Abstract: When there are several gross errors in the observations, the Pau Ta's criterion can only eliminate one at a time. If the mutual deviations among the gross errors do not satisfy certain conditions, the Pau Ta's criterion will be invalid. To solve this problem, a new approach named as data jump method based on the Pau Ta's criterion was proposed. It overcomes the limitation of the Pau Ta's criterion. The deformation monitoring data with gross errors can be eliminated in batches. The method is used to process some actual monitoring data, and a good result is obtained, which provides a theoretical basis and algorithm for the gross error determination and elimination using computer program.