Inferring upper-mantle temperatures from seismic velocities

Constraints on mantle temperatures are fundamental for a better understanding of the dynamics of the Earth. We evaluate how well upper-mantle thermal structure can be inferred from seismic velocities. Our forward calculation of seismic velocities and density includes uncertainties in anharmonic and anelastic parameters for various compositional models. The sensitivity of V P ( V S ) to temperature along a 1300 °C adiabat decreases with depth from −0.75±0.15% (1.3±0.30%) per 100 °C at 200 km, to −0.23±0.05% (−0.45±0.10%) per 100 °C at 800 km, with an additional reduction in the wadsleyite stability zone. Shear anelasticity introduces a significant non-linear dependence of seismic velocities on temperature in the upper-mantle. This means that seismic anomalies with similar amplitudes but different signs correspond to temperature anomalies of a different magnitude, and absolute seismic velocities are required for making a thermal interpretation. With depth, the importance of anelasticity decreases. Temperature derivatives of bulk sound velocity ( V φ ) and density ( ρ ) are a factor two to five lower than those of V P,S and do not depend strongly on temperature. Common mantle compositions cause velocity anomalies of about 1% throughout the upper-mantle. Above 400 km depth, the effect of composition on V P and V S is secondary to the effect of temperature, but it gains importance with increasing depth. For most upper-mantle compositions and depths, the relative sensitivity of V P , V S and V φ to temperature and composition is not different enough to distinguish the two factors from combined velocity models. But ∂ ln ρ / ∂ ln V S allows a separation of thermal anomalies from compositional anomalies for iron-depleted and subducted material. Spherically symmetric reference profiles underlie most seismic velocity models. However, small differences between common one-dimensional velocity models translate into disparate thermal interpretations. A better understanding of seismic uncertainties and the physical interpretation of reference models is necessary to interpret absolute velocities. Assuming seismic structure is well resolved and composition known, uncertainties in inferred thermal structure are ±100 °C above 400 km depth and ±250 °C in the shallow lower-mantle. This makes an inversion of seismic velocity models for temperature feasible, above 400 km depth. At larger depths, testing seismic velocities calculated for a proposed thermal and compositional structure against observations is a more promising interpretation approach.