A Python-based Framework for Ensemble-based Flow-like Landslide Forecasting
Abstract
Landslides are heavily occurring hazards worldwide that cause loss of life and devastation of economy. Process-based numerical modeling is widely used to forecast landslide run-out and dynamic properties, assess potential hazard, and develop mitigation measures. Such modeling typically requires a host of inputs, including digital elevation models (DEMs), release area, model parameters, etc. All of these are usually subject to uncertainty. In order to allow for an informed decision, it is necessary to study the impact of each of them on simulation results as well as their interactions. On a technical level, such study requires the capability to conduct uncertainty modeling of inputs (e.g. DEM uncertainty modeling), landslide process modeling, and usually requires a large number of simulation runs that leads to extensive post-processing work. Each step requires certain level of expertise.
In this study, we develop a python-based framework PSI-slide (Predictive Simulation of slides) to facilitate such study (first steps are reported in Zhao and Kowalski, 2018). PSI-slide integrates self-written python scripts and existing software and toolboxes, including SGeMS for DEM uncertainty modeling, WhiteboxTools for terrain analysis, RAMMS and r.avaflow for landslide process modeling, scripts for visualization and statistical analysis, etc. The advantage is that self-written scripts and individual software and toolboxes interact seamlessly. Workflow control is reduced to providing an input file that specifies inputs and corresponding uncertainties (e.g. root mean square error or higher accurate information of DEM error, probability distribution of model parameters, etc.), numerical solver, and required outputs (e.g. boxplots of flow dynamics at specific locations, probabilistic hazard map, etc.). The potential of PSI-slide is investigated by conducting a comprehensive flow-like landslide forecasting based on a historic landslide case. In the end, we discuss to which extent this computational modeling stack could be further developed into a new community tool. References Zhao, H. and Kowalski, J. DEM uncertainty propagation in rapid flow-like landslide simulations. Proceedings of the Second JTC1 Workshop, Hong Kong, 2018.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFMNH33D0944Z
- Keywords:
-
- 1810 Debris flow and landslides;
- HYDROLOGY;
- 1826 Geomorphology: hillslope;
- HYDROLOGY;
- 4306 Multihazards;
- NATURAL HAZARDS;
- 7212 Earthquake ground motions and engineering seismology;
- SEISMOLOGY