基于不同位移预测模型的地震滑坡危险性评估研究-以天水地区为例

近年来,Newmark累积位移分析方法经过不断的改进和应用成为国际主流的地震滑坡危险性评估方法之一,众多学者基于位移预测模型开展区域地震滑坡危险性评估,然而鲜有针对不同位移模型对评估结果影响的定量研究。以天水地区为例,基于不同的位移预测模型开展地震滑坡危险性评估,对比位移模型对地震滑坡危险性评估的影响,探讨建立适用于我国的Newmark位移预测模型。结果表明:基于不同位移预测模型评估所得的地震滑坡危险性结果整体趋势一致,均能区分区域地震滑坡危险性等级的相对差异,但在同样的危险性分级标准下,所得中、高危险区的分布范围有较大差异。这与位移模型的函数形式及其区域相关性有关,在引入Newmark累积位移分析方法开展地震滑坡危险性评估的同时,应尽快建立考虑地震动衰减特征和工程地质背景的Newmark位移预测模型,为中国潜在地震滑坡危险性预测评估、震后滑坡快速评估等提供技术支撑。      

Probabilistic modeling of seismically triggered landslides using Monte Carlo simulations

The 2004 Chuetsu earthquakes of Niigata (Japan) triggered numerous landslides, and the most widespread types of landslides were highly disrupted, relatively shallow slides and soil (debris) flows. This paper presented a method to evaluate slope instability using Newmark displacement on a pixel-by-pixel basis in a given area. The proposed method was able to integrate Newmark displacement modeling and Monte Carlo simulations within geographical information systems. In the modeling, an empirical attenuation relationship was utilized to calculate Arias intensity over this study area, and the variability of geotechnical parameters was taken into account to calculate coseismic landslide displacement. Before deriving the displacement from related inputs, the Monte Carlo simulations ran 1,500 times and generated 1,500 displacement values for each grid cell, and then means and standard deviations of displacement were calculated and probabilistic distributions can be obtained. Finally, given 10 cm as a threshold value of displacement, estimated probabilities of displacement exceeding 10 cm were shown as a map of seismic landslide hazards. The resulting hazard map was classified into four categories from very low to high level.     

蒙特卡罗模拟在区域地震滑坡灾害评价中应用

2004年10月23日,日本新泻中越地区发生史上最强地震之一,震级达到Mw6.6。这次地震诱发上千个规模大小不一、形式各样的滑坡,造成一定的灾害损失和人员伤亡。因此,为了有效地避免和减轻这方面的灾害,有必要有效地、科学地预测和评价地震诱发滑坡的空间发生规律。为此以一定的地区作为研究对象,提出了蒙特卡罗随机模拟和Newmark滑动体位移分析法在区域地震斜坡失稳概率分析中的应用。该法充分考虑到岩土参数和相关地震参数空间变异性,结合地理信息系统空间分析的优势,以网格模型为基础,分析该区斜坡滑动体位移分布规律,并以2cm为失稳临界值探讨网格斜坡破坏的概率。实例证明,所提出的模型有效地预测了滑坡发生的空间分布规律。     

Setting up evaluate indicators for slope control engineering based on spatial clustering analysis

Rainfall-induced landslides (RILs) have been a source of social and economic disruption in the mountainous Baguio area in northern Philippines. Prolonged heavy rainfall usually happens during tropical cyclone and southwest monsoon activity. A pragmatic approach to RIL mitigation is to develop rainfall-based early warning. We implemented a modified regression method to derive the empirical minimum intensity (I)–duration (D) threshold I = 6.46 D ^?0.28 and a normalized ID threshold NI = 0.002 D ^?0.28 for rainfall duration ranging between 24 and 264 h. Using a separate data set to evaluate the applicability of the threshold, 93% of the landslide-triggering rainfall events fell above the derived threshold. RILs also occurred when 24-h rainfall was 0.02–28% of the mean annual precipitation or after accumulating at least 500 mm of rainfall from the onset of the rainy season. The thresholds may be further refined as more landslide data become available in the future.     

Shear wave velocity imaging of landslide debris deposited on an erodible bed and possible movement mechanism for a loess landslide in Jingyang,Xi’an,China

The South Jingyang Plateau, with a total area of 70 km², is located in Shaanxi Province, China. Since 1976, more than 50 landslides of different types have occurred repeatedly on the edge slopes of the plateau due to the start of diversion irrigation on the plateau, resulting in great loss of lives and property. To better understand the initiation and movement mechanisms of these loess landslides, we surveyed them and carried out a detailed investigation of a large landslide in the Xihetan area. Our field survey results revealed that although most of these landslides had a long runout with high mobility, most of the landslide materials originating from the edge slopes may have been in an unsaturated state when the landslide occurred. This suggests that the materials at the toe of the edge slope as well as on the travel path along the river terrace might have played a key role in landslide movement. To examine how the materials on the travel path were involved in the landsliding, we used a multichannel surface wave technique and surveyed shear wave velocity (V _s ) profiles of the landslide deposits. We also examined the internal geometry of the deposits that outcropped on the right-side slope of the landslide foot. The longitudinal profile of V _s along the direction of movement showed that terrace deposits near the toe of the edge slope may have been sheared upward, indicating that at the toe, the surface of rupture might be located inside the terrace deposits. The V _s contours showed an A-shaped fold within the landslide deposits in the middle part of the travel path and became greater in the most distal toe part. The V _s profile across the deposits showed a U-shaped belt, in which the soil layers have smaller V _s . This belt may be the boundary between the sliding landslide debris and terrace deposits. The observed internal geometry of the landslide deposits indicates that a sliding surface developed within the sandy layer underlying the gravel layer. Therefore, we inferred that after failure, the displaced landslide materials overrode and sheared the terrace deposits along its main sliding direction, resulting in the formation of thrust folds within the terrace deposits, and greater V _s on the distal toe part of the landslide.     

Near-source strong motion database catalog for Iran

This paper discusses a newly developed high-quality integrated dataset of shallow earthquake ground motions that occurred in Iran, from 1976 to 2013. A total of 860 three-component strong motion records are processed from 183 earthquake events, moment magnitudes 5.0?≤?M _w ?≤?7.4, and rupture distances of R _RUP  ≤ 120 km. Strong motion data from Iran having special tectonic features and shallow earthquakes with depths less than 35 km are included. This paper presents a thorough procedure used to collect and to generate a database following the Next-Generation Attenuation-West research projects. This database can be used in the development and ranking of ground motion models and for seismological and engineering hazard and risk analyses. Unprocessed strong motion records are obtained from the Iranian Strong Motion Network (ISMN). The time series collected were thoroughly examined through several rounds of quality reviews. The newly generated database includes the peak ground acceleration, peak ground velocity, and pseudo-spectral acceleration for the 5% damped with periods ranging from 0.01 to 10 s. The database also includes ground motion information and source characterization and parameters. This study is the near-source compiled ground motion database that can be used for Iran, and it is consistent with standard worldwide databases.     

Regression models for estimating coseismic landslide displacement

Newmark's sliding-block model is widely used to estimate coseismic slope performance. Early efforts to develop simple regression models to estimate Newmark displacement were based on analysis of the small number of strong-motion records then available. The current availability of a much larger set of strong-motion records dictates that these regression equations be updated. Regression equations were generated using data derived from a collection of 2270 strong-motion records from 30 worldwide earthquakes. The regression equations predict Newmark displacement in terms of (1) critical acceleration ratio, (2) critical acceleration ratio and earthquake magnitude, (3) Arias intensity and critical acceleration, and (4) Arias intensity and critical acceleration ratio. These equations are well constrained and fit the data well (71% < R² < 88%), but they have standard deviations of about 0.5 log units, such that the range defined by the mean ± one standard deviation spans about an order of magnitude. These regression models, therefore, are not recommended for use in site-specific design, but rather for regional-scale seismic landslide hazard mapping or for rapid preliminary screening of sites.     

Bounding relationships between parameters of earthquakes and seismogenic clastic dikes (based on worldwide data)

This paper presents the first boundary equations describing the relationship between earthquake parameters (magnitude M _S and macroseismic intensity I _P at the observation point on the MSK-64 scale) and clastic dikes (having maximal thickness m _cd , visible height h _cd , and the index of manifestation intensity of dikes in the cross section I _cd ). As was expected, the maximal size of dikes grows with an increase in the earthquake magnitude and macroseismic intensity. Analysis of the dependences showed that it is better to use all three parameters for estimation of the minimal threshold M _S or I _P from clastic dikes, and, in the absence of data on seismogenic rupture, the maximal calculated value should be used. Some limitations in application and the advantages of the equations obtained are discussed with respect to characterizing earthquakes of the pre-instrumental period.     

Numerical evidence of site effects contributing to triggering the Las Colinas landslide during the 2001 <Emphasis Type="Italic">M</Emphasis><Subscript>w</Subscript> = 7.7 El Salvador earthquake

The Las Colinas landslide, which was triggered by the 13 January 2001, M_w?=?7.7 El Salvador earthquake, was highly destructive. The local site is composed of pyroclastic flows, brown cinders, soft pyroclastic fall deposits and a thin palaeosol and is characterised by steep slopes. The extremely high ground motions recorded near the landslide location are assumed to be both produced by site effects and responsible for the landslide. To characterise the ground motion amplifications due to site effects in terms of the variation in geometrical and geological settings, parametric studies were conducted with a linear elastic slope model, which was vertically subjected to the scattering SV wave of the Gabor wavelet. The results show that a maximum amplification is obtained when the model slope angle is approximately 30° (similar to the actual slope angle), and the maximum amplification is located approximately 20 m behind the crest, where the actual movement was initiated. Additionally, a slope with a height of approximately 160 m enhances ground motion amplification. The subsurface geology is found to induce a greater effect on amplification than that of the slope topography. In particular, a soft pyroclastic fall deposit is observed to contribute most of the ground motion amplification. According to the numerical results, it can be concluded that the local site conditions induced extremely high ground motions that then contributed to the slope movement. Although the thin, buried layer of palaeosol did not cause any significant amplification, its weak cohesion enhanced movement.     

The character of scaling earthquake source spectra for Kamchatka in the 3.5–6.5 magnitude range

The properties of the source spectra of local shallow-focus earthquakes on Kamchatka in the range of magnitudes M _w = 3.5–6.5 are studied using 460 records of S-waves obtained at the PET station. The family of average source spectra is constructed; the spectra are used to study the relationship between M _w and the key quasi-dimensionless source parameters: stress drop Δσ and apparent stress σ_a. It is found that the parameter Δσ is almost stable, while σ_a grows steadily as the magnitude M _w increases, indicating that the similarity is violated. It is known that at sufficiently large M _w the similarity hypothesis is approximately valid: both parameters Δσ and σ_a do not show any noticeable magnitude dependence. It has been established that M _w ≈ 5.7 is the threshold value of the magnitude when the change in regimes described occurs for the conditions on Kamchatka.     

基于简化 Newmark位移模型的区域地震滑坡危险性快速评估以汶川MS 8.0级地震为例

以汶川MS8.0级地震重灾区的11县市为例,初步提出了基于简化Newmark位移模型的地震滑坡危险性应急快速评估方法。利用汶川地震即时地震动参数、工程地质岩性经验分组及地形坡度数据,借助ArcGIS空间数据建模工具编制了地震滑坡危险性快速评估流程模块。计算了区域浅表层饱和岩土体斜坡的静态安全系数Fs、临界加速度ac,并借此分析了地震滑坡易发性。利用经验式获得了汶川地震Arias强度和区域滑坡位移DN分布,实现了汶川地震重灾区地震滑坡危险性的快速评估,为应急救灾决策提供了参考。通过对比评估结果和震后滑坡调查成果,可知数十处灾难性滑坡绝大部分位于-高危险区的龙门山主中央断裂带两侧约20km地带中,显示了评估方法的可靠性; 同时,分析指出了空间数据精度及更新不足导致局部评估结果欠佳的局限性,并提出了改进建议。     

Coseismic landslides triggered by the 8th August 2017 <Emphasis Type="Italic">M</Emphasis><Subscript>s</Subscript> 7.0 Jiuzhaigou earthquake (Sichuan,China): factors controlling their spatial distribution and implications for the seismogenic blind fault identification

On 8th August 2017, a magnitude M_s 7.0 earthquake struck the County of Jiuzhaigou, in Sichuan Province, China. It was the third M_s ≥?7.0 earthquake in the Longmenshan area in the last decade, after the 2008 M_s 8.0 Wenchuan earthquake and the 2013 M_s 7.0 Lushan earthquake. The event did not produce any evident surface rupture but triggered significant mass wasting. Based on a large set of pre- and post-earthquake high-resolution satellite images (SPOT-5, Gaofen-1 and Gaofen-2) as well as on 0.2-m-resolution UAV photographs, a polygon-based interpretation of the coseismic landslides was carried out. In total, 1883 landslides were identified, covering an area of 8.11 km², with an estimated total volume in the order of 25–30?×?10⁶ m³. The total landslide area was lower than that produced by other earthquakes of similar magnitude with strike-slip motion, possibly because of the limited surface rupture. The spatial distribution of the landslides was correlated statistically to a number of seismic, terrain and geological factors, to evaluate the landslide susceptibility at regional scale and to identify the most typical characteristics of the coseismic failures. The landslides, mainly small-scale rockfalls and rock/debris slides, occurred mostly along two NE-SW-oriented valleys near the epicentre. Comparatively, high landslide density was found at locations where the landform evolves from upper, broad valleys to lower, deep-cut gorges. The spatial distribution of the coseismic landslides did not seem correlated to the location of any known active faults. On the contrary, it revealed that a previously-unknown blind fault segment—which is possibly the north-western extension of the Huya fault—is the plausible seismogenic fault. This finding is consistent with what hypothesised on the basis of field observations and ground displacements.     

Effect of Slope on <Emphasis Type="Italic">p</Emphasis>-<Emphasis Type="Italic">y</Emphasis> Curves for Laterally Loaded Piles in Soft Clay

Pile foundations are often subject to lateral loading due to various forces on a variety of structures like high rise buildings, transmission towers, power stations, offshore structures and highway and railway structures. The present investigation is to study the effect of slopes on p-y curves (where p is the static soil reaction and y is the pile deflection) due to static lateral loading in soft clay (Consistency index I_c = 0.42). A series of laboratory model tests were carried out on the instrumented model pile on sloping ground (slopes of 1V:1H, 1V:1.5H, 1V:2H, 1V:3H and 1V:5H) and with varying embedment length to diameter ratio (L/D) of 20, 25 and 30. From the experimental results, the bending moment curves along the pile shaft are double differentiated to obtain the soil resistance (p) and double integrated to obtain the deflection (y) using curve fitting method. New p-y curves for piles located on crest of soft clay with different sloping ground surface under static lateral loading are developed. Moreover, the effect of sloping angles on proposed p-y curves was studied.     

Earthquake hazard in Northeast India — A seismic microzonation approach with typical case studies from Sikkim Himalaya and Guwahati city

A comprehensive analytical as well as numerical treatment of seismological, geological, geomorphological and geotechnical concepts has been implemented through microzonation projects in the northeast Indian provinces of Sikkim Himalaya and Guwahati city, representing cases of contrasting geological backgrounds — a hilly terrain and a predominantly alluvial basin respectively. The estimated maximum earthquakes in the underlying seismic source zones, demarcated in the broad northeast Indian region, implicates scenario earthquakes of M _W 8.3 and 8.7 to the respective study regions for deterministic seismic hazard assessments. The microzonation approach as undertaken in the present analyses involves multi-criteria seismic hazard evaluation through thematic integration of contributing factors. The geomorphological themes for Sikkim Himalaya include surface geology, soil cover, slope, rock outcrop and landslide integrated to achieve geological hazard distribution. Seismological themes, namely surface consistent peak ground acceleration and predominant frequency were, thereafter, overlaid on and added with the geological hazard distribution to obtain the seismic hazard microzonation map of the Sikkim Himalaya. On the other hand, the microzonation study of Guwahati city accounts for eight themes — geological and geomorphological, basement or bedrock, landuse, landslide, factor of safety for soil stability, shear wave velocity, predominant frequency, and surface consistent peak ground acceleration. The five broad qualitative hazard classifications — ‘low’, ‘moderate’, ‘high’, ‘moderate high’ and ‘very high’ could be applied in both the cases, albeit with different implications to peak ground acceleration variations. These developed hazard maps offer better representation of the local specific seismic hazard variation in the terrain.     

Impact of rainfall intensity on the hydrological performance of erosion control geotextiles

The impact of erosion control geotextiles on the surface runoff from slopes is quite variable and depends strongly on site-specific conditions (soil characteristics, slope morphology, climate, etc.), as has been shown in several earlier studies. In addition, little is known about the proportion of runoff reduction that is caused by the geotextile and the proportion that is caused by soil characteristics. To shed more light on this issue, an experiment was carried out to test the impact of 500 g m^?2 jute nets (J500) and 400 g m^?2; 700 g m^?2 coir nets (C400, C700) on the surface runoff from simulated rainfall of four different intensities (I ₁ = 18.7; I ₂ = 27.2; I ₃ = 53.6; I ₄ = 90.5 mm h^?1). Data on runoff volume, peak discharge and time to peak discharge were collected from 40 simulated rainfall events. An impermeable “no-soil” subgrade was used to examine the impact of the geotextile on runoff without any influence of soil. All tested geotextiles significantly reduced runoff (volume, peak discharge) at all rainfall intensities, with the exception of C400 and C700 during simulated rainfall intensity I ₄. J500 seemed to have the most effective runoff reduction performance at all rainfall intensities. In general, as the rainfall intensity increased, the effectiveness of the geotextiles decreased. Interesting behaviour was observed for J500 under simulated rainfall intensity I ₄—the effectiveness of the geotextile increased with the duration of the rainfall.     

The characteristics and failure mechanism of the largest landslide triggered by the Wenchuan earthquake, May 12, 2008, China

Strong earthquakes are among the prime triggering factors of landslides. The 2008 Wenchuan earthquake (M _w = 7.9) triggered tens of thousands of landslides. Among them, the Daguangbao landslide is the largest one, which covered an area of 7.8 km² with a maximum width of 2.2 km and an estimated volume of 7.5 × 10⁸ m³. The landslide is located on the hanging wall of the seismogenic fault, the Yingxiu–Beichuan fault in Anxian town, Sichuan Province. The sliding mass travelled about 4.5 km and blocked the Huangdongzi valley, forming a landslide dam nearly 600 m high. Compared to other coseismic landslides in the study area, the Daguangbao landslide attained phenomenal kinetic energy, intense cracking, and deformation, exposing a 1-km long head scarp in the rear of the landslide. Based on the field investigation, we conclude that the occurrence of the landslide is controlled mainly by the seismic, terrain, and geological factors. The special location of the landslide and the possible topographic amplification of ground motions due to the terrain features governed the landslide failure. The effects of earthquakes on the stability of slopes were considered in two aspects: First, the ground shaking may reduce the frictional strength of the substrate by shattering of rock mass. Second, the seismic acceleration may result in short-lived and episodic changes of the normal (tensile) and shear stresses in the hillshopes during earthquakes. According to the failure mechanism, the dynamic process of the landslide might contain four stages: (a) the cracking of rock mass in the rear of the slope mainly due to the tensile stress generated by the ground shaking; (b) the shattering of the substrate due to the ground shaking, which reduced the frictional strength of the substrate; (c) the shearing failure of the toe of the landslide due to the large shear stress caused by the landslide gravity; and (d) the deposition stage.     

Site response of the Ganges basin inferred from re-evaluated macroseismic observations from the 1897 Shillong, 1905 Kangra,and 1934 Nepal earthquakes

We analyze previously published geodetic data and intensity values for the M _s = 8.1 Shillong (1897), M _s = 7.8 Kangra (1905), and M _s = 8.2 Nepal/Bihar (1934) earthquakes to investigate the rupture zones of these earthquakes as well as the amplification of ground motions throughout the Punjab, Ganges and Brahmaputra valleys. For each earthquake we subtract the observed MSK intensities from a synthetic intensity derived from an inferred planar rupture model of the earthquake, combined with an attenuation function derived from instrumentally recorded earthquakes. The resulting residuals are contoured to identify regions of anomalous intensity caused primarily by local site effects. Observations indicative of liquefaction are treated separately from other indications of shaking severity lest they inflate inferred residual shaking estimates. Despite this precaution we find that intensites are 1–3 units higher near the major rivers, as well as at the edges of the Ganges basin. We find evidence for a post-critical Moho reflection from the 1897 and 1905 earthquakes that raises intensities 1–2 units at distances of the order of 150 km from the rupture zone, and we find that the 1905 earthquake triggered a substantial subsequent earthquake at Dehra Dun, at a distance of approximately 150 km. Four or more M = 8 earthquakes are apparently overdue in the region based on seismic moment summation in the past 500 years. Results from the current study permit anticipated intensities in these future earthquakes to be refined to incorporate site effects derived from dense macroseismic data.     

A seismic landslide hazard analysis with topographic effect,a case study in the 99 Peaks region,Central Taiwan

It has been known that ground motion amplitude will be amplified at mountaintops; however, such topographic effects are not included in conventional landslide hazard models. In this study, a modified procedure that considers the topographic effects is proposed to analyze the seismic landslide hazard. The topographic effect is estimated by back analysis. First, a 3D dynamic numerical model with irregular topography is constructed. The theoretical topographic amplification factors are derived from the dynamic numerical model. The ground motion record is regarded as the reference motion in the plane area. By combining the topographic amplification factors with the reference motions, the amplified acceleration time history and amplified seismic intensity parameters are obtained. Newmark’s displacement model is chosen to perform the seismic landslide hazard analysis. By combining the regression equation and the seismic parameter of peak ground acceleration and Arias intensity, the Newmark’s displacement distribution is generated. Subsequently, the calculated Newmark’s displacement maps are transformed to the hazard maps. The landslide hazard maps of the 99 Peaks region, Central Taiwan are evaluated. The actual landslide inventory maps triggered by the 21 September 1999, Chi-Chi earthquake are compared with the calculated hazard maps. Relative to the conventional procedure, the results show that the proposed procedures, which include the topographic effect can obtain a better result for seismic landslide hazard analysis. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

Thaw-induced slope failures and stability analyses in permafrost regions of the Qinghai-Tibet Plateau,China

The distribution of permafrost-related slope failures along the Qinghai-Tibet Highway from Wuddaoliang to Fenghuoshan correlates with ice content, slope gradient, and ground temperature. Slope failures are of two types. (1) Retrogressive thaw slumps result from icy permafrost being exposed by either man-induced excavation or fluvial-thermal erosion and undercutting of basal slopes. (2) Active-layer-detachment failures are caused by thaw of icy permafrost at the active layer-permafrost interface. After initial failure, active-layer-detachment failures can lead to retrogressive thaw-slumping and localized surficial landslide. Common trigger mechanisms for failure include high summer air temperatures and heavy summer precipitation. A third possible trigger mechanism for slope failure is earthquake occurrence. A geotechnical slope stability analysis was undertaken for an active-layer-detachment failure that had progressed into a retrogressive thaw slump. A safety factor (F _s ) of 1.24 for the natural slope was determined using in situ tested strength parameters. However, the slope would lose stability when either the groundwater level over the permafrost table exceeded 1.42 m or seismic acceleration reached, or exceeded, 0.03?g.     

Field monitoring of groundwater responses to heavy rainfalls and the early warning of the Kualiangzi landslide in Sichuan Basin,southwestern China

The Kualiangzi landslide was triggered by heavy rainfalls in the “red beds” area of Sichuan Basin in southwestern China. Differing from other bedrock landslides, the movement of the Kualiangzi landslide was controlled by the subvertical cracks and a subhorizontal bedding plane (dip angle < 10°). The ingress of rainwater in the cracks formed a unique groundwater environment in the slope. Field measurement for rainfall, groundwater movement, and slope displacement has been made for the Kualiangzi landslide since 2013. The field monitoring system consists of two rainfall gauges, seven piezometers, five water-level gauges, and two GPS data loggers. The equipments are embedded near a longitudinal section of the landslide, where severe deformation has been observed in the past 3 years. The groundwater responses to four heavy rainfall events were analyzed between June 16 and July 24 in 2013 coincided with the flood season in Sichuan. Results showed that both of the water level and the pore-water pressure increased after each rainfall event with delay in the response time with respect to the precipitation. The maximum time lag reached 35 h occurred in a heavy rainfall event with cumulative precipitation of 127 mm; such lag effect was significantly weakened in the subsequent heavy rainfall events. In each presented rainfall event, longer infiltration period in the bedrock in the upper slope increased the response time of groundwater, compared to that of in the gravels in the lower slope. A translational landslide conceptual model was built for the Kualiangzi landslide, and the time lag was attributed to the gradual formation of the uplift pressure on the slip surface and the softening of soils at the slip surface. Another important observation is the effect on the slope movement which was caused by the water level (H _w) in the transverse tension trough developed at the rear edge of the landslide. Significant negative correlation was found for H _w and the slope stability factor (F _s), in particular for the last two heavy rainfall events, of which the drastic increase of water level caused significant deterioration in the slope stability. The rapid drop (Δ?=?22.5 kPa) of pore-water pressure in the deep bedrock within 1 h and the large increase (Δ?=?87.3 mm) of surficial displacement were both monitored in the same period. In the end, a four-level early warning system is established through utilizing H _w and the displacement rate D _r as the warning indicators. When the large deformation occurred in flood season, the habitants at the leading edge of the landslide can be evacuated in time.