Archeomagnetic intensity variations during the era of geomagnetic spikes in the Levant

Observational records of rapidly varying magnetic fields strongly constrain our understanding of core flow dynamics and Earth's dynamo. Archeomagnetic analyses of densely sampled artefacts from the Near-East have suggested that the intensity variation during the first millennium BCE was punctuated with two geomagnetic spikes with rates of change of intensity exceeding 1 μT/yr, whose extreme behaviour is challenging to explain from a geodynamo perspective. By applying a new transdimensional Bayesian method designed to capture variations on both long and short timescales, we show that the data considered only at the fragment (thermal-unit) level require a complex intensity variation with no less than six spikes, each with an approximate duration of between 30 and 100 years. However, the nature of the inferred intensity evolution and the number of spikes detected are fragile and highly dependent on the specific treatment of the archeomagnetic data. No spikes are observed when the data are considered only at the level of a group of fragments from the same archeological context, with a minimum of three different artefacts per context. Furthermore, the number of spikes decreases to zero when increasing the error budget for the intensity at the fragment level within reasonable levels of 3–6 μT and the data age uncertainty up to 50 years. Of the six spikes found, the most resilient when increasing the error budget was dated at ~970 BCE. However, we show that even this spike sensitively depends on the age model proposed for data from the Levant archeological site Timna-30 and disappears when considering a single Gaussian age prior distribution for these data and a moderate minimum intensity error. Thus, depending on the choices made, the Near-Eastern data are compatible with a broad range of time-dependence, from six spikes at one extreme to zero spikes on the other. An error of 6 μT at the fragment level produces a spikeless model with strong similarity with the reconstruction from the SHAWQ-Iron Age global model with rates of change of ~0.2–0.3 μT/yr.