Introduction

Ground shaking is considered the primary cause of damage, loss of life and injuries due to earthquakes. However, earthquakes are one of the most relevant triggering factors of ground failures (e.g. landslides, liquefaction and surface fault rupture). In this sense, slope movements are often the second major cause of the greatest damage, human and material losses when an earthquake occurs (cf. Bird and Bommer 2004). For instance, about 56 % of the total cost of the damage caused during the 1964 Alaska earthquake was due to earthquake-induced slope failures (Youd 1978). While ground shaking is mainly related to structural damage and collapse of buildings in urban areas, earthquake-triggered landslides are less likely to cause structural damage, but are frequently the cause of major disruptions of life-lines (e.g. roads, railways, power lines, gas pipes, water channels, etc.) and, therefore, are crucial in permitting a rapid response of the emergency services in the aftermath of a seismic event. Landslides can also produce dramatic changes in the landscape even far away from the earthquake epicentre. In particular, catastrophic floods can occur by rupture of dams formed by landslides blocking a river valley. A recent example is the M w 8.0, 2008 Wenchuan earthquake in China (Li et al. 2008). Landslides, rock falls and debris flows induced by this earthquake damaged or destroyed several mountain roads and railways and buried buildings in the Beichuan–Wenchuan area, cutting off access to the region and hampering the rescue operations for several days. Landslides also dammed several rivers, creating more than 32 barrier lakes that threatened about 700,000 people upstream and downstream (Xu et al. 2009; Yin et al. 2009).

The most common procedures followed in earthquake-triggered hazard assessment at a regional-scale deal with the well-known Newmark’s sliding rigid-block method (Newmark 1965) were implemented in a geographic information system (GIS) (e.g. Jibson et al. 2000; Luzi et al. 2000; Romeo 2000; Carro et al. 2003; among others). The Newmark method is a simple and effective way to assess the relative potential for slope failure. Through comparison between the location and magnitude of estimated Newmark displacements and the location of known slope failures, we can check how well they correlate and conceive a landslide susceptibility assessment for a specific earthquake scenario. In Spain, there are very few works devoted to this subject (García-Mayordomo 1999; Mulas et al. 2003; Delgado et al. 2006; Rodríguez-Peces et al. 2008), even though the seismic activity in some parts of the country is moderate to high. A review of the main results of these studies can be found in García-Mayordomo et al. (2009) and Alfaro et al. (2012).

This article provides a first-order assessment of regional earthquake-triggered landslide hazard in the western part of the Sierra Nevada Range (Fig. 1), which is the area with the highest historical and instrumental seismic activity of Spain. In this region, significant earthquakes have occurred triggering numerous slope failures (Delgado et al. 2011). The first step required to develop a proper earthquake-triggered landslide hazard assessment is a slope-failure inventory of the study area. This inventory allows the identification of the most common slope-instability types in the Sierra Nevada. An earthquake-triggered landslide hazard map in terms of Newmark displacements has been obtained simulating the occurrence of a maximum magnitude earthquake related to the Padul Fault and taking into account the site effects (soil and topographic amplification). Finally, the Newmark displacement map has been compared to the distribution of the inventoried slope instabilities to identify the areas where the seismicity could contribute to reactivate old slope failures or generate new ones, and the involved landslide typology.

Fig. 1
figure 1

Distribution of seismicity recorded between the years 1431 and 2011 in the Central Betic Cordillera (S Spain). The location of the Sierra Nevada Range and Granada Basin is also shown. Only intensity ≥VI and magnitude ≥2.0 seismic events are plotted. Sierra Nevada is composed of metamorphic Palaeozoic and Mesozoic rocks of the internal Betic zones, which are constituted by three superposed metamorphic complexes: Nevado-Filábride, Alpujárride and Maláguide complexes (Martínez-Martínez et al. 2002). The Sierra Nevada Range comprises the Nevado-Filábride and Alpujárride complexes. The Nevado-Filábride complex is formed mainly by micashists and quartzites while the Alpujárride complex is composed of metapelites and carbonate rocks. The Sierra Nevada emerged in the Late Miocene, progressively isolating different intramontane basins (e.g. Granada and Guadix basins). Thin black lines represent the main Quaternary active faults in the area (García-Mayordomo et al. 2012). Padul Fault is highlighted by a thick red line. The black square depicts the area of the forthcoming figures

Full size image

Seismic activity and earthquake-triggered landslides

Seismic activity in southern Spain is concentrated along the western and southern borders of the Sierra Nevada Range where the Neogene-Quaternary sediments of the Granada Basin contact the rocks of the range. Since the beginning of the instrumental record in the 1920s, many earthquakes have been recorded in the Granada Basin, although all of them of low to moderate magnitude (m b < 5.5) (De Miguel et al. 1989). The largest earthquakes that have occurred in historical time in the Granada Basin took place during the period between the XV and the XIX centuries (Vidal 1986; López Casado et al. 2001; Feriche and Botari 2002): 1431 Granada (I EMS = IX), 1526 Granada (I EMS = VIII), 1806 Pinos Puente (I EMS = IX) and 1884 Arenas del Rey (I EMS = X). Historical reports show that most of the landslide phenomena occurring in the Granada Basin are related to some of these major historical earthquakes (e.g. 1884 Arenas del Rey). Some well‐documented examples are the Güevéjar landslide (Jiménez Pintor and Azor 2006; Rodríguez-Peces et al. 2011a), rock falls and rock avalanches in Alhama de Granada and Albuñuelas villages and even some liquefaction phenomena (Muñoz and Udías 1981; IGME and Diputación de Granada 2007). However, in the Sierra Nevada Range there is lack of studies attributing specific landslides to individual earthquakes. The scarceness of descriptions can be attributed to a bias of the available information towards the evaluation of building damage, rather than to describing the effects occurring away from populated areas.

From a seismotectonic point of view, the Sierra Nevada Range is delimited at its western border by NW–SE faults (Fig. 1). Many of these faults show Quaternary activity and they are potential seismic sources of earthquakes with magnitudes larger than M w 6.0 (Sanz de Galdeano et al. 2003; García-Mayordomo et al. 2012). Among the NW–SE active normal faults, the Padul Fault (also known as the Padul-Nigüelas Fault) stands out because it is one of the most active and closest to the Sierra Nevada Range (Sanz de Galdeano and Peláez 2010). This fault is about 15 km long and comprises two connected segments: Padul and Padul-Dúrcal, which mainly dip towards the SW. The seismic activity during the instrumental period in the Padul area is characterized by the occurrence of small-magnitude earthquakes. However, there is paleoseismic evidence of moderate- to large-magnitude earthquakes during the Quaternary (Alfaro et al. 2001). Hence, the occurrence of moderate- to high-magnitude earthquakes associated with this fault is likely in the future. Considering the Padul Fault as the most likely seismogenic source, the maximum magnitude of an earthquake related to a complete rupture of the fault is M w 6.6 (Sanz de Galdeano et al. 2003). This magnitude was estimated considering a total length of 15.2 km and an average slip rate of 0.35 mm/year over the last million years. The estimated recurrence interval of this event ranges from 7,000 to 10,000 years. Coincidentally, earthquake epicentres of very deep focus (h ~ 640 km) are located nearby the Padul Fault, with some of these having magnitudes larger than M w 6.0 (Buforn et al. 1991, 2004). However, the origin of these deep earthquakes is not related to the Padul Fault activity (Fig. 2).

Fig. 2
figure 2

Geological map of the study area, showing the lithological groups outcropping in the Sierra Nevada Range and Granada Basin sector. Padul Fault is depicted by a thick red line

Full size image

Landslides distribution

To measure the distribution and typology of slope instabilities over the western Sierra Nevada, we constructed a GIS database based on previous studies (IGME and Diputación de Granada 2007) and the interpretation of aerial photographs and field surveys. Landslides were classified according to the definitions of Varnes (1978), Keefer (1984) and Cruden and Varnes (1996). A total of 2,444 slope instabilities were mapped in the study area covering 1,959 km2. The slope failures affect 7.7 % of the total study area with an average instability density of 1.25 instability/km2. The inventory comprises a 35.5 % of debris flows, 28.8 % of falls (rock falls and rock slides), 27 % of landslides (translational and rotational slides) and 8.9 % of earth flows. Hence, the most common slope-instability types in the Sierra Nevada Range are debris flows, followed by rock falls and rock slides (Table 1). These shallow slope instabilities (<3 m depth) are the most common types of earthquake-triggered slope failures in the Betic Cordillera (Delgado et al. 2011), as well as worldwide (Keefer 1984, 2002), and can be triggered by earthquakes as small as M ~ 4.0. Considering the numerous seismic sources of the Central Betic Cordillera (Sanz de Galdeano et al. 2003), there is no doubt that many of the instabilities in the Sierra Nevada Range owe their existence to earthquakes. However, the presence of water is a major factor controlling the triggering of landslides. This condition is very common in the south of Spain due to the typical Mediterranean heavy rainfall regime. All the inventoried slope movements present non-permanent activity but some of them could have seasonal reactivation due to heavy rainfalls.

Table 1 Distribution of mapped slope failures types (IGME and Diputación de Granada 2007) related to total number of instability processes and affected area related to total study area
Full size table

We consider three primary determining factors characterizing slope stability: lithology, slope angle and slope aspect. We make the assumption that lithology directly dictates the differences in material strength for these slope-stability analyses. A lithological map was arranged using the 1:50,000 scale digital geological maps from the Institute of Geology and Mines of Spain (IGME). Debris flows are developed mainly in micashists, rock falls in micashists and marbles, landslides in micashists, and earth flows in conglomerates, sandstones and argillites (Table 2). The micashists and quartzites units have the most unstable lithology in the Sierra Nevada Range, comprising 64 % of the inventoried slope instabilities. This fact is significant because this lithological unit is the most frequent in the Sierra Nevada Range. Slope angle has a great influence on the susceptibility of a slope to landsliding. For this reason, it is the most commonly determined factor used in slope-stability assessment by GIS. Following Wasowski et al. (2002), a 20°–25° slope range seems to provide a separation between soil and rock failures, the latter generally occurring on slopes steeper than 35°. With regard to failures on rocky slopes, Keefer (1984) reports a threshold of 35° for disrupted slides (e.g. rock falls and rock slides). We have used a 10 × 10 m pixel size digital elevation model (DEM) of the study area to derive a slope map. In general, slopes angles for the areas affected by slope instabilities in the Sierra Nevada range between 15° and 25° (Table 3). However, most common slope angles related to debris flows and falls are slightly greater (25°–35°). These ranges of slope angle comprise about 53 % of the study area. The slope aspect can also influence landslide initiation. This factor is related to soil moisture and weathering, which are commonly greater on slopes oriented to the north, because of the lower insolation. However, inventoried slope instabilities do not show a preferred orientation (Table 4), so the slope aspect seems not to be a determinant factor regarding slope stability.

Table 2 Distribution of slope-instability typologies (percentages, %) related to each lithological group outcropping in the Sierra Nevada Range
Full size table
Table 3 Slope angle distribution (%) for each slope-instability typology
Full size table
Table 4 Slope aspect distribution (%) for each slope-instability typology
Full size table

Hazard assessment of earthquake-triggered slope instabilities

Calculation of safety factor and critical acceleration maps

The Newmark method simplifies the slope instability as a rigid-block sliding on a planar surface, where static and dynamic strength parameters are assumed equal and stationary. The Newmark method involves the following simplifying assumptions (Newmark 1965; Jibson 2011): (a) the potential failure mass is rigid plastic; (b) the dynamic response of the failure mass is not influenced by the permanent displacement along the failure surface; (c) permanent displacement accumulates in only the downslope direction; (d) the vertical component of the ground motion does not influence the calculated permanent displacement; and (e) the critical acceleration of the potential failure mass is constant. Under these crucial assumptions, the calculation of the displacement of the rigid block is a two-step procedure. First, the minimum seismic acceleration to overcome shear resistance and initiate the displacement of the slope is calculated by Newmark (1965) equation:

$$a_{\text{c}} = \, \left( {{\text{SF }} - \, 1} \right) \, g{ \sin }\alpha$$
(1)

where a c is the critical acceleration (in gravity units, 1g = 9.81 m/s2), g is the gravity acceleration, SF is the static safety factor and α is the slope angle. Hence, the critical acceleration is an expression of slope capacity to resist seismic acceleration and therefore, it can be regarded as an effective measure of the susceptibility to a slope to earthquake failure. Secondly, slope displacement is calculated considering an acceleration time-history (accelerogram) representative of the expected seismic input at the site and double integrating in the time intervals where the critical acceleration overcomes. Cumulative displacement calculated this way, i.e. Newmark displacement (D N), provides a fair estimation of the actual displacement, as it has been shown both in laboratory tests and in field case studies (cf. Wilson and Keefer 1983). However, in a regional-scale approach Newmark displacements can be estimated by regression equations based on single strong ground motion parameters, such as the Arias intensity (I a) or peak ground acceleration (PGA) (cf. Jibson 2007).

Estimation of the static safety factor map is a first step required to produce the critical acceleration map. We calculated the safety factor map (Fig. 3a) assuming a simple infinite-slope limit equilibrium model following the Mohr–Coulomb criterion (Jibson et al. 2000):

$${\text{SF}} = \frac{{c{^\prime }}}{{\gamma {\kern 1pt} t{\kern 1pt} \sin \alpha }} + \frac{\tan \phi ^\prime }{\tan \alpha } - \frac{{m\gamma_{\text{w}} \tan \phi ^\prime }}{\gamma \tan \alpha }$$
(2)

where c′ is the effective cohesion, ϕ′ is the effective friction angle, α is the slope angle, γ is the unit weight of slope material, γ w is the unit weight of water, t is the normal thickness of the sliding plane and m is the degree of saturation of the failure surface. The simplified infinite-slope model is strictly only applicable to planar translational slides. One of the difficulties in the application of this model in real situations is that the failure surface is generally more complex than planar. However, the infinite-slope model has been also widely used in prior researches (Jibson et al. 2000; Luzi and Pergalani 2000; Luzi et al. 2000; Romeo 2000; Carro et al. 2003; Rodríguez-Peces et al. 2011c; among others) as a reasonable good approach to evaluate the stability of non-planar failure surfaces (e.g. rotational slides). The predictive capability using this model could be of 70–80 % of the actual seismically induced slope failures (Luzi and Pergalani 2000; Carro et al. 2003; Dreyfus et al. 2013). Moreover, it is the only practical model suitable for calculating slope stability on a pixel basis, and therefore is most suitable to be used at a regional scale in a GIS based on raster data. The rigid-block method has shown good results when applied to shallow instabilities in stiff material, which means fairly shallow slides, rock and debris falls (cf. Jibson 2011). These types of instabilities comprise the most common earthquake-triggered failures in natural slopes (Keefer 1984, 2002). However, larger and deeper slope instabilities, which are much less frequent, are not adequately modelled by this approach.

Fig. 3
figure 3

a Map of static safety factors. Red areas show lowest values, which are closer to the failure condition (SF <1.0). b Map of critical acceleration (g units, 1g = 9.81 m/s2). Red areas show lowest values, which are the most susceptible areas to earthquake triggering. See text for more details

Full size image

For this case study, we have considered dry conditions (m = 0) based on Eq. (2), as the climate of the Sierra Nevada Range is semi‐arid with alternating dry and wet periods on centennial, decadal and inter-annual scales (Rodrigo et al. 2000), being dry periods relatively more frequent and intense than wet periods (Jiménez Sánchez et al. 2008). In addition, the water table is deep (>20 m; IGME and Diputación de Granada 2007). For the sake of simplicity, we have assumed pore water pressure to be zero and ignored suction (negative pore water pressure). Hence, only total strength parameters have been used in the stability analysis. Considering these specific conditions, calculation of the static safety factor (SF) can be simplified as follows:

$${\text{SF}} = \frac{c}{{\gamma {\kern 1pt} t{\kern 1pt} \sin \alpha }} + \frac{\tan \phi }{\tan \alpha }$$
(3)

where c is the cohesion, ϕ is the friction angle, α is the slope angle, γ is the unit weight of slope material and t is the normal thickness of the sliding plane.

Average values of unit weight, cohesion and friction angle were assigned to each lithological unit based upon a geotechnical database (cf. Rodríguez-Peces 2010). The range of the shear strength parameters was very wide; particularly the cohesion values (Table 5). These shear strength parameters should be used as a first approach because actual values will likely vary from site to site, for a given rock type, and even across a single site. For this reason, a sensibility analysis of these parameters (unit weight, cohesion and friction angle) was performed. The aim of this analysis was to obtain the threshold values that provided the stability condition (SF >1.0) in nearly all the slopes in the study area. Following this approach, a few small areas remained statically unstable (SF <1.0) due to the presence of very steep slopes (70°–90°). These slopes are spatially limited, comprising only 0.002 % of the total area. Therefore, a safety factor of 1.01, close to stability condition, was assigned to these areas to avoid increasing strength parameters beyond realistic levels and also to preclude the occurrence of unrealistic negative critical acceleration values. Table 6 shows the adopted strength parameters values used in the forthcoming calculation of the safety factor. The depth of the failure surface has been set at 3 m based on field observations. The most common slope instabilities in the Sierra Nevada identified during field surveys are small rock slides and rock falls with block sizes of 1–6 m long. Our observations of shallow failure depths (~3 m) are in agreement with the typical value proposed by Keefer (1984, 2002). Furthermore, considering a deeper failure surface would increase the weight of the sliding block and the block would not meet the stability criterion (SF >1.0) for much of the range, which must be true unless there is an active slide.

Table 5 Typical values of shear strength parameters based on geotechnical bibliography and available laboratory tests (cf. Rodríguez-Peces 2010)
Full size table
Table 6 Lithological groups and strength parameters considered in the calculation of the safety factor map
Full size table

Finally, the safety factor and slope maps were combined using Eq. (1) to produce the critical acceleration map (Fig. 3b). We use this map to identify the most susceptible areas to seismic motion.

Input seismic scenario

We consider a deterministic seismic scenario based on the seismic potential of the Padul Fault. Strong ground motion related to the earthquake magnitude associated with a complete rupture of this fault was calculated using a selection of ground motion prediction equations (GMPEs) for the Mediterranean zone (Fig. 4). GMPEs have been selected from the literature to obtain an average PGA value as a function of magnitude and distance from the fault (Skarlatoudis et al. 2003; Ambraseys et al. 2005; Akkar and Bommer 2007; Bindi et al. 2010). Three main criteria were considered to select these GMPEs: (1) that they are derived from statistically significant data sets; (2) that they are widely used in European countries in a similar seismotectonic context (the European–African plate boundary) and (3) the magnitude scale is in terms of M w.

Fig. 4
figure 4

Map of peak ground acceleration (PGA) on rock for the deterministic seismic hazard scenario of a maximum magnitude earthquake (M w = 6.6) related to the rupture of the Padul Fault (PF)

Full size image

The output from this deterministic seismic hazard scenario is in terms of PGA on rock conditions. We considered soil amplification effects by assigning multiplying factors corresponding to each of the lithologic units (Fig. 5a). These factors have been adopted from previous work in the subject (Benito et al. 2010). Additionally, we developed an original GIS tool to estimate the topographic amplification effect based on terrain geometry variables and Eurocode-8 provisions (CEN 2003). The GIS tool first computes the slope and curvature maps and extracts the ridges from the digital elevation model. Subsequently, the relative height of the ridges is computed and then compared with the slope map. Finally, the topographic amplification factor (Fig. 5b) is assigned to each pixel according to the following possible cases: (a) slopes lower than 15° or ridges with a relative height of <30 m, amplification factor equal to 1.0 (no topographic amplification), (b) slopes between 15° and 30° and a relative height >30 m, amplification factor equal to 1.2 and (c) slopes steeper than 30° and a relative height >30 m, amplification factor equal to 1.4.

Fig. 5
figure 5

a Map of soil amplification factors. b Map of topographic amplification factors. Red areas show the highest values of seismic amplification. See text for further explanation

Full size image

Computation of Newmark displacement maps

The computation of Newmark displacements (D N) in regional hazard assessment is usually done making use of regression models based on basic earthquake parameters (magnitude and distance) and/or simple strong ground motion parameters. We have adopted the equation proposed by Jibson (2007):

$$\log D_{\text{N}} = 0.215 + \log \left[ {\left( {1 - \frac{{a_{\text{c}} }}{\text{PGA}}} \right)^{2.341} \left( {\frac{{a_{\text{c}} }}{\text{PGA}}} \right)^{ - 1.438} } \right]$$
(4)

where D N is the Newmark displacement (in centimetres), a c is the critical acceleration (in gravity units) and PGA is the peak ground acceleration (in gravity units). The R 2 and σ values are 84 % and 0.51, respectively. This equation was derived from a selected database of 875 records from earthquakes ranging from M w 5.3 to 7.6. Although the standard deviation is very high, this does not compromise the use of Newmark displacement values as an index of relative hazard at a regional scale (Jibson 2007). Newmark displacement maps were computed using Eq. (4) after PGA was corrected to account for soil and topographic amplification effects.

Newmark displacement values obtained in this work should not be considered as a precise measure of co-seismic slope displacement, but rather as an index of potential instability. The actual Newmark displacement that effectively triggers a landslide strongly depends on site-specific variables, particularly the way deformation is accommodated by the moving mass. In a regional-scale context, different authors have estimated that Newmark displacements >5–10 cm could potentially imply the occurrence of coherent-type landslides (landslides and earth flows) whereas smaller values could trigger disrupted-type landslides (rock falls, rock slides and debris flows) (cf. Wilson and Keefer 1983; Keefer 1984; Romeo 2000; Keefer 2002). In our analysis, we consider 5 cm as the minimum Newmark displacement required to induce coherent-type landslides based on the lower bound proposed by these authors. Studies in other mountainous areas of the Betic Cordillera have shown that smaller Newmark displacements (<2 cm) could potentially trigger disrupted-type landslides (Rodríguez-Peces et al. 2008, 2009; Rodríguez-Peces 2010; Rodríguez-Peces et al. 2011b, c).

Results and discussion

The scenario of a maximum magnitude earthquake along the Padul Fault results in a widespread distribution of Newmark displacements smaller than 2 cm, and locally larger than 5 cm, across the western Sierra Nevada Range (Fig. 6). These regions represent areas with significant hazard of seismically induced slope failure due to such scenario. Areas showing Newmark displacements comprise 7.8 % of the study area, which is coherent with the percentage of known slope instabilities (7.7 %). In general, these areas appear to be related to the strong incision of the rivers of the Sierra Nevada Range, which implies the slopes with both low safety factor and critical acceleration values. Furthermore, the steep slopes contribute to a significant topographic amplification of the seismic acceleration. Additionally, Newmark displacements appear concentrated near the first ~10 km around the fault trace, while Newmark displacements at longer distances from the fault are more scattered. This result is controlled by the high PGA values resulting from the GMPEs along the Padul Fault trace.

Fig. 6
figure 6

Newmark displacement map for the deterministic seismic scenario that considers the rupture of the Padul Fault (M w = 6.6). Displacement values should be interpreted as a measure of seismically induced landslide hazard

Full size image

As there are not yet case studies available in the working area, a rough validation of the results can be done by considering the distribution of the slope instabilities in the Sierra Nevada Range (Fig. 6). The Newmark displacement map (or seismically induced landslide hazard map) shows, in general, an acceptable correlation with the location of the slope instabilities (Fig. 6; Table 7). The percentage of the areas where the slope instabilities are located related to the area with seismically induced landslide hazard is 10.2 %. This result evidences seismic activity as a major control in reactivating former slope instabilities. An example of seismic reactivation, located close to Sierra Nevada Range, is the Güevéjar landslide that was triggered by both the 1755 Lisbon and 1884 Arenas del Rey historical earthquakes (Rodríguez-Peces et al. 2011a). The most frequent Newmark displacement value in the pre-existing slope-instability areas is 2 cm, or even less (Table 7). This low value agrees with the results obtained in other mountainous zones of the Betic Cordillera (Rodríguez-Peces et al. 2008, 2009; Rodríguez-Peces 2010; Rodríguez-Peces et al. 2011b, c). However, Newmark displacements larger than 5 cm are reached in some locations. Considering the threshold values proposed by different authors (Wilson and Keefer 1983; Keefer 1984; Romeo 2000; Keefer 2002; Rodríguez-Peces et al. 2008, 2009; Rodríguez-Peces 2010; Rodríguez-Peces et al. 2011b, c), the occurrence of both disrupted and coherent slope instabilities are plausible in this scenario. Landslides and falls (rock falls and rock slides) are the slope-instability types with the greatest concentration of areas with seismically induced landslide hazard (64 and 20 %, respectively), while earth flows show the lowest density of these hazardous areas (5 %). This fact suggests that the most likely earthquake-triggered slope instabilities in Sierra Nevada Range would be mostly rock falls and rock slides.

Table 7 Newmark displacements distribution (%) for each slope-instability typology
Full size table

On the other hand, some areas predicted as prone to earthquake-triggered slope instabilities do not match with the location of known instabilities, suggesting that the resulting hazard map is conservative since it predicts more slope failures than actually take place. The discrepancies between the hazard map and location of the actual slope instabilities could be related to the simplifications and assumptions made by the Newmark method applied to regional hazard assessment: (1) the resolution of the digital elevation model (DEM); (2) the spatial variability of geotechnical parameters; and (3) the simplifications assumed to estimate seismic shaking parameters (e.g. PGA) and seismic site effects (topographic and soil amplification).

The impact of using a DEM based on a pixel size much larger than the size of real slope instabilities produces both larger safety factors and critical accelerations values and, subsequently, smaller Newmark displacements than those obtained when using a better resolution DEM (Rodríguez-Peces et al. 2011c).

The second source of discrepancy is related to the infinite-slope limit equilibrium method used to estimate safety factors, as well as with the geotechnical parameters required for its estimation. It is clear that shear strength parameters vary across a geologic unit, although for the sake of simplicity a single value is assigned to that geologic unit. The variation of shear strength within a geologic unit is difficult to bind because geologic maps are based on lithology, depositional environment and age. One approach to cope with this problem is to measure the variability of strength parameters in the field, but this is extremely time consuming and requires collecting and testing a large number of samples, which is not practical in regional studies. Another more practical way to cope with this is considering a probabilistic distribution for modelling shear strength parameters, but a great number of data are required (Delgado et al. 2006). In addition, the SF values could be more accurate if more sophisticated limit equilibrium methods were used, particularly considering non-planar failures. However, these methods are very complicated to use in a regional scale using a GIS, being more suitable for site-specific studies. Nevertheless, the infinite-slope limit equilibrium method and the shear strength parameters used here are considered acceptable, since prior studies have evidenced that calculating safety factors at different scales produces very similar results even when different limit equilibrium methods are used (Rodríguez-Peces et al. 2011c).

Finally, the attenuation of strong ground motion is a major determinant of the hazard. PGA values were obtained using a selection of GMPEs developed for the Mediterranean region. The PGA and Newmark displacement values would be more credible when a statistically robust prediction equation for the particular case of south of Spain is available. Furthermore, the method considered here to account for soil and topographic site effects could underestimate seismic amplification. In particular, topographic amplification can have a more significant role in the generation of slope instabilities (Meunier et al. 2008). There is a need for developing studies focused on estimating seismic amplification factors based on the soil and topographic conditions of the region.

Conclusions

Earthquake-triggered landslide hazard of the Sierra Nevada Range has been analysed for the first time by means of computing a Newmark displacement map considering the rupture of the Padul Fault. This regional map is useful to identify areas with the highest hazard and also to infer the most common type of slope instability that could be triggered in relation to the occurrence of a great earthquake (M w = 6.6) close to the Sierra Nevada Range. In this sense, this map offers a first-order assessment on the possible interruption of life-lines and, hence, they could be used to improve emergency plans in the aftermath of an event.

From the landslide inventory developed in the Sierra Nevada, the most common slope instability types are the debris flows, followed by the rock falls and rock slides. Most frequent slope angles related to these instabilities are between 25° and 35°, which comprise more than 50 % of the study area. However, we noticed that the slope aspect is not a determinant factor. In addition, the most unstable lithological group is the micashists and quartzites. This fact is significant because these lithologies outcrop practically in the whole of the Sierra Nevada area.

Predicted seismically induced slope instabilities in the Sierra Nevada Range considering the rupture of the Padul Fault are mostly rock falls and rock slides, comprising a 8 % of the area. Moreover, the reactivation of old slope instabilities could be also possible in 10 % of the cases. In general, these types of instabilities can be triggered in locations where Newmark displacements of 2 cm or less are calculated, which is in agreement with the results obtained in other mountainous areas of the Betic Cordillera. However, the zones with the greatest hazard are the areas with Newmark displacements larger than 5 cm and, therefore, these areas should be of special concern for the possible interruption of life-lines.