Processing of the 1992, 1994 and 1997 Campaigns in the Northen GPS Deformation Traverse

Abstract: The Northern GPS Deformation Traverse was established to detect movements at two major shear zones now active in northern Sweden, namely the NW trending Bothnian-Senja zone and the N trending Bothnian-Seiland zone. So far three GPS campaigns have been observed the processing of these three campaigns, and to investigate and compare different processing and analysis strategies. It should be noticed that the investigations that are of a purely geodetic nature.The campaigns were processed in the Bernese software 4.0 using several different methods. In the first part of the work, more or less the standard method for long baselines described in the Bernese manual (Rothacher, 1996) was applied. In this case all situations in a session were first processed together, using the ionosphere-free linear combination L3, zenith tropospheric parameters, and correct modelling of the correlations. After that, the normal equations from the sessions were analyzed for the baselines between adjacent stations along the traverse, it was found that the results were very bad in the vertical component, the results are more promising in the horizontal, but the different solutions are nevertheless quite noisy. In addition, a few large errors were found also in the horizontal.    In the second part of the work, the relatively short individual baselines between adjacent stations were processed separately. The main purpose of this type of processing was to evaluate different processing methods, such as processing with and without zenith tropospheric parameters using either L1 or L2 plus an ionosphere model, and to find out what extent errors caused by antenna-mixing and site dependent effects are present in the data. Another aim was to find out if the baseline mode of processing is suitable for the present project in its own rights. The results show that elevation dependent errors, caused by antenna type mixing and/or site-dependent effects like scattering, are responsible to a considerable degree for the height errors found in the first part. As could be expected, because of the higher noise level for L3, the tests further indicate that the short baselines should be processed without the ionosphere-free linear combination, using for instance L1 and an ionosphere model instead. Also in the horizontal components, considerably better results are obtained in this way.In the final part, the campaigns were processed as the first part, but with the difference that L1 and an ionosphere model was used for the shortest baselines, while L3 was applied for the longer ones. The results in the horizontal indicate that the mix of frequencies has worked well. The short baselines are comparable to the best results from the processing of individual baselines, while the accuracy over longer distances is more or less independent of baseline length. At this point we also have investigated how large uncertainly we have because of the method that is used to estimate velocities from the final campaign solutions. Two different methods are compared. The first one is to utilize unweighted linear regression separately for each baseline and telocentric component, while the second is to make a simultaneous adjustment in the horizontal components, using the full cofactor matrices from the final campaign solutions for the weighting. In the latter case, all available correlations are modelled correctly. We conclude that we have significant uncertainties because of the method that is used in this step at the present stage, when only three campaigns are available.