This study used multi-temporal high-resolution LiDAR DEMs to map new landslides and to establish an empirical relationship between area and volume of landslides in a study area at southern Taiwan. The elevation changes are detected from 2 successive sets of LiDAR DEM in a forested land. They were acquired with a point-cloud density of more than 1.0 point per square meter in 2005 and 2010, respectively. The accuracy standards and survey procedures are based on the 2005 MOI guideline. Active landslides are successfully identified and mapped using differential DEM derived from the 2 data sets, using a change threshold of 0.50 m to remove noise from differential data and visual interpretation of typical landslide morphology with supplemental information from orthophotos. More than 95% of landslides were mapped under the threshold of 0.50m, in comparison to orthophotos. Subsequently, the coverage area and volume of each individual landslide are derived. An empirical relationship between landslide area A (m 2) and volume V (m 3), V = kA a with k = 0.099 and a = 1.396, is then proposed in this paper. The power law formula was obtained through linear fit of the landslides with coefficient of determination, R-squared = 0.837. It can be further applied for estimation of the total volume of landslide materials in a whole region. In addition, the differential DEM can be further applied to understand the regional sedimentation activities showing the source areas and the deposition areas.