Critical analysis of the Holocene palaeointensity database in Central America: Impact on geomagnetic modelling

Thanks to its rich archaeological heritage, Central America is a key region to recover the past secular variation of the geomagnetic field. This article presents 13 new palaeointensity data on Epiclassic (650 - 900 CE) pottery sherds from Central Mexico. Archaeointensities were determined using the Thellier-Thellier protocol with anisotropy and cooling rate corrections. Average results between 25 and 42 mu T reveals a fast secular variation in the second half of the first millennium CE but are not in agreement with global geomagnetic models that predict a higher geomagnetic field strength. To check the reasons of this discrepancy, we compiled all intensity data over the last millennia published in Central America. The Bayesian curve calculated from 194 data covering the last 4 millennia highlights a rapid succession of oscillations of the geomagnetic field strength between 20 and 80 mu T. But a critical analysis of the dataset shows a large influence of data quality, 74% of them having a poor cooling unit consistency and experimental quality. The small number of specimens per cooling unit and the anisotropy correction absent or incorrectly made increase the scatter between data, whereas the absence of cooling rate correction biases the dataset towards higher palaeointensity. Discarding these data results in a lower secular variation by removing most extreme values and several intensity oscillations. The weaknesses of the dataset are likely the main reason of the limitations of global models in Central America. Pending the acquisition of new high-quality data, archaeomagnetic dating seems premature in Central America.