In my PhD research, I use statistical models to analyze observed data on landslides and the factors that control them with the goal of better predicting rainfall-triggered landslides.  I am a geomorphologist of many interests and my research has previously focused on climate policy, cosmogenic nuclide geochemistry, past glaciations in the Andes, and long-term landscape response to climate and tectonics.

Current projects

Landslide seasonality in the Pacific Northwest

How probable are landslides in a given month? How many can we expect? How do these metrics vary from year to year?  In this study, we use Bayesian inference to combine information from five heterogeneous landslide inventories to learn the seasonal pattern of landsliding in the Pacific Northwest of the United States, one of the areas of the country most impacted by landsliding.

Global urban landslide rainfall thresholds

Rainfall-triggered landslides in urban areas cause fatalities and damage globally, but so far, we do not have a global overview of how much rain is needed to trigger urban landslides or how variable the needed rainfall is between cities.  In this study, we develop a global compilation of urban landslides and use both station-based and satellite-based precipitation datasets to estimate the distribution of intensity-duration (I-D) and event-duration (E-D) rainfall thresholds for landslide occurrence in cities worldwide. We apply a Bayesian approach that offers an objective way to determine rainfall thresholds and their uncertainty while explicitly modeling variation between cities.

Exploring the role of atmospheric rivers in triggering landslides in western North America

Atmospheric rivers are corridors of moisture that stretch over thousands of kilometers and are responsible for bringing heavy rainfall to the west coast of North America.  In this collaborative project lead by colleagues from the Potsdam Institute of Climate Impact research, we quantitatively explore the role of atmospheric rivers in triggering landslides across western North America.

Landslide initiation during intense rainfall in Southeast Alaska

Small towns in southeast Alaska are exposed to landslide hazards by virtue of their proximity to over-steepend glacial hillslopes.  In this collaborative project lead by colleagues from the University of Oregon and the Sitka Sound Science Center, we use both frequentist and Bayesian logistic regression to  estimate thresholds for landslide triggering-rainfall in Sitka, Alaska from landslide inventory data and station precipitation data.  

Incomplete landslide inventories.  Is it a problem, and what can we do about it?

Landslide inventory data is a form of presence-only data, meaning that while we know about landslides included in the inventory, we don't know if there were other landslides that weren't included in the inventory.  In this project, I use simulation to explore how missing landslide data might introduce bias into statistical model results for landslide susceptibility or rainfall thresholds, to illuminate when this is a problem and when it is not, and to investigate potential fixes for when it is.

Research interests

• Geomorphology
• Natural Hazards
• Landslides
• Machine Learning
• Bayesian Inference
• Cosmogenic nuclide geochemistry