Previous landslide prediction studies have focused on the assessment of location of landslides. Besides location, landslide geometric features (i.e., size and shape) are important factors that influence the distribution and dynamics of landslides. Statistical methods have been used to determine the frequency-size or frequency-volume relationships of landslides, through examining landslide inventories. However, the question of what sets their size and shape is unanswered. In this study, a landslide geometry generating algorithm (LsGA) is developed for quantifying landslide geometric features, including area, perimeter, upper length, lower length, average length and average width, with incorporating an existing landslide inventory and digital elevation model (DEM). The Kaoping watershed in Southern Taiwan is selected as the study area, and the landslide inventory prepared after Typhoon Morakot (August 2009) were applied for LsGA analysis. Landslide geometric features generated by LsGA were then used to correlate to geo-environmental factors, such as slope and contributing area (CA), in a logistic regression model. Preliminary findings are: (1) smaller landslides are generally longer than larger landslides, (2) the upper length of small landslides is relatively wider than large landslides, (3) small landslides are more likely to be observed over gentle slopes, and (4) small landslides are more likely to be observed over lower part of slopes (high CA value, near channels).