Landslide hazard assessment using a physics-based model informed by radar interferometry

Abstract:

Landslides pose a major hazard to the infrastructure systems and threaten the security, economy, safety, and health of the public, through both catastrophic failure and episodically accelerating creep. Here, we combine radar remote sensing observations, ground-based measurements, and physics-based numerical models to quantify the spatiotemporal hazard over two major landslides in Los Angeles (LA) metropolitan area: Portuguese Bend (PB) in southwest LA and Nimes Road (NR) in north LA (Fig. 1). We combine space-borne Synthetic Aperture Radar (SAR) imagery from Sentinel 1&2 with air-borne data set obtained from Uninhabited Aerial Vehicle SAR (UAVSAR) through an innovative wavelet-based Interferometric SAR (InSAR) processing algorithm to generate the three-dimensional (3D) landslide deformation time series during 2015-2020. The localized InSAR deformation history is then validated against the observations obtained through land surveying and GPS monitoring. Our results over the PB landslide (Fig. 2) suggest that an area of over 1 km² is undergoing creep with rate of up to 5 cm/yr in the satellite’s line of sight (LOS). Moreover, the time evolution of creep shows a periodic seasonal signal (Fig. 3), suggesting that elevated rainfall activity during winter and early spring might be a triggering factor for creep accelerations.

We further set up a two-dimensional Burgers-creep visco-plastic model based on the Finite Difference Method (FDM) to efficiently investigate and monitor the slow-moving time-dependent deformations and the failure mechanism of the landslides. The preliminary simulation results of the NR landslide indicate total displacements of up to 4 cm/yr (Fig. 4), in agreement with the in-situ inclinometer measurements. To calibrate the governing model parameters associated with the subsurface stratigraphy and geologic formations, we integrate the time-dependent landslide dynamics with the obtained 3D InSAR deformation time series through a non-linear model-updating procedure. This integrative approach enables establishing an operational landslide forecasting framework, capable of providing time-dependent landslide probabilities for infrastructure planning and regulatory agencies to aid disaster management and improve urban resilience.