Identification of the landslide deformation stage is a key aspect of landslide warning systems. However, many displacement curves lack the obvious characteristics of the three stages, making it challenging to identify the deformation stages of landslides. To overcome the defects of the Saito method, we proposed a method to extract the deformation stages based on a morphology analysis. A total of 1944 deformation stages were identified from the GNSS surface-displacement monitoring database to form a big data environment. Then, we analyzed the spatial distribution of the displacement stages of various lithologic zones in China and used cluster analysis to develop multi-stage warning criteria for deformation stages. The spatial probability and duration of each warning level were also discussed in this study. Analysis results indicated that the power law adequately described the distribution of daily deformation rate for each lithologic region, with daily deformation rates below 10 mm/d accounting for the majority of observations. Cluster analysis revealed that the number of deformation stages within the “none”, “blue”, “yellow”, “orange”, and “red” warning levels also followed the power rule for each lithologic region. It should be noted that the warning threshold of the deformation rate decreased from lithology 1 to lithology 4 at the same warning level. According to the duration analysis of the deformation stage, the longest-lasting daily rate was 0.30~2.14 mm/d with a duration of 120.22~160.96 days; whereas the shortest-lasting daily rate was 2 734.18~31 770.00 mm/d with a duration of 0.004 3~0.020 0 days. In this study, we analyzed the spatial distribution and early-warning criteria of landscape development stages in a big data environment, which could provide a scientific foundation and direction for the development of an intelligent landslide warning model.