Abstract: Slope collapse distributes widely in south China with huge damages. The distribution and dynamic changes of slope collapse is very helpful to controlling the slope collapse development and formulating the policy for slope collapse prevention, however at present few study on the accuracy of slope collapse investigation has been conducted. By comparing the characteristic data such as the spatial distribution, perimeter, and area collected from artificial survey and extracted from remote sensing images, we found that the data collected from artificial survey could generally reflect the actual location of slope collapse. But the shape and boundary differ greatly, with the average perimeter and area reduce 11.70%, and 9.07% respectively than the actual situation. If the number of incorrect slope collapse is considered, the total perimeter and area reduce less. In short, the slope collapse data from artificial survey reflects the actual value very correctly, and could support the database for scientific research and decision-making. However, more financial, human resource, and time are required in artificial survey, and long-term dynamic monitoring could not be carried out. In addition, its accuracy is affected by different investigators.

Key words: slope collapse, artificial survey, remote sensing image, relative error, spatial distribution, policy, topographic map