This paper presents Auckland Transport's (AT) innovative approach to enhancing resilience in the Auckland road network through the development of a landslide risk framework using advanced technology. Previously, asset management of inland assets relied on reactive responses to natural hazards due to a lack of comprehensive analytics for proactive maintenance and resilience works.
To achieve this, cross-sections were generated at 5m intervals along the 7500km road network, extending 250m on either side of the centreline. Data was obtained from various open-source platforms, and Python scripts were developed to automate the collection of geospatial data, slope angles, embankment and cut dimensions, geological information, existing structures, overland flow paths, and fault lines.
The framework also involved training an AI model to identify historical slope failures by using polygons of landslide and non-landslide extents overlaid on a 1m Digital Elevation Model (DEM) in an iterative process until an acceptable level of accuracy was achieved in its predictions. A similar process was employed to detect vegetation coverage using satellite imagery. The model utilized a Random Forest machine learning algorithm to analyse input data and identify patterns associated with landslide occurrences. Building multiple decision trees, the model aims to find patterns in the input data that result in the presence or absence of a landslide.
The outcome is a GIS layer that visualizes risk ratings from 0 to 1 at 5m intervals along the road centreline. This tool enables a shift from reactive repair strategies to proactive landslide prevention, significantly reducing disruption and repair costs associated with extreme weather events. The landslide risk layer serves multiple purposes, aiding asset management, route resilience, long term planning and supporting Auckland Emergency Management in early warning and communicating with communities. Ultimately, this framework enhances the resilience of Auckland's transport infrastructure and informs adaptation planning for future climate hazards.