Megatsunamis Induced by Volcanic Landslides in the Canary Islands: Age of the Tsunami Deposits and Source Landslides

: Evidence for frequent, large landslides on the ﬂanks of the volcanic ediﬁces forming the Canary Islands include outstanding landslide scars and their correlative submarine and subaerial rock and debris avalanche deposits. These landslides involved volumes ranging from tens to hundreds of km 3 . The sudden entry of large volumes of rock masses in the sea may have triggered tsunamis capable of affecting the source and neighboring islands, with the resulting huge waves dragging coastal and seabed materials and fauna and redepositing them inland. Here, we present new geological evidence and geochronological data of at least ﬁve megatsunamis in Tenerife, Lanzarote, and Gran Canaria, triggered by island ﬂank megalandslides, and occasionally explosive eruptions, during the last 1 million years. The exceptional preservation of the megatsunami deposits and the large area they cover, particularly in Tenerife, provide fundamental data on the number of tsunami events and run-ups, and allow proposals on the sources and age of the tsunamis. Tsunami run-up heights up to 290 m above coeval sea level, some of the highest known on Earth in recent geological times, were estimated based on sedimentological, geomorphological, paleontological, and geochronological data. The research results made it possible to estimate the recurrence of tsunamis in the archipelago during the last hundreds of thousands of years, and to establish relationships between tsunami deposits and the probable triggering island ﬂank landslides.


Introduction
Megatsunami waves higher than 40 m have been attributed to voluminous subaerial and submarine landslides (particularly to volcanic islands giant flank landslides), explosive volcanic eruptions, and asteroid impacts. The most outstanding documented examples of megatsunamis are those caused by an earthquake-triggered rock avalanche in Lituya Bay [1], Alaska, 1958, generating the highest wave run-up in recorded history (525 m); the Storegga submarine slide [2,3], Norwegian Sea, 8000 y BP; volcanic island flank megalandslides in Hawaii [4][5][6][7], Canary [8][9][10][11] and Cabo Verde archipelagos [12][13][14], throughout Another outstanding case (at a much smaller volume) is the most recent flank landslide of the island of Stromboli (Aeolian archipelago), dated at 5-6 ka, and involving a volume of 0.7 to 1.8 km 3 [29,30]. Outcrops of volcanoclastic breccias on the SE flank of the island, extending to an elevation above 120 m above sea level (a.s.l.), are tentatively attributed to a giant tsunami generated by the so-called Sciara del Fuoco landslide [30].
In historical times, the largest volcanic landslides occurred on island arc volcanoes, characterized by rapid failure and tsunami generation, with volumes <3 km 3 (magnitudes significantly smaller than prehistoric events on intraplate volcanic islands, as is the case of Hawaii and the Canary Islands) and a recurrence of~100 years [31]. The most recent and largest landslide occurred in Ritter island (Papua New Guinea) in 1888 involving a volume~2.4 km 3 [32], which generated a tsunami with observed maximum run-up heights of 15 m on nearby shores [33]; this event ranks among the most disastrous volcanic events in Papua New Guinea in modern times and caused more than 1500 casualties [34]. At Oshima-Oshima, a small volcanic island in Japan Sea, a landslide of similar volume [35] in 1741 generated a tsunami with maximum wave heights of 10 m, being the most destructive tsunami ever originated in the Japan Sea. Both are the only two large volume (>1 km 3 ) landslides to have occurred on island volcanoes, according to historical and written records [35].
The only documented historical megatsunami caused by volcanic island flank failure occurred at Mount Unzen (Kyushu island, Japan) in 1792, involving a volume of~0.34 km 3 , producing wave heights up to 25 m in the nearby shoreline [36,37], the worst and largest volcanic disaster in the history of Japan, causing more than 15,000 deaths. The maximum tsunami run-up height reached 55 m [38,39] due to the particular topography of the coasts. Approximately 80 years later, in 1871, the landslide of Mount Ruang (Indonesia) caused a tsunami with waves reaching up to 25 m in height [40].
Recently, in 2018, the failure of the flank of the Anak Krakatau volcano (Indonesia) mobilized a volume of 0.22-0.3 km 3 (one order of magnitude smaller than those of Ritter and Oshima-Oshima), producing a tsunami with run-ups of up to 13 m on the adjacent coasts that caused more than 437 fatalities, the largest death toll from a volcanically induced tsunami since the catastrophic eruption of Krakatau in 1883, and the flank landslide of Ritter Island in 1888 [41].
Thus, even relatively 'small-volume' flank landslides (<1 km 3 ) have caused destructive tsunamis in historic times, emphasizing the importance of a tsunami hazard associated to this type of event. Landslide tsunamis depend largely on the failure of a submarine portion of the volcanic flank, which may involve much greater volumes of rock than the subaerial portion. In the outstanding cases of Oshima-Oshima and Ritter islands, the landslides were mostly submarine (~2 km 3 in the first case [42], an order of magnitude larger than the subaerial portion), probably due to the low elevation of the islands, with maximum heights of 700 and 780 m a.s.l., respectively. In both cases, there is evidence of the occurrence of a single large landslide, but not of sequential multiple landslides [42,43].
Although not comparable in volume, the typology, failure mechanism, and subaerial and submarine morphological features observed in recent small-scale volcanic island landslides are similar to prehistorical megalandslides in Hawaii or the Canaries, which produced proximal tsunami run-up heights of tens to hundreds of meters on neighboring coastlines [4][5][6][7][8]11].

Flank Megalandslides and Tsunami Deposits in the Canary Islands
The Canary Islands ( Figure 1) were affected in recent geological times (the last 1 million years) by relatively frequent large flank landslides [44][45][46][47]. The destruction of large portions of volcanic edifices, not necessarily associated with explosive eruptions, is part of the geological evolution of the archipelago. Some younger islands (<12 Ma), such as Tenerife, La Palma, and El Hierro, present subaerial and submarine conspicuous geomorphological features, evidencing large recent landslides [48][49][50][51][52]. Intense volcanic activity caused rapid growth of the volcanic edifices in a few million years, reaching thousands of meters a.s.l., as in the cases of Tenerife (3715 m) and La Palma (2426 m). The near-limit equilibrium conditions of the flank slopes in these islands together with the low strength of the materials that underlie the emerged portion of the volcanic edifices, added by volcanic processes acting over time, combine to cause flank instability and trigger huge rock mass failures, probably some of the largest on Earth. Such landslides and their sudden and violent entry into the ocean must certainly have triggered tsunamis that struck the coasts of the source and nearby islands. Such huge tsunami waves bear the potential to drag seabed and coastal materials and fauna and to redeposit them inland during extensive and cataclysmic inundations. As a considerable number (at least 10) of large landslides occurred during the last 1 million years in the Canary Islands, megatsunamis should also have been relatively frequent. Deposits attributed to megatsunamis were described for the first time in the island of Gran Canaria [8]. Marine deposits on the island of Lanzarote were studied by [53], suggesting a high-energy emplacement mechanism, compatible with a possible tsunami. For the fossiliferous deposits from Teno, in northwest Tenerife, an origin related to a punctual and momentary sea level rise or waves was suggested [54], based on their particular characteristics.
In the GRANDETEN research project, on the causes and mechanisms of the Canary islands megalandslides carried out by the authors of this paper [55][56][57], the megatsunami deposits of Tenerife and Lanzarote were described and identified as tsunami deposits, and new sites were described in Tenerife and Gran Canaria [9,10,58,59]. The excellent preservation condition of these deposits and their distinctive sedimentological and paleontological features with mixed marine and terrestrial fauna, constitute an exceptional source of information on megatsunami deposits. The characteristics of the sediments indicate a high energy source (they were deposited at tens-even hundreds-of meters above present sea level) and a high speed mechanism, suggesting a relation with the large flank landslides that affected the Canary Islands during the Pleistocene, with giant rock avalanches suddenly entering the sea [10].
Recently, Teno deposits were also described [11] and data on the megatsunami deposits in the Canary Islands were included in review and compilation articles [60,61].

Materials and Methods
This work deals with marine conglomerate deposits occurring in three of the Canary Islands, describing the characteristics that undoubtedly allow their attribution to a tsunamigenic origin and their relation to the geologic evolution of the islands, namely the occurrence of large flank landslides during the last 1 million years as a source of the megatsunamis. One key goal was to determine the age of the tsunami deposits to relate them to the flank failure events.

Field Work
Marine chaotic conglomerate outcrops were analyzed, described, and photographed in the field, and logs and sections were drawn. A detailed map of the tsunami deposits from the Teno platform was produced in the field. Outcrop coordinates were obtained with a hand-held GPS and altitude retrieved from detailed topographic maps and an altimeter. The stratigraphy, texture, and structure of the deposits were recorded, and both sediment and fossiliferous content were sampled. Special attention was paid to the stratigraphic relation with underlying or intercalated formations, which correspond to terrestrial sediments, paleosols, or subaerial volcanic units, and to the nature of contacts between different geologic units. The relation with the local geomorphology was also considered.

Amino Acid Racemization/Epimerization (AAR) Dating
The AAR method is based on the racemization of amino acids preserved in fossilized biominerals, in this case from Glycymeris shells abundant in the tsunamigenic conglomerates. Through time, L-amino acids racemize or epimerize to their D-isomer form; the ratio D/L measures the extent of epimerization.
Glycymeris shells were collected in the four sites to establish their age. The number of samples from each locality is shown in the Supplementary Data. A total of 105 specimens were collected, grouped by locality and collection point, of which 23 were rejected for dating because of contamination by modern amino acids.
For AAR, a hollow diamond drill was used to remove a discoid sample (8 mm in diameter) from an area close to the beak of the shells. Peripheral parts with visible weathering (approximately 20-30%) were removed after chemical etching with 2 N HCl. Afterwards, 10-20 mg of samples were taken.
Amino acid concentrations and D/L values were quantified in the Biomolecular Stratigraphy Laboratory (Madrid, Spain) by high performance liquid chromatography (HPLC) following the sample preparation protocol described in [62]. Samples were injected into an Agilent-1100 HPLC equipped with a fluorescence detector. Excitation and emission wavelengths were programmed at 335 nm and 445 nm, respectively. A Hypersil BDS C18 reverse-phase column (5 µm; 250 × 4 mm i.d.) was used for the separation.
The obtained mean D/L values of the five amino acids (isoleucine, aspartic acid, alanine, valine, and glutamic acid) in Glycymeris shells are presented in the Supplementary Data.

Thermoluminescence Dating (TL)
Samples of calcareous crusts were collected at La Aldea site, in Agaete, on calcrete crusts developed upon the tsunami deposit, with the objective of determining their absolute ages using the thermoluminescence technique and the additive dose method in fine siltsized quartz particles incorporated in calcrete lining of clasts. The age is determined by means of measuring the accumulated radiation dose, of the time elapsed since sediment containing crystalline minerals was exposed to sunlight. Four samples were collected corresponding to two different calcrete levels, at 74 and 71 m above sea level. The measures and results are presented in the Supplementary Data.

U-Th Dating of Coral Fragments
Fragments of corals from the tsunami deposits at Teno and Piedra Alta were collected for dating purposes. Subsamples for U-series age determinations were milled from sections of the corals. All samples were (previously) mineralogically and geochemically analyzed, including X-rays, to determine their compositions in the laboratories of the Geological and Mining Institute of Spain, IGME.
Three different groups of analysis were carried out. Two included 6 + 10 samples from Teno and Piedra Alta, respectively, at the laboratories of the Geological Survey of Israel, and a group of 3 samples from Piedra Alta at the laboratories of the Institute of Earth Sciences Jaume Almera, Spanish Research Council (CSIC).
As corals might have suffered diagenetic alteration and opening of the U-Th system, the analysis consisted of analyzing several samples from the same coral; a sequential leaching of the samples was performed to dissolve the bulk aragonitic (or aragoniticcalcitic) coral in weak acid and then redissolved the insoluble residue. For the first group of corals, the measured activity ratios of the individual samples could not provide reliable ages; the same happened for the corals from Teno of the second group. For the corals from Piedra Alta, when looking on the data together, an age of~120 ka can be assumed, which can be interpreted as the time of diagenesis of the corals by groundwater. The measures and results are presented in the Supplementary Data.

Results: Description of the Deposits
Megatsunami deposits that are recognized in several field locations of the Canary Islands share the following common features: • Sediments correspond to fossiliferous heterometric breccias and conglomerates, both matrix and clast-supported, displaying massive structure and chaotic fabric; clast sizes vary from 1 cm to more than 1 m; • They are poorly sorted, with a mixture of angular and subrounded clasts, sometimes showing crude normal or reverse grading; frequently two or more layers can be distinguished; • The sediments present marine or mixed (terrestrial and marine) facies and in all cases overlie terrestrial deposits (paleosols, colluvium, alluvial sediments, eolian sands, subaerial basaltic lava flows), supporting a subaerial emplacement above coeval mean sea level; • A sharp and erosive character of the unconformity at the base of the tsunamigenic layers; • The deposits overlie elevate lava platforms or are plastered onto irregular subaerial sculptured slopes, incompatible with long-term marine abrasion related with former sea level high-stands, as would be expected on raised beach deposits; • A polygenic composition of conglomerates that includes marine-sourced lithic and biogenic elements (rounded beach gravel and boulders, blocks of beachrock, marine bioclastic sand, and fragments or whole marine fossils), subaerial elements (e.g., sub-angular clasts of basaltic nature and terrestrial fossils, including pulmonate gastropods), fragments of unconsolidated materials dragged from the flooded surface, and rip-up clasts of paleosols; • Marine fossil content decreasing in abundance landwards, where terrestrial fossils (including bird or reptile bones) may occasionally be found mixed with marine shells; • Marine fossils corresponding to organisms living in a wide depth range, from the shallow littoral to more than 200 m depth (bivalves, gastropods, bryozoans, rhodoliths, corals, and echinoderms); the fossils, frequently broken and lacking surface re-working, were never found in living position; • Presence of imbricated large and small clasts, indicating seaward and/or landward orientations.
Elevation of the deposits ranges from 2 to nearly 200 m above present-day mean sea level (m.s.l.), providing grounds to estimate the run-up of coeval inundations. Maximum observed distances of the deposits from the present-day coast up to 1.5 km, with large blocks weighing more than 1.5 ton found higher than 50 m.s.l., and more than 500 m inland.
In the following sections, the specific features of the tsunami deposits in each location are described.

Teno Lava Platform (Tenerife)
Teno deposits are located on the Teno coastal platform, at the NW corner of Tenerife ( Figure 1). The platform corresponds to Pleistocene lava deltas, and offers the flat morphology and space required for deposition and preservation of the materials washed onshore by tsunami waves. These deposits were described in the literature as Pleistocene conglomerates and cemented marine sands, overlying a quaternary marine terrace at 18 m a.s.l. [63]. However, their peculiar sedimentological and compositional characteristics suggest an origin related to punctual and momentary sea level rise or wave action, possibly linked to catastrophic phenomena derived from intense volcanic activity [54].
The Teno fossiliferous detrital sediments ( Figure 2) crop out from about 2 m at the coast up to 60 m a.s.l. further inland. They extend over 1.3 km 2 , their landward limit lying about 500 m, locally up to 800 m, from the coastline. They are overlain, largely, by a 60-70 cm-thick pedogenized red alluvial fan deposits.
The lava delta and overlying deposits form a poorly dissected surface sloping 3-5% seaward. In places, the original morphology of the lava flows is well preserved and clinckery lava crests crop out from the sedimentary cover. The main stream draining the area is Barranco de Itobal, where the fossiliferous conglomeratic deposits clearly filled an incised and deep alluvial ravine that should have been around 3-4 m deeper than present at the time of tsunami sediment deposition. Subsequent stream re-incision and slope erosion partly removed and truncated the sedimentary infill. The preserved deposits crop out in two sections along both the left and right banks of the creek.
Tsunami sediments present an irregular and patchy distribution ( Figure 2), resulting from variable exposure thickness determined by erosion of the alluvial cover. In addition to the common features described above, the Teno deposits show the following characteristics The lower unit is composed exclusively of angular and less abundant rounded basaltic clasts from the underlying lava deltas arranged in an open-fabric, largely clast-supported texture with scarce sand matrix ( Figures 3C and 4B). The upper unit is composed of finer, but highly heterometric angular fragments of basalt, syenite, phonolite, orange-colored hydrothermally altered lithic clasts, and pumice, embedded in a light-colored ash matrix ( Figure 3D). Rounded clasts are absent. In places, the upper unit is less varied and entirely composed of brownish to yellow volcanic ash, containing dispersed angular clasts of fibrous pumice; • Both units contain fossils of marine and terrestrial origin, including fish bones, vermetid tubes, mollusks, corals, echinoderms, bryozoans, rhodoliths, terrestrial gastropods, and one lizard bone. The marine fossils observed in the breccia include whole shells and fragments of bivalves, with abundant disarticulated valves of Glycymeris; fossils are more abundant in the upper unit ( Figure 3E-I), and in both units their abundance decreases landwards. In general, thick shells are more commonly found near the base of the breccias, where fossils are also more abundant; • Both the lower and upper unit display normal grading. The clasts are mostly of small size (≤10 cm), although some elements with long axis reaching 40-50 cm occur dispersed within the sediment. In places, the deposit presents ill-defined lamination roughly parallel to the cross section of the gullies; • Clasts in the upper unit are embedded in a pyroclastic sand-sized matrix presenting some consolidation. In places, flatter clasts are imbricated in agreement with seaward flow, but less frequent and ill-defined landward flow imbrications were also noted. Largest clasts concentrate closer to the thalweg of the gullies; • Where the fossiliferous deposits filled deeply incised gullies, as in Barranco de Itobal (Figure 2), the breccias are coarser and the deposits increase in thickness up to 1.8 m.
The overall structure suggests rapid emplacement of 4 to 5 stacked depositional units, with a very short time interval between successive depositional pulses; once more, this is confirmed by no marked discontinuities separating the superposed units ( Figure 4B). The lower unit (20-40 cm thick) is composed exclusively of angular basalt clasts, with few bioclastic sand matrix; the upper units are similar in composition to those of the Teno platform. At this site, the conglomerates overlie whitish fine and well sorted sands (climbing dunes?) that grade laterally to a pedogenized semi-consolidated brownish sand, which is herein interpreted as representing a paleosol originally developed in eolian sediments (either dune or interdune deposits) ( Figure 4C); • Several large rounded basalt blocks (up to 1.7 m in diameter) stand isolated on the platform; they sit at 40 m a.s.l. and 560 m landward from the present-day coastline and >200 m from the toe of the paleo sea-cliff limiting the lava platform to the east, suggesting inland transport from the coast.

Isla Baja Lava Platform
Isla Baja area is located in NW Tenerife coast, to the east of Teno platform (Figure 1), and separated by tall coastal cliffs cut in Miocene lavas. As in Teno, an extensive littoral platform developed in relation with subaerial lava flows, provided a low and flat surface that favored deposition and preservation of materials washed inland by tsunami waves. The platform is intensely modified by human activity, mainly for agricultural and urban development, and the tsunami deposit can only be observed at some coastal sites and occasionally further inland. Texture, composition, and depositional architecture of both Teno and Isla Baja deposits are strikingly similar, just as their distribution in elevation and location at the coastal ribbon. These characteristics and the proximity to Teno provide grounds to infer that the deposits of both areas are correlative.
The main outcrop is located in Playa de las Arenas ( Figure 5) on the western edge of the platform at 3-4 m a.s.l. and is exposed on a vertical road cut some tens of meters long. The surfaces upon which the conglomerates rest correspond to subaerial lava delta morphologies, slightly modified by subaerial erosion. The main characteristics of the event-deposits are listed below: • Tsunami deposits are up to 2.5 m-thick and are covered by sandy-clayey colluvium, containing abundant terrestrial fossils (pulmonate gastropods); • The deposit consists of two stacked conglomeratic layers ( Figure 6A,C). Both layers contain marine and terrestrial fossils (bird bones), and are separated by an undulating, locally erosive, surface, without any evidence of a significant time gap; • The lower layer is at least 2 m thick and highly heterometric, with scarce sandy matrix ( Figure 6A). It is composed exclusively of basalt clasts entrained from the underlying lava flows and is characterized by a chaotic fabric. The dominant clast size range is 5 to 10 cm, occasionally reaching up to 1 m; numerous large and elongated blocks standing in vertical position were observed. This layer shows the largest (meter-sized) elements on top, suggesting crude reverse grading. The fossil content includes fragments of marine gastropods and bivalve shells (such as disarticulated valves of Glycymeris) and ahermatypic corals; • The top layer consists of a 1.5 m-thick breccia containing clasts of the same range and lithologies found in the upper layers of the Teno outcrops: basalt, syenite, phonolite, orange-colored hydrothermally altered debris, and pumice. A small number of flat clasts present imbrication indicating seaward flow. Clasts are much finer than in the underlying layer, predominantly in the 1-5 cm size-range, and only occasionally reaching 20 cm ( Figure 6B). The strong contrast in modal particle size observed immediately below and above the surface separating the lower from the upper layers, together with the erosive nature of that surface, highlights the signature of two distinct depositional episodes.  Other outcrops of breccias containing marine fossils, showing compositional characteristics shared with the Playa de las Arenas upper layer, were identified both near the coast (at Lomo de las Campanas and El Rayo sites, at~2 km and~3 km, respectively, to the east of Playa de las Arenas outcrop, where the deposits present the same characteristics; Figure 5), and on scattered outcrops further inland. Field observations show that the tsunami deposits extend for about 7 km along the coastline, and reach more than 1.5 km inland where they stand at elevations as high as, or higher than, 150 m a.s.l. At some points on the platform, where soil and vegetation cover has been removed by erosion, or small slopes have been excavated by human action, there are outcrops of the upper heterolithic layer, in some cases containing marine fossils (Glycymeris valves). However, except at Cenizales ( Figure 5), neither the complete section nor the basal layer are entirely exposed. At the highest observed points (>150 m), Glycymeris shells were not found, and the massive heterolithic deposits exhibits a reworked and re-deposited appearance, indicating a different origin from the subaerial volcanic deposits with similar lithological content present in the area (the so-called blast deposit).

Piedra Alta (Lanzarote)
Tsunami deposits at Piedra Alta, on the SW coast of Lanzarote (Figure 1), stand at 15 to 21 m a.s.l., overlying an elevated coastal platform formed by lava flows. The marine conglomerates were firstly described as a sequence of uplifted marine levels [64], and later assigned to an abrupt and high-energy emplacement mechanism, compatible with a tsunami violently striking this coast during an interglacial period [53]. The deposit extends for several kilometers landward [65], and is partially covered towards the north by a subsequent lava flow. The sediments are well-exposed in several outcrops scattered along the coast for 2.5 km, the longest section being about 200 m ( Figure 8A); the deposits can also be observed at the bottom and lateral slopes of some small gullies, up to 15 m inland.
The Piedra Alta's most extensive and representative outcrop is 4 m thick ( Figures 8B  and 9A,C). Aeolian sand and a red clayey paleosol containing abundant terrestrial fossil fauna, such as insect oothecae and terrestrial gastropods, occur sandwiched between the tsunami deposits and the basement lava flow ( Figure 9C,D), indicating subaerial deposition.
The deposits consist of chaotic breccias with angular, frequently broken, clasts of basalt and abundant lithified bioclastic sand matrix. Clast sizes range from 1 cm up to 1.5 m, with blocks larger than 0.5 m predominating in the upper part of the deposit, many of them fractured. Some of the large and flat-shaped blocks are imbricated or oriented, indicating landward flow ( Figure 9B); some blocks are covered by a thin layer of marine sand ( Figure 9E), indicating a source other than the deposit itself. Fragments from the underlying red colored paleosol (rip-up clasts) are present among the clasts in the lower and coarser layer.
In general, two layers of conglomerates can be clearly distinguished, with marine and terrestrial fossils embedded in a sandy-silty cemented matrix. Between the two layers there is a lenticular one composed of a high proportion of lava blocks and scarce matrix. Fossils are abundant, and mainly occur in the cemented conglomeratic layers. No significant variation in their distribution or content has been observed landwards. A mixture of marine and terrestrial fossils is present, never observed in living position, including a remarkable abundance of coral fragments of various species. In addition, marine fossils range from deep-sea to infralittoral species. Besides the groups of fossils common to other deposits, the marine fauna also includes brachiopods and crustaceans. Part of the fauna corresponding to the littoral depth-range (corals and bryozoans, Figure 9F,G) is typical of a warm interglacial period and a paleoclimate similar to the present [53]; moreover, some mollusc species in these sediments have never been described in the Canary Islands, but are presently found on the Atlantic coasts of Africa (Senegal, Guinea Gulf), southwards of the Canaries [53].

Agaete (Gran Canaria)
Chaotic marine conglomerate deposits, lenticular in shape and up to 3.5 m thick, occur as several isolated and scattered outcrops within the Agaete valley, at the NW of Gran Canaria (Figures 1 and 10). This broad and steep-sloped valley was excavated in Miocene times and its bottom was partially filled by Pleistocene lava flows. Conglomerates crop out from 40 to 188 m a.s.l. and have been described and assigned to a tsunamigenic origin [8]. Agaete tsunami deposits were revisited and provided new data on the sediments [9], including two previously unknown outcrops of tsunamigenic deposits to the existing data set. Furthermore, evidence for the presence of three distinct tsunami inundations was found.
At Llanos de Turman and Berrazales sites (Figure 10), field evidence indicates the presence of two distinct tsunami deposits. The marine conglomerate layers are separated by subaerial deposits (colluvial and fluvial conglomerates with pedogenized tops). The development of the paleosols on these terrestrial sediments indicates that a significant time interval separates the deposition of the tsunamigenic sediments, confirming the presence of two distinct events on those outcrops. At Llanos de Turman, the structure of the upper marine conglomerate varies laterally. In the center of the outcrop, it is composed of two units: a finer lower unit and a coarser upper unit both reversely graded. Although a well-defined discontinuity separating the two units is absent, there is a sudden textural change marking the transition between the two superposed sedimentary layers highlighted by the deposition of secondary pedogenic carbonates. To the west, the conglomerate is coarser and a third unit is present at the top of the sequence, which is dominantly composed of rounded and finer boulders supported by a sand matrix. The fossil content is abundant including bivalves (Venus, Glycymeris), gastropods (Mitra, Bolma), and rhodoliths. Fragments of fossiliferous beachrock are also present. This marine conglomerate covers a torrential deposit whose top is pedogenized. This 50 m-long lenticular torrential deposit is truncated to the east by the overlying tsunami deposit; here the upper marine conglomerate stands directly on top of the lower conglomerate. The contact between the two is marked by carbonate incrustations. The lower deposit is very rich in large (meter-sized) rip-up clasts of paleosol developed on the underlying Pleistocene volcanic sequence ( Figure 11F).  At the La Ruina site, the upper deposit (t3; see Figure 11B) is a 3.2 m-thick chaotic conglomerate presenting reverse grading; the lower part of the deposit displays a clast supported texture, while the upper 2 m are coarser with clasts up to 70 cm in diameter dispersed in a coarse sand matrix ( Figure 11D). The clasts are basaltic, and the white sandy matrix is biogenic, although intensely calcretized; fossils are rare and include marine gastropods and bivalve shell fragments. Although there is no evident discontinuity in the deposit, the texture variation may correspond to the two layers observed in the near-by road-cut outcrops of La Aldea (28 • 05 46 N-15 • 41 57 W; Figure 11A). The outcrop of the upper marine conglomerate, which is certainly correlative of La Aldea sediments, stands at an elevation 10 m lower than the highest 1.8 m-thick road outcrop. It is thicker and much coarser (both in terms of clasts and matrix) and may represent an up-slope (and landwards) fining and thinning of the same depositional unit. This conglomerate stands on an 80 cm-thick pedogenized colluvium deposited on top of another 3.5 mthick marine chaotic conglomerate (t2; Figure 11B). This layer is finer (clasts 20-50 cm in diameter) and the pedogenized biogenic sand matrix presents a yellowish color. The fossil content is rare and includes fragments of bivalve shells. One distinctive feature of this conglomerate is the presence of abundant rip-up clasts of the friable underlying colluvium (similarly to the lower tsunami deposit of Llanos de Turman). Laterally, this marine conglomerate displays two units: a finer lower unit, covered by a coarser unit (this part of the outcrop is inaccessible). This marine conglomerate rests on a 3.5-4 m-thick sequence of fluvial deposits composed of a thick mudflow (a muddy brownish conglomerate with floating large clasts, presenting a paleosol developed on top) that overlies a lenticular thinly stratified sandy stream flow deposit. Both fluvial deposits cover (and partially erode) a paleosol developed on top of another marine sequence (t1; Figure 11B). This is composed of two layers: a lower conglomerate with a white sandy matrix containing marine fossils, overlain by a homogeneous sandy layer apparently without fossils. This marine deposit is lenticular in shape and extends horizontally for 12 m. All three deposits display evidence of erosive bases represented by large rip-up clasts of friable material (deposits t2 and t3; Figure 11B), by erosive truncations of the underlying units (t1 locally truncates t2 and the overlying colluvium), and by downwards injected clastic dikes on their bases (t2 and t3; Figure 11B). One of the dikes is planar and strikes 306 • , an azimuth that is probably coincident with the inundation direction, which is certainly controlled by the general trend of the valley. These clastic dikes (also observed on the base of t3 on the road outcrops) can penetrate up to 6 m into the underlying fluvial layers.
The remains of another marine conglomerate outcrop were found on the road access to the Agaete graveyard (El Cementerio site: 28 • 06 13.2 N-15 • 42 11.8 W, 49 m a.s.l.; Figure 10). This outcrop shares the characteristics of every other tsunami deposit of Agaete and covers a 1.75 Ma-old basaltic lava flow (see below for data on ages of event-deposits and lava flows). The sandy matrix is yellowish, resembling the t2 deposit of La Ruina outcrop, and the deposit contains rare marine fossils (rhodoliths).

Age of the Tsunami Deposits
The ages of the different tsunami deposits here described in Tenerife, Gran Canaria, and Lanzarote are discussed below using new data obtained in this study together with previously published ages of lava flows and fossils. Numerical and relative dating of the tsunami deposits was carried out by different methods, summarized in Table 1 and Figure 12. The most consistent ages were obtained by amino acid racemization/epimerization (AAR) and thermoluminescence (TL), in contrast with the U/Th radiometric dating in corals, which strongly depends on the diagenetic evolution and mineral transformation of the fossils. The overall results indicate emplacement of the tsunami deposits during the middle Pleistocene: younger than 168 ka in Teno and Isla Baja; <200 ka in Agaete; and <218 ka in Piedra Alta.
The ages of the lava flows making up the deltas of Teno and Isla Baja in Tenerife, underlying the tsunami deposits, were dated as 178 ± 6 ka and 194 ± 8 ka, respectively [11,66], which provide a maximum age limit for the emplacement of tsunami sediments. AAR performed in Glycymeris shells from the fossiliferous conglomerates in Teno yielded ages of 145 ± 23, 141 ± 27 and 126 ± 17 ka that are consistent with the lava flow ages. The ages correspond to three sets of shells collected at different moments with 15, 7, and 11 samples respectively. At Isla Baja, the AAR results from four different localities (with 2, 6, 4, and 6 shell samples) yielded ages within a similar interval, between 141 ± 3 ka and 119 ± 13 ka, except for the age of 74 ± 2 ka from shells collected at Cenizales-2 site, an 'anomalous' value, since another shell from the same area (Cenizales-1 site) provided an alternative age of 123 ± 8 ka. Considering the uncertainties, the ages obtained by AAR in both sites in Tenerife are ≤168 ka. We analyzed two sets of six coral fragments, each from the deposits at Teno by U-series dating method. Unfortunately, no real ages could be obtained, as the corals underwent diagenesis to calcite, influencing the results, so useful slopes on the "property-property" diagrams could not be obtained [67,68].
At Lanzarote, the tsunami deposits from Piedra Alta lie on lava flows with radiometric age of 820 ± 160 ka [69], and are overlain inland by lava flows dated at~196 ka [53] and 160 ± 20 ka [64]. Based on paleontological and paleoclimatic criteria, they were attributed to Marine Isotope Stages MIS 11 or MIS 13, between~365 and~500 ka, based also in a non-decisive U-series age of coral of 481 ± 39 ka [53,70]; later, the deposits have been attributed to MIS 11c, 400-410 ka [65]. Although a wide age range is represented between 820 ka and 160 or 196 ka, both are compatible with new ages of 218 ± 3 ka and 181 ± 27, obtained by AAR in Glycymeris shells (two groups of seven samples each) from the tsunami deposit, but these ages are far from the 400-500 ka, to which the paleontological criteria point out ( Table 1). As in the case of Teno, we analyzed two sets, 10 fragments of coral each, collected from the deposit at Piedra Alta for U-series. Unfortunately, most of the coral fragments did not provide reliable ages, as they probably correspond to open systems. They show 234 U/ 238 U isotope ratios that indicate interaction with freshwater, and provide no reliable age information. Only the second set provided ages ranging from 136 ± 6 to 358 ± 48 ka [67,68]. Other additional U-series dating analysis performed for this study provided two radiometric ages of 193 ± 18 ka and >350 ka (Table 1), the former in the same range as those obtained by AAR, but again, far from ages based on paleontological and paleoclimatic criteria. Considering the uncertainties, the ages provided by the numerical methods (AAR and U-series) in Piedra Alta lay between 154 and 221 ka, except for that of >350 ka.  In Gran Canaria, the two younger tsunami deposits cropping out on the Agaete valley rest upon flows from a Pleistocene lava sequence, dated at 1.8 ± 0.03 Ma at Llanos de Turman site [71]. At the El Cementerio site, the marine conglomerates overlie a lava flows dated to 1.75 ± 0.03 Ma [71]. These stratigraphic relations indicate a maximum age bound of 1.75 Ma for the two younger tsunamis, far from our more reliable AAR results obtained from 17 Glycymeris shells contained in three different tsunami deposits at La Aldea and Llanos de Turman (Table 1): 164 ± 23 ka, 160 ± 8 ka and 180 ± 21 ka, ranging betweeñ 140 and~200 ka when considering the uncertainties. An upper boundary for the age of the younger tsunami deposits from La Aldea site is provided by four ages determined by TL analysis of quartz grains included in two different calcrete crusts developed upon the tsunami deposit, of 143 ± 19 ka and 62 ± 7 ka (Table 1); thus, the higher value is the upper age bound for the tsunami, which is consistent with ages of 140 to 200 ka obtained by AAR analysis. The age data quoted above are from outcrops with elevations between 50 and 80 m a.s.l (La Aldea and Llanos de Turman); for the outcrops at higher elevations in the Agaete valley, no age data are available.

Sea Level during the Tsunami Events and Estimated Run-Up Elevations
Previous data on the elevation of the tsunami deposits and inundation distance refer to the current sea level. However, variations of sea level during the middle Pleistocene should be accounted for to constrain the run-up heights.
Global sea level curves reconstructed for the last hundreds of thousands of years [72,73] indicate that sea level coeval with the emplacement of the Teno-Isla Baja deposits, assuming an interval 120-170 ka, was in the range of 80 to 120 m below present sea level (b.p.s.l.).
In Piedra Alta, the available data do not allow a reliable age range to be assigned to the tsunami deposits. Two main ranges can be proposed depending on whether the ages obtained by numerical methods (150 to 250 ka) or by paleontological criteria are considered. In the first case, sea level varied between 20 and 100 m b.p.s.l. [72,73]; in the latter case, if a MIS 11 age is considered, then the corresponding interglacial sea level would be 10 to 50 m b.p.s.l. or about 16 to 30 m b.p.s.l. should the deposits be assigned for MIS 13 [65]. In the case of Agaete, sea level at the time of the megatsunamis was most probably between 100 and 80 m b.p.s.l., if an age range of 140-200 ka is considered.
These data indicate that, in all cases, the sea level was tens of meters below present level and, therefore, the run-ups of the tsunami inundations are significantly higher than the highest elevation of the tsunami outcrops, especially in the case of Teno, where more reliable ages were obtained.
Considering that the Canary Islands were stable over the last few millions of years, and have neither undergone neither subsidence nor uplift [74], the minimum run-up elevation (referred to sea level coeval of their emplacement time) reached by the tsunami waves may have been 140 and 230 m for Teno and Isla Baja, respectively, 45 m for Piedra Alta, and 268 m for Agaete. Even considering the occurrence of uplift movements of a few meters in Gran Canaria and Lanzarote in the last hundreds of thousands of years [70], this would virtually have no influence on the run-up heights indicated above. In the north of Tenerife, in addition to the lack of uplift movements [75], the tsunami deposits occur from the coastline, thus indicating that the position of coeval sea level would have been lower than present.

Tsunami Sources
The identification of at least four noteworthy middle-Pleistocene tsunami deposits on three islands of the Canary Archipelago and the occurrence of several tsunami deposits on the same site, corresponding to different events (at least three distinct tsunamis in Agaete), highlight the relatively high frequency of tsunamis generated by the abrupt entry of very large volumes of rock masses into the sea. Such huge mass movements correspond to large-scale volcanic flank landslides, capable of generating giant waves that would have struck the coasts of both the source and nearby islands.
At least 10 large landslides occurred over the last 1 million years in the Canary Islands, with those of Icod, La Orotava, Güímar, and Micheque in Tenerife (the latter has no morphological features on the surface, as the valley was completely filled in by successive lava flows), Cumbre Nueva (La Palma) and El Golfo (El Hierro) involving outstanding volumes, in the order of tens to hundreds of km 3 (Figures 1 and 13). None, however, has taken place on the remaining islands of the archipelago during this period.
The relationship between the tsunami deposits here described and the most recent landslides on the islands of La Palma and El Hierro, the westernmost islands of the archipelago, can be practically ruled out since they are far apart, at opposite ends of the archipelago. On La Palma, the last megalandslide, named Cumbre Nueva, occurred 540-560 ka ago [49,77,78]. On El Hierro, four to five megalandslides have occurred during the last 200 ka [49], including the most recent of the Canary islands, in the El Golfo area, where at least two superimposed large flank failures have been identified by subaerial and submarine morphological evidence [51,79]. They have been assigned controversial ages, around 15-20 ka and 130-136 ka [49,50,78], or between 39 ka and 87 ka [80]. The main argument ruling out the La Palma and El Hierro landslides as a source for the identified tsunamis is that those landslides were directed towards the north and west, in opposite directions to the location of the islands where the tsunami deposits have been identified. Another consideration is wave attenuation imposed by distance between the potential tsunami sources and the impacted areas, even though tsunamis can have extremely high initial wave heights of hundreds of meters at the source point and nearby coasts.
The Icod landslide, in Tenerife island, at around 165-175 ka (see Table 2), is related to a large phonolitic explosive eruption (blast) triggered in response to the sudden decompression of the magma chamber caused by the northern flank landslide of the Las Cañadas edifice [81][82][83]. The resulting massive explosive deposits, rich in heterolithic angular fragments, are distributed throughout the northwest area of the island, presenting a thickness of more than 3 m at the Isla Baja platform.
The data presented above and field observations in Teno and Isla Baja (Tenerife) suggest that the two well-differentiated stacked tsunami units are related to the Icod massive landslide and the ensuing volcanic blast, both events separated by an extremely short time interval (a few minutes?). The first inundation pulse was triggered by the large-scale flank landslide that preceded the blast, and was followed by a second pulse (each pulse consisting of one or several large waves) produced by the rapid entry of voluminous pyroclastic flows into the sea. This chronology and rapid succession of events is imprinted in the sedimentary record at Teno and Isla Baja ( Figures 4B and 6A), where the lowermost tsunami deposits contain angular basalt clasts from the lava deltas representing an early inundation of the lava platforms during the initial pulse. This inundation also transported marine coastal sand and shells landward, mixing them with terrestrial material (angular clasts, bones of pulmonate organisms, such as lizards; Figure 3E) collected from the subaerial portion of the lava delta. The mixture of terrestrial fossils with marine shells is to be expected in a landward transport context. The mixed sedimentary load would then have been released suddenly as the water flow reduced intensity, just before or during withdraw to the sea. Landward-dipping imbricated clasts occasionally found in the deposits, testify deposition during out-wash and the extremely chaotic nature of the deposits in terms of particle size, sorting, and directional properties is compatible with a single, extreme-energy marine-borne inundation event. It is noteworthy that in the tsunami deposits at Isla Baja, just at the western boundary of the Icod landslide, and in the nearby outcrop of Barranco de Itobal, in Teno platform, a greater development of the lower layer with large basalt boulders occurs, while on the large extent of the Teno lava platform, farther away and located outside the direct impact of the waves caused by the landslide, this layer shows smaller thickness and size of the basalt clasts. The second inundation pulse, which probably occurred within a very short timeinterval after the first wave, essentially transported and delivered lithic and juvenile fragments (pumice and ash) to the coast. These materials were sourced from the pyroclastic flow violently entering the sea and were eventually mixed with the remaining marine shells and sediments of the coastal fringe that had been previously swept and depleted of coastal sediment during the earlier inundation phase ( Figure 3D,I; Figure 4B; Figure 6B). The second pulse may have encompassed several waves. Evidence for this is the presence of three to five layers containing material from the pyroclastic flow found at the Teno platform, though it is not clear if every layer represents an inrush event. The composition of the deposits shows their relation to the volcanic blast event, and the absence of internal structure and the massive chaotic character of each sedimentary unit are consistent with high-energy, short-lived transport. The inundation of the subaerial portion of the lava deltas facilitated the mixture of terrestrial (e.g., Helicidae shells and vertebrates bone fragments) with marine fossils.
The age obtained for the Teno and Isla Baja deposits through AAR is ≤168 ka, a result within the same age range attributed to the Icod flank landslide (165-175 ka ago) and the associated explosive eruption [52].
The thickness of the marine coarse-detrital tsunami deposits decreases very rapidly westwards, along some 9 km of coast. It exceeds 3.5 m at Playa de las Arenas (Isla Baja), decreases to about 1.8 m thick at Barranco de Itobal, and is just 0.4-0.6 m at Teno platform. This can be interpreted as a very fast lateral (alongshore) westward decrease in amplitude of the tsunami waves produced by a source located further east on the same coast. The Icod landslide and pyroclastic flow(s) were probably directed towards the north, focusing most energy on that direction, with a rapid loss of wave energy occurring along the coast and away from the source point.
At Piedra Alta, the ages provided by the numerical methods (AAR and U-series) suggest an age range of 154 to 221 ka, considering the uncertainties, while from biogeographical/paleoclimatic criteria an age of 365-500 ka has been assigned [53,65]. A plausible correlation can be proposed with two of the large landslides on Tenerife, the nearest island and most likely source of the tsunamigenic landslide. Considering the age of the Tenerife landslides, the Icod landslide (165-175 ka) is compatible with the AAR age interval, while one of the landslides that probably occurred in the La Orotava valley at~500 ka (Table 2), is within the higher age range proposed from biogeographic/paleoclimatic criteria. Since an uncertainty range is not provided for the 196 ka age of the post-tsunami lavas [53], it is not possible to assess the consistency with the age of the Icod landslide; however, the alternative age of 160 ± 20 ka for the same lavas [64] does supports this correlation. If either the Icod or La Orotava landslides, which affected the northern flanks of Tenerife, were the source of the Piedra Alta tsunami, the distance between the two islands would explain the 'low' run-up height (≥45 m, a much lower height than the run-ups observed at the coasts of north Tenerife) observed in Lanzarote.
The tsunami deposits that occur in Agaete had previously been attributed to the Guimar landslide, based upon their location in front of the landslide scar on the opposite coast of Tenerife [8,84]. The massive landslide(s) directly facing the neighboring island of Gran Canaria must have generated large waves travelling directly towards that target area. However, this relationship is not supported by the new 160-180 ka age interval provided by AAR analyses in Glycymeris (Table 1).
The age of the Güímar landslide has been estimated at~830 ka, or somewhat older, as the first lava flow filling the landslide depression has been dated to around 1 Ma ( [52] and references therein), which is consistent with the lower limit imposed by the 1.75 Ma age of the lavas underlying the tsunami deposits. On the other hand, the stratigraphic evidence for three different tsunamis events in Agaete would point to the occurrence of several time-differentiated landslides in the area of Güímar, a hypothesis previously proposed by [83], based on morphological features of the landslide scars (tsunami deposits from prehistoric megalandslides with layers deposited by multiple tsunamis have also been described in the Hawaiian islands [6,85]). Neither the AAR nor TL ages obtained in this study for the marine conglomerates agree with the postulated age for Güímar at around 0.83-1 Ma [83]. In this respect, it should be noted that the age determinations here presented correspond to the two younger tsunami deposits located at the lowest elevations (50 to 80 m a.s.l.), at La Aldea and Llanos de Turman sites. The deposits at higher elevations in the Agaete valley, up to almost 200 m, may have an older age, corresponding to the older tsunami exposed at La Ruina or to other tsunami events.
The existing ages for both the landslides and tsunami deposits in the Canary islands, preclude a consistent relationship between specific landslides and tsunami deposits to be established, except for Teno and Isla Baja, although these phenomena can be constrained to a broad time-interval corresponding to the middle Pleistocene (~800 ka to~160 ka ago).

Conclusions
The main results obtained from the geological, geomorphological, paleontological, and geochronological investigations carried out on the tsunami deposits identified in Tenerife, Gran Canaria, and Lanzarote, as well as their possible source areas and frequency, are summarized below.

1.
Fossiliferous marine deposits composed of chaotic conglomerates have been identified and described in detail at several sites in Tenerife (Teno and Isla Baja), Gran Canaria (Agaete), and Lanzarote (Piedra Alta). In all of them, common sedimentological, geomorphological, and paleontological features unequivocally indicate a tsunamigenic origin. 2.
According to the age determinations and stratigraphic relations of the investigated deposits, at least four to five tsunami events (or up to six if we consider the two successive tsunamis in Icod as individual events) have occurred, some of them presenting several inundation pulses.

3.
A total of 144 age determinations were carried out, of which 105 correspond to amino acid racemization (AAR) analysis, 35 to U-series dating in corals, and four to the thermoluminescence (TL) technique. The overall results show that the tsunami deposits were emplaced during the middle Pleistocene, with an age younger than 168 ka for Teno and Isla Baja, between 140 and 200 ka in Agaete, and within the 154-221 ka age interval at Piedra Alta.

4.
The maximum tsunami run-up heights associated with these deposits have been calculated according to their present elevation, estimated ages, and the coeval sea level position. Maximum tsunami wave run-ups of 180 and 270 m for Teno and Isla Baja, respectively, 290 m for Agaete, and 125 m for Piedra Alta are proposed. 5.
The megatsunami waves are attributed to large flank landslides of the Canaries volcanic edifices. Attempts were made to establish relationships between the tsunami deposits and potential source landslides, based on their estimated ages and spatial distribution. The age for Teno and Isla Baja tsunami deposits (<168 ka), is in the same range of the Icod flank landslide and ensuing explosive eruption (165-175 ka), showing a strong correlation between the landslide/volcanic event and the tsunami inundations of the north shore of Tenerife. This is also supported by the specific lithological and sedimentological characteristics of the deposits. 6.
The available geochronological data for the Agaete and Piedra Alta tsunami deposits precludes the establishment of reliable bi-univocal correlations with the potential landslide sources. In the case of Agaete, ages younger than 1.75 Ma (based on the age of underlying lavas), or between 160 and 180 ka according with AAR and TL results, have been proposed. These results prevent a possible relation with the admitted age for a single landslide at Güímar (at around 830 ka or older). However, the presence of at least three different tsunamis deposits in Agaete suggests that more recent landslides may have been generated within the Güímar valley. 7.
Paleontological and biogeographical markers for Piedra Alta deposits suggest an age range for their emplacement between~400 and~500 ka, which is compatible with the age interval for landslides at La Orotava valley (~500 to 560 ka). However, AAR ages in the 154 to 221 ka range, do not preclude a correlation with the Icod event.
8. According to the global age results for the deposits, the main sources for tsunamis in the Canary Islands during the last 1 million years may have been megalandslides that affected the flanks of volcanic edifice of Tenerife, where, at least, seven large events have occurred in this period: one in the Icod valley, three at La Orotava, two in Güímar, and one in Micheque. 9.
Considering the frequency of megalandslides during the last 1 million years in the Canaries, with possibly several overlapping landslides on the same island flank, it can be stated that correlative tsunami events have also occurred with a relatively high frequency during the Middle Pleistocene in the archipelago, with an average interval of 80 ka. Funding: This research was carried out with financial support from the Ministry of Science of Spain (CICYT-project GRANDETEN II n • CGL2008-01423) and the Geological Survey of Spain (IGME).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available in this article and the Supplementary Data.