WHY WAS THE AUGUST 2010 ZHOUQU LANDSLIDE SO POWERFUL?

. On August 8, 2010 in the northwestern Chinese province of Gansu, rainstorm-triggered debris flow devastated the small county of Zhouqu. A modeling study, using a new multiple-phase scalable and extensible geo-fluid model, suggests that the cause is the result of an intersection of several events. These were a heavy rainstorm, not necessarily the result of global warming, which triggered the landslide and followed a drought that created surface cracks and crevasses; the geology of the region, notably the loess covering heavily weathered surface rock; and the bedrock damage, which deepened the surface crevasses, inflicted by the 7,9 magnitude Wenchuan earthquake of May 12, 2008. Deforestation and topsoil erosion also contribute. The modeling results underscore the urgency for a high priority program of re-vegetation of Zhouqu county, without which the region will remain exposed to future disastrous, ‘progressive bulking’ type landslides. NPP quantifies the of atmospheric by NPP ecosystem constraints NPP perturbation NPP


INTRODUCTION
Landslides occur irregularly and future research is concerned with developing more accurate predictions about their timing (when), location (where) and size (how big they will be), and in developing procedures that convey risk and warnings to the public to mitigate loss of life and damage to infrastructure and ecosystems [ The Zhouqu landslides were preceded by an extreme precipitation event which occurred around midnight of August 7, 2010 ( Fig. 1). Both the precipitation intensity of 77,3 mm/hr near 104,42E, 33,78N, and total rainfall amount of 96,3 mm in 24 hours are the highest recorded for the period since the May 2008 Wenchuan magnitude 7,9 earthquake. From a longer perspective, the Zhouqu rainfall event had a 20-year probability of occurrence under the present climatology, considering ongoing, significant climate change. The hills around Zhouqu have been well-known for their long history of landslides [http://news.sciencenet.cn/ htmlnews/2010/8/235921-1.shtm; Ma & Qi, 1997; p. 187 of Bolt, Horn, Macdonald & Scott, 1975]. However, this event is unique in its unprecedented magnitude, involving ~ 2,05×10 6 m 3 of sliding material. Because the landslide produced significant loss of life and great economic cost, it has generated intense discussion about the possible cause of the slide: 1. Previous drought conditions caused surface crevasses; 2. Unprecedented intense precipitation; 3. Wenchuan earthquake loosened slopes; 4. Historical earthquakes (century ago) and debris leftover; and 5. environmental consequences as population increases (living to previously un-occupied places). Climate warming often is seen as the major cause, by contributing to the severity . Rain events with rainfall totals > 30 mm can trigger significant landslides in the Zhouqu region (see inset histogram). Moreover, in that period, there are 7 events with rainfall totals > 60 mm, which therefore are rainstorms as intense as that which preceded the August 7, 2010 Zhouqu mudslides (the event indicated by the red arrow in the left inset). Thus, extreme precipitation alone does not explain the magnitude of the Zhouqu mudslide of the rainstorm; others argue that it was the recent drought, which produced cracks in the soil mantle. In addition to its geological uniqueness, because it occurred so soon after the 2008 Wenchuan earthquake, the presence of earthquake-broken bedrock also is cited as a factor contributing to the size of the landslide. A number of factors other than extreme precipitation therefore have been suggested as responsible for magnifying the Zhouqu landslide to its unexpected great size [Ma & Qi, 1997;Yu, Yang & Su, 2010].
To investigate quantitatively the relative importance of the possible causal factors, above-mentioned SEGMENT-Landslides model was applied to the event. SEGMENT-Landslide is a fully three-dimensional dynamical landslide model that incorporates not only soil/rock mechanical properties but also the hydrological and mechanical effects of vegetation on storm-triggered landslides. The model requires a wide variety of input variables, such as land cover, land use and geological data, which were provided by a research group of Beijing Normal University. The digital elevation data were from the Shuttle Radar Topography Mission, SRTM [http://srtm.mgs.gov/], at 90 m resolution. To reproduce historical landslides, we used precipitation forcing from the satellite-based National Aeronautic and Space Administration (NASA) Tropical Rainfall Measuring Mission (TRMM) which has 3-hourly data on a 0,25 by 0,25 degree resolution grid. For surface biomass loading, we used the Moderate Resolution Imaging Spectroradiometer (MODIS) products [Zhao & Running, 2010;Zhang & Kondragunta, 2006]. A team survey of the area also provided a 300 m resolution vegetation mask. To investigate possible mechanisms, we performed several sensitivity experiments with assumed vegetation conditions. Our selected region is 33,66-34,06°N; 104,26-104,66°E. The hilly terrain of this area is composed mainly of metamorphosed limestones, interspersed with altered clay layers. The ground surface rocks range from highly-to completelyweathered. The weathered rocks date from the Paleozoic (primarily the Permian period) and Mesozoic eras, the yellowish interbedded sandstone and siltstone date from the Silurian period, and the grey limestones dates from the Triassic period. The infiltration of rainfall through macro-pores, which are well-developed in the soil and rock mass of the Zhouqu region, plays a critical role in slope stability. The hills intersect with canyons in which increased erosion occurs during the highly regular rainy season.

METHODS
Landslides occur irregularly and future research is concerned with more accurate predictions about their timing (when), location (where) and size (how big they will be), and in developing procedures that convey risk and warnings to the public to mitigate damages to infrastructure and ecosystems. Such an effort is critical, particularly in anticipating the effects of climate change on areas prone to instability. Our study is a bellwether in this research direction as it uses a sound physically based predictive technique to assist understanding and informed decision-making. For mudslides, our model will help answer the following key questions. How and when will a particular landslide be initiated? How large will it be? How fast will it move? How far will it travel?
SEGMENT-landslides has been extensively documented by Ren et al. Ren et al., 2011]. Here we present the governing equations to provide context for the above discussion. For the sliding material, we solved a coupled system for conservation of mass under the granular rheological relationship, with viscosity parameterized as where ρ is bulk density, V is velocity vector (u, v and w are the three components), σ is internal stress tensor, and g is gravity acceleration. Here ν is viscosity, S is the spherical part of the stress tensor σ, μ 0 and μ 1 are the limiting values for the friction coefficient μ, e ε is the effective strain rate and 0.5 , I 0 is a constant depending on the local slope of the footing bed as well as the material properties, and I is inertial number defined as I = e ε d/(S/ρ s ) 0,5 , where d is particle diameter and ρ s is the particle density. Soil moisture enhancement factor on viscosity is assumed varying according a sigmoid curve formally as Eq. (9) of Sidle [1992] but with the time decay term replaced by relative saturation.
For considered granular material resting on vegetated slopes, the cohesion provided by the roots are implemented in the full internal stress σ's. The root mechanical properties are prescribed according to the vegetation types of the Zhouqu area. There are different ways to decompose the full stress into spherical and deviatoric components. Only the deviatoric part is assumed to proportional to the strain rate through viscosity, which unfortunately depends also on normal pressure.
As a derivative from Eq. (1), the prognostic equation for surface elevation (h(x, y)) is where X| top indicates evaluated at the free surface elevation. In the case with slope movements, Eq. (4) is solved regularly to update the sliding material geometry. It is also from this equation that we estimated the sliding material involved in the simulated landslides (e.g., Fig. 4).
The viscous term in Eq. (2) implies an energy conversion from kinetic energy to heat. To make a full closure of energy, we need the following thermal equation: where c is heat capacity (J/kg/K), T is temperature (K), κ is thermal conductivity (W/K/m), and σ e is effective stress (Pa). The last term is 'strain heating' , converting of work done by gravity into heat affecting the sliding material by changing viscosity or causing a phase change.
For quantitative predictions of storm triggered landslides, a numerical modeling system like SEGMENT is needed. However, some of the requirements of SEGMENT, especially the input and verification data, generally are not available even in modern geological maps. These parameters include vegetation loading and root distributions in soils and weathered rocks. The extension of the SEGMENT landslide model to other regions is limited primarily by the lack of these high resolution input datasets. The landslide features implemented in SEGMENT, if adopted by the relevant community, hopefully will encourage the collection of such vital information in future surveys. Figure 2a shows the SEGMENT-Landslide simulated unstable areas, as indicated by the maximum obtainable surface sliding speed. Under the current vegetation regime, the most significant scar is that near the Sanyanyu Valley (33,81-33,87N; 104,36-104,42E). The particular sliding is a characteristic "progressive bulking" [Iverson, 1997] type (Fig. 3). The accumulation area spreads up to 3500 m elevation, in a fan shape with the fan "handle" extending down to the Bailongjian River. The surface runoff essentially is clear water above the 3500 m elevation contour, but at lower elevations gradually becomes turbid and entrains small stones and coarse granular material into the slide streams. These creeks are usually dry except during rainy periods. Figure 4 is an enlargement of the Sanyanyu gully, showing the surface elevation changes at two times in the sliding process: the beginning (Fig. 4a) and the cessation (Fig. 4b). At the cessation, the areas indicated by the two red arrows have little elevation change, despite the massive total mass in the slide. They acted like a pathway for the sliding materials at higher elevations. For example, at point A there is a break in the slope where some of the sliding material originated. Over 70% of the sliding material came from the gully banks. Below 2300 m, the solid form of sliding material is continuous in nature and the entrainment effects are so significant that boulders (>50 cm in diameter) are relocated down the slope.

RESULTS
The thick mud has a viscosity of about 100 Pa•s and the peak sliding speed reaches as high as 2 m/s. A total of 2,05×10 6 m 3 of solid sliding material was involved in this slide and was spread over an area of about 3,2×10 6 m 2 . Figure 4 uses the actual vegetation coverage of the area. If we assume the entire region is bare of vegetation, SEGMENT-Landslide identifies the following "hot-spots" as unstable (Fig. 2b): (33,773N,  This self-propagating mechanism has no lower limit before leveling the slope to below the granular material repose angle. In the Zhouqu area, the shear zone depth is variable and depends on the quantity of water penetrating into the crevasses. For bare ground (e.g., covered by previous landslide deposits or rockfalls caused by historical earthquakes), runoff readily drains into the crevasses, moistens the granular material and forms a shear zone at the bottom (the lowest reaches of the crevasses). Vegetation cover reduces surface runoff through canopy interception. Roots also assist in the retention of water within the rhizosphere. Thus, with vegetation cover, runoff water cannot be effectively channeled into the crevasses and much less sliding material will be involved in the landslide (Fig. 5). The cohesion of the granular particles (loam soil, pebbles from fractured grey lime-stones, and sands) are of the order of 0,1 Kpa, far less than the root strength (~10 Kpa).
The presence of aboveground vegetation introduces the following effects: aboveground biomass loading (gravitational), growing season soil moisture extraction by live roots (hydrological), fortification of the soil within its extension range (mechanical), changing chemical environment of the soils through life processes (e.g., respiration, absorption of minerals selectively and secretion of organic substances) and therefore the bond strength among unit cells (chemical), and wind stress loading (meteorological). The overall effects are the interaction of the above factors and it is difficult to generalize before a detailed analysis is carried out that is specific to a certain situation. For example, the fortifying roots have yield strength larger than dry soils and the existence of roots is commonly thought to increase the resistance of soils. However, the presence of roots, especially when there is precipitation, also facilitates water channelling into deeper depths. After the soil is moistened, the cohesion between soil and the root surface is reduced greatly (to negligible strengths <0,001Mpa, [Lawrence et al., 1996]) and the root strength cannot be effectively exerted. Also, the effect of roots is to 'unify' the soil particles within root distribution range. Once the entire rhyzosphere soil layer is saturated, the fortifying effects will be totally lost.

Fig. 5. Diagram showing vegetation effects on storm-triggered landslides. Panel (a) is the case with vegetation and panel (b) is the case void of vegetation. All else being equal, with vegetation, a proportion of rainfall goes into canopy interception (canopy runoff) and runoff and surface ponding are reduced as a consequence. Vegetation also effectively prevents water infiltration into deeper depth. Thus less sliding mass is involved in the vegetated case (the blank arrows show the magnitude of bulk resistive and driving stresses). In the illustration, R means resistive stress and τ d means driving stress
Thus, a more accurate statement would be "the reinforcement effects from roots are an effective mitigating factor for shallow stormtriggered debris flows". In SEGMENT-Landslide, because roots occupy a small fraction of the soil volume, the root reinforcement can be factored in as an added stress over the case of no root presence. As not only tensile strength but also the cross-sectional areas (thickness of roots) are critical, we propose the following 'allometric' approach that uses 'root weight density' and vegetation type to characterize the added tensile strength to the soil medium shear strength and elevated yielding criterion.
where σ + is the root mechanical reinforcement (kPa), C smc is soil moisture control (0-1), index i differentiates woody transport roots, woody supporting roots and ephemeral absorbing roots, σ i (kPa) is a root species-dependent reference value (i.e., tensile strength of xylem of roots, varies from 10 to 30 MPa for most plants), NPP is net primary production (kg/m 2 /yr), P a is annual precipitation, T air is annual mean air temperature, and I veg is species-dependent reference NPP value. F varies from 0 to 1 represents the weight fraction of roots in a unit volume of soil (within range of influence of the roots). Value of F usually is close to 0,2%. The functional form of C smc can be tabulated according to soil type. The modifications from climate conditions are necessary because same plants have very different strategy in allocating biomass when living in different clime zone.
Thus the mechanical effects of the roots also contribute to slope stability. We performed an additional set of sensitivity experiments to further investigate the importance of vegetation in reducing the magnitude of landslides. If the Sanyanyu Basin had been covered 70% by shrub of negligible biomass loading, with root strength of 0,1 Mpa, and coarse root (diameter >1 mm) density of 2 m -2 all residing within the top 2 m of soil, the amount of sliding material would be only 1,1×10 6 m 3 , or about half the actual volume involved. If there is a closed cover (that is, 1,0 vegetation fraction), the sliding material can be further reduced to 10 4 m 3 , and primarily involves only pebbles and protruding boulders at lower elevations. These experiments underline the critical role of vegetation in reducing the magnitude of the "progressive bulking" types of storm-triggered landslides.
Importantly, loss of vegetation has occurred not in recent 10 years and there actually  46° 29'37'' N, 11° 39'25'' E)  are clear signs that local vegetation cover has been increasing (Fig. 6, also in [Zhao & Running, 2010]). Because the 2008 Wenchuan earthquake has deepened the crevasses within the soil mantle and the bedrock, the criteria for storm-triggered landslides are significantly lowered. Large landslides did not occur before 2008 because, previous storms, although can be equally intense and have even larger total amount (e.g., Sept. 4, 2001), could not infiltrate into deeper shear zone, without the help of the earthquake's tearing of the bedrocks. Large landslides did not occur during the past two years because the threshold precipitation intensity and total was not reached (the past two years are relatively dry as indicated by total annual precipitations: 500 for 2008 and 480 mm for 2009 respectively, see Fig. 6). The landcover in August 2010 therefore was unable to prevent landslides caused by an intense rainstorm, owing to the legacy of the 2008 Wenchuan earthquake. The sealing of the cracks caused and/or deepened, by the Wenchuan earthquake is slow process occurring on a timescale of several decades. Thus, a program of rapid restoration of the vegetation cover over the Zhouqu area is urgently required for re-building that region. The climate of that region, with an annual precipitation over the last 40 years is only 435 mm/yr, suggests the priorities are the restoration of forest on the north facing slope and of a seamless grass cover for the south facing slopes.
The Sanyanyu deep valley has much coarse granular sliding material, particularly stones and boulders, because of a self-accumulation mechanism originating from its specific topographical features and because its loamy soil mantle is more easily dissected by running water. Topographically, the creeks in the valley have 'graded river beds' because the upper parts (near peaks) are steeper than the lower parts (close to the toes). Thus the upper river bed slopes are larger than the lower river bed slopes. For lighter precipitation events, the stones and pebbles cannot roll directly to the toe, stopping at mid-slope and creating natural barriers to the sliding material that follows (see the red blobs in Fig. 4). These accumulations apply to small slides, typically caused by low to moderate precipitation events. The massive Zhouqu landslide of August 2010 was caused by an extreme precipitation event, but was magnified by the Wenchuan earthquake of May 2008 which greatly deepened the pre-existing cracks (either from historical earthquake or more gradual erosion processes) in the ground surface. For such surfaces, intense precipitation events favor the channeling of runoff water to greater depths than usual, creating sliding surfaces at those depths. Thus, more sliding material was involved than for a less intense rainstorm. Vegetation is very effective at holding drainage water in the rhizosphere and reducing drainage into deeper levels, but the severe vegetation loss in the Zhouqu region prevented the vegetation cover from playing a protective role in reducing the critical impact of the hydrological process of deep level drainage.
The modeling results underscore the urgency for a high priority program of re-vegetation of Zhouqu county, without which the region will remain exposed to future disastrous, 'progressive bulking' type landslides. A direct cause of the large magnitude of the 2010 debris flow is the loss of historical deposits and the undercutting of loose gully bed. Re-vegetation of the areas with historical deposits is a priority. Thus, engineering approaches, such as installing check dams, slope protectors, and leveling gullies, should be followed by re-vegetation, because, restoring the current vegetation cover to its natural, much denser state is the most effective long-term approach to landslide mitigation.

CONCLUSION
The Earth's climate currently is in an interglacial period that possibly will continue for another 50 kyr [ In the following two decades he was responsible for the research development and operational implementation of the Short Range Prediction Group. In 1994 he took a chair professorial position at The University of New South Wales, Sydney. The following year he founded and became Director of the Center for Environmental Modeling and Prediction at The University of New South Wales. He has received a number of awards for research and teaching, the most significant of which is the Max Planck Research Prize, awarded in Bonn, Germany (1994). He has over three hundred publications in his long and successful career.