Upper-plate deformation in response to flat slab subduction inboard of the aseismic Cocos Ridge, Osa Peninsula, Costa Rica

Flat slab subduction of aseismic ridges has generated considerable scientific interest because of the profound impact it has on the tectonic, magmatic, and landscape evolution of the upper plate that can extend hundreds of kilometers inland from the trench. Subduction of aseismic ridges generally leads to rapid outer forearc uplift over a broad region immediately inboard of the trench (Corrigan et al., 1990; Gardner et al., 1992; Hsu, 1992; Macharé and Ortlieb, 1992; Sak et al., 2009); shortening in the inner forearc with development of a fold-and-trust belt (Fisher et al., 2004; Sitchler et al., 2007); changes in the properties of the volcanic arc and cessation of arc magmatism (Carr et al., 1990, 2003; von Huene et al., 2000; Patino et al., 2000; Ramos et al., 2002; Phipps Morgan et al., 2008; Alvarado et al., 2009); or even broad uplift of the magmatic arc (Gräfe et al., 2002; Fisher et al., 2004; Morell et al., 2012) with uplift even extending into the backarc (Coates et al., 1992; Collins et al., 1995; McNeill et al., 2000; Ramos et al., 2002) and the foreland basin (Espurt et al., 2007). Taken together, these tectonic processes modify regional drainage patterns and broadly control landscape evolution (Espurt et al., 2007; Morell et al., 2008; Regard et al., 2009).

Along thinly sedimented convergent margins, subduction of rough crust and seamounts associated with aseismic ridges controls local margin erosion, subsidence, and deformation of the forearc (Ballance et al., 1989; Cloos, 1992; Gardner et al., 1992; Geist et al., 1993; Dominguez et al., 1998; von Huene et al., 1995, 2000; Gardner et al., 2001; Laursen et al., 2002; Bilek et al., 2003; Sak et al., 2004a, 2009; Pedley et al., 2010; Wang and Bilek, 2011). It has been demonstrated, under some conditions, that uplift and subsidence rates and topography of the outer forearc are directly related to seamount volume, relief, and the convergence rate between the upper and lower plates (Mann et al., 1998; Gardner et al., 2001; Meffre and Crawford, 2001; Sak et al., 2004a, 2009; Taylor et al., 2005). In some places, bathymetry of the subducting lower plate can affect the geometry of foreland fold-and-thrust belts up to a hundred kilometers from the trench (Fisher et al., 1994, 1998, 2004; Marshall et al., 2000, 2003; Espurt et al., 2007). Furthermore, subduction of bathymetric features such as seamounts, oceanic plateaus, and aseismic ridges can cause erosion of the upper plate (Ranero and von Huene, 2000), leading to significant trench retreat (Clift et al., 2003; Vannucchi et al., 2003; Clift and Vannucchi, 2004).

First-order control on the flat slab subduction of aseismic ridges is either buoyancy related to the thickness and age of the downgoing plate (Protti et al., 1995; Gutscher et al., 1999, 2000; Gutscher, 2002; Alvarado et al., 2009) or trenchward motion of thick cratons over the incoming slab (Manea et al., 2012). First-order controls on the deformation of the upper plate that result from aseismic ridge subduction are plate convergence rate (Fisher et al., 2004; Sak et al., 2004a), relative plate material strengths, basement cover sediment thickness and porosity (von Huene and Ranero, 2003; Sak et al., 2004a; Taylor et al., 2005; Ryan, 2012), and volume and relief of the subducting asperity (Meffre and Crawford, 2001; Sak et al., 2004a; Taylor et al., 2005). In the outer forearc of southern Costa Rica (Figs. 1A and 1C), the relative plate motion rate is rapid (∼85 m/k.y.), the subduction angle is shallow, 3° to ∼10° (Fig. 1C), and the upper plate is tapered to less than 10 km (Protti et al., 1994, 1995; Kolarsky et al., 1995), and is composed of a deformed, accreted complex of seamounts (Denyer et al., 2006; Vannucchi et al., 2006; Buchs et al., 2009). Importantly, the downgoing plate has a thin sediment cover (Fig. 1B), but it also has considerable small-wavelength roughness, related to seamounts and plateaus, that is superimposed on the overall long-wavelength bathymetry associated with the aseismic ridge. Where aseismic ridges subduct, the strain rate in the thin, overriding margin wedge is determined by the curvature of the incoming bathymetry, while the uplift/subsidence rate is determined by the slope, and the duration and amount of uplift are determined by the relief of the underthrusting asperities (Sak et al., 2004a).

The western edge of the Caribbean plate at the Middle America Trench is a classic, thinly sedimented, erosive margin (von Huene and Scholl, 1991; Clift and Vannucchi, 2004) extending from central Mexico (Clift and Vannucchi, 2004) and Guatemala (Vannucchi et al., 2004) through Nicaragua (Ranero et al., 2000) and Costa Rica (Vannucchi et al., 2003) to the border with Panama (Gardner et al., 1992; von Huene et al., 2000; Sak et al., 2004a; Morell et al., 2011), the eastern boundary of the subducting Cocos plate. Along this margin, trench retreat rates from subduction erosion range from 0.9 m/k.y. in Guatemala (Vannucchi et al., 2004; Clift and Vannucchi 2004) to 3–3.6 m/k.y. in northern Costa Rica (Vannucchi et al., 2001; Scholl and von Huene, 2007), the southernmost limit of currently available data. Net arcward migration of the Middle America Trench ranges from 20 to 30 km in Guatemala to 50–60 km in northern Costa Rica (Scholl and von Huene, 2007). Because subduction erosion leads to subsidence, the margin wedge, the outermost forearc, and the updip limit of the seismogenic zone are seldom exposed subaerially, critically limiting observations and direct measurements of tectonic processes, structural deformation, and stratigraphic relationships. However, in southern Costa Rica, flat slab subduction of the aseismic Cocos Ridge, a large bathymetric feature on the Cocos plate, results in the greatest uplift and unroofing of the upper plate anywhere along the margin. On the Osa Peninsula, the outermost forearc is subaerially exposed above the subducted portion of the Cocos Ridge, allowing for detailed structural and stratigraphic mapping of this critical upper-plate region.

In this paper, we integrate 10 previously published radiocarbon ages from coastal sections on the Osa Peninsula (Gardner et al., 1992) with eight new radiocarbon ages and five new optically stimulated luminescence (OSL) ages from the interior of the peninsula. New field mapping of Quaternary marine deposits along the central coast and within the interior of the peninsula allows for significant revision of the distribution and facies architecture of the Marenco formation, as originally described by Sak et al. (2004a), and a revised geologic map for the Osa Peninsula. These new data sets allow for construction of E-W and N-S structural cross sections along and across the peninsula. This was not possible from previous mapping and geochronology, which were confined to eastern (Gardner et al., 1992) and western (Sak et al., 2004a) coastal outcrops. From these new structural cross sections and revised stratigraphy, we constrain (1) rates of deformation across the peninsula, (2) the spatial and temporal distribution of those rates, and (3) the size of independently uplifting and subsiding blocks. We show that the basement rocks of the Osa Peninsula have experienced pervasive, brittle deformation related to subvertical faults that extend from the plate interface to the surface. Based on these constraints, we propose a more detailed model for deformation in the outer forearc along this part of the margin, where bathymetric relief on the subducting Cocos Ridge dominates over plate-boundary friction (i.e., basal shear stress) in causing permanent deformation of the upper plate.

In southern Costa Rica, the Cocos plate subducts under the Caribbean plate at a rate of ∼91 m/k.y. nearly orthogonal to the Middle America Trench (DeMets et al., 1990; DeMets, 2001; Bird, 2003; Jin and Zhu, 2004) and under the Panama microplate at 85 m/k.y. (Morell et al., 2012; Fig. 1A). The Cocos plate is sharply segmented and changes significantly along strike of the Middle America Trench due to hotspot volcanism, which intrudes oceanic lithosphere that was created at the East Pacific Rise and the Cocos-Nazca spreading center (Barckhausen et al., 2001). Seismic imaging of the Wadati-Benioff zone along this margin indicates that the Cocos plate decreases from a steeply dipping (70°) slab at the Nicaragua–Costa Rican border to a very shallowly dipping slab (>10°) off southern Costa Rica (Protti et al., 1994, 1995; Fig. 1C). This shallowing of the subduction angle is caused by an increase in buoyancy resulting from a decrease in age of the plate (Barckhausen et al., 2001; Fig. 1A) and a significant increase in thickness of the plate (von Huene et al., 2000; Sallarès et al., 2003; Walther, 2003; Fig. 1B).

The most prominent morphological feature on the Cocos plate is the Cocos Ridge, which is formed by movement of the Cocos plate over the Galapagos hotspot. The Cocos Ridge, an aseismic ridge with a maximum crustal thickness approaching ∼20 km along the ridge axis (Walther, 2003; Sallarès et al., 2003), rises over 2.5 km above the neighboring abyssal plain, with local bathymetric relief along the ridge crest in excess of 1 km. The Cocos Ridge is cut on its eastern edge by the north-south–striking Panama fracture zone, a right-lateral transform fault that forms the boundary between the Cocos and Nazca plates (Figs. 1A and 1B). Subduction of the Cocos Ridge dramatically increases uplift rates from the Nicoya Peninsula (Marshall and Anderson, 1995; Gardner et al., 2001) in the northwest to the Osa Peninsula (Gardner et al., 1992; Sak et al., 2004a, 2009) and Burica Peninsula (Corrigan et al., 1990; Morell et al., 2011) in the southeast (Fig. 1C). The global positioning system (GPS)–derived surface velocity field in the outer forearc overriding the ridge is consistent with strong coupling along the plate boundary (Norabuena et al., 2004; La Femina et al., 2009), with horizontal velocities up to 41 m/k.y. relative to the magmatic arc along an azimuth parallel to Cocos-Caribbean plate motion. An observed gradient in the magnitude of horizontal velocity arcward (La Femina et al., 2009) is consistent with buildup of elastic strain in the upper plate (Fisher et al., 2004). Geodynamic modeling of the distribution and rates of uplift within the outer forearc are predicted most accurately when the subducted part of the Cocos Ridge contains a steep, near-vertical eastern leading edge due to truncation against the Panama fracture zone (Gardner et al., 1992) and when the subducted part of the Cocos Ridge behaves as a shallowly subducting rigid indentor (Corrigan et al., 1990; La Femina et al., 2009).

Estimates for the time of arrival of the Cocos Ridge at the Middle America Trench offshore of the Osa Peninsula and the initiation of rapid, upper-plate deformation are wide ranging and have been controversial. Estimates range from ca. 1 Ma (Lonsdale and Klitgord, 1978; Gardner et al., 1992) from modern plate reconstructions; ca. 1 Ma (Corrigan et al., 1990) and 3.6 Ma (Collins et al., 1995) based on paleobathymetry from benthic foraminiferal stratigraphy; <3 Ma (Morell et al., 2012) to as much as ca. 5–7 Ma (Gräfe et al., 2002) from unroofing of the Cordillera de Talamanca; to ca. 2–3 Ma (MacMillan et al., 2004), ca. 5 Ma (de Boer et al., 1995), and ca. 8 Ma (Abratis and Wörner, 2001) from the age and distribution of volcanic and igneous rocks in the magmatic arc inboard of the Cocos Ridge.

We favor a younger age, ca. 1.5–3 Ma, for the initiation of Cocos Ridge subduction offshore the Osa Peninsula for several reasons. First, the younger age is consistent with modern plate reconstructions, which predict that subduction of the Cocos Ridge is constricted to a time window ca. 1.5–3 Ma for the region offshore the Osa Peninsula (Lonsdale and Klitgord, 1978; Gardner et al., 1992; MacMillan et al., 2004; Morell et al., 2012). Second, river longitudinal profile analyses and the preservation of a low-relief landscape atop the magmatic arc restrict the onset of increased rock uplift induced by Cocos Ridge subduction to <3 Ma, given the tropical climate and high erosion rates (Morell et al., 2012). Third, the distribution of the youngest radiometric ages from plutons of the Costa Rica–Panama volcanic arc (e.g., de Boer et al., 2005; Wegner et al., 2011) suggests that the cessation of arc volcanism in Costa Rica may be unrelated to Cocos Ridge collision (Morell et al., 2012). This interpretation supports a younger Cocos Ridge arrival, given that many of the older estimates for the onset of Cocos Ridge subduction (e.g., 8 Ma; Abratis and Wörner, 2001) are derived from estimates related to the timing of arc inactivity. Therefore, the deformation we describe on the Osa Peninsula from subduction of the Cocos Ridge is most likely Quaternary in age.

The upper plate in southern Costa Rica is the Panama microplate (PM, Fig. 1A) and was first identified by Vergara Muñoz (1988a, 1988b) on the basis of seismicity, earthquake focal mechanisms, and bounding structural zones. The Panama microplate is separated from the Caribbean plate by a series of east-west–striking offshore thrust faults located north of the isthmus (North Panama deformed belt of Silver et al., 1990), and a diffuse zone of deformation that runs through central Costa Rica (Central Costa Rican deformed belt of Marshall et al., 2000). Offshore of the Osa Peninsula, the Cocos plate subducts under the Panama microplate along the relative plate motion vector at ∼85 m/k.y. (Fig. 1A). Here, we focus on the thin outer forearc of the Osa Peninsula along the southern edge of the Panama microplate where Cocos Ridge subduction has exhumed the Late Cretaceous to middle Miocene Osa igneous complex and Osa mélange and exposed the overlying late Neogene and Quaternary marine sediments.

Stratigraphy on the Osa Peninsula directly inboard of the Cocos Ridge ranges from Late Cretaceous oceanic basement rocks to late Neogene slope cover turbidites and finally Quaternary shallow marine, estuarine, and alluvial sediments (Fig. 2). A prominent, high-relief unconformity (Ivosevic, 1977; Lew, 1983; Barritt and Berrangé, 1987; Sak et al., 2004a) separates the basement rocks from the overlying late Neogene and Quaternary sedimentary units (Fig. 2, bold unconformity).

The outer forearc basement of the Osa Peninsula is composed of two major rock bodies, the Osa igneous complex, an exotic sliver(s) accreted to the margin of the Caribbean plate in the Paleocene and Eocene (Buchs, 2003; Denyer et al., 2006; Buchs et al., 2010) on the landward side, and the Osa mélange on the seaward side. Basement rocks on the seaward side of the Osa Peninsula are composed of the Osa-Caño accretionary complex (Di Marco et al., 1995), which is now called the Osa mélange (Buchs, 2003; Sak et al., 2004a; Vannucchi et al., 2006; Buchs and Baumgartner, 2007; Buchs et al., 2009). The mélange, characterized by a block-and-matrix fabric, is exposed along coastal wave-cut platforms, sea cliffs, and in river gorges that are incised below the basement-cover unconformity. The Osa mélange sequence consists of sheared sandstone-mudstone, marble, pelagic sediments (radiolarian chert), basalt, and gabbro. Locally, the mélange consists of intensely sheared greenstones with web-like arrays of scaly fabrics that reveal polished, striated surfaces. In many places, the boundary between clasts and matrix is a shear zone or fault. Foraminifera in the mélange matrix indicate a latest Cretaceous to middle Miocene age (Lew, 1983; Di Marco et al., 1995). The Osa mélange is regionally homogeneous, but locally heterogeneous and disorganized at a scale of <50 m (Denyer et al., 2006), with offsets along subvertical faults that define independently moving blocks <∼5 km on a side (Ivosevic, 1977; Barritt and Berrangé, 1987; Sak et al., 2004a).

Basement rocks on the Osa Peninsula are unconformably overlain by Neogene and Quaternary, locally derived, semilithified, greenish gray to orange, graywacke-type marine and continental conglomerates, sandstones, siltstones, and claystones (Ivosevic, 1977; Barritt and Berrangé, 1987). The unconformity occurs on either unweathered Osa mélange with a sharp contact into the overlying sediment or a deeply weathered paleosol locally incorporating angular, colluvial clasts of Osa mélange (Berrangé, 1989). Relief on the unconformity can exceed 10 m at the outcrop scale, but locally relief can exceed 100 m. Abrupt variations in facies architecture and thickness of the overlying sediments on the Osa Peninsula indicate that block faulting was coeval with sedimentation (Ivosevic, 1977; Barritt and Berrangé, 1987; Berrangé, 1989). Facies immediately above the unconformity can vary from subaerial colluvial and alluvial sediments to high-energy beach deposits consisting locally of molluscan-bored Tertiary limestone cobbles or shallow-water estuarine or open-marine siltstone and mudstones. Thickness of the sedimentary cover above the unconformity varies from nearly zero to almost 500 m across buried, steep paleotopography that bounds active tectonic basins and paleovalleys (Berrangé, 1989; this study).

The late Neogene and Quaternary stratigraphy of the outer forearc was initially described on the Burica Peninsula, ∼30 km southeast of the Osa Peninsula (Fig. 1C), from molluscan fauna, planktonic foraminifera, and calcareous nanofossil distributions, and it was assigned to the Pliocene Charco Azul Formation or the Pleistocene Armuelles Formation (Coryell and Mossman, 1942; Olsson, 1942; Terry, 1941, 1956). On the Burica Peninsula, a transitional, time-transgressive unconformity separates the deep-water turbidites of the Charco Azul Formation from the overlying and rapidly shallowing-upward early Pleistocene to Holocene Armuelles Formation (Corrigan et al., 1990). These units are truncated by an extensive flight of late Quaternary marine terraces of the Monteverde formation (Morell et al., 2011). The Charco Azul and Armuelles Formation were formally applied to the Osa Peninsula by Sprechmann (1984) and subsequently used by many later researchers (Fig. 2). However, some researchers, have preferred to use the informal designation Punta La Chanca formation (Lew, 1983) or Osa group (Barritt and Berrangé, 1987; Berrangé, 1989; Buchs and Baumgartner, 2007) instead of the Charco Azul Formation (Fig. 2) and the Puerto Jimenez Group (Barritt and Berrangé, 1987; Berrangé, 1989) or the Marenco formation (informal designation; Sak et al., 2004a; this study) instead of the Armuelles Formation because of significantly different tectonic histories and facies architecture between the two peninsulas. Osa group sediments have been assigned to the middle–late Pliocene based on planktonic and benthic foraminiferal assemblages and were deposited as turbidite fans in water depths ranging from 200 m to as much as 1200 m (Lew, 1983; Berrangé, 1989).

Mapping and radiocarbon dating of the late Quaternary Marenco formation have been restricted to the southeastern (Gardner et al., 1992) and coastal sections along the northwestern part (Sak et al., 2004a) of the Osa Peninsula. Facies analyses and radiocarbon dating defined two chronostratigraphic units (Fig. 2). Chronostratigraphic sequence I is composed of unconsolidated beach ridge, beach, shallow marine, and estuarine facies of mid- to late Holocene age (marine isotope stage [MIS] 1, 0–10 ka). Chronostratigraphic sequence II (Gardner et al., 1992) or the Marenco formation (Sak et al., 2004a) is composed of poorly lithified, shallow marine to estuarine facies of late Pleistocene (MIS 3, ca. 27–60 ka) age. Those initial investigations (Gardner et al., 1992; Sak et al., 2004a) provided the first radiocarbon ages for the late Quaternary stratigraphy. Gardner et al. (1992) proposed a simple tectonic model requiring regional down-to-the-northeast tilting with minor block faulting. More detailed mapping and radiocarbon dating allowed Sak et al. (2004a) to extend the late Quaternary Marenco formation to the northwest corner of the Osa Peninsula and to identify small, independently moving blocks that experienced uplift and subsidence in response to bathymetry on the subducting Coco Ridge.

The Osa Peninsula has rugged terrain with high topographic relief and steep slopes. Dense jungle vegetation covers most of the land surface. Access to the interior of the peninsula is limited to stream bottoms, footpaths, and a few dirt roads with limited access to the interior. Outcrops are small (generally less than 10 m in width or height) and occur mostly along wave-cut coastal cliffs, stream banks, and landslide scars along valley side slopes. Under these conditions, geologic mapping is limited to generally small outcrops, and correlation to neighboring outcrops is based on radiometric ages, similar facies types, or projection along strike. Most faults were poorly exposed and observed only over short distances. Elevation changes of 50–100 m over short distances for samples of similar ages allowed for projection of faults along strike.

The Quaternary sedimentary cover on the Osa Peninsula is sparsely fossiliferous. Samples used for radiocarbon dating consist of collections of articulated mollusks, disarticulated but unabraded, thin-walled mollusks, and unabraded, thin-walled gastropods. Fossiliferous samples were subjected to scanning electron microscope (SEM) analysis to look for evidence of recrystallization, and any recrystallized samples were discarded. Organic, woody debris, consisting of mangrove seed pods, twigs, and leaves and unidentified bark, was air and oven dried immediately after collection. All woody debris and unrecrystallized shelly material (after SEM analysis) that we observed in any outcrop was submitted for radiocarbon dating.

OSL samples were collected from high-energy beach facies and above-wave-base rippled fine sand facies by driving 40-mm-diameter polyvinyl chloride (PVC) or aluminum tubes into cleaned, vertical surface exposures more than 2 m below the ground surface. Sampling those facies maximized the potential for full exposure to sunlight during sediment transport and deposition. Laboratory and analytical procedures for OSL analysis are given in Appendix 1.

The Marenco formation consists of poorly lithified, poorly consolidated, predominantly marine sediments. Facies are assigned to depositional environments and water depths by comparison to modern environments around the Osa Peninsula (Gardner et al., 1992; Sak et al., 2004a). The most conspicuous deposits are high-energy, sand to pebble and cobble beach and tidal facies (Fig. 3A). This facies is assigned a water depth of 0.0 m ± 1.2 m. Pebble composition is predominantly basalt, with mixtures of chert, limestone, and lithic fragments derived from the Osa mélange or Osa igneous complex. Limestone cobbles are frequently bored and locally comprise most of the cobbles. Pebbles vary from well rounded to angular (Fig. 3B), with basalt clasts being the most angular. The coarsest and most angular clasts tend to occur in high-energy deposits closest to the unconformity where paleorelief is highest. These facies can locally contain a diverse assemblage of thick-walled mollusks and gastropods.

High-energy beach and tidal facies interfinger laterally and vertically into fossiliferous, shallow shelfal, wavy to ripple-bedded, interbedded, brown, often bioturbated, fine sands and silts (Fig. 3C). These sediments were deposited above wave base and are assigned a facies depth of −4.5 m ± 4.5 m. In many locations away from paleotopography, sediments in the Marenco formation are plane parallel-bedded, jointed, gray to olive green mudstones and siltstones (Fig. 3D) that may contain a more limited assemblage of thin-walled gastropods and articulated mollusks. These facies are assigned to a water depth below wave base of >−9 m. Locally, mangrove-lagoon-swamp facies with woody debris and leaf impressions on bedding planes are well preserved, but are volumetrically insignificant. This facies is light- to dark-gray clay with in situ mangrove stumps and rooted horizons.

Here, we report eight new and 10 previously published radiocarbon ages (Table 1) and five new OSL ages (Table 2) for the Marenco formation in the central part of the Osa Peninsula. These ages revise and elaborate on published radiocarbon ages for coastal sections of the Marenco formation on the northwest (Sak et al., 2004a) and southeast (Gardner et al., 1992) part of the Osa Peninsula. The new ages allow us to subdivide the Marenco formation into three chronostratigraphic members (Fig. 2, last column; Tables 1 and 2): the Jiménez member (MIS 1; modern to ca. 10 ka), the Tigre member (MIS 3; ca. 27 ka to ca. 60 ka), and the Rincón member MIS ≥5.1; older than ca. 80 ka). Importantly, we now recognize an older member, the Rincón member. These new ages for the Marenco formation and new mapping within the interior of the peninsula allow us to produce a revised geologic map for the Osa Peninsula (Fig. 4) that significantly reduces the areal extent of the Osa mélange as mapped by Sak et al. (2004a) and the Osa group as mapped by Buchs and Baumgartner (2007) while expanding and refining the areal extent of the Marenco formation, mapped by Coates et al. (1992) and Vannucchi et al. (2006) as their Armuelles Formation (Fig. 2).

Radiocarbon ages compare favorably within outcrops. Radiocarbon samples from the same outcrop give similar ages, and stratigraphically overlying beds yielded younger radiocarbon ages within statistical error. For example, radiocarbon samples 5 and 6, adjacent to each other, (Table 1; Fig. 5, N-S cross section) yield ages of 35.2 ka ± 1.8 ka and 35.5 ka ± 0.5 ka for upper beds in the Tigre member, which are statistically indistinguishable. Samples 8 and 10 (Table 1) from stratigraphically lower beds in the Tigre member yield ages of 39.9 ka ± 1.2 ka and 40.6 ka ± 0.6, also statistically indistinguishable, but older than ages for the overlying beds.

Where radiocarbon samples yield infinite ages, optical results give finite ages slightly older than the minimum radiocarbon age. For example, radiocarbon samples 13 (older than 34.62 ka), 15 (older than 42.2 ka), and 16 (older than 40.04 ka), from the lowest part of the Tigre member just above the unconformity with the Rincón member (Table 1; Fig. 5, N-S cross section), yield different infinite ages (depending on sample size), but the optical age for that outcrop, sample 21, is 53 ka ± 11 ka (Table 2). Similarly, radiocarbon sample 14 yields an infinite radiocarbon age (older than 38 ka; Table1), but the optical age for that outcrop, sample 22, is 53 ka ± 12 ka (Table 2). There was one inconsistent age between dating techniques. Radiocarbon samples 11 and 12 from lower beds in the Tigre member (Table 1; Fig. 5, N-S cross section) give ages of 42.6 ka ± 0.5 ka and 47.8 ka ± 0.7 ka, but the optical age for that outcrop, sample 19, is statistically younger, 33 ka ± 7 ka (Table 2). However, another optical analysis, sample 20, from an outcrop immediately overlying samples 11, 12, and 19 gave an optical age of 39 ka ± 8 ka (Table 2), i.e., statistically younger than the radiocarbon samples.

Soils formed on the Tigre member have B horizons that are poorly expressed with brown, 5-7.5 YR Munsell colors. The thicknesses of weathering rinds on basaltic clasts are generally >2 mm. These values are consistent with soil development and rind thicknesses on basaltic clasts for MIS 3 terraces on the Osa Peninsula (Bullard, 1995), along the central Pacific coast of Costa Rica (Murphy, 2002; Sak et al., 2004b; Fisher et al., 2004), and on the Burica Peninsula (Morell et al., 2011), further supporting the radiocarbon and OSL ages.

The oldest optical age, sample 21 from the Rincón member, is 109 ka ± 28 ka (Table 2; Fig. 5, N-S cross section). Given the large analytical error, we assign this sample to MIS 5.5 at ca. 125 ka, which would be the highest and longest lived of the MIS 5 eustatic sea-level highstands. This age assignment is also consistent with (1) the development of moderately thick, red (Munsell 10R-2.5 YR) soil B-horizon and (2) weathering rind thicknesses of ∼2 cm in basaltic clasts, similar to other MIS 5.5 deposits in southern Costa Rica (Bullard, 1995; Murphy, 2002; Sak et al., 2004b; Fisher et al., 2004). It is not possible to resolve the Rincón member into separate, older marine isotope stages with the available OSL dating.

Samples 3 and 4 from a freshwater lacustrine deposit within the Tigre member yielded radiocarbon ages of 30.9 and 32.1 ka (Table 1; Fig. 5, N-S cross section at km 4.5). These samples are well-preserved, delicate leaves on bedding planes in thin, horizontally bedded muds. Because these samples are from lacustrine deposits that formed behind a landslide dam, they are not useful for uplift rate calculation because the elevation at the time of deposition is unknown.

We interpret the distribution (Fig. 4), facies architecture, thickness, and structure (Fig. 5) of the late Quaternary Marenco formation on the Osa Peninsula to be controlled by interaction among eustatic sea-level fluctuations, paleotopography on the major unconformity with basement rocks and the Osa group, and the nature of block faulting (see next section). For deposits of the Marenco formation within the range of Quaternary eustatic sea-level fluctuations, between ∼+6 m and –∼125 m, eustatic sea-level highstands (MIS 1, 3, 5, etc.) allow for transgressions and deposition in topographically low, subsiding, fault-bounded basins around emergent paleotopographic highs.

Eustatic sea-level lowstands and regression produced internal unconformities between members (Fig. 2, last columns).

Geologic mapping and new radiometric ages throughout the central part of the peninsula allow us to construct two structural cross sections across the peninsula (Fig. 5). This has not been possible in previous studies (Gardner et al., 1992; Sak et al., 2004a) because of the geographic limits of those studies to mostly coastal sections.

Bedding in the sedimentary cover above the unconformity dips generally to the north-northeast around 10° to 30°, although dip can be locally quite variable in direction and steepness (Fig. 4B). Dips tend to be steeper in the older Osa group sediments than in the younger Marenco formation. Bedding strikes generally west to northwest, but it also can be highly variable. The general northeast dip and northwest strike probably reflect shortening from out-of-sequence, landward-dipping thrust faults offshore (Barritt and Berrangé, 1987; Vannucchi et al., 2006) that formed during initial subduction of the Cocos Ridge.

Mapped faults form an intersecting network that controls the topography and the orientation of streams and valleys (Fig. 4). However, most faults cannot be traced along strike very far in the dense jungle vegetation. The dominant set of faults strike northwest parallel to the trench and northeast perpendicular to the trench (Lew, 1983; Meschede et al., 1999; Sak et al., 2004a; Vannucchi et al., 2006; Buchs et al., 2009; this study). Faults extend through all members of the Marenco formation to the surface (Fig. 3E). These subvertical, through-going faults (Barritt and Berrangé, 1987) record an active period of pervasive brittle deformation (Sak et al., 2004a; Vannucchi et al., 2006; this study) and, as we will suggest later, extend through the Osa mélange to the plate boundary (Vannucchi et al., 2006). These faults are the first-order control on the structural relief of the basement Osa mélange (Fig. 5). Fault motion is predominantly dip slip with both normal and reverse sense of motion, similar to that observed on the northwest coast (Sak et al., 2004a) and central coast (Lew, 1983; Vannucchi et al., 2006). Fault slip indicators show that some faults are reactivated with the opposite sense of motion (Vannucchi et al., 2006), consistent with inferred up-down motion that controls deposition of the Marenco formation (Sak et al., 2004a).

These active faults cut the peninsula into a set of independent, actively deforming blocks. Most blocks are less that 5 km on a side. One such block locally ponds drainage in the northern interior of the peninsula (Fig. 4A, Laguna Chocuaco). The north-south cross section (Fig. 5) contains numerous faults that cut the section into at least six distinct, independently moving blocks. One of the larger, less-elevated blocks defines the Laguna Corcovado swamp and central graben, one of the more prominent topographic features on the peninsula (Fig. 3F; Fig. 5, E-W cross section). Higher topography on the most elevated blocks is underlain by the older (MIS ≥5) Rincón member of the Marenco formation. Distribution of the younger (MIS 3) Tigre member of the Marenco formation is confined to lower elevations on less-elevated blocks. As we will show later, this topography and vertical deformation on the Osa Peninsula correlates directly to bathymetry on the subducting Cocos Ridge, immediately offshore at the Middle America Trench only 20 km to the south.

We define the block structure on the Osa Peninsula by outcrop distribution of the Marenco members, fault boundaries, and differences in uplift rates between adjacent blocks (Figs. 4, 5, and B6B). The N-S cross section (Fig. 5) can be divided into six independently moving blocks with distinctly different rates of uplift. Rate of uplift is greatest, 5.5–8.5 m/k.y., on fault block 3 (Fig. 6B) in the central part of peninsula. Here, the Tigre member is inset into the Rincón member. Although this fault block has the highest time-averaged uplift rate, it is not the topographically highest block on the peninsula. The topographically highest block in the N-S cross section is fault block 2, which is capped by the Rincón member. On fault block 5, the thickness of the Tigre member, >∼100 m, is in excess of the sea-level rise from MIS 4 to MIS 3, i.e., ∼80 m, requiring subsidence of at least 20 m during MIS 3. Subsidence of individual blocks has also been reported along the northwest coast (Sak et al., 2004a).

Rates of uplift decrease systematically away from the central part of the peninsula toward both coasts in both cross sections. However, in the E-W cross section, the Rincón member is dropped down several hundred meters in the central graben (Fig. 5, E-W cross section) relative to adjacent blocks, which have the highest topography on the Osa Peninsula, in excess of 500 m to 700 m. On the E-W cross section, uplift rates decrease toward the coast, where the Tigre member (Gardner et al., 1992; Sak et al., 2004a) and locally where the Jiménez member (Gardner et al., 1992) are exposed. We will show in the next section that this topography, structural relief, and distribution of uplift rates are closely correlated to the bathymetric relief on the subducting Cocos Ridge.

A first-order model for uplift of the Osa Peninsula through time can be constructed from the uplift rates and eustatic sea levels that are appropriate for the age of a sample. In this model (Fig. 7), we assume a time-averaged, spatially uniform, maximum uplift rate of ∼4 m/k.y. These model assumptions do not reproduce the subtleties of individual block movements nor reproduce exact facies depths for individual samples, but the model does predict the general uplift and timing for exposure of the Osa Peninsula above sea level and does reproduce the distribution of facies in Marenco formation rather well. Samples and facies from field observations (Fig. 3) are located on each model time step. The model begins (Fig. 7A) when the Osa Peninsula becomes subaerially exposed at MIS 5 during deposition of the upper part of the Rincón member. The high-energy, angular to rounded, basalt cobble beach facies (Fig. 3B) can be seen developing close to the small islands that first appear above sea level, exactly where such a facies would be deposited. Early MIS 3 (Fig. 7B) and late MIS 3 (Fig. 7C) times produces expanding islands with a large tidal channel between the two major islands. The high-energy, sandy to pebbly beach and tidal facies of the Tigre member of the Marenco formation (Fig. 4A) is found along one big island in the rapidly constricting tidal channel between the islands. This is exactly where one would expect to find such facies. With continued uplift, the islands expand, and the opening to the sea constricts (Fig. 7C), reducing wave energy and allowing for deposition of wavy, ripple-bedded fine sands and silts (Fig. 3C) and plane parallel-bedded mudstones and siltstones (Fig. 3D) from the Tigre member of the Marenco formation. The Osa Peninsula reaches its maximum subaerial extent during MIS 2 (Fig. 7D), severely restricting the Gulfo Dulce and allowing for development of an unconformity on the Tigre member. During the sea-level transgression to MIS 1, the Jiménez member is deposited along the coast overlapping the Tigre member (Gardner et al., 1992).

The spatial distribution of deformation rates, topography, and structural fabric are, to the first order, controlled by bathymetry on the subducting Cocos Ridge. In fact, a stunning relationship exists between bathymetry on the subducting Cocos Ridge and topography and deformation on the Osa Peninsula (Fig. 8). This is true, in part, because the upper plate is relatively thin under the Osa Peninsula and can consequently deform in response to small (>0.5 km relief) bathymetric features on a stronger subducting plate. Estimates of the dip of the Wadati-Benioff zone under the Osa Peninsula, and thus the thickness along the seaward edge of the Osa mélange, are poorly constrained because of the diffuse seismicity (Protti et al., 1995). Estimates range from as low as 3° (3 km thickness; Kolarsky et al., 1995) to ∼10° (∼10 km thickness; Protti et al., 1994, 1995) to possibly as much as 20° (15 km thickness; Corrigan et al., 1990; DeShon et al., 2003; Norabuena et al., 2004). However, estimates of plate dip derived from historical seismicity by DeShon et al. (2003) and Norabuena et al. (2004) are northwest of the Osa Peninsula along the central Pacific coast, where the plate boundary is better imaged and defined and more steeply inclined. Therefore, the thinner estimates, ∼3 km to 10 ∼km, are more reasonable.

The thin basement that makes up the bulk of the margin wedge on the Osa Peninsula must deform around subducting bathymetric features. This leads to a strong correlation between topography and bathymetry. Topographically low areas on Osa (the central graben, the eastern lowlands, and the opening into the Gulfo Dulce) line up along the relative plate motion vector with the axial and lateral grabens on the Cocos Ridge. The western uplands and the central ridge on Osa line up with the flanking ridges on the western lateral graben (Fig. 8). The highest peak on Osa, the unnamed peak at 780 m elevation, lines up with the shallowest bathymetry immediately offshore on the Cocos Ridge (Fig. 8, cross-sections A-A′ and B-B′). However, linear features on the Cocos Ridge are not parallel to the relative plate motion vector between the Cocos plate and Panama microplate. Based on the obliquity and orientation of the Middle America Trench, these features move northwestward through time (Fig. 8, white arrows). This is clearly seen as a scarp along the margin wedge immediately inboard of the trench where the eastern flanking ridge of the western lateral graben is subducting (Fig. 8). On Osa, this northwestward drift of the flanking ridge is also manifested in the curvature of the coast between the central ridge and eastern lowlands (Fig. 8).

Bathymetry on the Cocos Ridge also correlates closely to structure and uplift rates on the Osa Peninsula. The structurally highest areas are along the central ridge where the Osa mélange reaches elevations approaching 400 m (Fig. 5, E-W cross section). The central graben, which is inboard of the lateral western graben, is structurally lower that the bounding central ridge and western uplands (Fig. 5, E-W cross section). The highest uplift rates are recorded along the central ridge (Figs. 5 and 6, fault block 3) and decrease toward the coast, both along strike of the ridge (N-S cross section in Figs. 5 and 6B) and perpendicular to the ridge (Fig. 5, E-W cross section, locations A and B).

Fault block 3 has the highest time-averaged uplift rate, 5.5–8.5 m/k.y. (Fig. 6A), but it is not the topographically highest block on the peninsula (Fig. 5). The topographically highest block in the N-S cross section is fault block 2, which is capped by the Rincón member. Two possible explanations are possible for this distribution of uplift rates on these two blocks. In the first possibility, block 2 has been uplifting at a slower rate for a longer time, producing a higher elevation than fault block 3. However, it is also possible that fault block 2 did, in the past, experience a rate of uplift that is similar to the current rate of uplift on fault block 3, but then began to subside with passage of a subducting seamount, in essence reversing the sense of motion. In this case, the long-term, average uplift rate would be lower for block 2. We prefer the second scenario because previous stratigraphic (Sak et al., 2004a) and structural (Vannucchi et al., 2006) studies have shown that fault motions can reverse, and blocks with a history of uplift can subsequently subside—similar to upper-plate deformation from seamount subduction in the New Hebrides and Solomon Arcs (Taylor et al., 2005). We interpret these vertical motions as a wave of rapid uplift followed by subsidence that propagates across the peninsula, parallel to the relative plate motion vector along the plate boundary, and tracks variations in the highest bathymetry on the subducting Cocos plate. Structures such as faults record permanent deformation as the upper plate deforms to override the curvature in the subducting plate, but the topography is ephemeral, rising and collapsing in response to subducting bathymetry on the Cocos Ridge. Given a Quaternary uplift rate distribution that is largely due to variations in the bathymetry of the underthrusting plate, the permanent deformation and uplift on the Osa Peninsula should occur coseismically when the roughness of the downgoing plate is episodically forced under the margin wedge as described next.

In the outer forearc, where the upper plate is only several kilometers thick, the margin wedge must deform to allow for subduction of rigid asperities on the incoming plate (Fig. 8; Fisher et al., 2004; Sak et al., 2004a). Flanking ridges and seamounts on the Cocos Ridge are composed of basaltic crust that is stronger than the much thinner and extensively faulted Osa mélange in the upper plate (9A–9D). In such a system, the strain rate in the upper plate is a function of the relative plate motion rate and straining distance (Knipe, 1985), a distance that is a function of the curvature of subducting bathymetric features (Sak et al., 2004a). Straining regions (Fig. 9E, in red) experience up-shearing and down-shearing expressed as subvertical faults. We believe these active faults, which are extensive on the Osa Peninsula (Figs. 4 and 5), extend from the surface to the plate interface because the upper plate is thin, 3–10 km thick, and because of the close correspondence of bathymetry on the Cocos Ridge to topography, uplift rates, and structural relief on the Osa Peninsula.

The rate of uplift or subsidence is determined by the slope of the asperity parallel to the relative plate motion vector and the rate of relative plate motion (Fig. 9F). Assuming an average slope of ∼7° for a flanking ridge that is at the trench (Figs. 8 and 9F) and a relative plate motion rate of ∼50 m/k.y. (the difference between the Cocos–Panama microplate relative plate motion rate and the rate of out-of-sequence thrusting in the Fila Costeña; Sitchler et al., 2007), we calculate a time-averaged uplift rate of ∼4 m/k.y. This value, used in Figure 7 to produce the time-averaged emergence of the Osa Peninsula above sea level, reproduces the regional uplift of the Osa Peninsula over the last glacial cycle and agrees well with the distribution of facies and ages of sediments. It is generally consistent with values for rates of uplift ranging from ∼2 m/k.y. to 6 m/k.y. and subsidence up to ∼6 m/k.y. along the northwest coast (Sak et al., 2004a) and uplift rates ranging from ∼2 m/k.y. to 6 m/k.y. along the eastern coast (Gardner et al., 1992). Therefore, we interpret the rates of vertical motion across the overriding outer forearc to be a function of the rate of relative plate motion and the complex surface geometry of the underthrusting plate. Under these circumstances, a margin may experience rapid uplift during subduction of the leading edge of a bathymetric high followed by subsidence along the trailing edge (Ballance et al., 1989; Dominguez et al., 1998; Mann et al., 1998; Sak et al., 2004a), with the amount of uplift or subsidence related to the volume of the bathymetric feature (Meffre and Crawford, 2001). Uplift is due not to shortening, but rather to the positive slope of the underthrusting ridge or seamount. Given the thin upper plate, even small asperities (<0.5 km height) will deform the upper plate.

We do allow for some transfer of mass from the upper plate (subduction erosion) along hanging-wall shortcuts that can develop at sharp curvatures around subducting bathymetric features (Fisher et al., 2004; Sak et al., 2004a). Similarly, we allow for some transfer of mass to the upper plate by underplating at the point arcward of where the thickness and strength of the upper plate are sufficient to shear off asperities (Fig. 9D). However, very little current uplift in the outermost forearc is attributable to permanent shortening, as occurs inboard of the d’Entrecasteaux Ridge in the New Hebrides Arc (Taylor et al., 2005). In the inner forearc, where the upper plate is thicker than in the outer forearc, small-wavelength bathymetric features have no effect on the uplift and shortening patterns, and the lateral distribution of shortening related to inversion of the Térraba basin is largely determined by the width and thickness of the Cocos Ridge and the regional plate-boundary stress related to flat slab subduction (Sitchler et al., 2007; Figs. 9C and 9D). Underplating and subduction erosion ultimately result in downdip smoothing of the plate boundary, but these are currently secondary processes with regard to the uplift pattern in the outer forearc.

The keys on a player piano serve as a good mechanical analogy for the independently moving blocks on the Osa Peninsula. The keys (blocks on Osa) move up and down in response to teeth on the metal spool (bathymetry on the Cocos Ridge) that rotates and activates the keys on the player piano. The key is depressed or elevated by the length of the tooth and rate of rotation of the music spool with very little transfer of mass from one plate to the other. The thin, outer wedge represented by Osa mélange deforms in response to strain that accumulates along northeast- and northwest-trending, high-angle faults that extend from the surface to the plate boundary (Fig. 5). Small blocks, several kilometers on a side (Fig. 5; Sak et al., 2004a; Vannucchi et al., 2006), uplift and subside (riffling; von Huene et al., 2004) in response to asperities on the subducting Cocos Ridge. Uplift and subsidence occur as a passive response to variations in relief on the Cocos Ridge (Adamek et al., 1987). Along this region of the margin, there was an initial period of protracted uplift, possibly as much as 2–3 km (Corrigan et al., 1990; Gardner et al., 1992), as the blunt front end of the Cocos Ridge collided with the outer part of the margin wedge. Superimposed on this uplift are shorter time scale variations related to the along-axis roughness of the ridge itself. These uplift “events” are ephemeral in the sense that topographic collapse follows uplift as three-dimensional bathymetric features along the crest of the ridge subduct underneath the Osa Peninsula.

New radiocarbon and OSL ages allow for refinement of the Quaternary stratigraphy and provide constraints on outer forearc deformation along the Middle America Trench in southern Costa Rica. On the Osa Peninsula, the Marenco formation is subdivided into three members: the Jiménez member (MIS 1), the Tigre member (MIS 3), and the Rincón member (MIS ≥5). Sediments in all members of the Marenco formation are predominantly marine. The mapped distribution, facies architecture, and structure of the late Quaternary Marenco formation on the Osa Peninsula are controlled by interaction among eustatic sea-level fluctuations, paleotopography on the major unconformity with basement rocks and the Osa group, and block faulting, with thickness of the Marenco formation ranging from 0 m on paleotopographic highs to nearly 300 m in fault-bounded basins.

The spatial distribution of deformation rates, topography, and structural fabric in the outer forearc are, to the first order, controlled by short-wavelength, high-relief bathymetric features on the downgoing Cocos Ridge. This is true, in part, because the upper plate is relatively thin, ∼3 km to ∼10 km, under the Osa Peninsula. The dominant set of active faults strike parallel (northwest) and perpendicular (northeast) to the trench and extend from the surface to the plate boundary. These faults cut the peninsula into a set of independently deforming blocks, ∼5 km on a side, and have been active since the arrival of the Cocos Ridge at the Middle America Trench (ca. 1–3 Ma). Time-averaged uplift rates vary rapidly across individual blocks, ranging from 1.7 m/k.y. to 8.5 m/k.y. Variations in uplift amount, uplift rate, and topography on the Osa Peninsula are primarily tied to variations in the bathymetry of the Cocos Ridge. The deformation on the Osa Peninsula is ephemeral in the sense that topographic collapse and subsidence follow uplift as three-dimensional bathymetric features continue to subduct. In the outer forearc, where the margin wedge is thin, with very little sediment on the subducting plate, subducting bathymetry is more important for the deformation of the overriding plate than any permanent deformation that accrues during interseismic locking of the plate boundary and buildup of elastic strain.

The samples were processed under subdued red light, with the 90–125 µm quartz fraction extracted for dating using standard procedures (e.g., Galbraith et al., 1999). A single-aliquot regenerative-dose protocol was used to calculate equivalent doses (Murray and Roberts, 1998; Galbraith et al., 1999; Murray and Wintle, 2000).

Approximately 100 aliquots per sample, each composed of single grains of quartz, were preheated at 240 °C for 10 s and optically stimulated for 2 s at 125 °C by green (532 nm) light from a solid-state laser beam attached to an automated Risø OSL/TL-DA-15 apparatus. Ultraviolet luminescence was detected using a photomultiplier tube with a 7.5 mm U-340 filter. Samples were then given applied doses using a calibrated ⁹⁰Sr/⁹⁰Y beta-source and re-stimulated to record their regenerative OSL signals. OSL sensitivity changes in the quartz crystals between the natural and regenerative cycles were monitored after each optical stimulation, using test doses of 10 Gy following a 160 °C cut-heat.

Output from the Risø apparatus was analyzed using Analyst version 3.21 software (Pirtzel, 2006). OSL data were corrected for any sensitivity changes, and dose-response curves were constructed using six regenerative dose points. Estimates of equivalent dose were obtained from the intercept of the regenerated dose-response curve with the natural luminescence intensity. Optical ages were derived from weighted mean equivalent dose using the central age model of Galbraith et al. (1999). The maximum age model of Olley et al. (2006) was used to estimate the greatest finite equivalent dose for sample CR07–05.

K, U, and Th concentrations were measured using instrumental neutron activation analysis (INAA) by Becquerel Laboratories, Mississauga, Ontario, Canada, and converted to beta dose rates using the conversion factors of Adamiec and Aitken (1998). A beta attenuation factor of 0.93 ± 0.03 (Mejdahl, 1979) was assumed. Gamma dose rates were measured in the field using a portable spectrometer and converted to dry values by oven drying sediment from the sample location for samples CR05–01, CR05–03, and CR05–4. Gamma dose rates for samples CR07–05 and CR07–09 were derived from the INAA values using the conversion factors of Adamiec and Aitken (1998).

Cosmic-ray dose rates were determined from established equations (Prescott and Hutton, 1994), allowing for sample depth, sediment density, and site altitude and latitude. Present-day field-moisture contents of the sediments were measured immediately after sample collection from a subsample and were considered broadly representative of long-term averages and used to correct attenuation of beta and gamma rays by water (Aitken, 1998).

We are grateful to Liz and Abraham from Bosque del Rio Tigre for their kind generosity, lodging, and great food. D. Scholl, D. DeVecchio, and an anonymous reviewer offered constructive comments on a previous version of this manuscript. This research was supported by National Science Foundation grants EAR-0337456 (Fisher) and EAR-0337467 (Gardner).