The role of plate tectonics in the deep carbon cycle is crucial for understanding Earth’s climate, planetary life, and atmospheric CO2 over geological time; however, knowledge of carbon sources and sinks is dependent on reconstructed plate tectonic histories. Intra-oceanic subduction (i.e. subduction of an oceanic plate beneath another oceanic plate) is a recognized knowledge gap in plate tectonic reconstructions, especially within Pacific-Panthalassa, because continuous recycling of oceanic plates into the mantle leaves fewer geological traces. We examine the role of unassessed intra-oceanic subduction on paleoclimate reconstructions since the Mesozoic. We compare two global plate reconstructions: ”Tomopac” version 1, which integrates tomography and other constraints to enhance intra-oceanic subduction histories within Pacific-Panthalassa; and a widely-used model that implements less intra-oceanic subduction (Matthews et al., 2016). We model global seafloor ages, estimate total and areal subducted carbon since 200 Ma, and input subduction histories into the COPSE biogeochemical model (Lenton et al., 2021) to compare predicted global atmospheric CO2 and mean surface temperature histories. Tomopac with more intra-oceanic subduction shows a ~5% increase in global subduction zone lengths but a ~8% decrease in global subducted area and slab flux since 200 Myr. Overall, contrasted intra-oceanic subduction histories since 200 Ma alter estimates of subducted carbon by ~15-20%. Incorporating previously unassessed intra-oceanic subduction from Tomopac in COPSE reduces global mean surface temperatures up to 2°C and reduces atmospheric CO2 up to 300 ppm from 200 Ma to present, highlighting the importance of recognizing intra-oceanic subduction in modulating Earth’s long-term paleoclimate.

Pacific-Panthalassa plate tectonics back to Jurassic times are recognized as the most challenging on Earth to reconstruct, due partly to a large (>9000 km length) unconstrained area between the Pacific and Laurasia (now NE Asia) during the Early Jurassic. We built four contrasted NW Pacific-Panthalassa global plate reconstructions and assimilated their velocity fields into global geodynamic models. We compare our predicted mantle structure, synthetic geoid and dynamic topography to Earth observations. P-wave tomographic filtering of predicted mantle structures allowed for more explicit comparisons to global tomography. Plate reconstructions that include intra-oceanic subduction in NW Pacific-Panthalassa fit better to the observed geoid and residual topography, challenging the Andean-style subduction along East Asia. Our geodynamic models predict significant SE-ward lateral slab advections within the NW Pacific basin lower mantle (~2500 km from Mesozoic times to present), which can confound “vertical slab sinking”-style restorations of past subduction zone locations.

The plate tectonic history of the hypothesized ‘proto-South China Sea’ (PSCS) ocean basin and surrounding SE Asia since Cenozoic times is controversial. We implement four diverse PSCS plate reconstructions into global geodynamic models to constrain PSCS plate tectonics and possible slab locations. We consider: southward versus double-sided PSCS subduction models; earlier (Eocene) or later (late Oligocene) initiation of Borneo counterclockwise rotations; and, larger or smaller reconstructed Philippine Sea plate sizes. We compare our modeling results against tomographic images by taking into account mineralogical effects and the finite resolution of seismic tomography. All geodynamic models reproduce the tomographically-imaged Sunda slabs beneath Peninsular Malaysia, Sumatra and Java. Southward PSCS subduction produces slabs beneath present Palawan, northern Borneo, and offshore Palawan. Double-sided PSCS subduction combined with earlier Borneo rotations uniquely reproduces sub-horizontal slabs under the southern South China Sea (SCS) at ~400 to 700 km depths; these models best fit seismic tomography. A smaller Philippine Sea (PS) plate with a ~1000 km-long restored Ryukyu slab was superior to a very large PS plate. Taken together, the four end-member plate models predict PSCS slabs at <900 km depths under present-day Borneo, the SCS, the Sulu and Celebes seas, and the southern Philippines. Regardless of plate models, we predicted passive mantle upwellings under Indochina during late Eocene-Oligocene times, and downwellings under the SCS during the late Cenozoic that do not support a deep-origin ‘Hainan plume’. Modeled Sundaland dynamic topography depends strongly on the imposed plate model, varying by several hundred meters.