Slow-downs and speed-ups of India–Eurasia convergence since 20 Ma: Data-noise, uncertainties and dynamic implications

India–Somalia and North America–Eurasia relative motions since Early Miocene ( 20 Ma) have been recently reconstructed at unprecedented temporal resolution (o1 Myr) from magnetic surveys of the Carlsberg and northern Mid-Atlantic Ridges. These new datasets revamped interest in the convergence of India relative to Eurasia, which is obtained from the India–Somalia–Nubia–North America–Eurasia kinematics (i.e. stage Euler vectors) appear to be surprisingly unusual over the past 20 Myr. In fact, the Euler pole for the India–Eurasia rigid motion scattered erratically over a broad region, while the associated angular velocity underwent sudden increases and decreases. Consequently, convergence across the Himalayan front featured significant speed-ups as well as slow-downs with almost no consistent trend. Arguably, this pattern arises from the presence of data-noise, which biases kinematic reconstructions—particularly at high temporal resolution. The rapid and important India–Eurasia platemotion changes reconstructed since Early Miocene are likely to be of apparent nature, because they cannot result even from the most optimistic estimates of torques associated, for instance, with the descent of the Indian slab into Earth’s mantle. Our previous work aimed at reducing noise in finiterotation datasets via an expanded Bayesian formulation, which offers several advantages over arbitrary smoothing methods. Here we build on this advance and revise the India–Eurasia kinematics since 20 Ma, accounting also for three alternative histories of rifting in Africa. We find that India–Eurasia kinematics are simpler and, most importantly, geodynamically plausible upon noise reduction. Convergence across the Himalayan front overall decreased until 10 Ma, but then systematically increased, albeit moderately, towards the present-day. We test with global dynamic models of the coupled mantle/ lithosphere system how a previously proposed uplift of the Tibetan plateau by 1–2 km, as well as the break-up of the Indo-Australian plate into the India and Australia units, impact the kinematic record from 20 to 10 Ma. Our modelling results indicate that the convergence slow-down until 10 Ma resulted chiefly from the Indo-Australian break-up. However, the tectonic scenario yielding the best fit to kinematic reconstructions is indeed one where the Tibetan mean elevation increased by 1–2 km before the Indo-Australian break-up was accomplished. & 2013 Elsevier B.V. All rights reserved.