility of offsets in the coordinate time series (see §6.2.4) for Strategy 2. Introducing the additional IGS station data also removed the need for the re–scaling of the daily standard errors prior to 13 September 1998 (§4.6).

Chapter 6. Analysis of the Continuous GPS Network 142

With the introduction of Strategy 2 at the start of the re–analysis of the 25 CGPS station data set, the ocean tide loading (OTL) (§3.4.4) model was also updated, to the more recent FES99 (Lef`evre et al., 2001) as opposed to the FES94.1 (Le Provost et al.,

1994). The OTL parameters for the FES99 model were obtained fromScherneck and Bos

(2001_{) (see §}4.3.3). With the introduction of the new model, it was also decided by the author to include modelling of the effect on the horizontal coordinate components, which was not carried out in Strategy 1.