Paleomagnetism and dating of a thick lava pile in the Permian Bakaly formation of eastern Kazakhstan: Regularities and singularities of the paleomagnetic record in thick lava series

Abstract

Paleomagnetic results on thick lava series are among the most important sources of information on the characteristics of ancient geomagnetic fields. Most paleo-secular variation data from lavas (PSVL) are of late Cenozoic age. There are far fewer results from lavas older than 5Ma. The Central Asia Orogenic Belt that occupies several million square kilometers in Asia is probably the world’s largest area of Paleozoic volcanism and is thus an attractive target for PSVL studies. We studied a ca. 1700m thick lava pile in eastern Kazakhstan of Early Permian age. Magmatic zircons, successfully separated from an acid flow in this predominantly basaltic sequence, yielded an Early Permian age of 286.3±3.5Ma. Oriented samples were collected from 125 flows, resulting in 88 acceptable quality flow-means (n⩾4 samples, radius of confidence circle α95⩽15°) of the high-temperature magnetization component. The uniformly reversed component is pre-tilting and arguably of a primary origin. The overall mean direction has a declination=242.0° and an inclination=−56.2° (k=71.5, α95=1.8°; N=88 sites; pole at 44.1°N, 160.6°E, A95=2.2°). Our pole agrees well with the Early Permian reference data for Baltica, in accord with the radiometric age of the lava pile and geological views on evolution of the western part of the Central Asia Orogenic Belt. The new Early Permian result indicates a comparatively low level of secular variation especially when compared to PSVL data from intervals with frequent reversals. Still, the overall scatter of dispersion estimates that are used as proxies for SV magnitudes, elongation values and elongation orientations for PSVL data is high and cannot be fitted into any particular field model with fixed parameters. Both observed values and numerical simulations indicate that the main cause for the scatter of form parameters (elongation values and elongation orientations) is the too small size of collections. Dispersion estimates (concentration parameter and standard angular deviation) are more robust, and their scatter stems from other sources, which may include non-stochastic features of datasets like clusters, loops etc., or non-stationary behavior of secular variation magnitude over time intervals of many million years.