This paper investigates the application of Physics-Informed Neural Networks (PINNs) to inverse problems in unsaturated groundwater flow. PINNs are applied to the types of unsaturated groundwater flow problems modelled with the Richards partial differential equation and the van Genuchten constitutive model. The inverse problem is formulated here as a problem with known or measured values of the solution to the Richards equation at several spatio-temporal instances, and unknown values of solution at the rest of the problem domain and unknown parameters of the van Genuchten model. PINNs solve inverse problems by reformulating the loss function of a deep neural network such that it simultaneously aims to satisfy the measured values and the unknown values at a set of collocation points distributed across the problem domain. The novelty of the paper originates from the development of PINN formulations for the Richards equation that requires training of a single neural network. The results demonstrate that PINNs are capable of efficiently solving the inverse problem with relatively accurate approximation of the solution to the Richards equation and estimates of the van Genuchten model parameters.
Uncertainties in parameters of landslide susceptibility models often hinder them from providing accurate spatial and temporal predictions of landslide occurrences. Substantial contribution to the uncertainties in landslide assessment originates from spatially variable geotechnical and hydrological parameters. These input parameters may often vary significantly through space, even within the same geological deposit, and there is a need to quantify the effects of the uncertainties in these parameters. This study addresses this issue with a new three-dimensional probabilistic landslide susceptibility model. The spatial variability of the model parameters is modeled with the random field approach and coupled with the Monte Carlo method to propagate uncertainties from the model parameters to landslide predictions (i.e., factor of safety). The resulting uncertainties in landslide predictions allow the effects of spatial variability in the input parameters to be quantified. The performance of the proposed model in capturing the effect of spatial variability and predicting landslide occurrence has been compared with a conventional physical-based landslide susceptibility model that does not account for three-dimensional effects on slope stability. The results indicate that the proposed model has better performance in landslide prediction with higher accuracy and precision than the conventional model. The novelty of this study is illustrating the effects of the soil heterogeneity on the susceptibility of shallow landslides, which was made possible by the development of a three-dimensional slope stability model that was coupled with random field model and the Monte Carlo method.
Water-induced landslides pose a great risk to the society in Norway due to their high frequency and capacity to evolve in destructive debris flows. Hydrological monitoring is a widely employed method to understand the initiation mechanism of water-induced landslides under various climate conditions. Hydrological monitoring systems can provide relevant information that can be utilized in landslide early warning systems to mitigate the risk by issuing early warnings. These monitoring systems can be significantly enhanced, and wider deployments can be achieved through the recent developments within the domain of the Internet of Things (IoT). Therefore, this study aims to demonstrate a case study on an automated hydrological monitoring system supported by the IoT-based state-of-the-art technologies employing public mobile networks. Volumetric water content (VWC) sensors, suction sensors, and piezometers were used in the hydrological monitoring system to monitor the hydrological activities. The monitoring system was deployed in a case study area in central Norway at two locations of high susceptible geological units. During monitored period, the IoT-based hydrological monitoring system provided novel and valuable insights into the hydrological response of slopes to seasonally cold climates in terms of VWC and matric suction. The effects of rainfall, snow melting, ground freezing, and thawing were captured. The current study also made an attempt to integrate the collected data into a physical-based landslide susceptibility model to obtain a more consistent and reliable hazard assessment.