Instability mechanisms and evolution of a rocky cliff on the Atlantic coast of Spain

Predicting the response of rocky coasts to different erosional agents remains a great challenge at present. The episodic and discontinuous nature of the instability processes typical of hard bedrocks makes it difficult to make predictions based on observations over short research periods. This work aims to contribute to the understanding of the geomorphological evolution of rocky cliffs by means of a case study of a geologically complex cliff (developed on quartzite and slate) located on the Atlantic coast of Spain. The analysis of high-precision topographic models and orthophotographs, the use of geomatics techniques and geomorphological characterization have made it possible to define a model of the cliff behaviour. The results indicate that the structure of the bedrock determines the type of instability processes affecting the cliff and the morphology of the associated deposits. Lithology is the other main conditioning factor: while slate is easily eroded, quartzite offers greater strength and its detached blocks act as an effective natural defence element protecting the cliff and slowing down the coastal retreat. The evolution model established for this cliff explains the absence of retreat in the study period (2003–2022) and confirms the important role of local factors in cliff evolution.


Introduction
Predicting the response of rocky coastal cliffs to different erosional agents remains a great challenge at present. The episodic and local nature of the instability processes characteristic of hard rocks make it difficult to make predictions based on observations made during the usual research periods (Kennedy et al. 2017), usually of a few years. In addition to the marine action and the weathering of the cliff (Trenhaile 1987;Sunamura 1992;Masselink and Hughes 2003), the influence of the lithology and geological structure of the bedrock must be considered (see, for example, Duperret et al. 2004;Mortimore et al. 2008;De Pippo et al. 2008;Del Río and Gracia 2009;Swirad et al. 2016). In the current context of climate change, an impact in the medium and long term on the current coasts and their geomorphology because of sea level rise (e.g., Church et al. 2013;Kopp et al. 2014; Le Bars et al. 2017) and marked changes in climate (general increase in temperatures, marked change in the rainfall pattern, increased frequency of storms, changes in the wind regime, etc.; see Dufresne et al. 2013;Noël et al. 2021) is expected. There is extensive literature on the effects that these changes will have on low-lying coastal areas (e.g., Trenhaile 2002;Davidson-Arnott 2010;Nicholls and Cazenave 2010;Nicholls et al. 2021;Slangen et al. 2012;Ranasinghe 2016;Bon de Sousa et al. 2018;Sanjosé Blasco et al. 2018;Lauzon et al. 2019;Scardino et al. 2020). However, the effects on rocky coasts have received much less study, even though they make up more than half of the world's coasts (Griggs and Trenhaile 1994;Young et al. 2009;Young and Carilli 2019). The reason for this is the slow recession of these coasts in comparison to the rapid retreat of sandy shores (Moura et al. 2006;Swirad et al. 2016;Trenhaile 2016).
It is known that the rates of rocky cliff retreat vary widely as a result of the many variables involved in their evolution. Prémaillon et al. (2018) proposes median erosion rates of 2.9 cm yr −1 for hard rocks, 10 cm yr −1 for medium rocks and 23 cm yr −1 for weak rocks. Since the 1980s, the north Atlantic coast of Spain (Fig. 1a) has been the subject of studies focused on sandy bays, estuaries, and marine terraces, especially in its central zone (e.g., Flor and Flor-Blanco 2006;Codrón and Rasilla Álvarez 2006;Flor-Blanco et al. 2015Sanjosé Blasco et al. 2020). Recently, the results obtained in cliffs formed in Paleozoic and Jurassic rocks in this sector have shown that they evolve mainly by rotational and planar landslides, with estimated coastline retreat velocities in specific areas at Cabo Peñas (55 km East of the studied cliff) of up to 2.19 m in lutite and limolite for 2006(Domínguez-Cuesta et al. 2020a or active mass movements of more than 70,000 m 2 extension in cliffs that have developed in Jurassic marl and sandstone located 100 km East of the cliff that is the object of this study Domínguez-Cuesta et al. 2020b). However, instabilities and rates of retreat of cliff faces on the west Cantabrian coast that have formed in other rocks, such as relatively hard Ordovician metamorphic rocks, remain unknown. The western and central sectors of the Cantabrian Coast are also locally affected by port activities, whose related structures are also subject to marine erosion (Flor et al. 2006), and have an important cultural heritage (e.g., Iron Age and Roman hillforts located on cliff peninsulas) that is being destroyed by marine erosion (Camino Mayor 1995;Jiménez-Sánchez and Ballesteros 2017;Domínguez-Cuesta et al. 2020a).
This research aims to contribute to the understanding of the geomorphological evolution of coastal cliffs that have developed in hard rocks. A rocky cliff located in Luarca (Fig. 1), which is representative of cliffs formed in Ordovician slate and quartzite on the Western Cantabrian Coast, has been selected as a case study. Given its complex geological features, it can serve to reveal the "attacking" or "defensive" role played by structural, lithological, and geomorphological factors in cliff retreat. This requires a necessary characterization of the instability processes, including their spatial and temporal delimitation, as well as the quantification of the contribution of these processes to the general retreat.

Setting
The cliffs of Luarca (extending for 4,500 m 2 ) are located on the Northwest Atlantic Coast of Spain (Fig. 1a), a sector dominated by rocky cliffs interrupted by small sandy beaches and narrow estuaries (e.g., Flor-Blanco et al. 2015). The base of the cliff is connected to the current abrasion platform, its upper edge being the northern limit of ancient marine terraces, locally known as "rasas". These surfaces extend, with a width of 2-6 km, along more than 200 km of the coast and have an elevation of between 100 and 150 m in their eastern part, which progressively decreases towards the West (e.g., Moñino et al. 1988;Álvarez-Marrόn et al. 2008;Flor and Flor-Blanco 2014;Ballesteros et al. 2017;Domínguez-Cuesta et al. 2015;López-Fernández et al. 2020). Rasas and the corresponding cliffs are the result of the emersion of the old abrasion platform, which originated during the Pleistocene, with an uplift rate since then of 0.1-0.2 mm·year −1 (Jiménez-Sánchez et al. 2006;Álvarez-Marrón et al. 2008).
The approximately 570 m long section of cliff in this study is located East of the village of Luarca (ca. 6,000 inhabitants in 2021). The section shows a general N43°E direction and reaches a height of 67 m (Fig. 1a). In general, the upper part of the cliff is notably steep, sometimes vertical, while its lower part is inclined towards the sea with a variable slope (58°NW on average). The slope is partially covered by detached fallen blocks. Those blocks located at the cliff foot, which cover the current abrasion platform (inclined 3°N), are directly affected by marine action (Fig. 1b). The upper edge of the cliff is the North limit of the rasa, which in this area has a 3 km width and dips 0.5°N. The slope is covered by small, scattered plant communities. These are shrubby and are in the middle and upper part of the cliff, as they are more protected there.
The climate of the study site is oceanic, temperate and without a dry season. Rainfall is regularly distributed throughout the year and averages 1,404 mm according to the State Meteorological Agency of Spain-Agencia Estatal de Meteorología (AEMET). The annual average temperature is 13.7 °C, oscillating between average minimum and maximum values of 9.6 and 18.0 °C. Prevailing winds come from the Southwest, although those responsible for modelling the coastline, given the orientation of the coast (West-East), are those of the West and Northwest component, generally very intense. The dominant waves are from the Northwest, with an average annual wave height of 1.5 m, with the median at 1.5-2 m and the P95 at 4-4.5 m. In this coastal sector, waves with heights significantly greater than 1.7 m are considered stormy, with maximum values of around 10-12 m, associated with storms with a West and Northwest component, frequent in winter (Izaguirre et al. 2011). The average annual wave period is 9 s, and 95% of the time it is less than 12 s. The coast shows semi-daily tidal variations between 1.00 and 4.65 m, although in extreme circumstances they reach 4.95 m. In general, the high tide allows the swell to impact the rocky deposits at the foot of the cliff and even the base of the cliff during storm events. On this coast the waves and tides cause a generalized transport of sediments to the East (Flor and Flor-Blanco 2005), with the consequent local topographic effects motivated by the presence of capes and inlets.
The area above the Luarca cliffs, as in other areas of the Atlantic Coast, is mostly subject to urban development. Beyond a point less than 5 m from the edge of the cliff, dozens of houses, a secondary school and a primary school, and tourist complexes have been built in the area in recent decades, progressively displacing the traditional agricultural use.
The cliff bedrock resulted from a complex geological history derived from the overlapping of two orogenies (Variscan and Alpine) and the different rheological response of the metamorphic rocks. The bedrock of the central sector of the cliff is quartzite, while slate outcrops at its eastern and western ends; both are Middle Ordovician in age and belong to the Luarca Formation (Marcos 1973;Gutiérrez-Marco et al. 1999). Quartzite layers adopt a Northeast-Southwest direction, dipping Northwest and are intensely fractured. Slate layers outcropping in the cliff are oriented North-South with a subvertical dip and present a tectonic foliation whose orientation and dip vary substantially depending on the folds and faults affecting these rocks. The slate is fine-grained, black, and locally rich in iron sulphides. This rock presents an average weight density of 26.4 kN/m 3 and its uniaxial compressive strength is between 20 and 35 MPa (according to our tests); therefore, the slate is classified as a soft rock (according to ISRM 1981) or a very low-resistance rock (according to Bieniawski 1989). The quartzite exhibits medium to coarse grain size and intense recrystallization; its average density is 26.5 kN/m 3 , showing a tensile strength of 20-28 MPa and an average uniaxial compressive strength of 135 MPa (according to our previous tests); thus, the quartzite is considered a very hard rock (ISRM 1981) or a high strength rock (Bieniawski 1989). In the sector studied, the slate areas correspond to coves and beaches, while the quartzite gives rise to a small promontory, with detached blocks accumulating at the foot of the cliff. Covering the bedrock, less than 3 m of Quaternary marine deposits appear on the rasa and are mainly made up of sand, rounded and subrounded quartzite pebbles and cobbles (Flor and Flor-Blanco 2014).

Methods
The methodology of the Luarca cliff study includes geomorphological mapping assisted by Remotely Piloted Aircraft (RPA), evaluation of the lithological and structural influence and the compilation of temporal evolution evidence.

Geomorphological mapping assisted by RPA
To identify the instability zones and processes, a detailed geomorphological map of the study cliff was created at a 1:2,000 scale in a Geographical Information System (GIS, ArcGIS Desktop v10.3), combining field evidence and data acquired by RPA. The cartography exhibits the spatial distribution of the features and deposits linked to instability processes or marine action. The upper edge of the cliff, the limit of the current abrasion platform and the scour zone at the foot of the cliff have also been delimited. The RPA provided a Digital Elevation Model (DEM) (8 cm spatial resolution) and an orthophotograph (3 cm resolution) of the study area corresponding to two photogrammetric surveys carried out in November 2018 and December 2020.
The RPA was operated by the company INGECOR Geomatics (Fig. 2a) using a CÁRABO S3 (ICOM3D), a quadcopter with a Sony A5100 with Exmor® CMOS Sensor type APS-C (23.5 × 15.6 mm), and a 20 mm fixed lens. From a flight height of 111 m, 147 images were taken (ground resolution of 2 cm / pix), establishing 9 control points (fixed targets). As observational instruments, two GNSS Topcon GR3 receivers (Antenna TPSGR3 NONE) were used, and the data collected from a permanent station located at 100 m (Antenna LEIAX1203 + GNSS NONE) were also incorporated. Image processing was carried out with Agisoft Metashape v.1.5.2 software. The maximum error calculated from the control points was 3.52 cm. For each flight, a textured 3D cloud (126,462 points) with an average point density of 157 points/m 2 was generated. The Cloud Compare v.2.10.2 open-source program (available at http:// www. cloud compa re. org) was used to analyse the point cloud obtained with the RPA. This software made it possible to identify and measure the bedrock discontinuities. The application of this technique allowed the acquisition of structural measurements in areas inaccessible to traditional field methods. It also makes it possible to carry out comprehensive digital analyses and optimize the time spent on data collection.

Geological mapping and structural analysis
To establish the contribution of the lithological and structural factors to the stability of the cliff, a geological map was made at a scale of 1:2,000, based on field work and the information provided by the RPA. Geological structure was complemented by four geomechanical stations; two stations on the slate and two on the quartzite bedrock. In each station, discontinuity families (stratification, joints, faults, foliations) were identified, detailing their dip, dip direction, spacing, opening, roughness and infill. Furthermore, structural data were collected at 57 inaccessible points on the cliff surface from the point cloud obtained with the RPA through the Compass plugin (Thiele et al., 2017) of the Cloud Compare software. This powerful program allowed the identification, positioning and measurement (dip/dip direction) of flat surfaces of geological origin (Fig. 2b). The DIPS v.8.020 (Rocscience) program was used for the geometric and kinematic analysis of the discontinuities. With this software, the discontinuities were analysed using stereographic projection (equal area, lower hemisphere), allowing us to determine the potential formation of wedges, rockslides, and other instabilities. Stereographic projection is a type of azimuth conformal projection that allows the representation of three-dimensional objects (e.g., joints, bedding) on a two-dimensional surface and the performance of linear, angular, and surface measurements.

Cliff monitoring
The Luarca cliffs were monitored to survey block displacements and coastline retreat. The information used for this was collected through historical orthophotos (taken in 2003, 2006, 2009, 2011 and 2014; National Plan for Aerial Orthophotography of Spain, available at http:// centr odede scarg as. cnig. es/ Centr oDesc argas), orthophotos acquired via RPA (2018-2020) and field campaigns using a total station (2018-2022). All images have a spatial resolution of 0.25 m, except for the 2003 and 2018/2020 flights, whose resolution is 0.5 m and 0.08 m, respectively. All the orthophotographs were integrated into GIS to undertake a comparative study for the period 2003-2020, taking the upper edge of the cliff and the spatial location of detached and identifiable blocks as a reference.
A high-precision quantitative comparative study was carried out between the DEM obtained by RPAs (Fig. 2a) in November 2018 and August 2020. This allowed us to identify the geomorphological changes that the cliff underwent in that period using the M3C2 plugin of the Cloud Compare program. This tool calculates the local distance between twopoint clouds along the surface normal direction that tracks 3D variations in surface orientation (Lague et al. 2013).
Moreover, topographical monitoring of the slope was carried out, consisting of the installation and surveying of 20 control points distributed along the cliff (Table 1). Seven feno markers were installed on the upper edge of the cliff (Fig. 2c). Each marker consists of a 50 cm long galvanized bar with a resin head and triple extensible anchor, driven into the ground by striking. Another eight topographic nails were drilled in rocky detached blocks in the eastern cliff (Fig. 2d), together with the opening of a vertical crack in the upper escarpment of the same sector, where five control nails were installed (Fig. 2e). The measurements were taken from June 2018 to April 2022, and eleven measurements were made at approximately five-month intervals (Table S-1, supplementary material). A Leica TC-407 total station was used (angular measurement precision = 7 "; distance measurement precision: + 2 mm + 2 ppm) (Fig. 2f). Displacements less than 3 cm are not considered significant considering the terrain configuration and the methodology used.

Results
The results are summarized below, including spatial distribution and geomorphological evidence of cliff slope instability, characterization of the lithology and structure of the area and the temporal evolution of the cliff.

Cliff slope instability: spatial distribution and geomorphological evidence
Geomorphological mapping (Fig. 3a, b) shows the existence of mass movements in 72% of the cliff. Translational landslides are the most widespread process, followed by rockslides and topples, which are mainly concentrated in its eastern part and, in part, in the central sector. In the eastern sector, small flows of debris, debris slides, and rock falls associated with two high angle faults affecting the bedrock are also observed. At the cliff foot, parallel to its base and approximately 300 m in length and 40 m wide, there is a main strip of accumulation of rocky blocks mobilized from different parts of the cliff, which are currently being affected by marine action (Fig. 3a, b). A scour zone, whose morphology responds to an origin linked to wave action, is observed on these deposits, located 20-25 m from the base of the rocky cliff. In the western and eastern sectors of the cliff there are two beach deposits, made up of sand and pebbles; the former is 150 m long and 45 m wide, while the latter covers an area 100 m long and 40 m wide.

Cliff structure and lithology
The geological map shows the general N27°E orientation of the more than 300 m thick quartzite succession that forms the main part of the study cliff (Fig. 3c). The layers range from a few decimetres to 4 m thick and dip between 35° and 60° to the Northwest. The base and top of the quartzite succession exhibit two successions of slate, with a similar orientation to the quartzite and with a dip of between 40° and 80° towards the West-Northwest. The contact between the two lithologies is net. The slate is hundreds of meters thick and occurs in massive form, it being possible to identify the stratification planes (S0) only within sandy layers. In addition, the slate shows a marked tectonic foliation (S1) caused by the orientation of the phyllosilicates and quartz during the Variscan Orogeny.
Five families of discontinuities have been recognized in the quartzite: the bedding (S0) and four families of joints (J1 to J4), whose main characteristics are detailed in Table 2 and Fig. 4. The J1 family is the most abundant, followed by J2 and the other families. In general, the J1 J2 and J3 families include mainly subvertical fractures-oriented NE-E-SE, while the J4 family comprises fractures with NE strike and a mean dip of 42°. Near the eastern contact with the slate, two faults (parallel to the bedding and inclined 65°NW) are observed in the quartzite. In this sector the bedrock shows (in a band of approximately 50 m width) intense fracturing, associated with these structures.
In the slate, the foliation (S1) is the predominant discontinuity, whose disposition is controlled by the folds that affect these rocks; thus, a range of orientations from E-W to NE-SW with dips between 29 and 42° can be observed. Locally, some bedding planes and a family of subvertical discontinuities with NE strike are identified.

Cliff evolution in 2003-2020 based on orthophotographs
The analysis of the sequence of seven orthophotographs corresponding to the period 2003-2020 confirmed the overall stability of the upper edge of the cliff, both in slate and quartzite. Only a single change can be observed in the upper part of the vertical escarpment located in the easternmost part of the cliff, in the contact zone between  2018-11-jul, 2018-29-oct, 2019-18-jan, 2019-26-apr, 2019-12-jul, 2019-05-dec, 2020-18-jun, 2020-18-dec, 2021-24-may, 2021-16-jul, and 2022-29-apr    The poles of each plane and their probability distribution according to von Mises-Fisher (Tanabe et al., 2007) are also shown. The right side shows the field appearance of the quartzite bedrock at the NE end of the cliff and of the slate at EG-1 quartzite and slate (Fig. 5). Between 2003 and 2011, the detachment of a sector 17 m long and 5 m wide, mobilizing an area of 85 m 2 in extension, was observed (red arrows in Fig. 5b). The detached material, mainly centimetre and decimetre-sized blocks, was quickly reworked by marine action. Based on this observation, a maximum retreat rate of 0.47 m·year −1 could be estimated in this small sector for the period 2003-2020. In this zone, the lower part of the cliff was affected by three mass movements: (i) a landslide (300 m 2 ) located in the eastern cliff and first observed in 2011, showing displacements of up to 6 m in 2014 and up to 4 m between 2014 and 2018; (ii) a second movement (198 m 2 ) is located in the westernmost area and had already been observed in the first orthophoto analysed (2003), showing displacements of 7 m over three years (130 m 2 ), and of 6 m in the following five years (68 m 2 ); and (iii) a new movement (140 m 2 ) was observed to the East in 2018. As in the previous case, the detached material (centimetre and decimetre size) was quickly reworked by the marine action that mainly affects the base of the cliff during storms.
In the central cliff, the movement of a rock block that slid 16 m westward following a stratification plane, from 23 to 7 m elevation between 2011 and 2014, is observed (black arrow in Fig. 5c). This block, very resistant to marine action given its size, is now part of the deposit of metric-sized blocks accumulated at the foot of the cliff. Likewise, between 2018 and 2020, centimetre-size blocks were detached from an escarpment located further to the Northwest (Figs. 5 and 6). The detached material remains deposited on the lower part of the cliff and is only affected by major marine storms. At the bottom of the western part of the cliff, translational sliding, and rotation of two blocks (3 and 7 m diameter) was observed between 2014-2020 (yellow arrows in Fig. 5d). Additionally, decimetric rockfalls were observed in the western cliff from the slate escarpments between 2003 and 2020. These blocks are now deposited on top of beach sediments, forming the western boundary of the block deposit at the foot of the quartzite cliff. Finally, during this period, no significant movement of the quartzite blocks located at the cliff bottom was observed, not even of the blocks most directly exposed to marine action.

Cliff evolution in 2018-2020 based on a quantitative comparison between high precision DEM
The comparison made using the M3C2-CloudCompare plugin between high-precision DEMs (8 cm pixel) acquired using RPA in November 2018 and in November 2020 has made it possible to quantify the evolution of the cliff over two years. For this analysis, considering the characteristics of the point cloud and the cliff, as well as the expected movements, a normal diameter of 1 m, a projection diameter of 1 m, and a projection depth of 6 m were used. Almost the entire cliff shows stability during this period, except for small occasional changes. These changes occur in the central and lower part of the cliff (dashed red box in Fig. 7), corresponding to debris deposits detached from a vertical escarpment in 2018-2020 (see Fig. 6), as shown by the sequential analysis of orthophotographs (see Fig. 5). Variations associated with translational sliding of blocks dislodged by marine action between 2014 and 2020 are also observed at the base of the western cliff (dashed white box in Fig. 7; Fig. 5d). Finally, in the western part, slate rock falls are observed (dashed orange box in Fig. 7). In the rest of the studied area, the differences observed between the two DEM correspond mainly to changes in vegetation, isolated blocks and beach deposits.

Cliff evolution in 2018-2021 based on topographic monitoring
The detailed analysis of the seven monitoring campaigns confirms that the measured movements of the control points between July 2018 and April 2022 (Table 3. Complete data in Table S-1, supplementary material; see location in Fig. 3) are within the error range of the technique used, no displacements in the same orientation > 3 cm being observed. Greater movements are only visible at points F1 and F6, which are explained by damage to the control point and by a compression in the level of the topsoil since its installation, respectively. There was also no increase in the opening of the monitored vertical crack (Fig. 2e) located in the upper escarpment-oriented E-W, in the studied period.

Discussion
The influence of the conditioning factors (lithology and structure) on the variability of slope instabilities developed in the Luarca cliffs is discussed here. The spatial and temporal distribution of slope instabilities is also analysed in terms of their contribution to the cliff evolution.

Lithological factor
It is well known that lithology is a major conditioning factor in cliff evolution (e.g., Duperret et al. 2004;Lim et al. 2010). The resistance conferred by the properties of each type of lithology limits erosion caused mainly by wave action (Summerfield 1991;Sunamura 1992;Bird 2000;Duperret et al. 2005;Trenhaile and Kanyaya 2007;Wolters and Mϋller 2004) and abrasion generated by the friction of rock fragments (Trenhaile 1987(Trenhaile 2002Griggs and Trenhaile 1994;Komar 1998). The evolution of the Luarca cliffs is conspicuously controlled by the lithology, which acts in combination with the geological structure. The quartzite shows a relatively high resistance to erosion and the cliff that has been formed in it is mostly adapted to the bedrock structure. The quartzite blocks detached and accumulated at the cliff foot act as a natural defence against marine action (owing to their high abrasion resistance and high angle of internal friction). Direct wave action acts mainly on this block deposit even at high tide, only reaching the cliff base during the strongest sea storms. An incipient scour zone has been detected in these blocks located at the foot of the cliff through geomorphological mapping (Figs. 3a, 5a and 6b). Thus, these deposits are slowing down the cliff retreat, even in the part most exposed to sea storms from the NW. Slate is substantially less resistant to marine erosion than quartzite, which leads to the development of vertical faces and even overhangs. The detached slate blocks and cobbles are easily reworked and eroded by marine action, allowing the waves to impact directly on the cliff bottom; here sandy beach deposits appear.

The role of discontinuities
The distribution of the discontinuities (bedding, faults, joints, and foliation) that affect the bedrock, together with the orientation of the coast, condition the instability mechanisms that affect the Luarca cliffs in a different way. The slate (with a more ductile behaviour against tectonic deformation) shows only three families of discontinuities (S0, S1 and J1), whereas five families of discontinuities were recognised in the quartzite (S0, J1 to J4). In most of the cliff, the position of the quartzite stratification (inclined 30-35° towards the sea) controls the occurrence of translational block slides, the main instability process affecting the cliff (Fig. 8a). Along the cliff there are numerous detached blocks showing slope instability movements that we interpret as episodic according to their distribution by slope (Figs. 5a,  c, 8b and 9a, b). The detached blocks remain for an indeterminate time (depending on several variables) on the cliff slope until they reach the base of the cliff and become part of the deposit at the foot of the cliff. This deposit is already affected by marine action (Figs. 5d and 9a, b). The individualization and detachment of the prismatic rocky blocks is a consequence of the intersection of the discontinuity families (mainly S0, J1, J2), while their size (with an average thickness of approximately 1 m and dimensions that reach 7 × 5 m) is determined by the spacing of each joint family. In the upper part of the cliff, tension cracks with centimetric openings are frequent, delimiting rocky blocks that are in an incipient phase of sliding. Marine action, which in this area mainly involves the action of marine waves coming from the NW and orthogonal to the cliff, acts on the discontinuities, favouring their opening (due to the impact of the waves on the air/water inside) and the progression of weathering. Gradually, the cliff bottom is undermined, and the quartzite layers are undercut, which allows and favours the blocks' sliding. This process, which progresses towards the cliff top and causes its retreat, is slowed down as a result of the natural defence against marine action provided by the accumulation of quartzite blocks at the cliff foot.
In areas where the cliff has a subvertical slope, the intersection between discontinuities leads to the formation of wedges and their subsequent topple/detachment by free fall and, sometimes, planar sliding towards the base of the escarpments (Figs. 5a, b, and 8b). In some sectors (such as the easternmost area), there is a high density of bedrock fracturing, resulting in the formation of centimetre to decimetre-sized blocks that accumulate at the foot of the cliff as result of debris slide and debris flow of small dimensions (< 10-15 m wide). The brecciation observed in the bedrock seems to be related to the existence of two faults (parallel to the strike of the stratification but with a steeper dip, 65° NW); these structures with an original reverse fault movement (according to the drag folds that can be seen in the quartzite layers), act as sliding surfaces favouring the instability that affects the central cliff sector (Figs. 8b and 9b).
In summary, the disposition, continuity and opening of the discontinuities of the bedrock of the Luarca cliffs are factors that control their erosion, which would be much less efficient in a massive or slightly fractured bedrock. The results obtained in the Luarca cliffs are analogous, for example, to studies on coastal cliffs on the Italian island of Favignana (Iannucci et al. 2017), on the Norwegian coast (Braathen et al. 2004), on the Scottish coast (Ballantyne et al. 2018), or in Portugal (Epifânio et al. 2014. Moreover, Donelan et al. (1996), Davidson-Arnott (2010), Limber andMurray (2011), Terefenko andTerefenko (2014) and Gerivani and Savari (2019) reported similar results on the role played by the orientation of the cliff and the massif against the erosive efficiency of marine action. De Lange and Moon (2005) concluded that a higher number of discontinuities causes a higher rate of cliff retreat. However, the discontinuities of the Luarca cliffs play a double role: (i) facilitating the erosion of the cliff and (ii) supplying blocks that form a deposit at the foot of the cliff and which, given their high resistance and metric size, form a natural breakwater protecting the cliff from the marine erosive action of the onslaught of waves.  2003,2006,2009,2011,2014,2018 and 2020 shown on the 2020 orthophotograph. The scour zone is shown as a dashed purple line; (b) Cliff top retreat between 2003 and 2011; (c) y (d) Translational slip trajectories of quartzite blocks between 2014 and 2020 ◂ Spatial and temporal distribution of instabilities: contribution to cliff evolution The Luarca cliffs have not shown a significant retreat between 2003 and 2022, as can be deduced from the analysis of the orthophotographs and topographic studies carried out, although they are affected by local instability processes. As shown by geomorphological evidence, 60% of the cliff surface presents translational landslides, which constitute the main erosion mechanism. This process affects the centralwestern half of the studied sector, where the direction of the layers (NE-SW) is subparallel to the coastline (perpendicular here to the main component of the prevailing storms, NW); these layers dip 30-35° towards the sea. Rockfalls are the main erosive processes within vertical slope sectors and, consequently, they are concentrated in the eastern and westernmost parts of the studied cliff (affecting 12% of the cliff), although they are also observed in the middle and upper part of the central section. At the eastern cliff foot small debris flows/debris slides affected the detached materials, which occupy an area of approximately 500 m 2 .
The negligible and localized data of retreat obtained in the studied period agree with the values obtained for rocks with similar resistance to quartzite (> 20 MPa) by Sunamura (1983Sunamura ( , 1992 or Tsujimoto (1987), who determined rates of recession ranging from 0.001 to 0.01 m·year −1 . In three sectors of the west coast of the Iberian Peninsula, similar low values were obtained on rocky coasts, including: 0.2 m·year −1 for two years (Anfuso et al. 2007(Anfuso et al. ), and 0.017 m·year −1 (1999(Anfuso et al. -2003 and 0.014 m·year −1   (Marqués 2006 Neves (2008) and Nunes et al. (2011) on the West coast of Portugal, over several decades. These data confirm the important role played by local factors in the retreat of rocky shores.
As might be expected in a rocky cliff, all these mass movements are discontinuous in space and intermittent  Sunamura (1992), Bird (2000), or Young et al. (2009). This episodic and sudden character means that these movements escape the resolution of the land surveying techniques used in this investigation in the period of 4 years. However, this does not invalidate the application of these techniques, used successfully by other authors (Wangensteen et al. 2007;Lateh et al. 2011;Katz et al. 2011;Salvini et al. 2013;Janeras et al. 2015, Swirad andYoung 2021, among others) to characterize millimetric and centimetric movements on rocky slopes, although it is necessary to address broader study periods. Thus, the comparative analysis techniques of orthophotos and DEM acquired by means of RPA corresponding to different moments in time have been revealed as more efficient in this case.

Contribution to cliff retreat research
The results obtained in this research contribute to the understanding of the instability mechanisms involved in the retreat of cliffs in hard rocks. They also contribute to understanding how local factors (mainly lithological, structural and geomorphological features) condition an episodic and sudden occurrence of instability processes (as opposed to periods of long-term stability) that affect rocky cliffs as opposed to more continuous processes, expected in homogeneous and more easily erodible lithological beds. The structural configuration and lithological nature of the bedrock play a dual role, (i) favouring the geomorphological activity that causes the cliff to retreat and (ii) slowing down marine action by providing resistant blocks of rock that form a deposit at the foot of the cliff that acts as a conservative barrier. This defensive role, conditioned by the amount of detached material, the size of the blocks and their angularity, has been studied by Trenhaile (1989Trenhaile ( , 2016 and Moses and Robinson (2011), among others. The residence time of the detached blocks on the slope and the time of reworking and erosion of the block-deposits that protect the cliff are some of the questions to be answered by future research.

Conclusions
This work has shown that the 570 m-long quartzite and slate cliffs studied in Luarca show high geomorphological activity and abundant recent evidence of instability throughout their extension. Local lithological and structural factors play an important role in the cliff evolution, which, despite being affected by intense geomorphological activity, does not show a generalized retreat in the period 2003-2022. The presence, orientation, and characteristics of the different families of discontinuities  landslide movement in favour of two faults (parallel to the strike of the stratification but with a steeper dip) that involves the central zone of the cliff. The J1 family has an arrangement roughly parallel to the scheme definitively condition the instability mechanisms that develop in the cliff, as well as the size of the generated blocks. Slate, affected by a smaller network of joints than quartzite, gives rise to more vertical cliffs and, given their weak nature, the detached blocks are quickly reworked and eroded by the sea, so there are no accumulations of blocks at the bottom. With respect the quartzite, structural factors condition the appearance of individualized blocks in favour of planes of weakness that are mobilized through different instability mechanisms. The blocks that reach the foot of the cliff form a deposit that acts as a natural defensive element against the erosive action of the waves and, consequently, slows down the retreat of the cliff. The time of permanence of the detached blocks on the cliff slope and the reworking and erosion of the quartzite deposits greatly exceeds the period addressed in this research, this being one of the unknowns to be cleared up in future studies.
Funding Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. The research leading to these results received funding from (1)   Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.