Paleomagnetism provides the only quantitative tool to reconstruct the position and motion of tectonic plates and continents in deep geological time. These reconstructions often rely on apparent polar wander paths (APWPs), which describe the past position and motion of plates relative to the Earth’s spin axis. In addition to providing a paleomagnetic reference frame for paleogeography, these paths serve to quantify relative tectonic displacements with other lithospheric blocks. A long-discussed problem is that conventional approaches to computing polar wander do not propagate key sources of uncertainty in the underlying data, such as data scatter or age uncertainty, but proposed solutions remained mostly qualitative. Recently, this problem became more urgent: Rowley (2019, Tectonics) showed that as much as 50% of the data that underlie the currently most widely used global APWP are statistically significantly displaced relative to that APWP. This implies that the resolution at which a geologically meaningful statistical difference between a paleomagnetic dataset and an APWP may be determined, is strongly limited, which undermines the current tectonic and paleogeographic applications of paleomagnetism. This thesis aims to examine the causes of dispersion of paleomagnetic data behind APWPs and to build a global APWP in which key data uncertainties are propagated, such that it may be used to determine geologically meaningful relative displacements. We show that the uncertainty of APWPs computed from paleomagnetic poles, which is the current standard, is mostly determined by the arbitrary choice of how many datapoints are used per pole. This thesis then develops a novel approach in which apparent polar wander paths are computed from site-level paleomagnetic data instead of paleomagnetic poles. In this approach, larger weight is assigned to larger datasets and temporal and spatial uncertainties in the paleomagnetic data are incorporated. We present a global apparent polar wander path for the last 320 Ma using the new statistical approach proposed in this thesis, as well as using a fully updated paleomagnetic database. We find that the first-order geometry of this path is similar to previous models but with smaller uncertainties. Moreover, we show that correcting for temporal bias in the paleomagnetic data allows improved quantification of polar wander rates, and that previously observed peaks in polar wander likely resulted from such bias. We introduce the open-source web application APWP-online.org that provides user-friendly tools to compute apparent polar wander paths using our new methodology as well as the computation of relative paleomagnetic displacements. The website also hosts a community platform for the continuous improvement of the path in the future through the addition of new high-quality paleomagnetic data. Finally, we provide new quantitative estimates of true polar wander during the last 320 Ma by comparing our new global apparent polar wander path with existing mantle reference frames. The computed true polar wander paths suggest that the true polar wander rotations occurred about two roughly orthogonal axes in the equatorial plane. We discuss the geodynamic implications of these findings and highlight future opportunities for the identification and quantification of true polar wander on geological timescales.
