Applying probability determination to refine landslide-triggering rainfallthresholds using an empirical "Antecedent Daily Rainfall Model"

Rainfall-triggered landslides constitute a serious hazard and an important geomorphic process in many parts of the world. Attempts have been made at v arious scales in a number of countries to investigate triggering conditions in order to identify patterns in behaviour and, ultimately, to define or c alculate landslide-triggering rainfall thresholds. This study was carried o ut in three landslide-prone regions in the North Island of New Zealand. Reg ional landslide-triggering rainfall thresholds were calculated using an emp irical "Antecedent Daily Rainfall Model." In this model, first introduced b y CROZIER and EYLES (1980), triggering rainfall conditions are represented by a combination of rainfall occurring in a period before the event (antece dent rainfall) and rainfall on the day of the event. A physically-based dec ay coefficient is derived for each region from the recessional behaviour of storm hydrographs and is used to produce an index for antecedent rainfall. Statistical techniques are employed to obtain the thresholds which best se parate the rainfall conditions associated with landslide occurrence from th ose of non-occurrence or a given probability of occurrence. The resultant regional models are able to represent the probability of occu rrence of landsliding events on the basis of rainfall conditions. The calcu lated thresholds show regional differences in susceptibility of a given lan dscape to landslide-triggering rainfall. These differences relate to both t he landslide database and the difference of existing physical conditions be tween the regions.