Towards global landslide detection from satellite radar

Overview

Following major earthquakes or storm events in mountainous regions, destructive and hazardous landslides are commonly triggered across large areas, posing risk to life and infrastructure. Mapping these landslides within days or weeks of such an event is critical to inform and support emergency response (Williams et al., 2018), but gathering this information quickly enough from optical satellites is often problematic due to cloud cover (e.g., Robinson et al., 2019). Radar satellites can overcome this problem as they can image the Earth’s surface through cloud, and recent work at Durham and elsewhere has demonstrated that satellite radar coherence techniques offer great potential for detecting triggered landslides on a timescale suitable for emergency response (e.g., Burrows et al., 2019, 2020). Ongoing work has also shown that it is possible to further improve landslide detection from radar data by incorporating additional empirical information that predicts which parts of the landscape are most likely to experience landsliding (such as topographic slope or land cover data), through a machine-learning framework.

It is now clear that satellite radar could provide useful information on the spatial pattern of landsliding within 72 hours of an earthquake; this is fast enough to be used in early-stage emergency response. However, there is still no operational system for landslide detection following major earthquakes. This is primarily due to a limited understanding of exactly where this approach will or will not work. In particular, it remains unclear how the inferred landslide signal in radar coherence is linked to the physical processes of landsliding, how and why the signal evolves over time, how temporal changes could be distinguished from other time-variable coherence signals (e.g. from soil moisture), and how these factors, and therefore also the performance of radar-based landslide detection methods, are likely to vary within and across different climatic environments.

This project aims to enable operational landslide detection from satellite radar, suitable for future emergency response worldwide. To do so, the student will address the issues above, applying the new techniques to a wider set of global case studies, further investigating the temporal signature of coherence change from landslides and other causes, and probing the underlying surface-change processes by comparing these data with other supplementary datasets.

Click on an image to expand

Image Captions

Figure 1: Bedrock landslide triggered by the 2008 Wenchuan earthquake near the Zipingpu Reservoir, Sichuan Province, China. Photo: A. Densmore

Figure 2: SAR-based landslide classification surfaces for landslides following the 2015 Gorkha, Nepal (a) and 2018 Hokkaido, Japan (c) earthquakes, calculated with ALOS-2 satellite radar data and (b, d) corresponding observed landslide density derived from optical satellite mapping. Adapted from Burrows et al. (2020).

Methodology

For a series of global locations across differing climatic environments (e.g., tropical-sub-tropical, arid, temperate), interferometric SAR (InSAR) from current (Sentinel-1, ALOS-2) and next-generation radar satellites (NISAR) will be used to systematically analyse radar coherence signals following major landslide-triggering events. The temporal evolution of landslide coherence in different locations will be compared to other supplementary datasets on surface processes for each region (including landslide inventories, optical satellite imagery, digital elevation models, and hydrological data), in order to better understand how and why radar coherence evolves through time following a landslide.

Through this work, the student will have the opportunity to employ and further develop advanced analysis methods developed by the supervisory team, including landslide-detection algorithms (e.g., Burrows et al., 2020) and machine-learning-based methods for the integration of radar data with empirical landslide prediction methods (Burrows et al., in prep.). These methods will be used to develop an operational global landslide detection model suitable for informing future emergency response.

Project Timeline

Year 1

Training will be provided in space geodetic techniques, in particular the processing and analysis of satellite radar coherence data. Training will also be provided on landslide processes, hazard and risk. Processing and analysis of radar coherence for selected example landslide-triggering events, and comparison to supplementary datasets. Evaluation of existing detection methods across a range of climatic environments.

Year 2

Investigation of further global landslide-triggering events with satellite radar, and determination of surface processes driving the temporal signature of landslide coherence signals. Work on development of landslide-detection methods based on these findings. Publication of this work in one or two peer reviewed journal articles.

Year 3

Completion of method development and evaluation, application of enhanced detection method for a chosen case-study event. This work should lead to an additional publication.

Year 3.5

Focus on combining the published outputs and associated material into a PhD thesis.

Training
& Skills

The student will receive training in space geodesy techniques, in particular the handling and processing of satellite radar coherence data. They will also receive training in measurement and theory of landslide slope failure, and in processing and analysis of other relevant remotely sensed and field data (e.g. optical, topographic, hydrological etc.).
Training in a wide range of essential skills (e.g. presentation skills, paper/thesis writing, and computational skills) important both for life as a PhD student and afterwards is provided by the Department of Earth Sciences and Durham University, and the student will also benefit from cross-disciplinary training provided as part of IAPETUS2.
The student will become a member of the UK’s Centre for the Observation and Modelling of Earthquakes, Volcanoes and Tectonics (COMET), benefitting from the shared expertise of Geosciences staff in several universities, and attending regular meetings where the research of these various groups is discussed. The student will have opportunities to work with other partners in the UK and internationally and to interact with parallel research projects at Durham on mountain hazards and risks. The student will travel to national and international scientific meetings to present their results, and there may be opportunities to spend time with international research partners; this will be decided in collaboration with the student and is dependent upon the precise direction of the project. We aim to see all students publish at least two papers in leading scientific journals during their PhD. Upon completion, the student will be well equipped for a career in academia or in a range of industries. Past students have gone on to academic fellowships, as well as to jobs in environmental consultancy or in government and non-government organisations tasked with hazard assessment and disaster risk reduction.

References & further reading

Burrows, K., Walters, R. J., Milledge, D., Spaans, K. and Densmore, A., 2019. A new method for large-scale landslide classification from satellite radar. Remote Sensing, 11(3), 237, doi:10.3390/rs11030237
Burrows, K., Walters, R. J., Milledge, D. and Densmore, A., 2020. A systematic exploration of satellite radar coherence methods for rapid landslide detection. NHESS, doi:10.5194/nhess-2020-168
Robinson, T. R., Rosser, N. and Walters, R. J., 2019. The spatial and temporal influence of cloud cover on satellite-based emergency mapping of earthquake disasters. Scientific Reports, 9, 12455, doi:10.1038/s41598-019-49008-0
Williams, J.G., Rosser, N.J., Kincey, M.E., Benjamin, J., Oven, K., Densmore, A.L., Milledge, D.G., Robinson, T.R., Jordan, C.A. and Dijkstra, T.A., 2018. Satellite-based emergency mapping using optical imagery: experience and reflections from the 2015 Nepal earthquakes. NHESS, doi:10.5194/nhess-18-185-2018

Further Information

Dr Richard Walters
richard.walters@durham.ac.uk
+44(0) 1913 341727

Apply Now