Análisis de la peligrosidad de movimientos de ladera en la Cordillera Bética mediante técnicas avanzadas

Resumen

Given the importance of the spatiotemporal impact that involves the phenomenon of landslides, the interest in risk reduction and the appropriate urban and civil-engineering planning is reflected in the numerous related research work and scientific publications. Many of these are focused on develop innovative methodologies for the estimation of the main components of the fundamental risk equation (susceptibility, hazard and vulnerability), or as guides and reviews of the most appropriate methods and techniques for the zoning of such components regarding the different types of landslides, scales or objectives. The level of risk as a consequence of landslides depends on the hazard associated; that is, on the recurrence or temporal probability with which one or more landslides-related processes affect a terrain unit. That is why the knowledge of the frequency with which these processes occur, or with which important activity changes arise, together with the different types of landslides and their sizes or intensity classes provide valuable information to predict the transient behavior linked to the destructiveness of these natural events. However, the hazard assessment appears rather as a challenge, especially if it is to obtain a quantitative estimate. This is mainly because the tasks are hampered by the complexity, requirements, restrictions and lack of data necessary for the application of methods that provide information about the chronology of the landslides activity at different geographical scales. This is the case of the dating of prehistoric and even historical events; sequentially obtaining landslides inventories as the basis for the analysis of events time-series; determining hydrological or hydrogeological parameters associated with the activation or reactivation of landslides, necessary for establishing cause-effect relationships; or establishing thresholds and their return period for these parameters, useful in the inputs of dynamic instability models or advanced hazard studies. The above difficulties, as in other cases, are found when assessing the hazard in the southern part of the Betic Cordillera, located in southern Spain; while the spatial potential or susceptibility to trigger landslides is in a more advanced state of researching. Thus, in the present work, the research has been addressed towards the achievement of objectives that aim to advance or improve those parts where there are still shortcomings of methodological or technical nature to generate critical information to enable the landslide assessment. For this purpose, the research work has been divided into three parts: the first one consisting of the detailed monitoring of the activity evolution of 7 landslides, taking advantage of the terrestrial LiDAR (TLS) technique, whose results can be correlated with variations in the rainfall record as the main cause of temporal changes in activity; the second one dealing with the landslides inventory at regional scale from sequential data obtained by the technique of airborne LiDAR (ALS), which enable the streamlining in the update of the information needed to estimate the spatio-temporal probability based on series of events detected within certain time periods (eg., mean no. of movements/year). The last part is dedicated to the development of an analysis methodology based on Partial Duration Series (PDS), enabling the distinction of accumulated rainfall thresholds associated with documented landslides, and whose results provide information about the cause-effect relationships of this phenomenon and its temporal recurrence. 6 In the first case, the products obtained from the TLS technique have enabled to reveal significant changes in the activity of some of the landslides studied, even showing incipient deformation. The small terrain changes detected have been interpreted on the accurate zoning of differential displacements resulting from the comparison of digital terrain models with resolutions on the order of decimeters, built using a refined methodology that seeks to minimize common constraints (hidden areas or "holes", accumulated error in the adjustment of scanned parts, unwanted objects for the analysis of results, supervised classification, etc.), enabling to establish the average noise threshold for remote detection of geomorphologic changes about 10 cm, working from distances between 500 m and 1 km. This methodology regularly applied to landslides on various monthly intervals shows the stability of some landslides against the dynamics manifested in others, according to their different stages of reactivation or changes in the displacement rate. Among these, in the translational slide "Almegíjar" a reactivation of the whole mobilized mass was detected, initiated in the wet period 2009-2010 and continuing with different velocity rates in the following analysis periods. This reactivation generated a strain with a shortening in its longitudinal dimension and an extrusion towards the outside; which was observed from an aerial perspective by means of the airborne equivalent technique, ALS, although with different precision magnitudes. According with the time intervals imposed by the sequential monitoring in the two first periods of the reactivation (2009-2010 and 2010-2011), the differential displacements reached maximum values between 1.3 m and 2 m, with minimum rates of 1.42 m/a, 1.64 m/a and 1.35 m/a for the head subsidence, downward displacement (approximately in the direction of the failure plane) and the advance at the bottom of the mobilized mass respectively. This deformation, which continued until 2011 with similar characteristics, showed a recession by 2012, with displacements of decimeter order; while in the short period from 04/2012 to 11/2012 no significant changes were observed. In the period 2012-2013 displacements closest to the threshold for the relative error (10 cm) were detected, which compared to stable areas and the accretion of debris lobes, these could mean a new acceleration. The slide "Borincho" also showed remarkable results in terms of functionality of the applied technique, showing a partial reactivation at the toe of the hillside that began in the same period (2009-2010) and continued until 2011. After an analysis of historical rainfall series, the most conservative thresholds that with greater possibility favor the dynamics of ¿Borincho¿ and¿Almegíjar¿ landslides were established in 284 mm (recurrence interval of 7.5 y) and 256.4 mm (6.4 y) respectively, setting a safety factor of 25%. As regards the other landslides, on the main body of the slide "Puente Viejo" a secondary debris slide was detected also in the interval 2009-2010; while in the landslides "Chuca" and "Budas" only small collapses at the toe or lateral flanks during the same period were detected. Although with lower resolution, altimetry data acquired by airborne LiDAR technique, treated by a specific methodology, enabled to identify various landslides events on the results of models of differences between different types of DEMs (DEMoD). In total 47 natural landslides and 50 slope-cuts failures were inventoried, starting from the comparison of the products obtained for 2 flights, the first conducted between August and November 2008 and the most recent in July 2010. Crossing the inventory of natural landslides with the lithological map, it is revealed a larger surface affected by translational landslides in phyllites with interbedded limestone (69.81%), followed by quartzites and quartzitic schist (15.16%), with the less influence on conglomerates, 7 calcareous crust and travertine. Unlike the above, slope-cuts failures largely affected quartzite and quartzitic schists, matching different road sections. Regarding the failure mechanisms, translational slide was the predominant type. For its part, the analysis on the distribution of landslides regarding the existing susceptibility map showed greater affected area for the moderate susceptibility class than for the high one; difference that, when comparing only the representative points (centroids) of each landslides results in 57.47% of them associated with the moderate susceptibility class and 42.55% with high one. However, this contrast must be interpreted with appropriate consideration of the difference between the temporal intervals of the inventory obtained with this methodology and the historical - prehistoric inventory used for the development of the susceptibility model. This effect would be minimized by updating the recent inventory, so their spatio-temporal distribution becomes increasingly comparable to that predicted by the susceptibility model. Otherwise, the updated inventory could be used to calibrate the model susceptibility. About the type of inventory obtained, regarding the covered time window, it can be typified as a multitemporal inventory; however, due to the occurrence of heavy and lasting rains of the period 2009-2010, whose effect on the ground was also determined by the results of the methodology based on the TLS technique, these are considered as a triggering event of the multiple landslides detected (MORLE). Therefore, this inventory rather may be classified as an event inventory. The availability of catalogues or inventories of dated-landslides events for an extensive and potentially active area provides the basis for calculating the spatiotemporal frequency and study of the causes which determine the occurrence of these processes; however, those based on remote sensing monitoring hardly provide the exact initiation or reactivation date of landslides. In this case, documental catalogues of events associated with this phenomenon represent the best database for assigning a more precise dating to these events, especially after the occurrence of a MORLE that awakens the socio-economic interest and increased the number of news or reports available in the literature. This fact has a direct involvement in hazard assessment, enabling not only the dating of recurrent events in time, but also the analysis of the causes or triggers characteristics. This is what has been treated in the third part of the investigation, in which the magnitudes of different rainfall variables both contemporaneous as antecedent to the dates of 20 landslides, geographically located and accurately dated from the documental review, have been assessed. All these events were generated in datesclose to particularly rainy periods in the winters of 1996-1997 and 2009-2010, which can be considered as MORLEs. Thus, taking into account the dates of the landslides events, and by developing a methodology consisting of an exhaustive frequency analysis, the information concerning the triggering factor could be extended to narrow down the different thresholds of antecedent and contemporary rainfall to the occurrence of each landslide. The basis of the methodology lies in the analysis of multiple Partial Duration Series (PDS) derived from the combination of the different pluviometric variables with a wide duration-range (1- 90 days). This procedure facilitates the displaying of anomalies on the graphic results of the continuous spectrum of both variables and their recurrence interval, either previously or during the generation of each landslide. As result of this retrospective analysis, previously nonexistent information on the behavior of precipitation, as main causative factor and also preparatory of the multiple landslides was gathered. Applying 8 the suitable criteria and, taking the advantage of observing the anomalies on the return-period curve, the antecedent cumulative rainfall over one or more anomalies (ADG1), equivalent to the precipitation collected during the last storm (EDG1) when no previous anomalies exist, was established as the most appropriate variable for obtaining recurrence intervals. Moreover, both variables were taken into account in determining the thresholds related to the landslides events. Analyzing the threshold distribution derived from the intensity-duration (I-D) combination for ADG1, it was possible to adjust a potential function of the form ¿¿¿¿ = ¿¿¿¿¿¿¿¿¿ ¿ ¿¿¿¿ , with ¿=88.005 and ß=0.69. By relating values of both variables, although with a low correlation coefficient (R = 0.11), a negative linear relationship was found with equation: ADG1 = -0.6617 EDG1 + 222.83. For this correlation, the case where both variables present low values can be attributed to an underestimate, possibly linked to measures of precipitation significantly lower than those prevailing in the landslide area. Despite the low number of landslides compiled, by means of the indirectly assigned frequency through this analysis, it was observed that the natural landslides initiated (or reactivated) with a lower average frequency that slope-cuts failures (16.4 years versus 9.7 years, respectively). Considering the different durations of the associated rainfall events, the first ones activated with average values of ADG1 intensity greater than the second ones (39.1 mm/d from 4.15 mm/d, respectively). Regardless of whether the slope is of the natural or artificial type, analyzing the various types of landslides, it is observed that slides respond to mean intensities of ADG1 (30.1 mm/d) lower than rock falls (39.3 mm/d) or complex landslides (42.5 mm/d). In contrast, the average recurrence period for the first ones is significantly higher (22.4 a), coinciding with a longer average duration of the associated rainfall events (23.6 d for slides against the 15.86 d for rockfalls, or 12 d for complex landslides). Among the total cataloged events, return periods of up to 3.6 years are assigned to the 50% of them, and only 25% of the antecedent rainfall (ADG1) magnitudes were for recurrence intervals longer than 24 years. Overall, the results have shown that the applied methodologies contribute to the progress in the quantitative-hazard assessment in the study area, either recording landslide events in large areas using the technique of ALS useful in time-series analysis to establish the frequency for the mean interval between sequential acquisitions; monitoring the evolution of the diachronic landslides activity in relation to triggers with the TLS technique, or addressing a retrospective study of cause - effect from catalogs of historical landslides and expanding the information provided by an intensive analysis based on PDS. However, the implicit temporal connotation in hazard assessment requires research continuity aimed to the compilation of new landslides events, as well as their intrinsic characteristics and those of their triggers. With this practice the outcomes in the spatiotemporal probability could be determined more precisely, and on the other hand, it could be obtained the required parameters to perform numerical or physical models at the advanced level of detail.