similar way, i.e. the scale being defined by SLR and VLBI solutions, and the origin being defined by SLR. From our experiences regarding the analysis and combination of space geodetic observations (solutions) we identified a number of deficiencies regarding current ITRF realizations, e.g.: - Systematic effects (biases) between solutions and/or techniques can be considered as one major limiting factor of the combined solutions. As an example for the current status we mention that the individual solutions submitted for ITRF2000 reach differences of up to 5 cm for the origin and a few ppb for the scale (see Altamimi et al. 2002, and http://lareg.ensg.ign.fr/ITRF/ITRF2000/T.gif and D.gif). Also differences up to a few centimeters for positions and up to one cm/yr for station velocities exist for some collocation sites with long time series (many years) of observations (Angermann et al., 2002). The reported differences often exceed their r.m.s errors by a factor of 5-10 or even more, indicating that the individual solutions are significantly influenced by systematic errors. - Another important issue is the reduction of a-priori datum constraints, which normally are included in the solutions. The input solutions submitted for ITRF2000 realization were classified to include loose, minimum, or removable constraints. We found remarkable contradictions with respect to these declarations, and for some of the ITRF2000 input solutions the constraints could not be removed. If one neglects these constraints, significant biases and systematic effects might be introduced into the combination results. - Conventionally the ITRF is realized by the adoption of a set of positions referred to a reference epoch and constant velocities for the ITRF network stations. The observed non-linear effects in position time series (e.g. due to seismic or volcanic effects, deformations at plate boundary zones, local effects) are in conflict with the assumption of constant velocities (see Angermann et al., this issue). This may produce errors and systematic effects in the individual solutions, which would also propagate into the TRF and degrade its accuracy.
14
- Collocation sites are a key element for combining TRF solutions provided by different techniques. Currently, only six ITRF sites exist with collocations of VLBI, SLR, GPS and DORIS, and none of these is fully satisfactory regarding the data quality and/or the time span for the various techniques. There are a number of SLR and VLBI stations that are not even collocated with a GPS receiver. In many cases, the intra-site vectors (local ties) are not well determined or dubious. The current situation regarding collocation sites and the accuracy and availability of local ties is not satisfactory. Another important aspect concerning collocation sites is how to handle velocity estimations obtained from different techniques at one site. Normally (for previous ITRF computations) the different estimations were forced to be identical. The results obtained from our latest TRF combinations indicate, that for some collocation sites the velocity estimations for differenent occupations seem to differ significantly, and in those cases the velocities should not forced to be identical (e.g. systematic biases between techniques do exist, local phenomena may cause occupation- and/or stationdependent effects). - A number of relevant aspects related to the combination methodology need to be studied in more detail, e.g. the level on which the combination should be performed (e.g. observation, normal equation or solution level), the weighting of solutions for the intra- and inter-technique combination, the handling of local tie information, the handling of non-linear site motions, and datum definition issues. 3. Recommendations for future ITRF combinations Finally we address some important aspects that should be considered for future ITRF realizations: - To achieve further improvements there is an urgent need to investigate the existing differences between different techniques and/or solutions regarding various aspects (e.g. modeling, parameterization, software-rela-