 2004 Geological Society of America. For permission to copy, contact Copyright Permissions, GSA, or editing@geosociety.org. Geology; October 2004; v. 32; no. 10; p. 913–916; doi: 10.1130/G20563.1; 3 figures. 913 Generic model of subduction erosion Roland von Huene* Department of Geology, University of California, Davis, Davis, California 95616, USA Ce ´ sar R. Ranero Institut fu ¨ r Meereswissenschaften, GEOMAR and SFB 574, Wischhofstrasse 1-3, 24148, Kiel, Germany Paola Vannucchi Dipartimento di Scienze della Terra, Universita ` di Firenze, Via La Pira, 4, 50121 Firenze, Italy ABSTRACT Erosion by high stress abrasion of convergent margins from horsts and grabens on the subducting plate is not shown in seismic images. In a proposed model, the frontal sediment prism is a dy- namic mass that elevates pore-fluid pressure. Overpressured fluid invades fractures in the upper plate and separates fragments that are dragged into a subduction channel along the plate interface. Removed fragments are smaller than surface ship seismic tech- niques have resolved and beyond the reach of past scientific ocean drilling; however, current drill capability and downhole geophysics can test the model. Keywords: continental margins, nonaccretionary, convergent, plate boundary, role of fluids, subduction processes. INTRODUCTION Convergent margins dominated by subduction erosion are the most prevalent type (von Huene and Scholl, 1991; Clift and Vannucchi, 2004), but the processes shaping them are scarcely understood. Erosion along the plate interface occurs beyond the depths at which outcrop- scale structure is resolved in seismic records, or was penetrated by past scientific drilling. Exposures of ancient erosional subduction zones are difficult to identify convincingly. Among the observations requiring subduction erosion, the clearest is large-scale long-term margin sub- sidence. However, subsidence was an unexpected discovery or collat- eral finding of drilling for other objectives, and the unforeseen finding left central aspects of erosion unexplored—hence the delay in under- standing erosional mechanisms. Here we integrate earlier ideas regarding subduction erosion with current observations supporting a proposed generic subduction-erosion model. Our objective is to summarize in a model some concepts that will help focus future investigation. PREVIOUS MODELS AND SUPPORTING OBSERVATIONS Concepts regarding subduction erosion include mechanisms of physical abrasion involving high stress and fluid-assisted abrasion in- volving low stress. Both concepts originated in the late 1970s with two investigations along the northern Honshu margin of Japan (von Huene et al., 1978; Hilde and Sharman, 1978). Margin subsidence recorded in Deep Sea Drilling Project (DSDP) cores was explained by thinning of the upper plate (Murauchi and Ludwig, 1980; Langseth et al., 1981). Elevated pore pressure was inferred to hydrofracture the upper plate, dislodging clasts that subducted. Alternatively, from horsts and grabens entering the Japan Trench axis, it was proposed that horsts act as strong teeth that rasp material from the underside of the upper plate (Hilde, 1983). Much better prestack time-migrated records acquired by an in- dustry seismic vessel lacked the lower-plate ‘‘chain saw’’ character (von Huene and Culotta, 1989). At the trench axis, grabens fill with loose material, smoothing the subducting surface, and the de ´collement is located 1 km above lower-plate relief. Such smoothing of lower- plate relief in the trench axis is common in seismic images, and a contact between subducted horsts and terminated upper-plate reflec- *E-mail: rhuene@mindspring.com. tions has not been imaged. Thus the seismic evidence for rasping is scant. Erosion modeled in sandbox experiments (Lallemand et al., 1994) appears to require high basal friction (Adam and Reuther, 2000). The northern Chile margin structure was used as a natural analogue, a mar- gin commonly cited as a high-stress and high-friction interplate end member (cf. Jarrard, 1986; Uyeda and Kanamori, 1979). The Mesozoic crystalline upper-plate basement rock is assumed to be strong. Across the empty trench axis, the converging ocean plate has a spectacular horst-and-graben topography and thin (100 m) sediment. Sediment transport from the south along the trench axis has probably been blocked for 12 m.y. (Yan ˜ez et al., 2001), and sediment from the Andes is trapped in forearc basins. Without a sediment interface to reduce interplate friction, strong coupling between igneous ocean crust and crystalline continental basement is a reasonable assumption. Conflicting information across the northern Chilean margin was clarified with multibeam bathymetry and seismic records (von Huene and Ranero, 2003). Despite sediment starvation, an interplate clastic layer was imaged. Debris from mass wasting forms a frontal prism and loose material fills grabens in the trench axis forming an interplate layer a few hundreds of meters over horsts and 1.0 km thick in grabens. Middle slope extensional faults are imaged deep into the upper plate. Active contractional structures are observed only in the frontal prism. Finally, a measured material strength that was input to taper analysis lowered the previously calculated upper-plate basal friction. Observations do not support a high-friction plate interface updip of the seismogenic zone. OBSERVATIONS SUPPORTING THE PROPOSED MODEL Three observations are central to the model: (1) erosion along the underside of the upper plate to explain subsidence, (2) a frontal prism that reduces friction and allows subduction of all trench sediment, and (3) a middle slope progressively deformed by normal faults until it breaks down at the frontal prism. Subsidence Drill cores at seven Pacific convergent margins show 3–5 km of subsidence during Neogene time (cf. von Huene and Scholl, 1991; Clift and MacLeod, 1999; Vannucchi et al., 2001, 2003, 2004). Shallow- water sediment deposited on an unconformity was recovered in cores from midslope to lower-slope depths. Subsidence requires upper-plate thinning, and since sedimentation continued during subsidence, most erosion occurs along the upper plate’s base. The average long-term rate of upper slope subsidence off Nicoya Peninsula is 250 m/m.y. and during the past 5 m.y. it was 400–600 m/m.y.; that off Guatemala was 140–280 m/m.y. (Vannucchi et al., 2003, 2004). If subsidence equals the upper-plate material removed, it requires a clastic layer along the Guatemala plate interface 160–180 m thick and 400–800 m thick be- neath Costa Rica where Cocos Ridge arrived at the trench 5 m.y. ago and accelerated erosion. Frontal Prism We use the term ‘‘frontal prism’’ because an accreted prism is composed mostly of sediment transferred from the subducting plate, whereas the frontal prism is not. A frontal prism occurs along most margins undergoing subduction erosion and is structured like a small