A simplified one-dimensional unsaturated hydrology model for rainfall-induced landslides

The Richardson-Richards equation is the standard approach to model transient unsaturated landslide hydrology. As a nonlinear partial differential equation, numerical solutions are needed and typically require fine grid sizes and high computing times; these requirements make utilizing the equation for spatially explicit landslide forecasting models cumbersome for storm periods exceeding a single pulse. Conversely, analytical models used to compute pore-water pressure evolution within landslides often assume linear diffusivity, which is generally applicable only under conditions where landslides are fully saturated and thus already close to failure. These observations point to a need for a physically representative yet numerically tractable approach for modeling the pore-water pressure evolution of the subsurface over broad areas and seasonal time scales.

With that goal in mind, here we develop a one-dimensional variably saturated hydrology model using a variant of the Soil Moisture Velocity Equation-based vadose zone solver (Ogden et al., 2015, 2017), called the Green-Ampt Redistribution Talbot-Ogden (GARTO) model (Lai et al., 2015), and couple it to a dynamic water table model to track the evolution of subsurface pore-water pressure in response to rainfall. While this modeling framework is currently limited to one dimension, advantages include relatively fast computing times, the ability to include additional hydrologic processes such as lateral fluxes and secondary flow to the model domain, and possible incorporation into 2-D saturated groundwater models (Ogden et al., 2015).

We apply our model to two end-member cases where we have data for material properties and landslide failure timing: 1) an extensive shallow landslide event on March 22, 2018 along the Tuolumne River canyon in Northern California from extreme rainfall by mesoscale effects associated with an atmospheric river; and 2) the time-series record of seasonal displacement for the Oak Ridge earthflow outside San Jose, CA due to annual precipitation variability. Comparable results to solutions using the USGS model VS2DT suggest that considering only the advection of pore water through the vadose zone to the water table may provide a path toward more efficient spatially distributed seasonal forecasting of rainfall-induced landslides.