The geodesy as an important technique is applied to monitor the land subsidence, and leveling is the classical observation means of the vertical movement. An algorithm to adaptively process the data of multi-periods is introduced when there are the outliers in level data. The equivalent weight is constructed with two segments functions, the influence of outliers will be weakened. Collocation is applied to model the vertical movement. The data set of second order level network in Shandong province is employed to test the new algorithm. The results show that the modified robust estimation can control the influence of outliers, and improve the accuracy of dynamic adjustment for level network. The accurate trend of crustal vertical movement for recent 10 years can been obtained, the downward trend exists in the northern and western of Shandong province, but a slight upward trend in the eastern and southeast of Shandong province occurred.