Analyzing the geomagnetic axial dipole field moment over the historical period from new archeointensity results at Bukhara (Uzbekistan, Central Asia)

Highlights

• We report 9 new archeointensity data from Central Asia dating from ~1560 to ~1800_.

• Our results agree with previous Triaxe data obtained in Western Eurasia.

• Taken together, these data support a non-monotonic evolution of the axial dipole between ~1600 and ~1840_.

• A minimum ∣g₁⁰∣ of ~29400 nT occurred during the second half of the 18th century.

• The subsequent increase of ∣g₁⁰∣ from that period to ~1840 is confirmed by Triaxe data in Brazil.

Abstract

Since the mid-19th century, direct measurements of both intensity and direction of the Earth's magnetic field have been available, allowing an accurate determination of its spatio-temporal variations. Prior to this time, between ~1600 and 1840, only direct directional measurements are available. Therefore, the construction of global field models over this period requires either a specific treatment of the axial dipole field component or the use of archeomagnetic intensity data. In this study, we use a regional approach based on the construction of an archeointensity variation curve in Central Asia. We analyze baked clay brick fragments sampled in Bukhara (Uzbekistan), dated between the end of the 16th century and the beginning of the 19th century. This city is of particular interest for archeomagnetism due to the well-preserved old buildings accurately dated by documentary archives. A series of archeointensity results is obtained using the Triaxe experimental protocol, which shows a decreasing trend in intensity from ~1600 to ~1750, with intensities during the 18th century lower than expected from global geomagnetic field models. These new data appear consistent with other Triaxe data previously obtained in western Europe and western Russia, when transferred to Bukhara using the field geometry of the gufm1 model. Together, these data are used to recalibrate the axial dipole moment evolution provided by this model. The resulting evolution appears non-linear, with a clear relative minimum in the magnitude of the axial dipole during the late 18th century. We illustrate the fact that at present this evolution can neither be satisfactorily confirmed nor refuted by other datasets available in western Eurasia (as well as at a wider spatial scale), mainly due to the significant dispersion of the data. Our interpretation relies on the accuracy of the field geometry of the gufm1 model, which appears less reliable prior to ~1750. Nevertheless, the minimum proposed in the 18th century seems to be a true feature of axial dipole behavior.

Introduction

Variations of Earth's magnetic dipole cover a wide range of timescales from a year or less to tens of millions of years. Three different frequency bands are evidenced by analyses of the dipole power spectrum from paleo- and geomagnetic data and simulations (Constable and Johnson, 2005; Ziegler et al., 2011; Olson et al., 2012; Panovska et al., 2013; Bouligand et al., 2016; Lesur et al., 2018): an ultra-low to low frequency band (UF), a transitional frequency band (TF), and a high frequency band (HF). The UF band comprises chrons and superchrons and is associated with the thermal evolution of the outer core. The TF band covers paleo- archeomagnetic secular variations and is associated with geodynamo processes. Finally, the HF band contains the shortest periodicities of the axial dipole's variations (as observed from satellite data). These bands are separated by two cut-off frequencies T_s (between UF and TF) and T_f (between TF and HF), estimated by Hellio and Gillet (2018) from recent field statistics as T_s = 100 kyr and T_f = 60 yr, for the purpose of constructing the COV-ARCH model (more on global models below). The axial dipole's power spectrum from numerical dynamo simulations corroborates these results (Olson et al., 2012; Bouligand et al., 2016), although the estimated characteristic timescale T_f is longer (T_f ~ 10² − 10³ yrs), which is probably associated with the convective timescale in the outer core of order 150 yr. While secular variations recovered from global archeomagnetic models are representative of the low-frequency TF band, regional variation curves spanning the last few millennia based on high-quality archeomagnetic data could be associated with the high-frequency band, on time scales on the order of the convective turnover time (e.g. Genevey et al., 2016, Genevey et al., 2019).

Studying past field variations requires the construction of time-dependant global field models from the compilation of direct (or instrumental) and/or indirect geomagnetic field measurements. One of the most widely used models is the gufm1 model, which covers the past 400 years (Jackson et al., 2000) from 1590 to 1990, and which was constructed from a large set of direct geomagnetic measurements obtained in land-based observatories and by mariners during their voyages across the seas (e.g. Jonkers et al., 2003), as well as from satellite data for the most recent period. However, our ability to instrumentally measure geomagnetic field intensities only dates back to the 1830s (Gauss, 1833). To overcome this lack of intensity data, Jackson et al. (2000), following Barraclough (1974), impose a linear decay rate of 15 nT/yr to the axial dipole component between 1590 and 1840, i.e. a rate corresponding to a crude extrapolation back in time of the behavior observed since ~1840. Since it is essential for the construction of the gufm1 model, and in general for our knowledge of geomagnetic field behavior during the historical period, this crude extrapolation has been tested against paleo- archeointensity data (i.e. indirect measurements) provided by the study of the thermoremanent magnetization carried by archeological artifacts and volcanic deposits (e.g. Gubbins et al., 2006; Finlay, 2008; Genevey et al., 2009; Hartmann et al., 2011; Suttie et al., 2011; Poletti et al., 2018). Hulot et al. (1997) indeed establish that the geomagnetic field can be recovered from directional data alone, up to a constant multiplier (the uniqueness of the sought-after solution being guaranteed by the existence of two, and only two, poles at Earth's surface). The multiplicative constant is in practice provided by independent intensity measurements, each Gauss coefficient entering the mathematical description of the field being renormalized to account for the intensity measured at the specific location of interest.

Gubbins et al. (2006) follow this line of reasoning and this is the first study to use the set of indirect intensity data available between 1590 and 1840 to recalibrate the axial dipole component provided by gufm1 by the ratio of measured to predicted intensities at intensity determination sites. Due to scattered data, they assume that a linear fit is indeed the most reasonable solution prior to 1840, but estimate that the axial dipole component between 1590 and 1840 had a rate of decrease of 2.28 ± 2.72 nT/yr, which is significantly lower than that proposed by Barraclough (1974) and used by Jackson et al. (2000) (15.46 nT/yr and 15 nT/yr respectively).

Next, Finlay (2008) combines both direct and indirect geomagnetic measurements to calculate a new geomagnetic field model between 1590 and 1840, without imposing a linear decrease in the axial dipole during this period (but with an artificial overweighting of the indirect records). He shows that this approach does not provide better results than those favoring no change in axial dipole during the 17th and 18th centuries.

Suttie et al. (2011) propose a radically different approach based on the statistical analysis of errors in the paleo- archeointensity data. In particular, the dataset available between 1840 and 1990 is used to estimate reasonable errors in the data, which are best assigned as fractions (~15%) of the field intensity values expected from gufm1. When applied to data prior to 1840, and again assuming a linear evolutionary trend in axial dipole over this period, they find a rate of decay (~11.9 nT/yr) close to what Barraclough (1974) found. They further show that if data errors are assigned as fractions of measured intensities, the decay rate is similar to that proposed by Gubbins et al. (2006) and Finlay (2008) (i.e., with either a slight change or no change at all in the axial dipole component over the 17th and 18th centuries). However, this observation is the result of a bias toward lower field values, as their uncertainties are lower when given as a proportion of measured intensities.

For the different methods above, dispersion of paleo- archeomagnetic data is such that it prevents overcoming the assumption of a linear evolution of the axial dipole component over the historical period. In addition, Suttie et al. (2011) demonstrate that the use of quality criteria on the dataset does not significantly change the conclusions. More recently, Poletti et al. (2018) also use a selected global dataset with strict criteria covering the historical period (1590 − 2009). After converting intensity data into corresponding axial dipole moments and performing linear regression computations for datasets covering various time intervals, they reach a conclusion favoring a linear decreasing trend of the axial dipole over the historical period of ~12.5 nT/yr, thus close to that advocated by Barraclough (1974) and Suttie et al. (2011).

Given the dispersion observed in the global compilation of intensity data regardless of the selection criteria considered, Genevey et al. (2009) explore a different approach using a single consistent regional intensity dataset to recalibrate the $g_1^0$ coefficient of gufm1. The principle remains the same as above (Hulot et al., 1997), which assumes that the geometry of the geomagnetic field as provided by gufm1 is correct. While it potentially avoids the problem of global data scatter, and the almost insoluble issue of selecting only the most reliable data, it does raise the pending issue of which dataset is sufficiently reliable to be used to recalibrate the Gauss coefficients (an evaluation that will surely vary from one author to another). Genevey et al. (2009) use the set of accurate and precisely dated archeointensity results obtained in western Europe (700 km around Paris, France). Instead of a linear decrease of the axial dipole magnitude over the historical period, they find a significant decrease between ~1590 and the second half of the 18th century, with a minimum magnitude during this period, followed by a moderate increase from ~1800 to ~1840 and then, the well-established linear decrease up to the present. As a follow-up to this first study, Hartmann et al., 2010, Hartmann et al., 2011 analyze precisely dated architectural brick fragments from southern and northern Brazil. Despite a significant non-dipole field effect between these two regions associated with the South Atlantic Anomaly (SAA), the results obtained appear to support the evolution in dipole field moment proposed by Genevey et al. (2009). As a new development, the present study carried out in Central Asia (Bukhara, Uzbekistan) focusing on the 1590 to 1850 period aims to further constrain the accuracy of the non-linear dipole moment evolution deduced from the western European dataset.