Physically-based landslide prediction over a large region: Scaling low-resolution hydrological model results for high-resolution slope stability assessment
Introduction
Rainfall-triggered shallow landslides are worldwide natural hazard (Hong et al., 2006; An et al., 2016), causing a large number of fatalities and property losses (Liao et al., 2010; He et al., 2016; Godt et al., 2009). Globally, landslide hazards cause approximately 1000 deaths and USD 4 billion losses per year (Pradhan and Youssef, 2010). China is severely affected by landslide hazards, which has led to 1100 fatalities and 5–10 billion US dollars since 2000 (Hong et al., 2017). Due to frequent fatalities and extensive property damages that landslide hazard may cause (Papathoma-Kohle et al., 2015), it is essential to investigate and predict landslide hazards to avoid the damages (Zhang et al., 2019). Methods for hazard assessment range from heuristic susceptibility approaches (Fookes, 1997; Guzzetti et al., 2000; Griffiths and Edwards, 2001; Griffiths, 2002) to lumped regional empirical rainfall thresholds (Bogaard and Greco, 2018; Glade et al., 2000; Caine, 1980; Guzzetti et al., 2007, 2008) to detailed physical-based coupled hydrological-slope stability models (He et al., 2016; Zhang et al., 2016; Wilkinson et al., 2002; Alvioli and Baum, 2016; Montrasio and Valentino, 2016a, b; Van Asch et al., 2007). However, most of the studies focused on a single slope or landslide event in a relatively small catchment, usually in the magnitude of 102–103 km2 in a resolution of 3 arcsec (~90 m). This resolution is also a prerequisite to validate model results against observed landslides from inventories (Tian et al., 2008). Yet, the corresponding computational burden jeopardizes the applicability of physically-based modeling to larger area (e.g., basins≫ 10,000 km2) (Bellugi et al., 2011; Camilo et al., 2017) and to derive necessary model parameters (Bardossy and Singh, 2008; Yao et al., 2012). In order to apply physically-based models in landslide prediction with high resolution, some computational expediency is needed in large basins.
Low-resolution but large scale hydrological models (typically applied at 1 km resolution or coarser) are widely available due to increasing availability of climate data and land surface datasets (Sun et al., 2017; Huang et al., 2019; Yao et al., 2019; van Beek et al., 2011; Wada et al., 2011; Luo et al., 2018). These models can be calibrated against observed discharge (Beck et al., 2016; Chao et al., 2018) so it captures the spatiotemporal variations across the basin. However, this coarse resolution does not fit the purpose for validation of landslide prediction because the size of a rainfall-triggered shallow landslide event is usually only tens or hundreds m2 (Chen et al., 2017; Zhang et al., 2018a). In this case, a downscaling method should be used in order to couple coarse-resolution hydrological models with fine-resolution slope stability models.
Soil moisture is an important component in water and energy balance, affecting subsurface flow, soil evapotranspiration, hydrological response of a catchment etc., as a linking variable between hydrological model and slope stability model (Bogaard and Greco, 2016; Zhang et al., 2019). For slope stability model, the degree of soil saturation affects ground water table response to rainfall and thereby slope stability (Talebi et al., 2007; Bogaard and Greco, 2016; Krzeminska et al., 2012). Several studies showed that topography could be a good indicator for spatial soil moisture distribution (Beaudette et al., 2013; Burt and Butcher, 1985; Sveditchnyi et al., 2003). The topographic wetness index (TWI), as one of the topographic indexes, shows a high correlation with soil moisture during wet conditions (Grayson and Western, 1998; Brocca et al., 2010). Therefore, it is reasonable to downscale the soil moisture via using TWI to link hydrological model and slope stability model under different spatial resolution.
The main objective of this study is to assess the potential of physically-based landslide hazard assessment on the scale of 100,000 km2. In this study we use topographic information to downscale the results of the coarse-scale hydrological model to the finer resolution slope stability model in an efficient and expedient manner. The hydrological model we used is the Coupled Routing and Excess STorage (CREST) model (Wang et al., 2011), a distributed hydrological model, which has been applied at a resolution of 1 km to the Shaanxi Province in Northwest China for an area of more than 200,000 km2 for the period of 2009–2012. We downscaled the soil moisture information via TWI and land surface conditions to a finer resolution of 90 m. For the landslide assessment, we used a physically-based landslide model, namely, the SLope-Infiltration-Distributed Equilibrium (SLIDE) model (Montrasio and Valentino, 2008; He et al., 2016), which used the result of downscaled soil moisture as input variable. The results of the landslide hazard assessment are compared to classical lumped regional rainfall thresholds to quantify the possible improvement.
This article is organized as follows. It firstly describes the study area and datasets in Section 2. Then a brief introduction of hydrological model and slope stability model along with the downscaling method are described in Section 3. Section 4 presents results of our research and hence a comparison between the results and the results of landslide prediction using classic rainfall threshold is discussed in section 5. Conclusions, limitations of the research and its potential improvements are described in the end.
Section snippets
Description of the study area
Our study area is Shaanxi Province, located in the middle land of northwest China (Fig. 1). It situates between 105°29′ E−111°15′ E and 31°42′ N-39°35′ N with a total area of about 205,800 km2. Elevation ranges from approximately 150 to 3800 m, which the highest elevation is in the south, followed by north and middle. Land cover is distinct from south to north: forest, cultivated land and grass land, respectively (Fig. 2b). The soil type is mainly loam in this region (Fig. 2c). The study area
Description of the CREST model
The CREST model (Wang et al., 2011), developed by the University of Oklahoma and NASA SERVIR project team, is a physically-based distributed hydrological model. The model simulation starts from canopy interception. After that, the rainfall, reaching soil surface, is divided into surface runoff and infiltration according to the Variable Infiltration Capacity curve (VIC), a concept originating from the Xinanjiang Model (Zhao, 1992) and later represented in the VIC model (Liang et al., 1994, 1996
Simulation of CREST model
The CREST model was used to simulate the hydrological processes for the period of 2009–2012 for the study area. The 2009–2010 period is used to calibrate the model, while the 2011–2012 as the validation period. In order to reduce the effect of initial conditions, the first three months were not considered in model evaluation. The model was calibrated manually. The processes of modeled and observed daily discharge for the four stations are shown in Fig. 4. The modeled discharges show generally
Discussion
In this study, we coupled the low-resolution hydrological model CREST with the high-resolution slope stability model SLIDE using a spatial downscaling method for soil moisture, which is a connection variable between these two physically-based distributed models.
High-resolution hydrological modeling has a better representation of spatial heterogeneity of land surface attributes (e.g. topography, soil, and land cover) (Wood et al., 2011), however, it is difficult and inefficient, sometimes
Conclusions
This study aims to predict landslides in a very large region by coupling a low-resolution hydrological model with a high-resolution slope stability model. The soil moisture is the coupling variable, which should be downscaled to use in the slope stability model. The topographic wetness index is the downscaling parameter linking the coarser-resolution soil moisture with the finer one.
The CREST model and the SLIDE model were applied in this study. Both of the two models performed well: the CREST
Author contributions
KZ, TB, LPHvB and SW designed this study; KZ and SW set up the models; SW,TB, and KZ conducted this study; SW and KZ wrote the manuscript; SW, KZ, TB, LPHvB and XT contributed to the discussion and revision.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
The first author is grateful to PhD candidate Changrang Zhou for her inspiring idea about algorithm coding. This study is supported by the National Key Research and Development Program of China (2018YFC1508101, 2016YFC0402701), National Natural Science Foundation of China (51879067), Natural Science Foundation of Jiangsu Province (BK20180022), Six Talent Peaks Project in Jiangsu Province (NY-004), and Fundamental Research Funds for the Central Universities of China (2018B42914, 2018B04714). SW
References (81)
- et al.
Parallelization of the TRIGRS model for rainfall-induced landslides using the message passing interface
Environ. Model. Softw
(2016) - et al.
Development of time-variant landslide-prediction software considering three-dimensional subsurface unsaturated flow
Environ. Model. Softw
(2016) - et al.
Effect of raster resolution and polygon-conversion algorithm on landslide susceptibility mapping
Environ. Model. Softw
(2016) - et al.
Handling high predictor dimensionality in slope-unit-based landslide susceptibility models through LASSO-penalized Generalized Linear Model
Environ. Model. Softw
(2017) - et al.
Global land cover mapping at 30 m resolution: a POK-based operational approach
ISPRS J. Photogrammetry Remote Sens.
(2015) - et al.
Radiance calibration of DMSP-OLS low-light imaging data of human settlements
Remote Sens. Environ.
(1999) An introduction to ROC analysis
Pattern Recognit. Lett.
(2006)- et al.
Towards areal estimation of soil water content from point measurements: time and space stability of mean response
J. Hydrol.
(1998) - et al.
Development of a coupled hydrological-geotechnical framework for rainfall-induced landslides prediction
J. Hydrol.
(2016) - et al.
Effects of spatial resolution on a raster based model of flood flow
J. Hydrol.
(2001)
Land information system: an interoperable framework for high resolution land surface modeling
Environ. Model. Softw
Effects of DEM sources on hydrologic applications
Comput Environ Urban
Regional mapping of human settlements in southeastern China with multisensor remotely sensed data
Remote Sens. Environ.
Impact of forest maintenance on water shortages: Hydrologic modeling and effects of climate change
Science of the Total Environment
A tool for the automatic calculation of rainfall thresholds for landslide occurrence
Environ. Model. Softw
Modelling rainfall-induced shallow landslides at different scales using SLIP - Part II
Procedia Engineer
Modelling rainfall-induced shallow landslides at different scales using SLIP - Part I
Procedia Engineer
Loss estimation for landslides in mountain areas - an integrated toolbox for vulnerability assessment and damage documentation
Environ. Model. Softw
Rainfall thresholds for possible landslide occurrence in Italy
Geomorphology
Problems in predicting the mobility of slow-moving landslides
Eng. Geol.
Landslide hazard and bioengineering: towards providing improved decision support through integrated numerical model development
Environ. Model. Softw
Statistical and hydrological evaluation of TRMM-based Multi-satellite Precipitation Analysis over the Wangchu Basin of Bhutan: are the latest satellite precipitation products 3B42V7 ready for use in ungauged basins?
J. Hydrol.
A priori parameter estimates for a distributed, grid-based Xinanjiang model using geographically based information
J. Hydrol.
Using multi-satellite microwave remote sensing observations for retrieval of daily surface soil moisture across China
Water Science and Engineering
Spatially explicit shallow landslide susceptibility mapping over large areas
Robust estimation of hydrological model parameters
Hydrol. Earth Syst. Sci.
Terrain-shape indices for modeling soil moisture dynamics
Soil Sci. Soc. Am. J.
Global-scale regionalization of hydrologic model parameters
Water Resour. Res.
Validation and evaluation of predictive models in hazard assessment and risk management
Nat. Hazards
Landslide hydrology: from hydrology to pore pressure
Wires Water
Invited perspectives: hydrological perspectives on precipitation intensity-duration thresholds for landslide initiation: proposing hydro-meteorological thresholds
Nat. Hazard Earth Sys.
Spatial-temporal variability of soil moisture and its estimation across scales
Water Resour. Res.
Topographic controls of soil-moisture distributions
J. Soil Sci.
The rainfall intensity - duration control of shallow landslides and debris flows
Geogr. Ann.
Applicability assessment of the CASCade Two Dimensional SEDiment (CASC2D‐SED) distributed hydrological model for flood forecasting across four typical medium and small watersheds in China
Journal of Flood Risk Management
Relationship between landslide size and rainfall conditions in Taiwan
Landslides
Downscaling Soil Moisture Using Topography: the Evaluation and Optimisation of a Downscaling Approach
Mapping city lights with nighttime data from the DMSP operational linescan system
Photogramm. Eng. Remote Sens.
Geology for engineers: the geological model, prediction and performance
Q. J. Eng. Geol.
The relationship of the unsaturated soil shear strength to the soil-water characteristic curve
Can. Geotech. J.
Cited by (99)
PHyL v1.0: A parallel, flexible, and advanced software for hydrological and slope stability modeling at a regional scale
2024, Environmental Modelling and SoftwareA decadal review of the CREST model family: Developments, applications, and outlook
2023, Journal of Hydrology XAn open-source and QGIS-integrated physically based model for Spatial Prediction of Rainfall-Induced Shallow Landslides (SPRIn-SL)
2023, Environmental Modelling and SoftwareRegional-scale modeling of rainfall-induced landslides under random rainfall patterns
2022, Environmental Modelling and Software