DETECTION OF OLD AGRICULTURAL TERRACES IN STEEP, VEGETATED TERRAIN USING AIRBORNE LIDAR: CASE STUDIES FROM HONG KONG

摘要

Old agricultural terraces and associated features are common on steep hillsides throughout the Hong Kong Special Administrative Region (HKSAR) and, since the late 1960s, most have been obscured by dense vegetation. Some of these terraces may date from a thousand or more years ago and some are still actively cultivated today. Based on aerial photograph interpretation (API) of older aerial photographs, particularly from 1963, and more recent field observations, generally these features can be classified as earthen step-form terraces, rubble wall terraces, and other forms of anthropogenic activity (e.g., military, prospecting/mining, quarrying, and ad hoc cuts/spoil). Their influence on slope stability is not yet well understood, however they need to be identified so that any relationship with landslide susceptibility can be better understood. Currently, these unknowns remain a potential source of residual risk for landslide hazards. With the aim of continuous improvement of natural terrain hazard assessments, suitable detection and mapping of the terraces and other features should first be systematically undertaken. A dense tropical vegetative canopy presents an obstacle to terrace detection. Much of the HKSAR has undergone revegetation due to changing land uses, government regulations, reforestation, and rapid natural growth over the last 30 - 40 years. Many terraces and other features are visible in 1963 aerial photographs, when hillsides were relatively bare of vegetation. Additionally, many of the oldest features are degraded and, when combined with the modern, dense vegetation, may be confused with natural landscape-forming processes when observed in the field. To assist in identification, a topographic position index (TPI) was applied to digital elevation models (DEM) derived from airborne LiDAR for three case studies. The result is an emerging, indirect mapping approach complemented by integrated API and field work to detect and map old terraces.