The eastern part of the oceanic Cocos Plate presents a heterogeneous crustal structure due to diverse origins and ages as well as plate-hot spot interactions which originated the Cocos Ridge, a structure that converges with the Caribbean Plate in southeastern Costa Rica. The complex structure of the oceanic plate directly influences the dynamics and geometry of the subduction zone along the Middle American Trench. In this paper an integrated interpretation of the slab geometry in Costa Rica is presented based on 3-D density modeling of combined satellite and surface gravity data, constrained by available geophysical and geological data and seismological information obtained from local networks. The results show the continuation of steep subduction geometry from the Nicaraguan margin into northwestern Costa Rica, followed by a moderate dipping slab under the Central Cordillera toward the end of the Central American Volcanic Arc. Contrary to commonly assumed, to the southeast end of the volcanic arc, our preferred model shows a steep, coherent slab that extends up to the landward projection of the Panama Fracture Zone. Overall, a gradual change in the depth of the intraplate seismicity is observed, reaching 220 km in the northwestern part, and becoming progressively shallower toward the southeast, where it reaches a maximum depth of 75 km. The changes in the terminal depth of the observed seismicity correlate with the increased density in the modeled slab. The absence of intermediate depth (> 75 km) intraplate seismicity in the southeastern section and the higher densities for the subducted slab in this area, support a model in which dehydration reactions in the subducted slab cease at a shallower depth, originating an anhydrous and thus aseismic slab.
The southeastern end of the Middle American convergent margin is characterized by the segmentation of the subducting oceanic lithosphere and a heterogeneous crustal basement on the overriding plate. The oceanic Cocos Plate is composed of crustal segments generated at two different spreading centers and affected by hot spot interaction (Barckhausen et al., 2001). This diverse morpho-tectonic environment on the oceanic plate causes heterogeneities that carry over to the subduction zone and affect the manner in which plate interaction occurs. Differences in the response of the lithosphere upon arrival to the trench are recognizable, starting from the structure of the outer rise, the morphology of the margin and the characteristics of the subduction zone (von Huene et al., 2000). This paper focuses on the subduction zone, geometry the density distribution of the subducting slab under Costa Rica, and on the relationship between the density structure and the seismicity distribution. The 3-D density modeling based on interpretation of the satellite-derived gravity data is constrained by seismological information from local networks. This joint interpretation of seismological and potential field data allows for an integrated analysis of slab geometry and its interaction with the overriding plate. Determining the slab structure is a major task in the study of subduction zones, which provides a platform to gain a better understanding of subduction processes. Further studies, such as thermal and stress modeling, geochemical and petrological models, and seismic hazard assessments, would benefit from detailed knowledge of the slab and overriding plate configuration.
The Central American Isthmus is located in the western edge of the Caribbean Plate (Fig. 1), a predominantly convergent margin defined mainly by the subduction of the oceanic Cocos Plate along the Middle American Trench (MAT). Costa Rica is located in the eastern portion of the isthmus, where subduction of the Cocos Plate along the MAT is bound by the Panama Fracture Zone (PFZ) – a right-lateral shear zone, acting as a transform boundary between the Cocos and the Nazca Plates. The overall structure of the southeastern part of the Cocos Plate is highly heterogeneous due to the multiple origins of the highly segmented crust (Fig. 1). Segments of oceanic lithosphere created at the East Pacific Rise (EPR) and at the Cocos–Nazca Spreading Center (CNS) are present off the coast of Costa Rica (Barckhausen et al., 2001). The latter shows direct influence on the Galapagos Hot Spot, adding to the heterogeneous nature of the plate and resulting in three different morphotectonic domains recognized by von Huene et al. (2000): a northwestern section with smooth relief, which contrasts with a central seamount province outlining a rough–smooth boundary, and a southeastern domain characterized by the Cocos Ridge. These morphotectonic domains correlate with significant changes in Moho depth for the Cocos Plate from 8 to 10 km for the EPR section (Sallarès et al., 2001), to 10 to 12 km for the seamount province (Walther, 2003; von Huene et al., 2000) and a maximum of 18.5 km for the crust of the Cocos Ridge (Sallarès et al., 2003).
The arrival of the Cocos Ridge to the trench and its effects on the subduction zone are still controversial subjects. Arrival time estimations range from 5–1 Ma, but the latest researches place the event at the early Pleistocene (Vannucchi et al., 2013) or late Pliocene (Morell et al., 2014). Hypothesis about the tectonic style include collision (LaFemina et al., 2009), flat subduction (Kolarsky et al., 1995), and steep angle subduction (Arroyo et al., 2003; Dzierma et al., 2011). Shallow or flat subduction is still commonly referred to explain forearc shortening, regional uplift, and Pleistocene decrease and termination of volcanic activity in southeastern Costa Rica (e.g., Fisher et al., 2004; Sitchler et al., 2007). However, tectonic reconstructions (MacMillan et al., 2004; Lonsdale, 2005) and recent results from local ocean drilling (Vannucchi et al., 2013) require further tectonic events –besides the arrival of the Cocos Ridge to the trench – to explain the evolution of this region.
Convergence rates are variable for the different domains with 85 mm yr
Tectonic setting of the Central American Isthmus. White
lines show plate boundaries and major tectonic structures. Location of
Quaternary volcanoes (black triangles) modified from Siebert and Simkin (2002). White arrows show direction and rate of plate motions relative to
the Caribbean Plate (fixed) according to DeMets et al. (1994, 2010) and DeMets (2001). Plate boundaries modified from
Coffin et al. (1998) and Lonsdale (2005). Black lines depict the
coastline and international borders. C.R.: Costa Rica; E.S.: El Salvador;
Guatm.: Guatemala; Jam.: Jamaica; CCRDB: Central Costa Rica Deformed Belt;
MPFS: Motagua–Polochic Fault System; ND: Nicaragua Depression; NPDB: North
Panamá Deformed Belt; PFZ: Panamá Fracture Zone; RSB: rough–smooth
boundary modified from Hey (1977); SITF: Swan Islands Fault; SR: Sandra
Rift (de Boer et al., 1988). White contour represents the
Subduction related volcanism occurs along the Central American Volcanic Arc
(CAVA), which begins at latitude 15
Seismic tomography studies have been carried out in Nicaragua and
northwestern Costa Rica by (Syracuse et al., 2008; DeShon et al., 2006),
in central Costa Rica by (Arroyo et al., 2009; Dinc et al., 2010);
likewise, a comprehensive study of the Costa Rican subduction zone was
performed by Husen et al. (2003b). A previous work on
the geometry of the subduction zone based on earthquake hypocenters was
published by Protti et al. (1994), who describe a segmentation of
the subducting slab in northwestern Costa Rica and interpret a change in its
dipping angle as a tear or sharp contortion. Moreover, those authors propose
a termination of the deep intraslab seismicity in southeastern Costa Rica
and interpret it as a shallow underthrusting of the Cocos Ridge. Like Protti
et al. (1994), Husen et al. (2003b) also observe a
gradual decrease in the maximum depth of the intraslab seismicity from
northern to southern Costa Rica, but they did not find evidence of a slab
tear. More recently, Dzierma et al. (2011) modeled a
steeply dipping slab to a depth of approximately 70 to 100 km based on
receiver function analysis for the northwestern part of the Talamanca
region. Furthermore, local earthquake data from a temporal network show
evidence of a steep slab down to
Eastward from the PFZ, the boundary between the Caribbean and the Nazca plates in southern Panama is marked by a segment of the trench which, according to Lonsdale and Klitgord (1978), has transitioned since the Late Miocene to a strike-slip boundary after a period of 18 to 20 Ma. of subduction. However, de Boer et al. (1988) propose a reactivation of the subduction 3.4 to 5.3 Ma. ago, and consider the existence of a recent low angle subduction.
The Caribbean coast of Costa Rica appears segmented into a northwestern section extending to the Hess Escarpment and the slope of the Nicaragua Rise, and a southeastern section defined by the North Panama Deformed Belt. According to Marshall et al. (2000), this fold-and-thrust belt is linked with the MAT through the Central Costa Rica Deformed Belt, a diffuse fault zone outlining the western boundary of the Panama Microplate.
Recent seismological studies (DeShon et al., 2003, 2006; Arroyo et al.,
2009) have been used to constrain the slab geometry of
the gravity model down to depths of
The earthquake catalogue from Husen et al. (2003a) consists of nearly 4000 events recorded in the period 1984–1997 by the two permanent networks in Costa Rica, RSN and the Costa Rican Volcanological and Seismological Observatory at the National University of Costa Rica (OVSICORI-UNA in Spanish). Those authors used the data to derive a tomographic model of the whole country.
The RSN records seismic activity since the early 1980s, mainly with short-period, vertical-component stations, up to 2010 (Fig. 2), when new broadband equipment was acquired. This study contemplates the period 2004 back to 1991, the year when digital recording started. During that interval, good station coverage was achieved only for central Costa Rica. Because of this, two additional local catalogues were examined in order to add to the amount of well-constrained seismicity toward the northwest and southeast. In northwest Costa Rica, the RSN operates the permanent network from Arenal and Miravalles Seismological and Volcanological Observatory (OSIVAM) since 1994 (Fig. 2). Originally intended to monitor the seismic activity around the Arenal and Miravalles volcanoes, the network has expanded in the last decade with two more permanent subnetworks around the Tenorio and Rincon de la Vieja volcanoes and several temporary projects (Fig. 2). The network configuration has varied accordingly over time. At present, the subnetworks consist of four to eight, three-component, short-period stations each, and one broadband station at the Tenorio Volcano. Other temporary stations have been placed at different times in the Nicoya Peninsula area and toward the border with Nicaragua. The catalogue from OSIVAM used in this study covers from October 2006 until December 2010, a period when all four permanent subnetworks were recording simultaneously, and includes readings from 44 stations (Fig. 2).
The temporary Boruca network was operated by RSN in southeastern Costa Rica between May 1998 and November 2001 (Arroyo, 2001). This network consisted of two vertical-component and five three-component stations, all equipped with short-period seismometers. Readings from five short-period stations from Panama were added to the data set, improving the coverage eastward (Fig. 2). Earthquakes deeper than 40 km recorded by RSN in southeastern Costa Rica during 2008–2013 were added to this catalogue, taking advantage of new permanent and temporary stations installed since 2008 (Fig. 2).
The EGM2008 combined geopotential model (Pavlis et al., 2012) was
used as the source of gravity data for the 3-D density model, and was
obtained through the International Center for Global Earth Models (ICGEMs,
available at
Density scatter plots of representative earthquakes for each
quality class:
In order to assess which individual data set was adequate for the density modeling, different anomalies were calculated and compared with surface station data. A detailed analysis of a data subset for Costa Rica was carried out by Köther et al. (2012). To provide a data set consistent with previous solid Earth modeling (Lücke et al., 2010; Lücke, 2014), in this study, the Bouguer anomaly was calculated from the gravity disturbance. The gravity disturbance is calculated for a given point on the Earth's surface; for the satellite-derived gravity disturbance data used for this study, the downward continuation takes place between the orbit and the surface of the Earth. Further discussion on the use of gravity anomalies and disturbances in solid Earth modeling has been presented by Li and Götze (2001). Onshore, the effect of the Bouguer plate was subtracted from the value of the gravity disturbance, using the orthometric height at the topographic level as station elevation, which results in Bouguer anomaly values. For offshore data, values equivalent to the free air anomaly were obtained for stations located on the ocean surface. Figure 3 shows a compilation of the onshore Bouguer anomaly and the offshore free-air anomaly for the study area.
Gravity field data for the Central American isthmus and surrounding areas. Onshore Bouguer anomaly and offshore free-air anomaly calculated for this study from gravity disturbance data from EGM2008 (Pavlis et al., 2012). Location of the cross sections shown in Fig. 4 and the area of the local scale density model for Costa Rica are indicated in black. Location of Quaternary volcanoes (triangles) modified from Siebert and Simkin (2002). International borders and coastline are shown as black lines.
The 3-D
The PDF solution includes location uncertainties due to the spatial relation between the network and the event measurement error in the observed arrival times, and errors in the calculation of theoretical travel times. Because error estimates are included through covariance matrices and the solution is fully nonlinear, location uncertainties may assume irregular and multimodal shapes. In such cases, traditional error ellipsoids fail to represent the location error (Husen and Smith, 2004). NonLinLoc also provides traditional Gaussian estimates, like the expectation hypocenter location and the 68 % confidence ellipsoid (Lomax et al., 2000).
The density model was prepared by means of 3-D interactive
forward modeling of gravity data using the IGMAS
In this study, the lithospheric density distribution was modeled in three dimensions for Costa Rica, to a depth of 200 km. Previous investigations by Tassara et al. (2006) achieved a 3-D lithospheric density model for the South American subduction zone through a similar method, from which the slab geometry and the density distribution were modeled to a depth of 400 km.
In addition to the relocated earthquake hypocenters, geophysical constraints on the geometry of the plate interface, the lithospheric segmentation and its physical properties, were previously included in the process of forward density modeling. Wide angle seismic and seismic reflection cross-sections by Ye et al. (1996), Stavenhagen et al. (1998), Sallarès et al. (2001), von Huene et al. (2000), Ranero et al. (2003), Walther (2003), and Sallarès et al. (2003) provide constraints on the geometry of the plate interface and the structure of the subducting and overriding plates. Local earthquake seismic tomographies by Husen et al. (2003a), DeShon et al. (2006), Syracuse et al. (2008), and Arroyo et al. (2009) provide 3-D constraints on the physical properties of the subduction system. Receiver functions by Dzierma et al. (2010, 2011), respectively provide insights into the structure of the overriding plate and the geometry of the plate interface for the northwestern end of the Talamanca Range. Additionally, magneto-telluric surveys by Brasse et al. (2009) and Worzewski et al. (2010) provide information on the electrical properties of the lithosphere for northwestern Costa Rica and the fluid cycle of the subduction system.
Analysis of probabilistic earthquake relocation uncertainties in 3-D
velocity models (Husen et al., 2003a; Husen and Smith,
2004) show that, in general, hypocenter locations with less than six
The RSN, OSIVAM, and Boruca data sets were relocated and classified into four
quality classes using NonLinLoc. The 3-D
The Oct-Tree importance sampling algorithm included in NonLinLoc was used to
achieve an accurate, efficient and complete mapping of the earthquake
location PDF in the 3-D space. The Oct-Tree method uses recursive
subdivision and sampling of cells to generate a cascade of sampled cells.
The density of sampled cells follows the PDF values of the cell center,
leading to a higher density of cells in the areas of higher PDF (A. Lomax
and A. Curtis, October-tree importance sampling algorithm,
available at
Large differences between the maximum likelihood and the expectation hypocenter locations can result from an ill-conditioned location problem (Lomax et al., 2000). Husen and Smith (2004) found that a difference greater than 0.5 km between both hypocenter estimations cause large uncertainties of several kilometers in epicenter and focal depth. Numerous scatter plots were investigated in order to confirm this observation for the data sets relocated in this work. Further following Husen and Smith (2004), the average of the three axes of the 68 % confidence ellipsoid was taken into account to refine location quality. Figure 2 shows example scatter plots for each quality class used in this study. The relocation uncertainties for the tomography earthquake data set are analysed directly by Husen et al. (2003a).
Earthquakes in class D have an rms larger than 0.5 s and were not used. Events with location quality A, with well-defined PDFs, have differences between maximum-likelihood and expectation hypocenters of 0.5 km and lower, and maximum location uncertainties of 4 km. Differences above 0.5 km but under 3 km between both hypocenter estimations, and average uncertainties lower than 4 km define quality B. Their epicenters and focal depths are still relatively well defined. Earthquakes with a maximum rms of 0.5 s and differences in hypocenter estimations higher than 3 km are classified as C. They show large location uncertainties, of which the confidence ellipsoid is a poor approximation. From all the relocated data sets, 65 % of the events have quality A, 16 % quality B, and 19 % quality C.
The geometry of the Central American subduction zone was determined on a
regional scale to serve as a reference model for the Costa Rican subset. In
order to constrain the regional geometry of the subduction zone, hypocenters
from the catalogue of the Central American Seismic Center (CASC) were taken
into account. The CASC catalogue includes earthquakes with magnitudes above
3 and recorded by at least two national networks. The events selected for
this study have a minimum of 8 readings, a coverage gap of less than 250
The regional scale model is intended to provide the 3-D general framework for the detailed interpretation of the Costa Rican
subduction segment presented here. This is necessary in order to eliminate
edge effects in the 3-D modeling by extending the model outside of the area
of interest up to a point in which the effect of the edges of the model is
no longer significant for the modeled gravity in this area of interest. The
regional scale model considers the segmentation of the Central American
crustal basement and the regional Moho structure for the Caribbean Plate
published by Lücke (2014). The geometry obtained from the regional
model shows a uniformly dipping slab for the segment between 91
and 86
Using the regional density model of Central America (Lücke,
2014) as reference, a subset with enhanced detail was modeled for Costa Rica
considering the better quality and availability of geophysical constraints.
Off the coast of Costa Rica, the Cocos Plate was modeled with a density of 2.80 Mg m
In order to achieve a 3-D interpretation of the geometry of the subducting slab, the earthquake hypocenter results were integrated into the density model. By means of 3-D visualization and projection of results onto 2-D cross sections (Fig. 4), the density model was modified interactively to achieve the best fit with the measured gravity data by considering the structure outlined by the seismicity. The joint interpretation allowed constraining the overall geometry of the density model, while simultaneously accounting for a more precise determination of the plate interface.
Vertical cross sections from the local scale 3-D density
model and relation to the integrated earthquake hypocenter locations for
this study. The upper panels show the fit between the measured (black) and
calculated (red) gravity anomaly. Circles show the earthquake hypocenters
color coded for classes: A (blue), B (yellow) C (magenta). The modeled
density distribution of the subducted oceanic crust is depicted in grey
tones: 2.80 Mg m
Slab depth contours for Costa Rica from the integrated interpretation of seismological and density modeling results. White arrows indicate direction and rate of plate motion from DeMets (2001) and DeMets et al. (1994). Earthquake depths are indicated in the inset. Tectonic features on the oceanic plate after von Huene et al. (2000) and Barckhausen et al. (2001). The red dented line represents the axis of the Middle American Trench. EPR: East Pacific Rise, CNS: Cocos–Nazca Spreading Centre, PFZ: Panama Fracture Zone. RSB: Rough Smooth Boundary. B.V.: Baru Volcano. Bathymetric data from global multi-resolution topography by Ryan et al. (2009). Open diamond symbols show the location of adakite samples from Hoernle et al. (2008) and Gazel et al. (2011).
Cross section of the alternative density model considering a slab detachment at 80 km for the Cocos ridge. The upper box shows the fit between the measured (black) and calculated (red) gravity anomalies, the red stippled line shows the calculated gravity anomaly for the alternate slab detachment model with the Moho structure from Lücke (2014). The lower box shows the alternative model assuming a slab detachment at a depth of 70 to 80 km under the Talamanca region. White stippled line represents the location of the Moho from Lücke (2014) constrained by receiver function results from Dzierma et al. (2010). Blue dots show the location of earthquake hypocenters obtained for this study.
The results of this joint interpretation are shown in Figs. 4, 5, and 6.
Considerable changes in slab dip as well as in the density distribution of
the subducted oceanic crust are observed. The Cocos Plate segmentation
observed off the coast of Costa Rica (von Huene et al., 2000) carries over
to the structure of the subducted slab. The northwestern section of the
Cocos Plate (originated from the EPR and CNS) was modeled with a crustal
thickness of 6–8 km and subducts at a 14
At the southeastern end of the Quaternary Volcanic Arc, the dip angle of the
slab beneath 80 km deepens from 40 to 60
Intraslab seismicity in the southeastern part of Costa Rica is interrupted by a 55 km long gap in the onshore area adjacent to the Coronado Bay (Fig. 5), and then resumes to the northeast of the Osa Peninsula (Arroyo, 2001). In this southeasternmost section of the MAT, the seismicity reaches 70 km in depth (Fig. 4d), and outlines a subducting slab consistent with the geometry observed westward of the seismic gap in the Coronado Bay region (Arroyo et al., 2003). In spite of the presence of such a gap, the gravity field does not show segmentation that could be related to major changes in the subduction geometry. Overall, the results of the density modeling suggest the existence of a continuous slab structure along the trench and in depth.
The vertical extent of the Wadati–Benioff zone seismicity correlates with
the section of the subducted oceanic crust with a density of 3.15 Mg m
The origin of the Wadati–Benioff seismicity has been a subject of discussion
since its discovery, because at depths greater than
Hydration of the downgoing slab seems to take place mainly near the trench,
where water percolation and mineral alteration occur at extensional bend
faulting in the crust and in the upper few kilometers of the mantle
(Kirby et al., 1996), or even deeper, 15–30 km into the mantle
(Peacock, 2001; Ranero et al., 2003). This bend faulting has
been imaged off the coast of Middle America by high-resolution seafloor mapping,
near-vertical reflection lines, and outer-rise seismicity (Hinz et al.,
1996; von Huene et al., 2000; Ranero et al., 2005; Grevemeyer et al.,
2007; Lefeldt and Grevemeyer, 2007). Ranero et al. (2005) propose a
model in which seismicity starts between 60–80 km depth, when dehydration of
the oceanic crust reactivates the upper segment of bend-faults, causing
concentration of earthquakes on the upper part of the slab. Deeper than
Tomographic studies (Arroyo et al., 2009; DeShon et al., 2006; Husen et al., 2003a) indicate the existence of a low-velocity hydrous oceanic crust in the northwestern and central Costa Rica subducting plate. The petrologic modeling from Husen et al. (2003a) shows a good correlation between the predicted locations of hydrous minerals in the oceanic crust and the hypocenters of intermediate-depth earthquakes, further supporting the hypothesis of the latter being enabled by dehydration.
In the following, the potential relationship between slab hydration at the outer rise, the vertical and lateral changes in the slab density model, and the terminal depth of the intraslab seismicity along the Costa Rican margin will be discussed. In addition, an alternative scenario in which a slab detachment is simulated beneath southeastern Costa Rica is analysed in Sect. 6.2.
In northwestern Costa Rica, the geometry of the subducting slab is
consistent with the trench-normal subduction of the 24 to 25 Ma. oceanic
crust generated at the EPR (Barckhausen et al., 2001). This
section presents a steeply subducting slab (Fig. 4a) with a geometry similar
to that observed for the Nicaraguan section of the MAT
(Syracuse et al., 2008). Between the northwestern part
of the volcanic arc and the central region of Costa Rica, a change in slab
dip is observed. For the central part, results show slab dip angles between
40 and 50
Southeast of the Central Cordillera of Costa Rica mountain range, seismicity within a steep slab is observed to a depth of approximately 75 km, extending southwest to the Coronado Bay. The presence of a steeply subducting slab for this region is supported by seismic tomography results from Dinc et al. (2010) to a depth of 40 km and receiver functions results by Dzierma et al. (2011), where a steep slab is imaged to a depth of 100 km.
The seismicity relocated in southeastern Costa Rica shows a subducting slab
dipping at a 50
Considering an estimated arrival of the Cocos Ridge at the trench during the
late Miocene (von Huene et al., 2000), and assuming even a
conservative convergence of half the current rate (estimated by
DeMets et al. (2010) at
Evidence of a relatively steep slab directly beneath southeastern Costa Rica
is presented for the first time with this study. This result is surprising
to some extent, but might be interpreted in the light of recent numerical
models (Mason et al., 2010) and analogue experiments (Martinod et al., 2013)
on aseismic ridge subduction. They show that effects observed inboard
of subducting ridges, such as diminution of the slab dip, indentation of the
overriding plate, inhibition of arc volcanism, rapid forearc uplift, and
back arc deformation, appear only progressively. At the beginning of ridge
subduction, the ridge is pulled downwards by the dense slab, and
Lateral changes in the density structure of the slab are limited to the
section of the subducted oceanic crust with a density of 3.15 Mg m
Within the conceptual model from Ranero et al. (2005), the degree of
bend faulting occurring in the outer rise would directly influence the
characteristics of the intraslab seismicity by determining both the extent
of hydration through serpentinization of the upper mantle, and the
segmentation of the slab by means of fragile deformation. In this sense, the
Costa-Rican subduction zone provides a wide spectrum of environments with
different degrees of bending-related faulting. The smooth segment in which
the oceanic plate presents pervasive, trench-parallel, bending-related
faulting, subducts off Central America and northwestern Costa Rica (Hinz
et al., 1996; von Huene et al., 2000; Ranero et al., 2003a; Grevemeyer et al.,
2007; Lefeldt and Grevemeyer, 2007). In the Central
Pacific (seamount) segment, however, the thicker oceanic crust promotes a
smaller outer rise and less-developed faulting (von Huene et al.,
2000), while the Cocos Ridge section in the southeast virtually lacks
bending-related faulting (Ranero et al., 2003a). Moreover,
Rüpke et al. (2002) show considerable differences in
serpentinization of the upper mantle for the Cocos Plate at its smooth
segment in comparison with the seamount province. In Nicaragua, the
subducting oceanic plate generated at the EPR shares characteristics with
the oceanic plate off the coast of northwestern Costa Rica in terms of origin,
composition, age and thickness (Barckhausen et al.,
2001; von Huene et al., 2000). Also, the tectonic fabric observed in
bathymetric data for Nicaragua indicates the same magnitude of normal
faulting of the outer rise described off the coast of Costa Rica by Ranero et al. (2003a). For
this section of the Cocos Plate, Rüpke et al. (2002) estimate a 10 km thick serpentinized mantle layer with a
5.5 wt % of water; whereas for central Costa Rica, a 5 km thick 2 wt %
H
This strong contrast in the serpentinization of the upper mantle via
bend-faulting on the outer-rise, indicates that the hydrated mantle may
reach greater depths in northwestern Costa Rica, thus accounting for the
greatest depth of the 3.15 Mg m
Southeastern Costa Rica is arguably the less well-understood region of the study area. It presents highly complex tectonics, which include the presence of the Cocos Ridge, an exposed forearc, the gap in the Quaternary volcanic activity, the uplifted Talamanca Range, and the outcrop of adakites. A prior model of the slab geometry by Protti et al. (1994), based solely on earthquake hypocenters, does not report intraslab seismicity eastward of the end of the Quaternary Volcanic Arc in Costa Rica. This apparent lack of deep seismicity, together with the presence of the gap in Quaternary volcanism and an OIB-type geochemical signature for the volcanic rocks in the area (Hoernle et al., 2008), has led to several interpretations for the source of the adakites found in southeastern Costa Rica and western Panama (Fig. 5) (de Boer et al., 1991; Drummond et al., 1995). The hypothesis imply the melting of the slab edge by interaction with hot mantle through either a slab window (Johnston and Thorkelson, 1997; Abratis and Wörner, 2001) or a slab detachment (Gazel et al., 2011). These models contrast with the receiver function results from Dzierma et al. (2011) showing a steeply subducting slab beneath the northern edge of the Talamanca Range. By testing several alternative tectonic scenarios against their receiver functions results, the authors conclude that a slab break-off at a depth shallower than 70 km is not plausible. However, Dzierma et al. (2011) admit that deeper than 70 km, scattered signals may contaminate the receiver functions producing results similar to those from a steeply subducting slab.
As pointed out before, the seismological data presented here shows the
termination of possible slab-related seismicity at a depth of approximately
70 km for the area located inland of the Osa Peninsula (Fig. 4d). The
hypocenters show the presence of a steeply subducting slab at such depth and
have been used to constrain the geometry of the 3-D density
model. However, in the primary density model, the continuity of the slab to
depths greater than
The absence of a downgoing slab deeper than 80 km causes a misfit of
10–20
Those inconsistencies further support the preferred model of this study,
featuring the presence of a continuous slab with a relatively steep angle in
southeastern Costa Rica, as described in Sect. 6.1. In such a scenario, the
melting may happen through interaction with the mantle at the downdip
projection of the Panama Fracture Zone, where the lithospheric segmentation
caused by the Cocos–Nazca boundary may allow the subducted oceanic crust to
come in contact with the mantle. The presence of the westernmost adakite
outcrops might be explained by the migration of the triple junction toward
the southeast proposed by DeMets (2001). According to
Gazel et al. (2011), this is on par with a 35 mm yr
The gravity modeling and local earthquake relocation results were
successfully interpreted in a joint model showing the 3-D structure of the slab beneath Costa Rica. The depth of the intraslab
seismicity varies from
The two plausible models for the southeastern Costa Rican subduction zone derived from the joint gravity and seismological interpretation are the following: (a) a steeply subducting slab, which becomes aseismic at a depth of approximately 75 km, due to less mantle hydration through serpentinization at the Cocos Ridge; and (b), a slab detachment occurring at a depth of 70 to 80 km and a shallow Moho underneath Talamanca. A significantly better fit between measured and calculated gravity values, together with previously constrained Moho depths and the inferred presence of a slab structure at the northern edge of the Talamanca Range interpreted from receiver functions (Dzierma et al., 2011), suggests that the presence of a continuous slab is more likely.
The authors would like to thank the National Seismological Network (RSN) and S. Husen for allowing the use of their earthquake catalogues for Costa Rica. W. Taylor, from Instituto Costarricense de Electricidad (ICE), who generously shared the OSIVAM catalogue, and C. Redondo for providing the CASC data. R. von Huene and S. Martínez-Loriente provided thoughtful comments and suggestions that greatly improved the manuscript. G. Alvarado, G. Soto and I. Grevemeyer contributed valuable discussions and encouraged the present study. O. Lücke thankfully acknowledges the support of the German Academic Exchange Service (DAAD) as well as the University of Costa Rica (UCR), and the Special Priority Program 1257 “Mass Transport and Mass Distribution in the Earth System” of the German Research Foundation (DFG). Edited by: V. Sallares