Earth and Planetary Science Letters 255 (2007) 106 – 116 www.elsevier.com/locate/epsl FENNOSTACK and FENNORPIS: Varve dated Holocene palaeomagnetic secular variation and relative palaeointensity stacks for Fennoscandia Ian Snowball a,⁎, Lovisa Zillén a , Antti Ojala b , Timo Saarinen c , Per Sandgren a a GeoBiosphere Science Centre, Department of Geology, Quaternary Sciences, Lund University, Sölvegatan 12, SE-223 63 Lund, Sweden b Geological Survey of Finland, PO Box 96, FIN-02151 Espoo, Finland c Department of Geology, University of Turku, FIN-20014 Turun yliopisto, Finland Received 23 July 2006; received in revised form 11 December 2006; accepted 11 December 2006 Available online 15 December 2006 Editor: R.D. van der Hilst Abstract A Holocene palaeomagnetic secular variation master curve (FENNOSTACK) and a relative palaeointensity curve (FENNORPIS) for Fennoscandia are presented. These curves were produced by stacking palaeomagnetic data obtained from six annually laminated (varved) lake sediment sequences and one non-laminated sediment sequence. The six independent varve chronologies were combined to form a timescale, which extends between the present year and 10,200 Cal yr BP. Smoothed inclination and declination curves show trends that are similar to a palaeomagnetic secular variation curve for the United Kingdom, except that the ages of most of the statistically significant features are a few hundred years younger in the Fennoscandian data set. These differences are most likely due to excessively old ages of the United Kingdom features produced by the radiocarbon dating of bulk sediments. The standardized relative palaeointensity stack shows that the intensity of the geomagnetic field in Fennoscandia was highest at 2300 Cal yr BP and lowest at 7000 Cal yr BP. Three palaeointensity peaks (PIPs) and three palaeointensity lows (PILs) are identified in FENNORPIS and indicate that the intensity of geomagnetic field varied significantly on millennial scales in this region during the Holocene, but there is much scatter in the reconstruction prior to 8000 Cal yr BP. Higher frequency palaeointensity features present in individual records are not significant in the stacked data set, which implies that these are a result of local environmental bias. Significant discrepancies exist between the data and the output of the CALS7K.2 geomagnetic field model for the last 7000 yr. In particular, the largest declination swing observed in the data at 2670 Cal yr BP (feature “f”) is not produced by the model, which is most likely due to the influence of an inaccurately dated Icelandic secular variation curve on the model result. Short term trends in standardized relative palaeointensity do differ from the model output, but it is difficult to make a comparison because of the relative nature of the sediment record. Better absolute palaeointensity data are needed to reduce uncertainties in reconstructions of past geomagnetic field strength. © 2006 Elsevier B.V. All rights reserved. Keywords: Holocene; palaeomagnetism; secular variation; relative palaeointensity; Sweden; Finland; varves ⁎ Corresponding author. Tel.: +46 46 2223952; fax: +46 46 2224830. E-mail addresses: [email protected] (I. Snowball), [email protected] (L. Zillén), [email protected] (A. Ojala), [email protected] (T. Saarinen), [email protected] (P. Sandgren). 0012-821X/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2006.12.009 I. Snowball et al. / Earth and Planetary Science Letters 255 (2007) 106–116 1. Introduction Palaeomagnetic reconstructions, which are based on the natural remanence acquired by igneous rocks, archaeological artifacts and sediments at the time of their formation, are the only independent means of reconstructing the pre-historical variability of the geomagnetic field. Recent achievements in ocean sediment recovery [1] have improved reconstructions of the geomagnetic field during the majority of the Bruhnes chron (0–780,000 yr BP). Reconstructions of field behaviour for the current Holocene epoch, however, are dominated by studies of continental lake sediments because they are relatively easier to recover intact and can often be dated with better accuracy and precision than marine sediments. Palaeomagnetic intensity data [2] and the output of time varying geomagnetic field models [3] have also been used to correct radionuclide production for the variable geomagnetic field, thus producing different reconstructions of solar activity in the Holocene [4–6]. The largest source of uncertainty in these reconstructions is our incomplete knowledge of how the geomagnetic field varied in the past. New data are, therefore, required to improve the confidence of palaeomagnetic reconstructions and constrain time varying geomagnetic field models. Two and a half decades ago, Turner and Thompson [7] reconstructed a Holocene palaeomagnetic secular variation master curve for Britain (United Kingdom) which was based on multiple sediment cores collected from three lakes. More recent studies of lake sediments [e.g. 8] and marine sediments [e.g. 9] continue to improve our understanding of changes in the direction of the geomagnetic field. However, it is still rare that reconstructions integrate data obtained from multiple sites within one region and include a complete reconstruction of the geomagnetic vector, i.e. of the direction and intensity, the latter as relative palaeointensity (RPI). Over the past decade we have made an intensive effort to reconstruct the behaviour of the Holocene geomagnetic field by studying a series of annually laminated (varved) lake sediments that exist in Finland and Sweden [10–14]. These studies have generated a set of seven varve dated records of palaeomagnetic secular variation, which include varve dated records of RPI that conform to established reliability criteria [15,16]. The study of an eight, partially varved site has provided a secular variation and RPI data set dated by the radiocarbon method and the transference of the PSV features from a varve dated master curve [17]. We present here stacked palaeomag- 107 netic secular variation and RPI master curves of seven of these sites, which are based on the multiple varve chronologies (two of the varve chronologies are supported by radiocarbon dating and tephrochronology). These data were not used to constrain the time varying CALS7K.2 geomagnetic field model [18] and thus can be used to test the validity of its results for Fennoscandia. 2. Regional setting The locations of the seven sites from which we have used data are shown in Fig. 1. The inset in Fig. 1 shows the location of these sites in relation to the three United Kingdom sites studied by Turner and Thompson [7]. The Fennoscandian network extends between 12 and 25° East (13° of longitude) and between 57 and 64° North (7° of latitude) and, therefore, covers a considerably larger area than the United Kingdom master curve (2° of longitude and 4° of latitude). According to the International Geomagnetic Reference Field (IGRF) the geographical spread of Fennoscandian sites amounts to about 5° in inclination and 4° in declination. We have not included inclination and declination data from Korttajärvi [14] because the sediment accumulation rate was much faster and more variable than in any of the chosen sites. With the exception of the Byestadsjön sequence, which is partially varved in the early Holocene, the sediments in the lakes are composed of continuous biogenic/clastic varves. These varves have formed since isostatic uplift during the Holocene caused these basins to be isolated from much larger freshwater and brackish basins that existed after the last deglaciation. Petterson [19] has reviewed the processes that lead to freshwater varve formation in this boreal environment. 3. Materials and methods The palaeomagnetic data sets used in this study are listed in Table 1, which includes information about the location, dating methods, the technique used to reconstruct relative palaeointensity, the number of cores from each site and the original reference. A total of 14 sediment piston-cores contribute to the compilation. All palaeomagnetic measurements were carried out on discrete samples (2 × 2 × 2 cm) with 2G-Enterprises superconducting rock magnetometers at the Department of Geology, University of Lund and the Geological Survey of Finland, respectively. With the exception of Byestadsjön, each site has an independent calendar year timescale that is based on varve counting. The varve 108 I. Snowball et al. / Earth and Planetary Science Letters 255 (2007) 106–116 Fig. 1. The map shows the positions of the seven lake sediment sequences in Sweden and Finland that contribute to this study. The inset shows the positions of these sites with respect to the three sites that form the basis of a palaeomagnetic master curve for the United Kingdom [7]. chronologies are based on sites where varves continue to form and are, therefore, fixed to the present year. Sediment accumulation rates have ranged between 0.3 and 0.7 m/ka in these sites. Henceforth, we use an annual timescale where the zero calendar year before present (0 Cal yr BP) is AD 1950, which facilitates comparison to calibrated radiocarbon ages. The total age span of the stacked record is between 10,200 and − 30 Cal yr BP (where − 30 Cal yr BP is AD 2000). As the aim of this study, however, was to produce a secular variation master curve based on multiple sites, our final results extend over a common base period for which there are reliable data from at least three of the sites (401–9714 Cal yr BP for inclination and declination, and 161–9845 Cal yr BP for relative palaeointensity). 3.1. Compiled data set The frequency histogram in Fig. 2 shows the distribution of the samples according to bins of 500 yr. The sample population (n) for the inclination and declination sub-set data set (Appendix A) is 1891, while Table 1 Sites and data references Site name Location Dating methods RPI method No. of cores References Sarsjön Frängsjön Nautajärvi Alimmainen Savijärvi Furskogstjärnet Mötterudstjärnet Byestadsjön 64°02′ N, 19°36′ E 64°01′ N, 19°42′ E 61°45′ N, 24°24′ E 61°48′ N, 24°41′ E 59°23′ N, 12°05′ E 59°40′ N, 12°40′ E 57°24′ N, 15°18′ E Varve chronology Varve chronology Varve chronology Varve chronology Varve chronology, 14C, tephra Varve chronology, 14C, tephra 14 C and PSV correlation NRM/ARM at 40 mT NRM/ARM at 40 mT NRM/ARM at 20 mT – NRM/ARM at 40 mT NRM/ARM at 40 mT Pseudo-Thellier and NRM/ARM at 30 mT 2 2 1 1 2 2 4 [12] [12] [11] [14] [13] [13] [17] I. Snowball et al. / Earth and Planetary Science Letters 255 (2007) 106–116 109 Fig. 2. Two histograms that show the number of palaeomagnetic sub-samples in each 500-year bin for the last 10,000 yr. (A) shows the number of samples for palaeomagnetic directions and (B) shows the number of samples for relative palaeointensity. More than 75 data points contribute to each 500-year bin during the period between 9000 and 500 Cal yr BP. the relative palaeointensity sample population is 1756 (Appendix A). The period between 8500 and 8000 Cal yr BP contains the most data points (∼ 125) because of relatively high sediment accumulation rates (thickest varves) that occurred in the first centuries after lake basin isolation, while the periods earlier than 9000 Cal yr BP contain less than 70 data points due to the restricted age of individual records. 3.2. Palaeomagnetic directions and statistics Published palaeomagnetic directions (inclination and declination) were available for 14 individual cores and the raw data from each site are shown in equal area stereo plots in Fig. 3. Lakes that are situated within 1° of latitude and longitude of each other were considered as single sites and these data form sub-sets. The cores were not recovered with respect to a fixed azimuth and the declination data were subsequently calculated as deviations from their respective core averages. References given in Table 1 contain details of de-trending corrections made for probable core rotations. Statistics relating to the individual data sets are shown in Table 2 (Palaeomag-Tools version 4.2a provided by Mark Hounslow, Lancaster University). No corrections have been made for potential shallow inclinations, which can be expected to affect detrital remanent magnetizations, because none of the inclination averages deviated more than 5° from the inclination expected by a geo-axial dipole (GAD) at the respective site latitude (Fig. 3 and Table 2). It is not possible to determine what correction, if necessary, should be applied and it is possible that differences in the distributions in Fig. 3 could be due to other processes (e.g. coring methods, sub-sampling bias and analytical precision). We do not present a stacked data set based on transformation to virtual geomagnetic poles (VGPs) because this procedure would be based on the assumption of a GAD and could provide a false image about geomagnetic pole movement. As the geographical spread of sites should lead to differences in absolute inclination and declination values of a few degrees the conversion to deviations from core averages was made before the four sub-sets were amalgamated. Thus, the inclination data were converted to deviations from their respective core average and then the data points for inclination and declination (n = 1891) were stacked according to their varve ages. 3.3. Relative palaeointensity (RPI) and remanence carriers The mineral magnetic characteristics of the Swedish varved lake sediments suggest that single domain magnetite crystals produced by magnetotactic bacteria 110 I. Snowball et al. / Earth and Planetary Science Letters 255 (2007) 106–116 Fig. 3. Equal area stereoplots that show the distribution of the characteristic inclination and declination data obtained for the seven sites. Declination data were available as deviations from the respective core averages. Data obtained for sites located within 1° of latitude and longitude are shown together (A) Sar–Fra = Sarsjön and Frängsjön, (B) Fur–Mot = Furskogstjärnet and Mötterudstjärnet, (C) Naut–Ali = Nautajärvi and Alimmainen Savijärvi, (D) Bye = Byestadsjön). carry the majority of the natural remanent magnetization (NRM) in the post-isolation sediments [13,20,21]. No detailed mineral magnetic investigations of the Finnish sites have been carried out, but the median destructive fields of NRMs and anhysteretic remanent magnetizations (ARMs) of pilot samples were approximately 40 mT, which indicates that similar sized magnetite particles carry the NRM in these sites as well. It is also notable that the NRM intensity of the samples from all the sites was of similar magnitude to the anhysteretic remanent magnetization (ARM) acquired in a direct current bias field of 0.05 mT superimposed on a peak alternating field of 100 mT, which indicates that these sediments are efficient recorders of the geomagnetic field. The RPI was determined in all cases by normalization of the NRM Table 2 Palaeomagnetic statistics Sub-stack N Sar–Fra Fur–Mot Naut–Ali Bye 549 495 387 458 Mean declination Mean inclination Fisher Fisher 0.3 358.5 359.6 1.1 76.9 69.6 70.1 68.7 Alpha 95 Mean inclination Alpha 95 0.5 0.5 0.6 0.5 76.0 69.2 69.4 68.3 0.6 1.0 0.8 0.5 Inclination GAD 76.3 73.3 74.5 72.0 I. Snowball et al. / Earth and Planetary Science Letters 255 (2007) 106–116 to the ARM (precise details of the normalization procedures are provided in the references in Table 1) and relative palaeointensity estimates have been published for 13 of the 14 sediment cores. These data (n = 1756) were standardized according to the mean and standard deviation associated with each core and then stacked according to their varve age. 3.4. Smoothing routine Scatter in stacked palaeomagnetic data obtained from sediments can originate from true changes in geomagnetic field direction and intensity, biased recording of the NRM caused by environmental changes and sedimentation processes, sediment coring, extrusion and analytical error. A further source of error is also introduced when multiple records are stacked according to the ages of individual samples, because no geochronology is perfect and each varve chronology used here contains some inaccuracy and imprecision. However, for the purposes of this study we implicitly trust the six independent varve chronol- 111 ogies and do not make any adjustments of the time– depth relations based on inclination, declination and RPI correlations (with the exception of the partially varved Byestadsjön sequence). Thus, we avoid as far as possible the potential reinforcement of high frequency features that were not caused by geomagnetic field changes. We reduced the high frequency scatter by using a smoothing routine designed specifically for palaeomagnetic data, which provides 95% confidence limits [22]. The two data stacks (one for directions and one for RPI) were smoothed using a running time window of 150 yr, which was chosen because it is equivalent to the maximum error in the individual varve chronologies (see references in Table 1) and 100–200 yr is considered to be a good estimate of the average lock-in delay of the NRM associated with the sediment sequences studied [e.g. 12]. The CALS7K.2 geomagnetic field model has a resolution of approximately 100 yr [18] and although the use of progressively longer period filters can improve confidence, it also introduces the risk that higher frequency secular variation is removed. Fig. 4. Relative declinations (A) and relative inclinations (B) obtained for the seven sites plotted against their varve age (in Cal yr BP, where the 0 yr is AD 1950). (C) shows the standardized RPI values obtained for six of the seven sites (see Table 1). The individual site data are shown as three point moving averages in D, E and G (site labels as Fig. 3). Note the difference in the declination scale between A and D. 112 I. Snowball et al. / Earth and Planetary Science Letters 255 (2007) 106–116 4. Results 4.1. FENNOSTACK and FENNORPIS: inclination and declination, and relative palaeointensity data stacks for Fennoscandia Fig. 4 shows the stacked declination (A), inclination (B) and standardized RPI (C) data sets plotted against calendar years. Three point moving averages of the individual site data are shown in Fig. 4D–G. The majority of the raw declination data does not vary more than ± 50° from the average of zero. Exceptions are related to periods of high inclination at 8000, 3000 and 1000 Cal yr BP, respectively. Apparently anomalous declinations at high latitude sites can be caused by relatively small vertical offsets in core penetration and sub-sampling, which can cause the declination to rotate to the opposite hemisphere (see Fig. 3). The raw inclination data (Fig. 4B) and individual site data (Fig. 4E) clearly indicate that the steepest values occurred during the last 3000 yr and the shallowest values at approximately 9000 Cal yr BP. The standardized relative palaeointensity data (C) appear more scattered than the directional data, but there is a distinct maximum between 3000 and 2000 Cal yr BP and a distinct minimum at approximately 7000 Cal yr BP. Fig. 5 shows the result of applying the 150-year smoothing routine to the stacked data, with 95% confidence limits indicated. We call the smoothed directional data set “FENNOSTACK” (Appendix A), and the smoothed RPI stack “FENNORPIS” (Appendix A). Despite a systematic age difference between the UK master curve and the North Sweden master curve [12] prior to 3000 Cal yr BP the pattern of secular variation Fig. 5. The result of smoothing the data shown in Fig. 4 with a 150year running window. Grey lines show the 95% confidence limits associated with the declination (A), inclination (B) and standardized RPI (C). Solid and dashed vertical black lines show the positions of maxima and minima, respectively, which have been labeled according to the nomenclature of the United Kingdom palaeomagnetic secular variation curve [7]. In (C) odd and even numbers represent significant RPI peaks and lows, respectively. over the common base period is the same as the United Kingdom master curve and, therefore, we apply the nomenclature of Turner and Thompson [7] to label the Table 3 Ages of inclination and declination features in the Fennoscandian PSV master curve, their UK counterparts and the deviation between them Declination feature a b c d e f g1 h i j Fennostack (Cal yr BP) UK (Cal yr BP) Fennostack–UK (yr) Inclination feature 850 2210 2670 5500 6610 7920 9410 150 450 600 1150 1950 2705 5550 7165 8400 – – – – − 300 − 260 − 35 − 50 − 555 − 480 α β γ δ ε1 ε ζ η θ ι K λ μ ν Fennostack (Cal yr BP) UK age (Cal yr BP) Fennostack–UK (yr) 1290 1840 2590 3090 – – 4720 5400 6290 7890 9000 9780 250 650 1250 1650 ? 3270 4000 4200 5200 6000 7000 8300 8800 – – – 40 190 −180 −480 −600 −610 −410 200 I. Snowball et al. / Earth and Planetary Science Letters 255 (2007) 106–116 statistically significant declination and inclination features that we identify in FENNOSTACK. Table 3 and Fig. 6 compare the ages of these features and show that the inclination features between 8000 and 4000 Cal yr BP are approximately 700 yr younger in FENNOSTACK compared to the United Kingdom master curve. The oldest inclination feature and the two youngest features are slightly younger in the United Kingdom master curve. We assign numbers to six significant relative palaeointensity features, where odd numbers denote palaeointensity peaks (PIPs) and even numbers palaeointensity lows (PILs). The ages of these features are shown in Table 4. We also label an earlier peak (PIP7) but this should be treated cautiously because the confidence is poor. 113 Table 4 Ages of relative palaeointensity features in FENNORPIS Palaeointensity peak (PIP) FENNORPIS (Cal yr BP) 1 900 1400 2300 5400 6300 7200 8800 (?) 3 5 7 (?) Palaeointensity low (PIL) 2 4 6 predictions for the last 7000 yr were standardized and linear interpolation was applied to the FENNOSTACK and FENNORPIS data to provide time points consistent 4.2. FENNOSTACK and FENNORPIS compared to CALS7K.2 Fig. 7 shows the FENNOSTACK inclination and declination data next to the results of the CALS7K.2 geomagnetic field model [18]. The modeled inclination and declination data obtained for the four site positions (Table 1) were converted to deviations from the average direction for the last 7000 yr, which produced four values for each time point at 10-year intervals (the range of values is shown by the blue shaded band in Fig. 7). The FENNORPIS data and the CALS7K.2 intensity Fig. 6. A comparison between the ages of the inclination (crosses) and declination (open circles) features identified in FENNOSTACK and the UK palaeomagnetic master curve [7]. The grey dotted line shows where the ages would be equal. The majority of the features identified prior to 3000 Cal yr BP are between 400 and 600 yr younger in FENNOSTACK compared to the UK data set (see Table 3). Fig. 7. A comparison between the output of the CALS7K.2 geomagnetic field model [3] and the FENNOSTACK and FENNORPIS data for Fennoscandia. Inclination (A) and declination (A and B) were calculated as deviations from their respective averages for the last 7000 yr, while the RPI data were standardized over the same period (C). CALS7K.2 predicts a false declination feature between 4500 and 3900 Cal yr BP, and it fails to predict the most significant declination feature (f) observed in FENNOSTACK at 2650 Cal yr BP. CALS7K.2 fails to predict the period of steepest inclinations between 3200 and 2700 Cal yr BP, and it underestimates inclination for much of the last 3000 yr. The multi-millennial scale trends in RPI are similar between FENNORPIS and CALS7K.2, but there are differences at higher frequencies. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 114 I. Snowball et al. / Earth and Planetary Science Letters 255 (2007) 106–116 with the model output. The four model predictions generated for each time point were averaged, and the differences between the model and the data calculated. In Fig. 8 we show the differences between the CALS7K.2 model and FENNOSTACK (including FENNORPIS) for the last 7000 yr. The average errors in the model results for inclination and declination are shown by dashed grey lines in Fig. 8A and B. Features that exceed these limits cannot be explained by uncertainties in the model or the data. Figs. 7 and 8 show that CALS7K.2 produces an easterly declination feature with maximum amplitude of about 30° between 4500 and 3900 Cal yr BP. This is the highest amplitude feature predicted by the model, but it is not observed in the FENNOSTACK data. The modeled declination does not produce a fit to the westerly declination feature f, which is the most significant feature in the data and it appears to overestimate the range of inclination changes for the past 3000 yr. Fig. 8. The calculated differences between the CALS7K.2 model output and FENNOSTACK declination (A), FENNOSTACK inclination (B) and standardized palaeointensity (C). The dashed grey lines in (A) and (B) indicate the average errors included in the CALS7K.2 model [18]. The solid black line in (C) shows the difference between the standardized global dipole moment predicted by CALS7K.2 and FENNORPIS. The largest differences between observed RPI and the model are reflected in both regional and global output. It is difficult to compare RPI data obtained from sediments to absolute values obtained from ceramics and igneous rocks. It is important to note, however, that the major differences between the standardized RPI data obtained from FENNORPIS and the standardized results of CALS7K.2 do not depend on the use of regional or global model output (Fig. 8C). The largest consistent difference occurs between 4500 and 3500 Cal yr BP. 5. Discussion FENNOSTACK is the first Holocene palaeomagnetic secular variation master curve that includes a relative palaeointensity master curve (FENNORPIS) based on multiple sites with independent dating. The timescale is based on six independent and continuous varve chronologies that are fixed to the present. The compiled timescale also includes two varve chronologies (Furskogstjärnet and Mötterudstjärnet) that are supported by tephrochronology and radiocarbon dating of terrestrial macrofossils [23] and we consider the combined timescale to have an accuracy of better than 2% at any time point. Potential users of the stacked data must consider that the geomagnetic signal was potentially smoothed during natural lock-in of the natural remanent magnetization and that we have used statistics to reduce the scatter in the stacked records. These two smoothing processes imply that the reconstructions underestimate the amplitude of the real changes. It is notable that many of the inclination and declination features observed in FENNOSTACK are significantly younger (by some centuries) than their United Kingdom counterparts. These differences are most likely due to the radiocarbon dating of bulk sediments applied by Turner and Thompson [7], which frequently leads to overestimated ages caused by a local carbon reservoir effect. The age differences are too large to be explained solely by westward drift of higher term components of the geomagnetic field. It was also noted that the interpolated calibrated bulk sediment radiocarbon ages of the PSV features in Byestadsjön were slightly older than ages based on PSV correlation to the North Sweden master curve, but that the majority of the differences could be explained by dating uncertainties [17] and that the differences were absent in the older laminated sediments. An alternative explanation is that the lock-in depth of the NRM is significantly deeper in bioturbated (and therefore non-varved) lake sediments compared to varved sites where bioturbation is not detectable, leading to apparently older ages. Despite these uncertainties, the FENNOSTACK/FENNORPIS master curves provide an alternative method of dating I. Snowball et al. / Earth and Planetary Science Letters 255 (2007) 106–116 lake and marine sediments in and around the Scandinavian and Baltic countries, but only if they contain a valid palaeomagnetic signal. There are significant differences between our stacked data set and the results of the CALS7K.2 model for the last 7000 yr (Fig. 8). These differences must be discussed because they show that the model cannot provide a good fit for an area near central- and western Europe, which can be considered to be relatively densely covered by data. The differences imply either that (i) the data sets which constrain the CALS7K.2 output are unsuitable due to poor recording of the geomagnetic field and/or dating errors or (ii) the model cannot solve the time evolution of the magnetic field sufficiently to make predictions for regions where there are limited data. Korte and Constable [18] do reject some data sets because they do not fit the model results, but this does not apply to Fennoscandia. Despite the poor fits of inclination and declination at certain times, we cannot logically reject any of the individual data used to produce FENNOSTACK because the varve chronologies are independent and geologically robust, the natural remanent magnetizations were strong and stable and the confidence limits for the compiled palaeomagnetic data sets are relatively small. Data from only two Fennoscandian lakes (Vuokonjärvi and Pohjajärvi) were used to constrain the CALS7K.2 model [18]. Of these two, Pohjajärvi was dated by varve counting but the record extends only to 3200 Cal yr BP [10]. The Vuokonjärvi timescale is based on pollen stratigraphy [24], which is not an absolute dating technique and we would not recommend that the palaeomagnetic data from this site are used to constrain future models, unless the timescale is improved. In the case of western Europe the United Kingdom results [7] fit our data well, except that some features are too old by a few centuries. Easterly declination feature “f” is characterized by a large change in declination and it is distinct in records that extend in longitude from the United Kingdom [7], through Fennoscandia [10] to Siberia [25]. This feature appears to be muted in the stacked record produced from maar lakes located in Germany further to the south [26]. However, our major concern is that the CALS7K.2 model result for Fennoscandia is negatively influenced by data obtained from the Icelandic lake of Vatnalsvatn [27], which are dependent on a chronology based on the radiocarbon dating of mineral rich bulk sediments. Thompson and Turner [27] dated two cores and obtained an age of approximately 3600 14C BP for feature “f”, which calibrates to an age range of 3600–4200 Cal yr BP. The FENNOSTACK varve age for feature “f” is 2650 Cal yr BP and the oldest age in an individual data set included in the stack is 3000 Cal yr BP [13]. We conclude that the 115 CALS7K.2 output for Fennoscandia is biased by Icelandic lake sediment palaeomagnetic data that contain significant dating errors. Alternatively, the model underestimates the degree of spatial variability in the geomagnetic field between 5000 and 3000 Cal yr BP. Inspection of Figs. 7C and 8C shows that there is good agreement between the millennial to multi-millennial scale trends in the observed and modeled standardized palaeointensity estimates. However, our palaeointensity data are relative and we are unable to discuss systematic offsets in absolute intensity values, such as those known to exist between the virtual axial dipole moment (VADM) obtained from a global compilation of archaeomagnetic data [1] and the dipole moment (DM) produced by CALS7K.2. On the other hand, our data do suggest that there are differences between the millennial scale trends observed in our data and the results produced by the model for Fennoscandia. Compared to the long term trend for the past 7000 yr, our data suggest that the model overestimates the intensity of the regional magnetic field prior to 6500 Cal yr BP and over the last 2000 yr. It significantly underestimates the geomagnetic field intensity between approximately 4500 and 3500 Cal yr BP. These differences are essentially the same between FENNORPIS and the global DM results of CALS7K.2. Global compilations of palaeointensity data are heavily biased by European data sets, but none of these have the high temporal resolution provided by FENNORPIS. The differences between the model and FENNORPIS need to be solved if the model output is to be used for the correction of atmospheric radionuclide production for geomagnetic field changes at sub-millennial scales [6]. We emphasize that our stacked data set does not include reproducible sub-millennial scale changes in RPI, which have been used to make inter-hemispheric correlations between individual marine sediment cores [28]. Correlation between individual cores located in different hemispheres is not justified by our RPI results, which indicate that sub-millennial scale features in RPI in individual sites are more likely to be a product of environmental bias than the geomagnetic field. We have presented an independently dated record of geomagnetic field behaviour that can improve geomagnetic field models and shown that the stacking of RPI data from multiple sediment sequences can remove high-frequency environmental bias often present in individual sediment sequences. Acknowledgements This work has been supported by the Swedish Research Council, the Nordic Council of Ministers, the Academy of Finland and the Geological Survey of 116 I. Snowball et al. / Earth and Planetary Science Letters 255 (2007) 106–116 Finland. Raimund Muscheler is thanked for the critical comments on parts of the submitted manuscript and Cor Langereis provided suggestions for improvement. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j. epsl.2006.12.009. References [1] S. Lund, J.S. Stoner, J.E.T. Channell, G. Acton, A summary of Bruhnes paleomagnetic field variability recorded in Ocean Drilling Program cores, Phys. Earth Planet. Inter. 156 (2006) 104–204. [2] S. Yang, H. Odah, J. Shaw, Variations in the geomagnetic dipole moment over the last 12 000 years, Geophys. J. Int. 140 (2000) 158–162. [3] M. Korte, C.G. Constable, The geomagnetic dipole moment over the last 7000 years—new results from a global model, Earth Planet. Sci. Lett. 236 (2005) 348–358. [4] S.K. Solanki, I.G. Usoskin, B. Kromer, M. Schüssler, J. Beer, Unusual activity of the Sun during recent decades compared to the previous 11,000 years, Nature 431 (2004) 1084–1087. [5] I.G. Usoskin, S.K. Solanki, M. Korte, Solar activity reconstructed over the last 7000 years: the influence of geomagnetic field changes, Geophys. Res. Lett. 33 (L08103) (2006) 1–4. doi:10.1029/2006GL025921. [6] R. Muscheler, F. Joos, S. Mueller, I. Snowball, How unusual is today's solar activity? Nature (2005) E3–E4. doi:10.1038/ nature04045. [7] G. Turner, R. Thompson, Lake sediment record of the geomagnetic secular variation in Britain during Holocene times, Geophys. J. R. Astron. Soc. 65 (1981) 703–725. [8] C.S.G. Gogorza, A.M. Sinito, J.M. Lirio, H. Nunez, M. Chaparro, M. Vilas, Paleosecular variations 0–19,000 years recorded by sediments from Escondido Lake (Argentina), Phys. Earth Planet. Inter. 133 (2002) 35–55. [9] S.P. Lund, L. Stott, M. Schwartz, R. Thunell, A. Chen, Holocene paleomagnetic secular variation records from the western Equatorial Pacific Ocean, Earth Planet. Sci. Lett. 246 (2006) 381–392. [10] T. Saarinen, High-resolution palaeosecular variation in northern Europe during the last 3200 years, Phys. Earth Planet. Inter. 106 (1998) 299–309. [11] A.E.K. Ojala, T. Saarinen, Palaeosecular variation of the Earth's magnetic field during the last 10 000 years based on the annually laminated sediment of Lake Nautajärvi, central Finland, Holocene 12 (2002) 391–400. [12] I. Snowball, P. Sandgren, Geomagnetic field variations in northern Sweden during the Holocene quantified from varved lake sediments and their implications for cosmogenic nuclide production rates, Holocene 12 (2002) 517–530. [13] L. Zillén, Setting the Holocene clock using varved lake sediments in Sweden, LUNDQUA PhD thesis, 50 (2003) 1–37. [14] A.E.K. Ojala, M. Tiljander, Testing the fidelity of sediment chronology: comparison of varve and paleomagnetic results from Holocene lake sediments from central Finland, Quat. Sci. Rev. 22 (2003) 1787–1803. [15] J.W. King, S.K. Banerjee, J. Marvin, A new rock-magnetic approach to selecting sediments for geomagnetic paleointensity studies: application to paleointensity for the last 4000 years, J. Geophys. Res. 88(B) (1983) 5911–5921. [16] L. Tauxe, Sedimentary records of relative paleointensity of the geomagnetic field: theory and practice, Rev. Geophys. 31 (1993) 319–354. [17] I. Snowball, P. Sandgren, Geomagnetic field intensity changes in Sweden between 9000 and 450 cal BP: extending the record of “archaeomagnetic jerks” by means of lake sediments and the pseudo-Thellier technique, Earth Planet. Sci. Lett. 227 (2004) 361–376. [18] M. Korte, C.G. Constable, Continuous geomagnetic field models for the past 7 millennia: 2. CALS7K, Geochem. Geophys. Geosyst. 6 (2005) Q02H16. doi:10.1029/2004GC000801. [19] G. Petterson, Varved sediments in Sweden: a brief review, in: A.E.S. Kemp (Ed.), Palaeoclimatology and Palaeoceanography from Laminated Sediments, Geol. Soc. Spec. Pub., vol. 116, 1996, pp. 73–77. [20] I.F. Snowball, P. Sandgren, G. Petterson, The mineral magnetic properties of an annually laminated Holocene lake sediment sequence in northern Sweden, Holocene 9 (1999) 353–362. [21] I.F. Snowball, L. Zillén, P. Sandgren, Biogenic magnetite in Swedish varved lake sediments and its potential as a biomarker of environmental change, Quat. Int. 18 (2002) 13–19. [22] S. Björck, P. Sandgren, B. Holmquist, A magnetostratigraphic comparison between 14C years and varve years during the Late Wiechselian, indicating significant differences between timescales, J. Quat. Sci. 2 (1987) 133–140. [23] L.M. Zillén, S. Wastegård, I.F. Snowball, Calendar year ages of three mid-Holocene tephra layers identified in varved lake sediments in west central Sweden, Quat. Sci. Rev. 21 (2002) 1583–1591. [24] P. Huttunen, J. Stober, Dating of palaeomagnetic records from Finnish lake sediment cores using pollen analysis, Boreas 9 (1980) 193–202. [25] U. Frank, N.R. Nowaczyk, J.F.W. Negendank, M. Melles, A palaeomagnetic record from Lake Lama, northern Central Siberia, Phys. Earth Planet. Inter. 133 (2002) 3–20. [26] H. Stockhausen, Geomagnetic palaeosecular variation (0– 13 000 yr BP) as recorded in sediments from three maar lakes from the West Eifel (Germany), Geophys. J. Int. 135 (1998) 898–910. [27] R. Thompson, G. Turner, Icelandic Holocene palaeolimnomagnetism, Phys. Earth Planet. Inter. 38 (1985) 250–261. [28] V. Willmott, E.W. Domack, M. Canals, S. Brachfeld, A high resolution relative paleointensity record from the Gerlache–Boyd paleo-ice stream region, northern Antarctic Peninsula, Quat. Res. 66 (2006) 1–11.
© Copyright 2026 Paperzz