What drives 20th century polar motion?

Astrometric and geodetic measurements show that the mean position of Earth's spin axis drifted through the solid crust toward Labrador, Canada at an average speed of 10.5±0.9 cm/yr during the 20th century. Understanding the origins of this secular polar motion (SPM) has significance for modeling the global climate, as it provides a link to ice mass balance and sea-level rise. A perplexing issue, however, is that while glacial isostatic adjustment (GIA) models satisfactorily explain the direction of SPM, the associated prediction of the amplitude is insufficient. Our Bayesian GIA analysis, with constraints from relative sea-level and vertical land motion data, reveals that this process only accounts for 33±18% of the observed SPM amplitude. This shortfall motivates a more broadly scoped reassessment of SPM drivers. To address this, we assemble a complete reconstruction of Earth's surface mass transport derived from recent advancements in modeling the global 20th century cryospheric, hydrologic, oceanic, and seismogenic mass exchange. The summed signals, nonetheless, cannot fully reconcile the observed SPM, even when considering the error statistics of each driver. We investigate an additional excitation source: changes in Earth's inertia tensor caused by mantle convection. Sophisticated models have recently been advanced in tectonic plate reconstructions, in conjunction with geoid and seismic tomographic models. Here we use these models to compute new estimates of SPM. While the convection-driven SPM has considerable uncertainty, the average direction of 283 recent models aligns with the residual SPM (within 2.7 ^∘±14.8 ^∘), significantly reducing the gap between observation and prediction. We assert that one key mechanism for driving 20th century SPM is long-term mass movement due to mantle convection.