Continuous reconstructions of past variations of the Earth's magnetic field are based mainly on paleomagnetic and cosmogenic 10 Be records in marine sediments. In both cases, the recording mechanisms can be affected by environmental processes. Climatic overprints are only partially removed by normalization procedures, so that stacking is used to further remove site-specific effects. Regionally or globally correlated artifacts, however, cannot be removed by stacking. Here we present a modified approach where geomagnetic records are complemented by environmental proxies representing processes that might affect the field recording mechanism. Geomagnetic and environmental records are jointly processed with principal component analysis to obtain a set of components supposed to represent true variations of the geomagnetic field and climatic overprints, respectively. After discussing the theoretical background of this new approach and its underlying assumptions, a practical example is presented, using a worst-case scenario based on a single 10 Be record from the North Atlantic with strong climatic overprints, covering the last 600 ka. The first two principal components, which represent the modulation of 10 Be by global climatic variations and by the geomagnetic field, respectively, explain 66.3% of the signal variance. Comparison of the geomagnetic principal component with global relative paleointensity stacks shows that the original climatic overprint can be reduced by a factor of 2, outperforming a 10 Be/ 9 Be stack obtained from two sites with little glacial-interglacial variability. The proposed method for removing climatic overprints can be applied to multiple sites more efficiently than conventional stacking.

Plain Language Summary Continuous records of the Earth's magnetic field rely on measurements of magnetic minerals or cosmogenic isotopes in sediments. Both types of records are also sensitive to environmental conditions and are thus affected by past climatic variations. These unwanted climatic overprints are difficult to remove: one strategy consists in stacking records from different sites; however, regionally or globally correlated artifacts cannot be completely removed by this technique. Here we present a new method for separating the geomagnetic signal from unwanted climatic overprint, which is based on the principal component analysis (PCA). The efficiency of this new method is tested with a worst-case example based on a single site located in the North Atlantic, which is characterized by strong glacial-interglacial variability. The first two principal components obtained from PCA represent the modulation of 10 Be by global climatic variations and by the geomagnetic field, respectively. Comparisons of the geomagnetic field component with reference data show that the original climatic overprint has been reduced by a factor of 2, outperforming a 10 Be/ 9 Be stack obtained from two sites with little glacial-interglacial variability. The proposed method for removing climatic overprints can be applied to multiple sites more efficiently than conventional stacking.