Predicting rainfall‐induced landslide motion is challenging because shallow groundwater flow is extremely sensitive to the preexisting moisture content in the ground. Here, we use groundwater hydrology theory and numerical modeling combined with five years of field monitoring to illustrate how unsaturated groundwater flow processes modulate the seasonal pore water pressure rise and therefore the onset of motion for slow‐moving landslides. The onset of landslide motion at Oak Ridge earthflow in California’s Diablo Range occurs after an abrupt water table rise to near the landslide surface 52–129 days after seasonal rainfall commences. Model results and theory suggest that this abrupt rise occurs from the advection of a nearly saturated wetting front, which marks the leading edge of the integrated downward flux of seasonal rainfall, to the water table. Prior to this abrupt rise, we observe little measured pore water pressure response within the landslide due to rainfall. However, once the wetting front reaches the water table, we observe nearly instantaneous pore water pressure transmission within the landslide body that is accompanied by landslide acceleration. We cast the timescale to reach a critical pore water pressure threshold using a simple mass balance model that considers variable moisture storage with depth and explains the onset of seasonal landslide motion with a rainfall intensity‐duration threshold. Our model shows that the seasonal response time of slow‐moving landslides is controlled by the dry season vadose zone depth rather than the total landslide thickness.
Empirical thresholds for landslide warning systems have benefitted from the incorporation of soil‐hydrologic monitoring data, but the mechanistic basis for their predictive capabilities is limited. Although physically based hydrologic models can accurately simulate changes in soil moisture and pore pressure that promote landslides, their utility is restricted by high computational costs and nonunique parameterization issues. We construct a deep learning model using soil moisture, pore pressure, and rainfall monitoring data acquired from landslide‐prone hillslopes in Oregon, USA, to predict the timing and magnitude of hydrologic response at multiple soil depths for 36‐hr intervals. We find that observation records as short as 6 months are sufficient for accurate predictions, and our model captures hydrologic response for high‐intensity rainfall events even when those storm types are excluded from model training. We conclude that machine learning can provide an accurate and computationally efficient alternative to empirical methods or physical modeling for landslide hazard warning.
more » « less- Award ID(s):
- 1831770
- NSF-PAR ID:
- 10378294
- Publisher / Repository:
- DOI PREFIX: 10.1029
- Date Published:
- Journal Name:
- Geophysical Research Letters
- Volume:
- 47
- Issue:
- 16
- ISSN:
- 0094-8276
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
Abstract -
Modelling seasonal landslide motion: Does it only depend on fluctuations in normal effective stress?more » « less
Abstract Landslide motion is often simulated with interface‐like laws able to capture changes in frictional strength caused by the growth of the pore water pressure and the consequent reduction of the effective stress normal to the plane of sliding. Here it is argued that, although often neglected, the evolution of all the 3D stress components within the basal shear zone of landslides also contributes to changes in frictional strength and must be accounted for to predict changes in seasonal velocity. For this purpose, an augmented sliding‐consolidation model is proposed which allows for the computation of excess pore pressure development and downslope sliding with any constitutive law with 3D stress evolution. Simulations of idealised infinite slope models subjected to hydrologic forcing are used to study the role of in‐situ stress conditions and stress rate multiaxiality. Specifically, a Drucker‐Prager perfectly plastic model is used to replicate frictional failure and shear deformation at the base of landslides. The model reveals that conditions amenable to the shearing of a frictional interface are met only after numerous rainfall cycles, that is, when multiaxial stress rates are suppressed. In this case, the landslide is predicted to move through a seasonal ratcheting controlled only by the effective stress component normal to the plane of sliding. By contrast, in newly formed landslides, the multiaxial stress evolution is found to produce further regimes of motion, from plastic shakedown to cyclic failure, neither of which can be captured by interface‐like frictional laws. Notably, the model suggests that a transition across these regimes can emerge in response to an aggravation of the magnitude of forcing, implying that (i) fluctuations in climate may alter the seasonal trends of motion observed today; (ii) our ability to quantify landslide‐induced risks is impaired unless proper geomechanical models are used to examine their long‐term dynamics.
-
more » « less
Abstract Soil moisture is a major driver of microbial activity and thus, of the release of carbon (C) into the Earth's atmosphere. Yet, there is no consensus on the relationship between soil moisture and microbial respiration, and as a result, moisture response functions are a poorly constrained aspect of C models. In addition, models assume that the response of microbial respiration to moisture is the same for all ecosystems, regardless of climate history, an assumption that many empirical studies have challenged. These gaps in understanding of the microbial respiration response to moisture contribute to uncertainty in model predictions.
We review our understanding of what drives microbial moisture response, highlighting evidence that historical precipitation can influence both responses to moisture and sensitivity to drought. We present two hypotheses, the ‘climate history hypothesis’, where we predict that baseline moisture response functions change as a function of precipitation history, and the ‘drought legacy hypothesis’, in which we suggest that the intensity and frequency of historical drought have shaped microbial communities in ways that will control moisture responses to contemporary drought. Underlying mechanisms include biological selection and filtering of the microbial community by rainfall regimes, which result in microbial traits and trade‐offs that shape function.
We present an integrated modelling and empirical approach for understanding microbial moisture responses and improving models. Standardized measures of moisture response (respiration rate across a range of moistures) and accompanying microbial properties are needed across sites. These data can be incorporated into trait‐based models to produce generalized moisture response functions, which can then be validated and incorporated into conventional and microbially explicit ecosystem models of soil C cycling. Future studies should strive to analyse realistic moisture conditions and consider the role of environmental factors and soil structure in microbial response.
Microbes are the engines that drive C storage and are sensitive to changes in rainfall. A greater understanding of the factors that govern this sensitivity could be a key part of improving predictions of soil C dynamics, climate change and C‐climate feedbacks.
Read the free
Plain Language Summary for this article on the Journal blog. -
more » « less
Abstract Evidence suggests that following earthquakes, landslide activity, and susceptibility remain elevated for months to years after cessation of shaking. In landslides constrained within the soil mantle, enhanced post‐seismic activity may be attributed to persistence of progressive failure surface development that results in delayed triggering. With subsequent post‐seismic hydrologic disturbance, a commensurate decay in activity over time reflects evacuation of damaged hillslope materials and mechanical healing of colluvial media and vegetation recovery. In this study, we propose a hydromechanical model for representing the interplay of these processes including trends in timing and magnitude of post‐seismic shallow landslide activity. The hydromechanical model considers seismic or hydrologic disturbances as drivers for failure surface development in the soil mantle over a hillslope with commensurate weakening through accumulation of internal compression, shear softening, and loss of root strength. In the absence of well‐instrumented post‐seismic landslide measurement of evolving stress conditions and deformation, predictions were made against a unique field case study where hydrologically induced hillslope failure kinematics and internal compression were measured. The model accounts for seismic disturbances, demonstrating that certain peak ground accelerations directly reduce subsequent rainfall‐induced landslide‐triggering thresholds relative to undisturbed hillslopes. The model framework can be used to describe progressive, post‐seismic landslide activity, demonstrating that consideration of root or soil healing may suppress subsequent landslide triggering and contribute toward recovery to pre‐seismic disturbance levels of strength. Factors such as shear softening, vegetation, and healing are explored in the context of the temporal evolution of landslide rates after a seismic event. The results of this paper show mechanistically how (1) progressive development or cessation of a failure surface in active, post‐seismic landslides may describe elevated post‐seismic landslide rates and (2) that hillslope healing plays a modest role in the magnitude and timescale of attenuated landslide activity.
-
more » « less
Abstract Variations in hillslope soil moisture control forest hydrologic fluxes and storage pools, yet sparse observations combined with the complexity and heterogeneity of water movement and storage in the vadose zone can make temporal and spatial patterns and processes difficult to predict. We used two years of field observations of volumetric soil moisture at three depths (15, 30, and 100 cm) across five topographic positions (riparian, toeslope, sideslope, shoulder, and ridge) along three hillslope transects to better understand how soil moisture changes with hillslope position and through time. As expected, we found higher values of soil moisture at all depths at the riparian and toeslope positions. Unexpectedly, we found that ridges were particularly wet during the wet winter months and dried quickly during the summer months, indicating that topography alone cannot account for mean wet season soil moisture in our Mediterranean climate field site. The variability in soil moisture across all soil depths and topographic positions was greatest when soils were dry and decreased under wet soil conditions; this variability remained high in the deeper soil horizons, regardless of season. Lastly, event analysis suggests that the response to early season rainfall was highly variable along the hillslopes and was likely dominated by localized controls such as microtopography and vegetation as well as soil texture, antecedent moisture conditions, and rainfall characteristics. Our results suggest that the drivers of wet and dry season soil moisture dynamics can vary across topographic positions along a hillslope and do not always follow topographic controls.
