
In a 2018 special report, the IPCC reported that “there is high confidence that changes in heavy precipitation will affect landslides in some regions”. In Canada and the US alone, damages from landslides have been estimated at almost $4 B dollars per year, and some scientists have estimated that landslides occurrences could increase by 30 to 70% by mid to end of century. These bulk of these damages are inflicted upon linear infrastructure operators, such as highways, railways and pipelines.
Landslide risk assessments are expensive, time-intensive tasks that often rely on local expert judgment. Regional-scale studies are rare, and the results are typically delivered as reports written in domain-specific jargon along with static maps. Conversely, the hazard needs to be thoroughly understood so that risk can be accurately assessed, and protection or mitigation measures put in place. For this reason, we created GAIA Landslide.
Simply put, GAIA Landslide helps users identify assets that are at risk from landslides; it is an AI-assisted application that reasons with numerous complex datasets, enabling users to evaluate landslide susceptibility, hazard and risk over large areas. Our cognitive AI system mimics human-expert reasoning, and provides explainable assessments, making it easy for non-experts to understand the results.
Minerva has successfully produced landslide susceptibility maps for British Columbia, Canada and Veneto, Italy. The Veneto project won the prestigious “INSPIRE Helsinki 2019 Data Challenge” for the most innovative practical use of spatial data in the domains of sea, weather and cities. Applications for wildfires and floods are currently in development.
Case Study: Sea to Sky Highway, British Columbia, Canada
The Challenge
The Sea to Sky corridor north of Vancouver, British Columbia, Canada is a mountainous area that has been glacially sculpted, forming deep valleys flanked by rocky mountains. Slopes along the valley walls are covered by thick forests, prone to soil slides, while the steep, unvegetated slopes at high elevation are typically associated with devastating rock avalanches and localized rockfall. The highly variable landscape means that susceptibility to different landslide types must be evaluated based on the specific geological and geomorphological attributes of each slope.
The Solution
Mimicking the traditional process that professional geologists use to evaluate slopes, landslide susceptibility maps displayed on the user interface represent a new means of interrogating natural hazard data. Our method is an effective screening tool to evaluate the relative susceptibility of slopes to various types of landslides.
Case Study: Veneto, Italy
The Challenge
INSPIRE is a legislated directive in the EU that guides member states on the standardization spatial data. The 2019 Helsinki Data Challenge called for innovative and valuable applications of INSPIRE-aligned geospatial data. The challenge was to showcase how INSPIRE provides a framework to conduct complex geospatial data processing in a scalable and interoperable format.
The Solution
GAIA landslide won first prize in the event. We deployed our cognitive AI solution and developed an interactive web-map application that assesses landslide susceptibility and hazard in the province of Veneto, Italy. We compared over 80,000 slopes and almost 10,000 creeks to three different expert-based landslide models, demonstrating how INSPIRE-aligned data can be leveraged for scalable AI applications.