Font Size: 

Landslides occur when rock and soil masses on slopes are disturbed by earthquakes, intense storms, human activities, or a combination of these factors. Landslides can involve movements of sliding, flowing, toppling, and/or falling. A landslide map records the location and, where known, the date of occurrence and the types of movements, and a landslide database is a systematic collection of landslide maps on past events. Taiwan is a mountainous island, with steep slopes and fragile geological formation. Located in the western part of the Pacific Ocean, Taiwan is struck by an average of 3 or 4 typhoons (tropical cyclones) every year, causing numerous landslides and debris flows during the typhoon season (May to October). This paper describes a landslide database project that is currently conducted by the Hydrotech Research Institute (HRI), National Taiwan University, for the Soil and Water Conservation Bureau (SWCB). The HRI research team has so far collected 11 annual landslide catalogues from 2004 to 2014 and 12 event-based landslide inventories from 2009 to 2011 in the database. In addition, the database includes debris flow affected areas, sediment transport data, landslide recovery projects, satellite images, air photos, UAV images, and other framework data such as drainage, transportation, and boundaries. More data will be added to the database as the project progresses. Vector data are stored in shp and raster data in tif or img. Landslide boundaries have already been cleaned by referring to air photos and digital elevation models, and landslide attributes have been appended. The HRI team has tested the applicability of the database for landslide susceptibility modelling and data visualization. A statistical method (logistic regression) and a data mining technique (decision tree) were used for landslide susceptibility modelling island wide, and the results were very good, with the overall prediction accuracy rate ranging from 72 to 80% for the logistic regression models and from 74 to 76% for the decision tree models. In support of OpenData / Taiwan, the HRI research team has developed a data sharing prototype based on Operation Dashboard for ArcGIS. Linked to the landslide database maintained on the SWCB server, the prototype allows users to display and query landslide related data in the database. More on the development of the database (e.g., migrating to a cloud platform) and its applications (e.g., a landslide warning system) will be discussed at the presentation.

geospatial data; spatial data infrastructure; disaster management