Agricultural irrigation has increased the groundwater level in the Heitai terrace (part of the Heifangtai terrace) by 20 m over nearly five decades, which causes 3–5 landslides each year at the edge of the terrace. The Heitai terrace is of great interest in the study of loess-related landslides; but there is no unanimous agreement on the types of either the landslides in this study site or the loess-related slope failures in general. On the basis of aerial images (res. 5 cm), Digital Elevation Model (res. 10 cm), and field investigations, we analyzed the distribution and failure mode of the landslides in Heitai. The geological structure and characteristics of 69 landslides (vol. 5 × 103 –6 × 106 m3 ) are studied. The preliminary results of groundwater recharge in the terrace and formation of the apparent spring lines on the slope surface are analyzed to better understand the failure modes. We divided the landslides in Heitai into two groups based on the location of the failure surface, i.e. loess landslide and loessbedrock landslide, of which the development is governed by the angle between the principle direction of slope deformation and the dip of bedrock bedding. We further analyzed the failure mode of each type observed in Heitai, defined as follows: loess-bedrock planar slide, loess-bedrock irregular slide, loess flowslide, loess slide, and loess flow. The proposed types of loess-related landslide are to be incorporated in the Varnes classification (Varnes, 1978) in consideration of the engineering properties of loess, and to provide backward compatibility for Heitai and potentially other regions in the Loess Plateau of China.

Introduction

The Heitai terrace is formed by the fourth river terrace of the Yellow River in Gansu Province, China (Fig. 1a). More than 50 gullies are formed at the edge of the terrace. The terrace has a surface area of 9 km2 with the maximum length of 5.2 km (W–E) and a width of 2.5 km (NeS). It is part of the Heifangtai terrace with high concentration of loess-related landslides due to surface irrigation started in 1968 (Derbyshire, 2001; Peng et al., 2016; Zeng et al., 2016; Zhou et al., 2014). Loess slope failure is common upon wetting due to its high porosity and collapsibility (Dijkstra et al., 1994; Jiang et al., 2014; Liu et al., 2016; Smalley et al., 2001; Wu et al., 2017; Yuan and Wang, 2009). Irrigation provides 4–6 cycles of groundwater recharge annually, which has increased the groundwater level by 20 m at an average rate of 0.18 m yr−1 (Xu et al., 2014) and caused about 200 landslides along the edge of the terrace (Fig. 1b). Study on the loessrelated landslides in Heitai sustained its popularity due to frequent failures and significant societal impacts. Landslide classification system forms a long-lasting research topic (Alimohammadlou et al., 2013; Casagrande, 1940; Cruden and Varnes, 1996; Görüm et al., 2011; Hungr et al., 2011; Hungr et al., 2014; Hutchinson, 1988; Qi et al., 2010; Varnes, 1978; Wang and Zhou, 1999; Xu et al., 2016). In many cases, defining the landslide type serves as a ‘baseline’ of the problem because it reflects the landslide material and behavior (Hungr et al., 2014). Landslides in loess provide a unique collection of landforms and failure mechanisms that are often undervalued in the global classification systems. Loess is recognized as a unique material, created by aeolian deposition in the Loess Plateau of China, comprising predominantly silt-sized particles that form cemented meta-stable structures with characteristically high void ratios. Landslides in loess are often generalized as a single type of landslide based on the material composition, such as loess flow by Varnes (1978), and loess flowslide by Hungr et al. (2014), under the category of earth flow. In the context of landslide behavior, loess-related slope failure was divided into different types based on either the movement and failure mechanism or the location of the failure surface.

Our understanding of the spatial patterns of landslides in Africa is limited with available landslide studies typically focusing on only one or a few study areas. Moreover, Africa is clearly underrepresented in terms of available landslide inventories. This study aims to produce a first continent-wide landslide susceptibility map for Africa, calibrated with a well-distributed landslide dataset. We reviewed the literature on landslides in Africa and compiled all available landslide inventories (ca. 10,800 landslides), supplemented by additional landslide mapping using Google Earth imagery in underrepresented regions (ca. 7,250 landslides). This resulted in a dataset of approximately 18,050 landslides. Various environmental variables were investigated for their significance in explaining the observed spatial patterns of landslides. To account for potential mapping biases in the dataset, we used Monte Carlo simulations that selected different subsets of mapped landslides to test the significance of the considered environmental variables. Based on these analyses, we constructed two landslide susceptibility maps for Africa: one for all landslide types and one excluding the known rockfalls. In both maps, topography is by far the most significant variable. We evaluated the performance of the fitted multiple logistic regression models using independent subsets of landslides, selected from the total dataset. Overall, both maps perform very well in predicting intra-continental patterns of landslides in Africa and explain about 80% of the observed variance in landslide occurrence. To further test the robustness and sensitivity to mapping biases, we also modelled landslide susceptibility while excluding regions with arid climates, as landslides in these environments are expected to be better preserved over time and therefore likely relatively overrepresented. Despite this potential bias, the effect on the landslide susceptibility model is limited. Based on the constructed database and our analyses we further discuss potential research gaps for landslide prediction in Africa and at continental scales. For example, analysis of the African countries’ mean landslide susceptibility shows a lack of landslide research in various countries prone to landsliding (e.g.: Guinea, Gabon, Lesotho, Madagascar). Apart from the intrinsic value of this landslide susceptibility map as a natural hazard risk management tool, the map and compiled database are highly promising for other applications. For example, we explored the potential significance of landslides as a geomorphic process by confronting our landslide susceptibility map with an available database of measured catchment sediment yield for 500 rivers in Africa. Overall, a significant positive, but relatively weak relation between landslide susceptibility and sediment yield is observed.

Introduction

Studies on landslide risks and fatalities indicate that landslides are present on all continents and are a global threat to humans, infrastructure and the environment (Dilley, 2005; Guzzetti et al., 2012; Haque et al., 2016; Kjekstad & Highland, 2009; Petley, 2012; Sassa & Canuti, 2009; Stanley & Kirschbaum, 2017). While this is certainly also the case for Africa, this continent remains strongly underrepresented in landslide research (e.g. Gariano & Guzzetti, 2016; Kirschbaum et al., 2015, 2010; Maes et al., 2017; Nadim et al., 2006; Petley, 2012; Reichenbach et al., 2018). Also global landslide susceptibility (LSS) maps rely on very few (or no) data of observed landslides in Africa for their calibration (e.g. Hong et al., 2007; Kirschbaum et al., 2009; Nadim et al. 2006; Stanley & Kirschbaum, 2017). Nonetheless, landslides are one of the deadliest natural disasters in Africa (Guha-Sapir et al., 2017). Moreover, their importance and impact are expected to increase due to climate change, with an increase in total precipitation and increasing frequency and intensity of rainstorm events (Gariano & Guzzetti, 2016). This was for instance shown for the Great Lakes region (Shongwe et al., 2011; Souverijns et al., 2016; Thiery et al., 2016), probably one of the most landslide susceptible regions in Africa (Hong et al., 2007; Stanley & Kirschbaum 2017). In addition, Africa is facing a considerable growth in population, which is projected to triple by the end of the 21st century (Gerland et al., 2014). This will likely affect both the frequency and impact of landslide events. For these reasons, earlier studies have urged to fill the gap on landslide research in Africa (e.g. Gariano & Guzzetti, 2016; Jacobs et al., 2016; Maes et al., 2017).

The role of forests in regulating landslide risks is well established but estimates of the economic value of this ecosystem service are limited. In order to incorporate the role of forests for landslide risk mitigation in spatial planning and other decision-making contexts, there is a need for spatially explicit information regarding the value of this service. We develop a methodological framework to combine bio-physical modelling of natural hazard risk and socio-economic exposure in a predictive model to estimate and map of the economic value of forest regulation of landslides. This method is applied in a case study of Adjara Autonomous Republic of Georgia to examine alternative scenarios for forest management and associated land cover change. The approach produces credible spatially explicit results to inform policy decisions regarding investment in forest management; and has the potential for replication in other data scarce regions.

Introduction

It is expected that damages from landslides will increase steadily over the coming years (Dai et al., 2002). Underlying this trend is an increase in human activity in landslide prone areas, which increases risk through two channels: development generally involves deforestation, which increases the probability of landslides occurring; and more human activity generally means more assets that are exposed to damage (Nadim et al., 2006). An additional factor that will drive increasing landslide damage is that climate change is predicted to cause increased precipitation in many areas already prone to landslides and may result in additional areas facing the risk of slope erosion and landslides (Dale et al., 2001; Ciabatta et al., 2016; Crozier 2010). The role of forests in regulating landslide risks is well established (Endo and Tsuruta, 1969; Megahan et al., 1978; Wu and Swanston, 1980; Preston and Crozier, 1999; Jakob, 2000). The economic value of this regulating service has, however, received limited attention in the economic valuation and ecosystem services literature (Chiabai et al., 2011; de Groot et al., 2012; Häyhä et al., 2015). Existing research has tended to focus on the value of forests for timber (Phan et al., 2014; Pohjanmies et al., 2017), non-timber forest products (Schaafsma et al., 2014; Mutoko et al., 2015), water supply (Ojea et al., 2012; Wang et al., 2017), recreation (Zandersen and Tol, 2009) and carbon storage (Triviño et al., 2015). The relatively limited number of studies that do estimate the economic value of landslide regulation by forests tend to be for small-scale study sites (e.g. Olschewski et al., 2012; Dominati et al., 2014). Information on the economic value of forests in regulating landslides is useful for informing forest management decisions (Langner et al., 2017). Quantification of the damage costs of deforestation (or avoided damage costs resulting from reforestation) provides input for the appraisal of investments in conservation and restoration. Mapping ecosystem service values delivers additional information to support decision making, particularly for land use policy development, spatial planning and resource allocation (Schägner et al., 2013; Nahuelhual et al., 2015). In order to incorporate the role of forests for landslide risk mitigation in spatial planning and other decision-making contexts, spatially explicit information related to the value of forest landslide regulation needs to be developed. Many regional governments, especially in the developing world, do not have the resources to create and gather such data in usable forms, in the absence of which, policy attention is not paid to investments in forests as a landslide mitigation measure. While reliable techniques for assessing landslide hazard often require detailed geotechnical information on existing conditions, the high cost of which means they are available only where high risk is already anticipated, more exploratory techniques of hazard and risk can provide frameworks for development planning and environmental protection measures (Gaprindashvili and Van Westen, 2016).

Rain-induced shallow landslide is the major type of landslide that happens on natural terrain in Hong Kong due to its high seasonal rainfall and deep weathering soil profile. The Geotechnical Engineering Office has been in a leading role to steer the risk management of natural terrain landslides in Hong Kong. Recently, the territory-wide rainfall-based landslide susceptibility model (in terms of landslide density per year) has been developed to predict the number of natural terrain landslides that may occur in an anticipated rainfall event. Subsequently, the storm-based landslide density has been transformed to the annual landslide frequency to compile the territory-wide landslide frequency map by incorporating the mean annual frequency of occurrence of different probable rainfall scenarios. The annual rainfall frequency of a particular rainfall scenario is derived from its return period, determined based on the abundant real-time rainfall data at a five-minute interval from 110 automatic raingauges across Hong Kong at an average density of 10 km2 /gauge. The transformation is discussed in this paper, together with the evaluation of the performance of the landslide frequency map. In addition, the potential applications of the map in Hong Kong and the pitfalls of the common evaluation methods are highlighted.

Introduction

“Natural terrain” is commonly used in Hong Kong to denote hillsides that are not significantly modified by human activities. On average, about 300 natural terrain landslides occur every year in Hong Kong and rain-induced shallow landslide is the major type of landslide due to the high seasonal rainfall and deep weathering soil profile. Heavy rainstorms in Hong Kong generally occur during the wet season from April to September and are associated with either low pressure trough or severe tropical cyclones. As a general reference, the mean annual rainfall at the Hong Kong Observatory (HKO)’s Principal Raingauge was 2398.5 mm between 1981 and 2010. Landslides from natural terrain can result in serious consequence to life and property in Hong Kong due to its dense urban development on hillsides. The problem is chronic for the impacts of extreme weather events, slope degradation and ever-increasing population and encroachment of new urban development on steep hillsides. Over the past 35 years, the Geotechnical Engineering Office (GEO) has been in a leading role to steer the risk management of natural terrain landslides in Hong Kong to reduce and contain landslide risk within an as low as reasonably practicable level that meets the needs of the public and facilitates safe and sustainable developments. The slope engineering practice and landslide risk management have evolved in response to experience and through continuous improvement initiatives and technology advances (Wong & Ko 2006).

In 1983, a catastrophic landslide occurred in Saleshan, Dongxiang, Gansu, China, which resulted in the death of 237 people and the destruction of 585 houses. Saleshan landslide was a typical rapid moving landslide in loess area, China, which moved with extremely high speed and long run-out distance on a gentle sliding surface. This paper reviews the characteristics of the landslide and its consequence, analyzes the possible triggers and the processes of development, and back-calculates the kinematic characteristics of the landslide. The digital elevation models (DEMs) of the study area before and after the event were used to analyze the geomorphological characteristics before and after the event. The authors inferred the possible deformation and failure processes of the landslide based on available observations. Direct shear tests were carried out to obtain the parameters of the soil involved in the failure and postfailure process, which was adopted in the following simulation using an energy-based runout model. The velocities of moving materials at a different time stage together with the variation of runout distance and maximum flow height are presented. The authors also compared the calculated results with field observations and previous simulation results. Sensitivity analysis was conducted to understand the effect of internal and basal friction angles on the kinematic characteristics. This study indicates that the Saleshan landslide was a fast-moving landslide with long traveling distance. The landslide lasted more than one minute with a maximum velocity of approximately 25 m/s according to the calculation using the energy runout model, which was consistent with the estimations based on the observation of eyewitness. This study also gives an insight on the progressive failure mechanism and explains the high mobility of Saleshan landslide in details, which provides a reference for hazard zonation for areas along loess platforms.

Introduction Landslides are an important process which posed significant hazards, resulting in fatalities, property damage and degrading farmland (Huang, 2009). From 2004 to 2010, 2620 fatal landslides were recorded, which resulted in a total of 32,222 recorded fatalities (Petley, 2012). In China, most of the landslides occurred in high-altitude regions such as Qinghai-Tibet plateau and southwest area in China (Huang, 2007). According to the statistical data provided by the Ministry of Land and Resources of the People’s Republic of China, landslide geohazards have caused 4,718 fatalities from 2004 to 2016 (Zhang and Huang, 2018). Every year, one-third of the geohazards in China occur on the Loess Plateau (Zhou et al., 2002), with more than 80% involving landslides with slip surfaces shallower than 10 m (Zhuang et al., 2017). The main types of landslides involving substantial amounts of loess include flow, fall and slide (Li et al., 2013; Hungr et al., 2014; Zhang et al., 2012). According to the field investigation, it was found that most of the landslides in the loess area occurred on a slope around 20º – 35º and the flowslides had a long runout distance due to soil liquefaction resulting in a low frictional angle (Zhang et al., 2009; Zhang et al., 2013; Zhang et al., 2014). Since 1900, more than 20 catastrophic landslides have been recorded whose volumes are greater than 20  106 m 3 , several of which occurred in loess area of China (Derbyshire et al., 2000; Wen et al., 2004; Huang, 2007; Wang et al., 2014; Xu et al., 2014; Zhang et al., 2014; Zhang and Wang, 2018) (Table 1). The landslides triggered by Haiyuan Earthquake 1920, are typical examples of these cases which in total resulted in more than 100,000 deaths (Huang, 2007; Wang et al., 2014).

Researchers have long attempted to determine the amount of rainfall needed to trigger slope failures, yet relatively little progress has been reported on the effects of climate change on landslide initiation. Indeed, some relationships between landslides and climate change have been highlighted, but sign and magnitude of this correlation remain uncertain and influenced by the spatial and temporal horizon considered. This work makes use of statistically adjusted high-resolution regional climate model simulations, to study the expected changes of landslides frequency in the eastern Esino river basin (Central Italy). Simulated rainfall was used in comparison with rainfall thresholds for landslide occurrence derived by two observation-based statistical models (1) the cumulative event rainfall–rainfall duration model, and (2) the Bayesian probabilistic model. Results show an overall increase in projected landslide occurrence over the twenty-first century. This is especially confirmed in the high-emission scenario representative concentration pathway 8.5, where according to the first model, the events above rainfall thresholds frequency shift from * 0.025 to * 0.05 in the mountainous sector of the study area. Moreover, Bayesian analysis revealed the possible occurrence of landslide-triggering rainfall with a magnitude never occurred over the historical period. Landslides frequency change signal presents also considerable seasonal patterns, with summer displaying the steepest positive trend coupled to the highest inter-model spread. The methodological chain here proposed aims at representing a flexible tool for future landslide-hazard assessment, applicable over different areas and time horizons (e.g., short-term climate projections or seasonal forecasts).

Introduction

Over the next decades, societies will face massive environmental changes potentially able to substantially alter life styles and priorities (Lenton et al. 2008). Climate change will possibly be the major driver of such changes. However, if on the one hand the scientific community unequivocally recognized over these past decades the connection between the increase in greenhouse gases (GHGs) concentration and the increase in global temperature (Bindoff et al. 2013), on the other hand, changes in the rainfall patterns as a second-order effect of the increased temperature is connoted by higher uncertainty. Such reservations concern the magnitude and frequency of extreme events, how they will scale across the globe in response to the increase in temperature (Seneviratne et al. 2012; Scoccimarro et al. 2013; Drobinski et al. 2016). Notwithstanding these uncertainties, a broad consensus prevails on the acceleration of the hydrological cycle in a warmer atmosphere (Trenberth 1999; Mariotti et al. 2002; Lorenz and DeWeaver 2007; Volosciuk et al. 2016); acceleration appears strongly affected by the local-scale morphological and orographic peculiarities. (Xoplaki et al. 2004; Walsh et al. 2014). Rainfall is a major trigger of landslides (e.g., De Vita and Reichenbach 1998; Guzzetti et al. 2007), one of the most widespread geohydrological hazards over the world, and an increase in rainfall frequency and intensity may directly affect landslides frequency and magnitude thus increasing ensuing damages and fatalities. Indeed, good progresses have been made in linking the amount of rainfall needed to trigger slope failures (e.g., Crozier 1996; Aleotti 2004; Guzzetti et al. 2007; Martinotti et al. 2017), yet the nexus between climate change and landslide initiation appears more complicated and requires further study (Dikau and Schrott 1999; McInnes et al. 2007; Crozier 2010; Coe and Godt 2012). The available literature indicates a relationship between landslides and climate change, but the sign and strength of this correlation remain uncertain and extremely influenced by the spatial and temporal horizon considered (Seneviratne et al. 2012; Gariano and Guzzetti 2016). Landslides, strictly depending on geological, geomorphological, and land cover contexts, represent a category of climate change-related impacts (Pisano et al. 2017), which understanding and management require specific considerations. Moreover, the implementation of effective landslide adaptation measures depends on the availability of plausible multi-scale information about future climate trends. Regional climate model (RCM) simulations are considered primary tools for climate impact studies focusing on rainfall and other key climate variables (Giorgi 1990; Giorgi and Mearns 1999; Giorgi et al. 2009; Hawkins and Sutton 2009; Jacob et al. 2014).

Te Himalaya is a 2500  km long belt of mountains, which are the result of the progressive under thrusting of the Indian Plate beneath Tibetan Plate along Main Himalayan Trust (MHT) (Zhao et  al. 1993; Molnar and Tapponnier 1977). Global Positioning System (GPS) measurements indicate 4–5  cm/year of convergence rate between these two plates (Banerjee and Bürgmann 2002) of which 18 mm/year is accommodated by thrust systems along the Himalayan arc. Tectonic deformation has resulted in the formation of four major south-verging thrust systems viz. South Tibetan Detachment (STD), which separates Tibetan sedimentary and Higher Himalayan meta-sedimentary sequences, Main Central Trust (MCT), which bounds the Higher Himalayan Crystalline rocks, and Upper Precambrian to Paleozoic rocks of Lesser Himalaya. MCT has not been observed to rupture the Quaternary deposits and is considered to be inactive (Nataka 1989; Valdiya 1992). Main Boundary Trust (MBT) marks the southern edge of Lesser Himalaya. Te southernmost thrust is Himalayan Frontal Trust (HFT), which has displaced Tertiary and Quaternary alluvial and molasse sediments and is expressed as discontinuous range front scarps (Nakata 1972; Valdiya 1980, 1992). Tese fault systems are the result of the southward propagation of the Himalayan Front through time and each fault merges into the basal detachment fault, i.e., Main Himalayan Trust (Zhao et al. 1993; Pandey et al. 1995). Based on the active fault topography, the MBT and HFT are considered as active (Nakata 1989; Valdiya 1980; Malik and Mohanty 2007). Historical archives indicate that Himalaya has experienced several damaging earthquakes of intensity >XI in the last millennium (Iyengar et al. 1999; Chitrakar and Pandey 1986). Instrumental records and geodetic data indicate that several earthquakes have occurred along the basal detachment and propagated southward to produce surface rupture at Himalayan Front such as 1934 Nepal–Bihar (Sapkota et al. 2013) (Fig. 1c), but few nucleate below the Higher Himalaya and produced ruptures in the higher reaches only, 1905 Kangra Earthquake is one such example which produced surface rupture along Kangra Valley Fault (KVF) (Fig. 2a) (Malik et al. 2015). Te variable nature of the rupture pattern raises several questions: (a) does all stored strain developed due to India–Tibet convergence is released during a single event? (b) Is the strain accumulation similar along the entire Himalaya, or (c) how does the variation in the Himalayan strike and segmentation efects the variable rupture pattern?

Abstract

This paper describes monitoring results of ground movements due to slurry shield tunnelling on the Docklands Light Railway Lewisham Extension (hereinafter referred to as DLR). The DLR tunnels (twin bored tunnels) run from Island Gardens Station to Greenwich Station in East London, passing under the River Thames, a distance of around 4.2 km. The construction of the tunnels, which were at shallow depth, needed an extensive appraisal of the potential damage induced by ground movements to the overlying and underground structures. Therefore, in order to effectively minimize the ground movements, a slurry shield machine with a diameter of 5.85 m was employed. Careful control of excavation was carried out during shield tunnel construction. As a result, the slurry shield tunnelling method in Woolwich and Reading Beds by well controlled monitoring was successfully conducted, leading to volume losses less than 1.0%. This paper focuses on the ground surface movements, particularly the transverse and longitudinal settlement profiles, and volume losses during shield tunnel construction, in relation to the ground conditions. Based on the monitoring results, combined with previous centrifuge model test results and other field monitoring data, practical methods to appropriately predict ground movements due to shield tunnelling are proposed.