Keywords

Introduction

Global changes would have impacts worldwide, but their effects should be even more exacerbated in areas particularly vulnerable, as it might be the case in mountain regions. Indeed, in these areas, a range of socioeconomic sectors (e.g., tourism, forest production, agro-pastoralism, ecosystem resources…) have experienced considerable change in the last two centuries, resulting in pressures on natural resources and traditions imposed by increasingly-industrialized societies. Some mountain regions have been extensively transformed, converting them from inaccessible and relatively poor areas into attractive destinations for the wealthy, sometimes excluding longtime inhabitants from economic benefits. In other cases, outmigration and an aging population have led to economic declines in agro-pastoral and forestry production.

Moreover, climate change occurring in mountains may imply future modifications in temperature and precipitation patterns; this may lead to changes in the balance between snow, ice and rainfall, which ultimately will result in changing the quantity and seasonality. As a consequence, natural processes controlled by hydro-meteorological triggers, and amongst them landslides, will in a future climate add further environmental pressures on both social and natural systems.

Landslide hazard may be affected by global change, according to different features. More specifically, the socio economical changes may have some impacts on landslides through the evolution of the vegetation cover.

Indeed, if vegetation influences very poorly the deep-seated landslides, this influence exists for shallow landslides, but remains difficult to address. Indeed, in the same time, the vegetation cover increases ground weights, and so tends to initiate the ruptures, but it also increases the shear strength. The relative importance between these two contrary effects varies according to the localization of the vegetation cover on the slope. The stability would be increased if the vegetation is present on the toe of the slope (Genet et al. 2010; Ji et al. 2012) but this stabilizing effect would be reduced if the vegetation is located on the upper part of the slope (Norris et al. 2008; Genet et al. 2010).

Landslides are also sensitive to the presence of water within the layers susceptible to move. As future trends in the climate may imply some modifications of meteorological parameters, such as precipitations and temperatures, the resulting underground water table level may evolve.

In this study, the global change impact on landslide hazard is analysed according to two features:

  • The evolution of the land use, which can be analysed as the evolution of the vegetation cover from the landslide point of view;

  • The evolution of the climatic conditions, resulting in the evolution of hydrogeological conditions.

Methodology

The evaluation of the global change impact on landslide activities is realized through the analysis of both vegetation cover and meteorological conditions on landslides activities; this has been conducted through the following methodology:

  • The assessment of future land use is addressed through the construction of four prospective socio-economic scenarios up to 2040 and 2100, which are then spatially modeled with land use and cover changes (LUCC) models.

  • The climate change inputs of the project correspond to 2 scenarios of emission of greenhouse gases. The simulations were performed with the Green House Gas (GHG) emissions scenarios (RCP: Representative concentration pathways, according to the standards defined by the GIEC) RCP 4.5 and RCP 8.5.

The impact of land use and climate change is then addressed through the use of scenarios into hazards computations. For that we use a large-scale slope stability assessment tool ALICE which combines a mechanical stability model, a vegetation module which interfere with the first model, to take into account the effects of vegetation on the mechanical soil properties (cohesion and over-load), and an hydrogeological model, which permits to provide water table level, based on meteorological temporal data, to be then integrated into the stability model (see Fig. 1). Then the methodology permits to obtain for two different future periods (2040 and 2100) the corresponding hazard maps with combining the socio-economical and climatic scenarios.

Fig. 1
figure 1

Global methodology for evaluating the global change impact on landslide occurence

Full size image

The following paragraphs go into details on the components of the methodology.

Hydromechanical Model

Slope Stability Model ALICE

ALICE (Assessment of Landslides Induced by Climatic Events) is a tool dedicated to the analysis of landslide hazard for areas ranging from slopes to department (Baills et al. 2011; Sedan et al. 2013). The mechanical model is based on a 2D slope stability analysis, for which the main physical characteristics of the soils and surfaces are quantified. These parameters are integrated in a mathematical model to calculate a safety factor for each pixel of the different profiles covering the whole area (Aleotti and Chowdhury 1999).

The software uses the Morgenstern and Price method (1965, 1967), using Zhu et al. (2005) algorithm, which is a finite slope stability model based on the equilibrium calculation between slices subdividing the landslide volume.

The geometry of the geological layers to be considered are designed based on a geomorphological approach, which combines field analysis, geological and surficial layers analyses. The geotechnical characteristics of the soil layers (cohesion (c), angle of friction (ϕ) and unit weight (γ)) are then taken into account thanks to the altitude maps of the interfaces between each soil layer, the highest limit corresponding to the topographic surface (DEM).

The model also considers the effect of the water table level into safety factor computations. In this study, we consider that the water table can fluctuate within the surficial layer. The water table is then automatically generated between a low and a high piezometric levels by setting a filling ratio.

The safety factor calculation also needs the landslide type (rotational or translational) and its dimensions (length and depth).

The probabilistic approach used allows taking into account heterogeneities and uncertainties by giving probabilistic distributions of the geotechnical parameters. Thus, the probability of having the safety factor below one represents the probability of landslide occurrence for a given triggering scenario (i.e. landslide geometry and water table level).

The dispersion of the distribution provides the uncertainty associated to the result. Unless the slip surface is previously known, the safety factors should be calculated for every potential slip surface of a hill slope. The slip surface for which the safety factor is the smallest is the critical slip surface, along which sliding is the most likely to occur.

Hydrological Model Gardenia

The global model, derived from GARDENIA (Thiéry 2003) considers a system of 3 tanks, which reproduces the 3 layers characteristics of the hydrological behavior of the soil (see Fig. 1):

  • The top zone, the first tens centimeters, where the evapotranspiration occurs,

  • The unsaturated zone, where runoff occurs,

  • The saturated zone.

It allows estimating the relationships between the local piezometric levels and global river discharge and the meteorological parameters, such as precipitation, temperature (to distinguish between snow and rain fall) and the potential evaporation. The water level is then converted into a pore pressure indicator to be integrated in the slope stability assessment equations.

Socio Economical Scenarios

The overall methodological approach for constructing socio-economical scenarios is adapted from Houet et al. (Accepted). It consists in co-constructing with stakeholders fine-scale socio-economic scenarios based on existing national or regional sectorial scenarios, while developing a spatially explicit local LUCC model dedicated to mountains land uses.

The method resorts to two participatory workshops aiming first at defining the storylines, and second at validating them and pre-identifying the areas of future land use changes. Meanwhile, the use of the LUCC model allows simulating the land cover changes induced by future land use changes, which in turn, can have feedbacks effects on land cover changes.

Thus, the narrative scenarios are defined so as to produce relevant inputs to the LUCC model while the model itself is developed so as to be able to represent the likely land cover changes identified in the narrative scenarios and provide quantitative outcomes to illustrate the narratives.

Four scenarios have then been defined according to this methodology, resulting in maps of future land covers for the four scenarios in 2040 and 2100:

  • abandonment of the territory;

  • sheeps and woods;

  • a renowned tourism resort;

  • green town;

Figure 2 shows the maps of existing and future land covers for the scenario 2 in 2040 and 2100.

Fig. 2
figure 2

Maps of existing and future land covers for the scenario 2 in 2040 and 2100

Full size image

Integration of Vegetation into Hydromechanical Model

So far, the vegetation has been integrated in the model in two ways: (i) with an additional apparent cohesion of the soil due to root reinforcement of the resistance to shear and (ii) with additional weights on the slices. Hence, the vegetation cover is an additional layer of the model, with the additional cohesion due to the roots. These modifications affect mostly shallow landslides.

Climate Change Scenarios

The climate change inputs of the project correspond to 2 scenarios of emission of greenhouse gases. The used simulations available on the portal DRIAS (http://www.drias-climat.fr) were performed with the GHG emissions scenarios RCP 4.5 and RCP 8.5 for the ALADIN-Climate model of Météo-France.

The climate model shows a tendency to the increase of extreme events of precipitations at short and long term, and for the annual cumulative precipitation. The results depend on the elevation of the area of interest. Indeed, for the highest points, the model shows an increase of cumulative precipitations. For the lowest points, the model indicates a slight increase at short term, and a small decrease at long term. Concerning the temperatures, the model clearly indicates a significant increase of the temperatures at short (+1.5 °C) and long term (+4 °C), resulting in large changes in precipitation pattern (balance between snow and rainfall).

Integration of the Water Table Level into Hydromechanical Model

Thanks to Gardenia model, the daily water table level is computed according to the daily meteorological parameters. In this way, it is possible to obtain the distribution of the water table level between a low and a high piezometric levels. The Fig. 3 shows the distribution of the classes of water level filling ratio for the current period (1981–2010), two future periods (2021–2050 and 2071–2100), and for 2 scenarios (RCP4.5 and RCP 8.5). This figure demonstrates the significant increase of the mean water table level for future periods, especially between 2071 and 2100 with the worse scenario (RCP8.5).

Fig. 3
figure 3

Evolution of the distribution of the classes of water level filling ratio

Full size image

Cauterets Site

The study site is located in the French Pyrenees and covers more than 70 km2 in the Cauterets municipality. The elevation ranges from 800 m and 2700 m, which results in a mountainous climate with high precipitations of around 963 mm/year. This territory is the place of the occurrence of several natural hazards, such as avalanches, rock falls, or torrential floods. Amongst them, several landslides occurred in this area as it is illustrated in Fig. 4.

Fig. 4
figure 4

Historical landslides and lithological map of surficial deposits

Full size image

This municipality is an increasing touristic station, initially linked to health resort, and then more largely developed with ski resort since 1964. It has a regulatory natural hazard risk map, which covers landslides, avalanches, torrential flood, rock falls, earthquake, inundations, and forest fire.

The Fig. 4 shows the geological map of the surficial layers, with the corresponding geotechnical parameters indicated in Table 1.

Table 1 Geotechnical parameters of the surficial deposits
Full size table

These parameters have been determined based on previous studies (Fabre et al. 1999; Lebourg et al. 2003).

Results

The current map of probability of Safety Factor (SF) <1 is shown on Fig. 5, with considering rotational landslide, with a length of 25 m, and a depth between 1 and 3 m. The water table level is at the top level (i.e. the topography). The historical landslides are also reported on this figure.

Fig. 5
figure 5

Impact of the vegetation on the landslide susceptibility—a probability of SF <1 with vegetation in 2010, b type of forest with the corresponding additional cohesions (in kPa); c probability of SF <1 without vegetation; d differences between the probability of SF <1 with and without vegetation

Full size image

The Fig. 5a corresponds to probability of SF <1 with considering the vegetation, Fig. 5b shows the type of forest considered in this simulation with the corresponding additional cohesions (in kPa); Fig. 5c indicates the probability of SF <1 without considering the vegetation; and finally Fig. 5d provides the differences on the probability of SF <1 without and with vegetation.

The prediction rate curve shown on this figure evaluates the performance of the spatial model in reproducing the historical landslides observed on the field (Chung and Fabbri 2003). This curve plots the proportion of the area considered as susceptible to the proportion of landslide recognized as susceptible on the entire surface. In particular, the curve which rises quickly to get to 100% of recognized landslide, indicates that the model is efficient. Then, with calculating the area under the curve (AUC), it is then possible to quantitatively estimate its performance. If the AUC is close to 0.5, the model is considered hazardous. At the contrary, if the area is close to 1 the model is considered valid. It is generally accepted that AUC greater than 0.7 shows that the model is efficient, as it is the case in this study with an AUC equal to 0.837.

These results show that the absence of the forest implies an increase of the probability of landslide occurrence of about 0.5; at the contrary, the presence of forest has a local stability effect on the slope.

Moreover, when regarding the evolution of land use designed by scenarios 2, we can observe some significant local variations of landslide susceptibility. Indeed Fig. 6 shows the (a) differences of the probability of SF <1 between 2010 and 2040, (b) the corresponding evolution of the additional cohesion (in kPa), (c) differences of the probability of SF <1 between 2010 and 2100, (d) the corresponding evolution of the additional cohesion (in kPa). The results clearly indicate the increase of the stability in the areas where the forest is developing, whereas we observe that the stability is dropping in the areas where the forest is disappearing.

Fig. 6
figure 6

Evolution of landslide susceptibility according to land-use change designed with scenario 2—a differences of the probability of SF <1 between 2010 and 2040, b the corresponding evolution of the additional cohesion (in kPa), c differences of the probability of SF <1 between 2010 and 2100, d the corresponding evolution of the additional cohesion (in kPa)

Full size image

Finally, Fig. 7 shows the evolution of landslide susceptibility according to land-use designed with scenario 2 and climate change with two different scenarios, with (a) the mean probability of SF <1 during the period 1981–2010, (b) the differences of the probability of SF <1 between 1981–2010 and 2021–2050 with RCP 4.5 scenarios, (c) the differences of the probability of SF <1 between 1981–2010 and 2071–2100 with RCP 4.5 scenarios, and (d) the differences of the probability of SF <1 between 1981–2010 and 2071–2100 with RCP 8.5 scenarios.

Fig. 7
figure 7

Evolution of landslide susceptibility according to land-use designed with scenario 2 and climate change with two different scenarios—a probability of SF <1 in 2010, b differences of the probability of SF <1 between 2010 and 2040 with RCP 4.5 scenarios, c differences of the probability of SF <1 between 2010 and 2100 with RCP 4.5 scenarios, d differences of the probability of SF <1 between 2010 and 2100 with RCP 8.5 scenarios

Full size image

The results point out that an increase of the water content of the soil (as it is described in Fig. 3) induces a reduction of the SF in a large part of the studied area. These changes are not uniform over the area, and are particularly significant for the worse scenario of RCP 8.5.

Conclusion

The present work shows a methodology to evaluate the influences of both vegetation cover and climate on landslide activities over a valley until 2100.

The results demonstrate the influence of the forest on slope stability. The results also indicate some future evolution of the land use, leading to significant modifications of the stability of the slopes. We observe some increases of the stability in the areas where the forest is developing, whereas the stability is dropping in the areas where the forest is disappearing. Finally the climate change may have significant impact with the increase of the water content of the soil; the results point out a reduction of the SF in a large part of the studied area. These changes are not uniform over the area, and are particularly significant for the worse scenario RCP 8.5.