Biostratigraphic Evidence for Dramatic Holocene Uplift of Robinson Crusoe Island, Juan Fernández Ridge, Se Pacific Ocean

Hotspot oceanic islands typically experience subsidence due to several processes related to migration of the oceanic plate away from the mantle plume and surface flex-ural loading. However, many other processes can interrupt subsidence, some of which may be associated with catastrophic events. A study of the biostratigraphy and sedi-mentology of Holocene deposits on Robinson Crusoe Island (RCI) on the Juan Fernández Ridge (JFR) indicated that dramatic uplift has occurred since 8000 years BP, at a rate of about 8.5 mm yr −1. This is evidenced by supratidal flats with tepee structures and sand layers containing marine gastropods (mostly Nerita sp.) that are now exposed ca. 70 m a.s.l. The active hotspot is located 280 km further west and the last volcanic activity on RCI occurred at ca. 800 000 years BP. Long-term subsidence is evidenced by deep submerged marine abrasion terraces at RCI. As no direct evidence was found for the existence of a compensating bulge generated by the present hotspot upon which RCI would be situated, it must be concluded that subsidence in the wake of the mantle plume beneath the migrating plate was interrupted by very rapid uplift, but on a scale that did not fully compensate for the previous subsidence. This can be attributed to large-scale landslides followed by isostatic rebound, although this is only vaguely reflected in the low-resolution bathymetry of the area. To determine if this mechanism produced the uplift, a detailed bathymetric survey of the area will be required. If such a survey confirms this hypothesis , it may have implications for the short-term dynamics of vertical variations of oceanic edifices and their related effects on ecosystems and human population.


Introduction
Oceanic hotspots are relatively mantle-fixed and localized volcanic sources, which transport rising magmas through oceanic lithosphere.As the plates move away due to seafloor spreading, oceanic volcanoes are extinguished and new volcanic edifices arise over the active hotspot, forming ageprogressive island chains such as the Hawaiian-Emperor seamount chain.
A conspicuous feature of hotspot oceanic islands is their complex history of vertical displacement (e.g.Ramalho et al., 2013).As noted by Charles Darwin in the 19th century, these vertical movements respond to a number of largescale processes known at present to be related to the growth and decay of the underlying swell (Ramalho et al., 2010a), flexural loading (Watts and ten Brink, 1989), isostatic rebound (Smith and Wessel, 2000), bulging effects resulting from loading of nearby islands and seamounts (e.g.Bianco et al., 2005), density changes in the mantle, intrusions at the base of the edifice (e.g.Klügel et al., 2005) and gradual cooling of the lithosphere (Stein and Stein, 1992).
Although most of the present oceanic islands are subsiding (such as Surtsey Island over the last decades; e.g.Moore et al., 1992), other processes such as those mentioned above and especially catastrophic events like giant landslides can also trigger sudden uplift, as inferred for archetypical hotspot volcanoes such as Hawaii (e.g.Smith and Wessel, 2000).It is therefore important to identify such mechanisms and their possible short-or long-term effects, especially for those few cases where they occur at very high rates.
Vertical movements along the Juan Fernández Ridge (JFR), located on the relatively fast-moving Nazca Plate (Gripp and Gordon, 2002) are still poorly known due to the general absence of detailed bathymetry and other known subsidence or uplift tracers.In this context, the record of local sea level changes is a tool to better understand the uplift/subsidence history in different geological settings (e.g.Toomey et al., 2013;Jara and Melnick, 2015).Here we present biostratigraphic and sedimentological evidence for a very high Holocene uplift rate of Robinson Crusoe Island (RCI) and discuss possible mechanisms with implications for the future evolution of this oceanic island.

Geological and geomorphological background
The JFR, located on the Nazca Plate in the Pacific Ocean off central Chile (Fig. 1a), is an 800 km long chain of seamounts and volcanic islands extending E-W at latitude 33 • S. It has been interpreted as the expression of a fixed hotspot (Von Heune et al., 1997;Montelli et al., 2006) related to a primary mantle plume in the sense of Courtillot et al. (2003) or as part of a "hot line" (Bonatti et al., 1977).
The JFR is largely formed by the Miocene (ca. 9 million years BP) O'Higgins guyot and seamount (Von Heune et al., 1997), with lavas dating back to ca. 4 million years BP on RCI and nearby Santa Clara Island (Fig. 1a), and ca. 1 million years BP on Alejandro Selkirk Island about 120 km away (Farley et al., 1993).The relief on the western, arid part of RCI is characterized by coastal cliffs bordering a terrace at about 70 m a.s.l., which is especially well developed in the southwestern panhandle (Fig. 1b).This terrace is formed on top of a middle Pleistocene age (ca.800 000 years BP), post-shield volcanic platform from which pyroclastic cones emerge, reaching a maximum elevation of 915 m (Lara et al., 2013).Holocene sedimentary deposits are restricted to the terrace in the vicinity of Bahía Tierra Blanca (Spanish for "White Land Bay").The latter name is applied to a succession described by Morales (1987) as poorly consolidated, calcareous sandstones at the base grading upward into tuffaceous sandstones with numerous fossils.In the transition zone are Acanthina and Lima fossils with Bryozoa fragments, whereas the tuffaceous sandstones host Luccinea, Distoechia, Bithynia, Orcula, Tropicorbis, Ena and Cyrena spp.indicating a Pleistocene-Holocene age, based on a similar fossil assemblage on the continent at this latitude (Covacevich, 1971;Valenzuela, 1978).The Bahía Tierra Blanca succession has its base at a variable elevation but generally at ca. 70 m above the present mean sea level, where active, incipient barchan dunes (Morales, 1987) partially rework the succession described above.A satellite image of the southwestern "panhandle" where the aerodrome is situated.White areas are those of the Bahía Tierra Blanca succession, where a well-exposed supratidal Holocene sequence was dramatically uplifted (see text for details).Sampling sites labelled with numbers (see Table 1).NZ: Nazca Plate; SA: South American Plate.
RCI has at least two submerged marine abrasion terraces, with edges at ca. 200 m and ca.500 m b.s.l.(Astudillo, 2014).The present depth of these originally shallow features suggests long-term subsidence of this volcanic edifice.The latter were collected mostly from sites 1 and 5 (Fig. 1b).Several specimens were hand-picked from bulk samples and three were selected for dating based on their stratigraphic position and systematics.AMS radiocarbon dates were first corrected for the global marine reservoir effect (e.g.Ulm, 2006) with the Marine IntCal09 calibration program (Reimer et al., 2009).For the localized reservoir correction a Delta-R value of 373±76 from a nearby site was used (Marine Reservoir Correction, 2012).Elevations were measured with a dGPS Trimble ® NetRS ® and barometric altimeters with respect to the current sea level and corrected for the regional sea level and daily variation (Tabla de Marea, 2012) with a nominal uncertainty of 5 m.In order to determine thresholds for the wind velocity capable of moving collected fossils we used the formula in Appendix A, mostly based on Le Roux (2005).

Depositional environments
Four lithostratigraphic units and three lithofacies were identified in the Bahía Tierra Blanca succession, which reaches a total thickness of between 2 and 4 m at any specific locality (Fig. 2).Unit 1 is largely composed of facies 1, which discordantly overlies weathered, basaltic lavas.It consists of very poorly consolidated, slightly calcareous, reddish brown to reddish purple deposits ranging in size from very fine sandstone to claystone.Their composition is made up of volcanic ash mixed with the underlying, weathered lava material.These deposits contain up to 2 % bioclasts (mostly marine bivalves) together with pellets.The most striking feature of this facies is ubiquitous tepee structures up to 1 m in diameter, which display prominent edges elevated 3-5 cm above the central parts (Fig. 2).The cracks have been filled in by sands from the overlying unit.Locally, shallow channels and rill marks are present.
While the reddish to purplish brown colour suggests a mainly subaerial environment, the presence of tepee structures with elevated rims indicate frequent flooding and drying cycles.These, as well as the occurrence of pellets, are typical of supratidal flats (e.g.Assereto and St C. Kendall, 1977), which concurs with the presence of shallow channels probably reflecting tidal creeks.The general scarcity of hard-shell fossils in this facies can be interpreted as representing a generally hostile environment subjected to frequent dry periods between spring high tides, followed by seawater flooding that would kill land-dwelling snails and other organisms.Marine shells washed in during spring high tides would probably accumulate along the shoreline.Soft-bodied forms more tolerant to such conditions, on the other hand, would not be preserved in such an oxidizing environment.
Facies 2 is present in units 2 and 3, which differ mainly in the darker brown colour of the latter due to a thin brownish film coating the grains.Both units 2 and 3 show large-scale, low-angle planar cross-bedding and horizontal lamination, but in unit 3, high-angle planar and trough cross-bedding are locally present.The 1-2 cm thick cross-beds are formed by alternating light and darker-coloured grains without any evident gradation.Rhizocretions are present in the uppermost parts of both units, where individual forms may reach 1.5 m in length (Fig. 2).Although rhizocretions and vertebrate burrows are generally rare, some parts have a fairly high density of the former.Unit 2 is capped locally by whitish calcrete indicating incipient pedogenesis.Gastropods such as Succinea, Fernandezia and Nerita occur in the middle to upper part of unit 2. Petrographically, the sandstone is well sorted with subrounded grains, lacking a matrix, and cement being only locally present.Bioclasts compose around 55 % of the rock, including brachiopod and pelecypod fragments, echinoderm spines, Bryozoa, red algae, foraminifers and subrounded pellets.The rest of the composition is made up of volcanic fragments and minerals such as K-feldspar, plagioclase, clinopyroxene and olivine, with rare quartz.Grain-size analysis of several samples from this facies showed a small traction load, a prominent and very well-sorted saltation load, and a medium-to well-sorted suspension load.This facies was interpreted as reflecting coastal aeolian deposits perhaps locally affected by weak wave action.This is supported by the reddish brown colour of the sandstones, their predominantly fine grain size with cumulative curves typical of wind-blown deposits, and the presence of the land-dwelling snails Succinea and Fernandezia, as well as root and burrow systems.The horizontally laminated strata probably formed in sand sheets between low dunes, which might have been parabolic in shape as suggested by the dominance of lowangle planar cross-bedding.Some were subsequently converted into dikaka dunes (Glennie and Evamy, 1968) by vegetation.Some low-angle cross-bedding might represent reworking by dissipated wave action during storms and spring high tides along the landward edges of wide supratidal flats.This could also explain the presence of thick-shelled Nerita (a marine species) in unit 2. The presence of fragmented marine invertebrates indicates a marine source for most of these sands, which suggests that they formed at a low elevation above sea level.
Facies 3 is composed of greyish white, medium sorted sandstones interbedded with gravel.The sandstones consist of bioclasts (45-57 %) mostly represented by marine shell fragments including bivalves, gastropods such as Succinea, Bryozoa, algae and foraminifers, together with lithic volcanic fragments (27-45 %) and volcanic minerals such as pyroxene, olivine and felsic minerals (10-17 %).The gravels are greyish brown and matrix-to clast-supported, with the clasts reaching up to 5 cm in diameter.They are mainly volcanic and angular.Locally, calcretes are present at the top of this facies.This facies clearly represents fluvial deposits, probably consisting of shallow, quick-flowing ephemeral streams with gravelly channels and sandy bars.These most likely drained exposed basalts on the fringes of the aeolian sand sheets, but also reworked the latter to incorporate the marine bioclasts.

Radiocarbon dating
Specimens of Nerita (Fig. 3) from the aeolian sandstones of unit 2 yielded calibrated radiocarbon ages between 8320 and 8030 BP (conventional radiometric age of 7860 ± 40 years BP).Values corrected for the global marine reservoir effect (with a local Delta-R of 373 ± 76) correspond to 7550 ± 90 years BP (see Table 1).These marine species were probably reworked from the supratidal flats of unit 1 and would thus represent the age of the latter.On the other hand, land-dwelling species as Succinea and Fernandezia (e.g.Odhner, 1922) from units 3 and 4 gave calibrated radiocar-bon ages between 5440 and 5090 years BP (4580 ± 30 conventional years BP) and 7680 and 7580 years BP (6790 ± 40 conventional years BP), respectively.

Discussion
The stratigraphic succession of the Bahía Tierra Blanca deposits suggests that supratidal conditions existed in the southwestern panhandle of RCI between 8320 and 8030 years BP (horizons with Nerita).It is unlikely that the tides reached more than 1-2 m above the mean sea level, because topographic tide-enhancing conditions such as funnel-shaped estuaries could not have existed due to the absence of large rivers on this part of the island.These supratidal flats were encroached upon by aeolian coastal plain deposits at around 5430 years BP (horizons with Succinea and Fernandezia) and finally fluvial sedimentation as the sea level receded further during the late Holocene Climatic Optimum (Davis et al., 2003;Koshkarova and Koshkarov, 2004), when the southwestern panhandle would have received more rain.The present elevation of the supratidal deposits on a marine terrace at 70 m a.s.l.indicates a very rapid relative sea level fall since that time.Furthermore, it can be assumed that the aeolian deposits of units 2 and 3 were also not more than a few metres above the tidal flats, as they had apparently been  reworked locally by waves.This is supported by the lowangle cross-bedding typical of beaches and the presence of reworked Nerita.The latter could not have been blown uphill by wind, considering that they reach up to 1 cm in diameter (Appendix A).Assuming that they were not more than 2 m above the tidal flats or beaches, a relative sea level fall of at least 8.5 mm yr −1 is implied.Eustatic sea levels have been well below the presentday level over the last 20 000 years (Bindoff et al., 2007;Fleming et al., 1998).In Tahiti and almost all other regions of the world where detailed records exist (e.g.Lambeck et al., 2002), there are indications that the sea level at 8000 years BP was about 15 m below that of the present (Fleming et al., 1998;Milne et al., 2005).This rules out an eustatic highstand at the time.A mean uplift rate of around 8.5 mm yr −1 is extremely high, considering that the average rate of uplift of the Andes has been only about 0.2-0.3mm yr −1 since the Late Miocene (Gregory-Wodzicki, 2000) and uplift rates of other oceanic islands were < 0.33 mm yr −1 (e.g. in the oldest Hawaiian islands as reported by McMurtry et al., 2004, and references therein).Oceanic islands with evidence of significant freeboard (e.g.Cape Verde) show uplift rates < 0.4 mm yr −1 (Ramalho et al., 2010a, b).This high vertical displacement rate is only comparable with the subsidence rate of the active Hawaii Island, which sinks at ca. 2.6 mm yr −1 (McMurtry et al., 2004).
The dramatic Holocene uplift of RCI cannot be explained as a flexural response to the loading exerted by the edifices created by the active hotspot.Isobaths (after Becker et al., 2009; see also Rodrigo and Lara, 2014) show that the sea floor north of the JFR descends from 3800 m b.s.l.northwest of Alejandro Selkirk to about 4000 m b.s.l.north of the latter, from where it declines further to reach 4200 m b.s.l.north of RCI and 4300 m b.s.l.northeast thereof.There is thus no direct evidence for the existence of a bulge upon which RCI would be situated.In fact, the negative anomaly of bathymetry for this part of the oceanic crust, as well as the deep abrasion terraces at RCI (Astudillo, 2014) suggest that subsidence did take place, but was partially reversed by a subsequent process commencing at about 8000 years BP.General subsidence normally occurs in the wake of a mantle plume beneath a plate migrating away from a particular area, as this part of the lithosphere would no longer be sustained by it, combined with the load exerted by the shield volcano.The generation of new islands and seamounts above a fixed mantle plume could also cause loading and subsidence of the crust accompanied by the formation of an adjacent, compensating bulge, and hence local uplift.However, a theoretical bulge caused by the youngest volcanism at the Friday and Domingo seamounts (250 km further west of RCI) is not enough to explain uplift at RCI if realistic values for elastic parameters are considered (e.g.Manríquez et al., 2013).For example, Watts and ten Brink (1989) proposed the existence of such a bulge 300 km from the present hotspot on Hawaii Island, which formed in response to subsidence of 1300 m at the latter locality over the last 500 000 years (McMurtry et al., 2010).Evidence of > 20 m uplift is found at Oahu in the now emerged coral reefs (McMurtry et al., 2010).Nevertheless, there is no evidence of recent Holocene volcanism P. Sepúlveda et al.: Biostratigraphic evidence of dramatic uplift of RCI further west at a distance short enough to promote uplift at RCI.In addition, 3-D modelling of the lithospheric flexure seaward of the trench (Manríquez et al., 2014) shows that even more complex loads (seamount loading, bending of the lithosphere near the trench and sedimentary fill inside the trench south of 34 • S) do not generate a flexural response beyond 350 km from the outer rise.
Intrusion at the base of the edifice, as proposed for the Canary Islands (Klügel et al., 2005) and Cape Verde (Madeira et al., 2010;Ramalho et al., 2010b) cannot be ruled out.However, because of the absence of volcanism younger than ca. 1 million years BP and the rapid displacement of the Nazca Plate, we have a reasonable doubt about the occurrence of this process in the Holocene.
Another possibility could be the development of largescale landslides.The southwestern part of the island is characterized by steep coastal cliffs, and the area lies opposite Santa Clara Island that is thought to have originally formed part of a larger island incorporating RCI (Danton, 2004).The region between the two islands might have experienced a large-scale landslide event (or events), which in turn may have caused isostatic rebound.The latter is thought to be larger on oceanic plates than on continental plates because of their more limited thickness.In hotspot environments and other high heat-flow areas such as spreading boundaries the asthenosphere should be less viscous, so that rebound rates may increase.Similar events have been reported in Hawaii during the last million years (McMurtry et al., 2004).Smith and Wessel (2000) calculated that the removal of 800 km 3 of material during the Alika landslide elevated the adjacent terrain by about 17 m, whereas McMurtry et al. (2004) calculated uplift of 109 m for a volume of 5000 km 3 removed during the Nuuanu landslide.Taking into account an elastic thickness of ca. 10 km (Manríquez et al., 2013), about 1000 km 3 of material (ca.15 % of the initial volume) would thus have had to be removed to account for ca.70 m of uplift at RCI.Such a large mass wasting deposit is not evident in the low-resolution bathymetry around the RCI, but the caldera-like structure open to the south and some rough relief on the distal flanks suggest that a landslide is a plausible hypothesis.

Conclusions
Biostratigraphic evidence for the exposure of former supratidal flats 70 m above the present sea level on RCI could be related to a large Holocene landslide not previously detected.Large-scale landslides around oceanic islands can probably be attributed to an increase in local slopes generated by the construction of volcanic edifices and the development of rifting.At RCI there has been no major surface volcanic activity since about 3 Ma, with only minor post-shield activity at 800 000 years BP (Lara et al., 2013).Nevertheless, the topography of RCI is even steeper than that of Hawaii, which could have allowed large-scale sliding to take place.As modelled by Smith and Wessel (2000), directed giant landslides generate isostatic rebound which is larger over the failed flank and spatially asymmetric.Catastrophic landslides can generate tsunamis in addition to directly affecting large parts of the island itself.The steep slopes on and around the island, its volcanic composition (notorious for rapid weathering and soil production), and the frequency of large earthquakes along the East Pacific rim, are factors that make such landslides more likely.It is therefore important that a detailed bathymetric survey of the area around RCI be carried out to detect possible Holocene landslide scars and deposits.If such a survey confirms this hypothesis, it may have implications for the short-term dynamics affecting vertical variations of oceanic edifices.These findings highlight the importance of biological markers as tools to better understand sea level changes and the complex evolution of oceanic islands.

Appendix A: Calculation of required wind speed
All equations can be found in Le Roux (2005).
Submerged density of shell in water:

Figure 1 .
Figure 1.(a) Location of Juan Fernández Ridge, with Robinson Crusoe Island (RCI) and Santa Clara Island (SCI) in a box.(b)A satellite image of the southwestern "panhandle" where the aerodrome is situated.White areas are those of the Bahía Tierra Blanca succession, where a well-exposed supratidal Holocene sequence was dramatically uplifted (see text for details).Sampling sites labelled with numbers (see Table1).NZ: Nazca Plate; SA: South American Plate.

Figure 2 .
Figure 2. (a) Exposure of sedimentary units as described in the text.Nerita dated at ca. 8000 years BP sampled from unit 1. Below are "tepee" structures in unit 1 (b), interpreted as part of a former supratidal flat.(c) A composite stratigraphic column from records at sites shown in Fig. 1.

Figure 3 .
Figure 3. Nerita shells found in aeolian deposits of unit 2. These are marine species, probably incorporated into dunes developed close to the supratidal flat shoreline.Visual field is 2.5 cm.

Sepúlveda et al.: Biostratigraphic evidence of dramatic uplift of RCI 1997Table 1 .
Radiocarbon dates for gastropods from Robinson Crusoe Island.Data obtained at Beta Analytic Inc., Miami, FL.Elevation computed from dGPS data with correction for daily variation of sea level and local height of the antenna.