A new approach of combining aerial photography with satellite imagery for landslide detection

Subsequent rainfall after a strong earthquake can easily trigger landslides. Aerial photography is always available after a strong earthquake but not always available in a timely manner after a subsequent rainfall following the earthquake. Sometimes, only panchromatic imagery is available because of its relatively low cost and large cover capacity. To detect multi-temporal landslides induced by earthquake and its post long-term effect, in company of other factors such as subsequent rainfall, traditionally, it needs to carry out image classification multi-times to calculate the variance information. Therefore, the accuracy will be affected by accumulated errors from multi-classification, and the process is very time-consuming. In this paper, a new semi-automatic approach combing aerial photograph with satellite imagery was proposed for rapid mapping of multi-temporal landslides. The approach can enhance the change information of each landslide event in one detection process. In addition, slope units were introduced to separate the detected conjoint landslides. Chenjiaba area, which located in the highest seismic intensity zone of Wenchuan earthquake in Beichuan, China, and had a strong rainfall 4 months later, was selected as a case study to demonstrate the usefulness of this methodology. Accuracy assessment was carried out by comparing those extracted ones with a manually prepared landslide inventory map. Correctly detected were 90.1 and 94.2 % for earthquake-induced landslides and new landslides, respectively. Results show that this approach is capable of mapping different temporal landslides efficiently and quickly.