Аннотация:Meaningful GPS station velocities should not depend on the choice of a specific reference frame since the origin translation rate (OTR) can be estimated from velocities in stable plate interiors, and an appropriate correction can be applied (R.W. King, private communication). A conventional approach is to translate and rotate the initial, loosely constrained geodetic solution in order to best fit the velocities of official ITRF2000. A very different approach used in this study consists of following steps: (1) Assume that tectonic plates do not move with respect to each other and that they are perfectly rigid. This assumption is equivalent to setting all a priori velocities to zero. (2) Translate and rotate the loose geodetic solution to best fit the zero velocities at stations uniformly distributed over the whole network. (3) Evaluate OTR from station velocities in stable plate interiors, obtained in step 2. (4) Correct the velocities for OTR and estimate plate rotation vectors. (5) Repeat steps 1-4 as many times as required, typically four iterations is adequate. Such approach was realized in recent versions of the GAMIT/GLOBK software which we used. We tested at step 1 various starting reference frames with nonzero a priori velocities, such as ITRF2000 and NUVEL1-A, with no change in the final solution at step 5. Our database included: all daily GAMIT solutions for positions of the IGS network performed at SOPAC in interval 1995.0 - 2005.5; our daily GAMIT solutions from observations of several tens of continuous and survey mode stations on the Eurasian, North American, and Pacific plates. We then combined by GLOBK (Kalman filter) all daily solutions in the database and produced a multiyear loose solution for positions and velocities. By iteratively applying procedure 1-5, we came to the constrained solution in terms of plate-residual velocities for all stations and a set of rotation vectors for eight major plates. 76 sites with zero velocity chosen for initial transformation at step 2 correspond to locations of the ITRF2000 (IGB00 version) core stations; 67 sites within plate interiors were used to estimate OTR at step 3. Differential OTR of ITRF2000 with respect to our solution is as small as 1.2 ±0.5 mm/yr, a value in agreement with the known uncertainty in translation rate of ITRF2000. Discrepancy (rms) in velocities at 76 common sites between ITRF2000 and our solution, after appropriate rotation and translation, is 0.7 mm/yr. The current plate kinematics inherent in our geodetic solution is significantly different from the geologic plate model NUVEL1-A. Some deviations, such as 10-mm/yr difference in the rate of indentation of India into Eurasia, 5-mm/yr difference in the rate of strike-slip motion along the San Andreas Fault, or splitting of Africa into Nubian and Somalian plates, were described in literature earlier. We present, however, a globally coherent pattern of discrepancy between current and geologic plate motions, or, probably, point to uncertainties in the geologic model. Plate-residual rms velocity components of stations in rigid plate cores are less than 1 mm/yr: from 0.3 mm/yr for South America to 0.9 mm/yr for India.