Objectives  With the development of modern millimeter level geodetic technology, velocity field is one of the indispensable basic elements in Geodetic work. Under the background that current velocity models have limitations such as small data, poor continuity and timeliness update, a method based on local seamless Delaunay triangulation with inverse distance weighting is presented to establish the velocity field model in Chinese mainland.

  Methods  Aiming at the limitations of the current velocity models, we obtain the high-precision coordinates and velocities of each CMONOC (crustal movement observational network of China) station, through processing the long-term CMONOC GNSS (global navigation satellite system) observation data of recent seven years by the GAMIT/GLOBK software. And we construct the velocity field of Chinese mainland grid with the inverse distance weighted model of locally seamless Delaunay triangulation. Compared with that calculated by NNR-NUVEL1A (no net rotation Nubia velocity 1A) Euler vector model and some other numerical fitting models, the velocity obtained by the new method in this research has the highest precision.

  Results  The fitting precision of horizontal velocity field constructed by this new model is better than 1.5 mm/a. By using more than 1 800 regional stations for external checking, it shows that the average of absolute value of velocity differences in East and North directions are 1.11 mm/a and 0.90 mm/a, respectively, and the corresponding mean square errors are 1.50 mm/a and 1.35 mm/a. In continental margin areas, the fitting precision of horizontal velocity filed by the method of local triangulation net is much higher than that from the whole triangulation net model. Furthermore, based on the method of local triangulation net, interpolation precision of horizontal velocity filed is higher, an improvement of about 0.3 mm/a on average, and the gross errors could be largely reduced.

  Conclusions  The new model, which is based on the method of local seamless Delaunay triangulation with inverse distance weighting, can effectively utilize both distance and orientation information of the adjacent points, depict subtler local features with high precision. And it overcomes the shortcomings of the large span of the whole triangulation network in the edge area and the discontinuity of the triangulation network at the edge of the secondary block.