The Temporal and Spatial Occurrence of Rainstorm-Triggered Landslide Events in New Zealand: an Investigation into the Frequency, Magnitude and Characteristics of Landslide Events and Their Relationship with Climatic and Terrain Characteristics

Abstract

This thesis represents an inductive analysis of the relationship between the temporal and spatial occurrence of landslides and their triggering and controlling factors. It offers the first attempt to investigate the rainfall-triggered landslide phenomenon by systematically examining trends in occurrence on a range of scales from the entire country to the specific landslide site. As a basis for this study, it has tried to extract all relevant records from a number of diverse sources, in an effort to identify real spatial and temporal trends. Previous work on the temporal and spatial occurrence of rainfall-triggered landslides in New Zealand has been based on local studies, case histories, and restricted data sources. While this has added substantially in the global literature on rainstorm-triggered landslides, the work has been limited by the existing database, data handling capabilities, and the urgency of reporting on individual events. This thesis summarises landslide information available in New Zealand, creates a comprehensive landslide inventory, and analyses this with respect to the precipitation record, landslide recognition, forest removal, and factors which influence the occurrence of landslides at the national scale (e.g. vegetation, soils, geology, terrain). This data base, combined with an analysis of regional climate using the entire record of the respective region has provided the first really comprehensive platform upon which to establish the relationship between landslide occurrence and triggering climatic conditions. Three study areas in the North Island of New Zealand - Hawke's Bay, Wairarapa, and Wellington – were analysed. Historical records indicate that landslides occur frequently in all three regions. Climatic triggering trends have been analysed using previous approaches to modelling triggering thresholds. These threshold models have been re-developed using physically justifiable parameters. This has involved the replacement of previously used arbitrary and broad regional parameters with values based on actual conditions and measured processes. The results show that thresholds vary strongly between regions, indicating different levels of susceptibility to the landsliding process. In addition, the return periods and probabilities of occurrence of triggering rainfall were determined. Results show that regions differ not only in the magnitude of rainfall required to trigger landslides, but also in the return periods and the probabilities of occurrence of this rainfall within a specific time period. Just as modern computer technology used in this study has allowed a comprehensive re-examination of climatic parameters, GIS and associated applications have also allowed a large database to be used to develop a number of spatial relationships between landform and landslide behaviour. These analyses include the relationship of spatial landslide occurrence with the distribution of rainfall and with the temporal probability of occurrence of a triggering rainfall. In addition, the spatial variables which determine landslide occurrence (such as slope angle, aspect" slope curvature, position on the slope, elevation, soils, stream order, drainage density, etc.) were used in multiple regression analysis. The resulting output describes the different probabilities of landslide occurrence at a given location. These probabilities were plotted as a regional landslide hazard map, which reflects both the temporal and spatial dimensions of landslide occurrence.