LandAware

The recording of the seventh webinar of the LandAware 2026 webinar series “What is wrong with shallow landslide susceptibility modelling? Insights from machine learning, numerical modelling and field investigation“, by Tobias Halter (ETH Zürich, Swizterland), held on 24 September 2026, is online on the LandAware YouTube channel.

The 7th webinar of the LandAware 2026 webinar series “What is wrong with shallow landslide susceptibility modelling? Insights from machine learning, numerical modelling and field investigation”, by Tobias Halter (ETH Zurich, Swizterland) is scheduled for 24 September 2026, 14:00 UTC.

Abstract

Landslide susceptibility maps have become essential tools for early warning and hazard management, aiming to predict the spatial deposition of slope failure across a landscape. In recent decades, machine learning approaches and high-resolution spatial data have steadily improved susceptibility assessments. Yet discrepancies between predicted susceptibility and observed landslide occurrence persist. In this webinar we explore these limitations through studies that combine machine learning, numerical modelling, and field observation. We examine soil profiles and geophysical measurements collected at landslide head scarps in Switzerland to understand what triggered these failures, and compare them with equivalent measurements from slopes that were predicted to have similarly high susceptibility, but did not fail. The results show that variations in soil texture, development, type, and depth strongly influence the mechanical and hydrological conditions governing slope stability — factors that large-scale models often cannot capture. Together, these findings offer insight into why susceptibility models fall short, and point toward what finer-scale, field-grounded data could contribute towards landslide early warning.

Bio

Tobias Halter is a doctoral student, soon to complete his PhD in Engineering Geology at the Swiss Federal Institute of Technology Zurich (ETH). His research focuses on integrating soil hydrological measurements into large-scale landslide early warning models and shallow landslide susceptibility mapping. His work combines data-driven machine learning analysis, physically based numerical modelling, and field observations, all directed toward a single goal: improving predictions of when and where landslides will occur in the future.

Through his involvement in the development of the Swiss national landslide early warning system, he gained first-hand experience of the challenges governmental agencies face in translating scientific advances into operational tools. This experience has reinforced his motivation to pursue research that not only advances scientific understanding but also addresses practical challenges in natural hazard management.

The sixth webinar of the LandAware 2026 webinar series “From Landslide Hazard Assessment to Early Warning: Integrating Geomechanical, Environmental, and Remote Sensing Approaches in Data-Scarce Mountain Regions”, by Danny Love Wamba Djukem (University of Liège, Belgium) is scheduled for 25 June 2026, 05:00 UTC.

Abstract

Landslide early warning systems are often developed in regions with extensive monitoring networks, long-term datasets, and well-documented landslide inventories. However, many mountainous regions around the world, particularly in Africa and other parts of the Global South, face a very different reality: limited monitoring, sparse environmental data, and incomplete records of past landslides.
This webinar explores how landslide hazard assessment can contribute to the development of early warning strategies in such data-scarce environments. Drawing on case studies from Cameroon, China,
Haiti, it highlights how geomechanical analyses, landslide inventories, physically based modelling, and remote sensing observations can be combined to better understand landslide processes and support forecasting efforts.
The webinar first examines the environmental and triggering factors controlling landslides in tropical mountain regions, using examples from Mount Oku in Cameroon. It then discusses the development and application of site-adaptive approaches for predicting earthquake-induced landslides across different tectonic and climatic settings. Finally, it explores how national landslide inventories, satellite observations, and integrated hazard assessments can provide practical foundations for future landslide early warning systems in regions where traditional monitoring networks remain limited.
The talk argues that effective early warning begins with understanding the landscape, its processes, and its triggers. By integrating geomechanical, environmental, and remote sensing approaches, it is possible to move from hazard assessment toward more operational and scalable warning frameworks, even in data-constrained mountain environments

Bio

Dr. Danny Love Wamba Djukem is an engineering geologist and geohazards researcher specializing in landslide processes, slope stability, earthquake-induced landslides, remote sensing, and hazard assessment in mountainous environments. She holds a Ph.D. in Geotechnics and Geohazards and was previously a Postdoctoral Researcher at Chengdu University of Technology, China. She is currently an Invited Researcher with the Georisk & Environment Group at the University of Liège, Belgium, and is based in Cameroon.
Her research focuses on understanding landslide triggering mechanisms and developing practical approaches for landslide hazard assessment and early warning in data-scarce regions. Her work combines geomechanical analyses, physically based modelling, remote sensing, and landslide inventory development, with applications in Africa, the Caribbean, and Asia.
Her recent studies include landslide hazard assessment in Cameroon, site-adaptive prediction of earthquake-induced landslides, and national-scale landslide inventory development to support disaster risk reduction and resilience planning.