About

Mass-movement events across the cryosphere are expected to increase in frequency as air temperatures continue to warm, putting infrastructure at risk. The CryoSlideRisk project will bring together a convergent team of researchers, policy makers, and local community representatives through two workshops to address infrastructure resilience and adaptation to increasing mass-movement risks across the cryosphere.

Join the network

Using AI for Landslide Detection and Forecasting

AI model

We compare satellite images before the landslide event (Pre-Image) and after the landslide event (Post-Image). The impacted area of the landslide event was highlighted by a human labeler (Human Annotated Label). An AI model was able to identify the landslide event from the images (Prediction) with high confidence (Confidence Map). The developed AI model may be used to compile a landslide database in the regions of interest.

Using AI models to predict debris flows triggered by heavy rainfall from September 9-13, 2013 in Colorado's Front Range with high accuracy

Predictive deep learning models can be created/trained to improve the prediction accuracy of where and when landslide hazards occur and the potentially-impacted areas based on the compiled database and other data sources, accounting for landslide contributing factors. This figure illustrates how AI models can be used to predict debris flows triggered by heavy rainfall from September 9-13, 2013 in Colorado's Front Range with high accuracy.

Goals & Activities

Project activities center around two workshops that are planned to take place in the spring of 2022 and spring of 2023. We will use these workshops to bring together network members to identify gaps in our current understanding and develop methodologies to predict cryospheric mass-movement events.

Our specific goals are to:

  • develop more automated hazard mapping tools through remote sensing and machine learning
  • compile and make accessible relevant datasets and databases, and encourage interactions with potential users to enable the development and validation of AI algorithms for forecasting landslide risk
  • formulate more flexible and adaptive approaches to enhance infrastructure resiliency and adaptation

Workshops

Meet the PIs

  • Penn State logo
  • university of Alaska Fairbanks logo