Quantifying slab sinking rates using global geodynamic models with data-assimilation

Diandian Peng, Lijun Liu

Research output: Contribution to journalReview articlepeer-review


Recent tomography-based tectonic (tomotectonic) reconstructions provide important new insights on past subduction processes. However, some of these exercises violate key geological observations. We suggest this is due to two common simplifications: 1) slabs sink vertically after subduction with their present geographic locations marking past trenches, and 2) slabs sink at a constant rate throughout the mantle with their present mantle depths linearly correlating with their subduction ages. In this study, we investigate the 4D evolution of subducted slabs using global data-assimilation models that successfully reproduce observed mantle slabs along all major subduction zones. We find that slabs can migrate laterally up to 6000 km while descending toward the core-mantle boundary (CMB). This enormous displacement mostly reflects strong and geographically variable horizontal mantle flow. The model results further show that the vertical sinking rate of slabs varies with subduction duration, depth, and geographic locations. The slab sinking rate generally decreases with increasing depth, ranging from >2 cm/yr above 1600 km depth to zero at ~500 km above the CMB. The sinking rate locally peaks at the asthenospheric and mid-mantle depths and diminishes at the base of the transition zone. We further find that surface plate motion strongly affects the horizontal migration of slabs but less affects their sinking rate. These results suggest cautions on inferring past tectonic events based on seismic tomography.

Original languageEnglish (US)
Article number104039
JournalEarth-Science Reviews
StatePublished - Jul 2022


  • Geodynamic modeling
  • Horizontal migration
  • Plate reconstruction
  • Sinking rate
  • Subduction

ASJC Scopus subject areas

  • Earth and Planetary Sciences(all)


Dive into the research topics of 'Quantifying slab sinking rates using global geodynamic models with data-assimilation'. Together they form a unique fingerprint.

Cite this