Magnetic characterisation of present-day deep-sea sediments and … · 2009-09-14 · netic properties may allow identi¢cation of the sources and transport pathways of detrital mag-neticgrainsor,intheabsenceofsigni¢cantdetri-tal

Magnetic characterisation of present-day deep-sea sedimentsand sources in the North Atlantic

S.J. Watkins, B.A. Maher �

Centre for Environmental Magnetism, Lancaster Environment Centre, Geography Department, Lancaster University,Lancaster LA1 4YB, UK

Received 25 March 2003; received in revised form 18 July 2003; accepted 18 July 2003

Abstract

The North Atlantic plays a key role in global climate through formation of North Atlantic Deep Water, whichdrives the thermohaline circulation. To understand past and future climate change, it is essential to understand theprocesses occurring within the present-day North Atlantic but ocean-wide studies of present-day sediment distributionand sources are relatively sparse. Here, we use magnetic measurements to characterise the surface sediments of theNorth Atlantic and identify the major climatic and oceanographic controls on their magnetic signatures. The magneticdata, and subsequent cluster analysis, identify distinct spatial patterns of sediment sources and transport pathways.Much of the sedimentary magnetic signal appears to be controlled by detrital inputs, especially of windblown dust,and ice-rafted debris (IRD), with a range of different sources. These sediment transport pathways can be validated bymaking direct comparison of the sediment magnetic properties with source rocks and soils, and with iceberg trajectoryand observation data. The spatial distributions of the IRD-dominated sediments substantiate those mappedpreviously using lithological tracers, provide additional spatial information on sediment sources and pathways, andsuggest that deep water currents are less significant than proposed in controlling present-day sediment mineralogy anddistribution. The data form a present-day basis for comparison with glacial-stage IRD and dust distributions, andmodes of ocean circulation.1 2003 Elsevier B.V. All rights reserved.

Keywords: North Atlantic circulation; sediment magnetic properties; sediment sources

1. Introduction

The North Atlantic plays a key role in globalclimate through formation of North Atlantic

Deep Water (NADW), which drives the thermo-haline circulation and controls the transport ofheat polewards from the equator. Palaeoclimatedata and ocean/atmosphere circulation models in-dicate that variability of convective activity in theNordic and Labrador Seas can result in rapidclimate change, on timescales of 1^105 years.Any future climatic change may be controlledby, and will certainly a¡ect, the oceanographyof the North Atlantic. In order to understandboth past and future change, it is essential to

0012-821X / 03 / $ ^ see front matter 1 2003 Elsevier B.V. All rights reserved.doi:10.1016/S0012-821X(03)00422-9

* Corresponding author. Tel. : +44-1524-593169;Fax: +44-1524-847099.E-mail addresses: [email protected]

(S.J. Watkins), [email protected] (B.A. Maher).

EPSL 6791 5-9-03

Earth and Planetary Science Letters 214 (2003) 379^394

R

Available online at www.sciencedirect.com

www.elsevier.com/locate/epsl

mailto:[email protected]

mailto:[email protected]

http://www.elsevier.com/locate/epsl

understand the processes occurring within thepresent-day North Atlantic.

Despite the large number of sediment cores tak-en in the North Atlantic during the last 40 years,ocean-wide studies of present-day sediment distri-bution and sources are sparse (e.g. [1,2]). Mostprevious studies have concentrated on selected At-lantic sub-regions, for example, the NE Atlantic[3], the Nordic Seas [4] and the Kara and LaptevSeas [5]. Kissel et al. [6] have examined sedimentproperties along the path of NADW. Bond et al.[7] mapped ice-rafted debris (IRD) in the northernNorth Atlantic, based on petrological analysis ofs 120 core tops; they identi¢ed three distinct IRDsources and sinks based on the percentages ofhaematite-stained grains, volcanic glass and detri-tal carbonate.

It is also notable that whilst palaeo-oceano-graphic conditions have been examined in detail(especially, for example, during Heinrich events),a detailed picture of present-day conditions hasyet to be de¢ned. For example, IRD is clearlyan important sediment source in the polar andsub-polar North Atlantic but the modern-dayIRD distribution has yet to be reconstructedover the whole ocean basin [8,9].

This study uses magnetic measurements tocharacterise the surface sediments of the NorthAtlantic and identify the major climatic andoceanographic controls on their magnetic signa-tures. It also aims to provide geological data forground-truthing of iceberg trajectory models [10].It directly complements both the study of Schmidtet al. [11], who used the magnetic characteristicsof surface sediments to identify major sedimentsources and pathways in the South Atlantic, anda similar magnetic data set for a smaller numberof North Atlantic surface samples (Robinson, per-sonal communication). Compared with other an-alytical methods, magnetic measurements havethe advantage of being non-destructive, rapidand can be carried out on small (V1 g) samples,enabling conservation of often scarce surface ma-terials.

The magnetic mineralogy of sediments can re-£ect distinctive detrital inputs (aeolian, ice-rafted,bottom water-transported), authigenesis via insitu formation of magnetite by magnetotactic bac-

teria and/or the e¡ects of post-depositional dia-genesis. Diagenetic e¡ects can mostly be excludedfor these surface sediments, speci¢cally obtainedfrom box cores and trigger weight cores; they areoxic in nature and show no evidence of the pres-ence of iron sulphides. Thus, investigation ofthese sediments using a number of di¡erent mag-netic properties may allow identi¢cation of thesources and transport pathways of detrital mag-netic grains or, in the absence of signi¢cant detri-tal sources, magnetically distinctive, bacteriallyproduced magnetosomes.

To represent the ocean-wide sample area, 321surface samples have been obtained, through theco-operation of international core archives (seeAcknowledgements). In order to identify potentialsource areas (PSAs), a number of rock, sedimentand soil samples were also obtained from the cir-cum-Atlantic region and their magnetic signaturescompared directly with the sediments.

Individual magnetic parameters such as mag-netic susceptibility can be a¡ected both by varia-tions in the concentrations of magnetic mineralsand/or by the e¡ects of biogenic dilution (e.g. bydiamagnetic carbonate). Measurement of mag-netic susceptibility (at more than one frequency)and an additional range of magnetic remanenceparameters enable calculation of a number of in-terparametric ratios, in which such concentrationand/or dilution e¡ects are removed. Given theremay be possible overlap between sample magneticproperties (perhaps re£ecting mixing of di¡erentmagnetic sources), multivariate methods of anal-ysis are appropriate. Fuzzy c-means cluster anal-ysis is used here, as applied recently to a numberof di¡erent environmental data sets, including:magnetic mapping of pollution in soils [12], anal-ysis of the climatic/magnetic connections for sedi-ments from the Azores region [13], and the pre-viously cited magnetic study of surface sedimentsin the South Atlantic [11].

2. Samples and methods

Core locations for the 321 North Atlantic sam-ples are shown in Fig. 2 (and following). Surfacesamples were mostly obtained from box and trig-

EPSL 6791 5-9-03

S.J. Watkins, B.A. Maher / Earth and Planetary Science Letters 214 (2003) 379^394380

ger weight cores (piston core samples were onlyused when comparison of the piston and triggerweight core indicated surface sediment waspresent in both). Samples from PSAs were alsocollected, including: soils from the North Africanregion; Icelandic basalts and volcanic ash; Devo-nian and Triassic redbeds from Spitsbergen andeast Greenland, respectively; Caribbean carbon-ates containing bacterial magnetite; granitesfrom north Bylot Island; and a range of otherlithologies from the circum-Atlantic area (Fig.1). Wet samples were dried overnight at 40‡Cand sample weights measured to allow for correc-tion of measurements to a dry mass-speci¢c basis.Samples were ¢rmly packed, to avoid internalmovement of the sample within the holder, into10-cc plastic pots. A suite of susceptibility andremanence measurements was applied to all sam-ples with the aim of characterising their magneticmineralogy, concentration and magnetic grain size(domain state). All ¢elds were applied and rema-nences measured along the same axis. The Appen-dix details the instrumentation used. In order toremove variations caused by changes in magneticconcentration (due to possible biogenic dilutionand/or variations in sedimentation rates), normal-ised, interparametric ratios were calculated fromthe susceptibility values (measured at low andhigh frequencies) and the anhysteretic and rema-nent magnetisations. Carbonate values were ob-

tained, either as published values or interpolatedfrom the data of Balsam and McCoy [1], for 60%of the sample set and spanning the whole mea-sured susceptibility range. Geographic plots of themagnetic data were generated using ArcView,each sample point being represented by a shadedarea of 300 km radius.

Given the reasonably large, multi-parametermagnetic data set produced from the sedimentand source samples, univariate or bivariate anal-ysis of individual magnetic parameters may beine¡ective in discriminating between possible sedi-ment sources and any mixing between sources. Toprovide such information, multivariate methodsare required. Two multivariate methods are ap-plied here: fuzzy c-means clustering and non-lin-ear mapping, using the program of Vriend et al.[14] in which the clustering algorithm was adaptedfrom Bezdek et al. [15] and the non-linear map-ping algorithm based on Sammon [16]. In tra-ditional clustering methods (hierarchical andk-means), a sample is forced to belong to an in-dividual cluster. This may be a disadvantage forthis magnetic data set, where a number of pro-cesses may be responsible for the magnetic signaland source mixing is a possibility. Fuzzy cluster-ing does not force a sample to belong to a speci¢ccluster; instead, it calculates a membership value,ranging from 0 (no similarity) to 1 (identical), foreach sample to each cluster. Following Hanesch etal. [12], samples are classi¢ed here as ‘belonging’to a cluster, if the ratio of the highest membershipto the second highest membership is greater than0.75; if this condition is not satis¢ed, the sampleis unclassi¢ed. This method does not require thenumber of clusters to be known before clusteringand the program iteratively performs the cluster-ing for two to nine clusters. The ‘best’ solution iscalculated by minimising the distance between asample and its cluster centre and maximising thedistance between the cluster centres [13]. Thesedistances are represented by two statistics pro-duced by the program: the partition coe⁄cientF and the classi¢cation entropy H. The ‘best’number of clusters is given by the highest F andlowest H value [12]. Non-linear mapping, a multi-dimensional scaling method [12], was also used toevaluate the results of the fuzzy c-means cluster-

Fig. 1. Bar chart showing susceptibility data (Mlf , U1036 m3

kg31) for measured PSA samples.

EPSL 6791 5-9-03

S.J. Watkins, B.A. Maher / Earth and Planetary Science Letters 214 (2003) 379^394 381

ing [17]. It provides a means of plotting the multi-dimensional relationship between samples andclusters in two dimensions.

Fuzzy clustering was run using the followingmagnetic parameters : (a) the high-¢eld remanence(the HIRM, as a percentage of the saturationremanence, SIRM), to identify the presence ofthe high-coercivity minerals, haematite/goethite;(b) frequency-dependent magnetic susceptibility(Mfd), to identify ultra¢ne-grained (6V20 nm),superparamagnetic (SP) ferrimagnets, such asmagnetite and maghaemite; (c) the ratio of theanhysteretic remanence normalised to magneticsusceptibility (MARM/Mlf ), to identify ¢ne-grained(V30^50 nm), single-domain (SD) ferrimagnets ;and (d) the ‘soft’ remanence fraction, i.e. that ac-quired at the relatively low magnetic ¢eld of20 mT (the IRM20 mT/IRM100 mT), to identifylow-coercivity, multidomain (MD)-like ferrimag-nets. Prior to cluster analysis, the magnetic pa-rameters were investigated using the non-paramet-ric Spearman test to ensure that they were notautocorrelated.

All parameters except HIRM% were found tohave a log-normal distribution and were thus log-transformed. Nine outliers were identi¢ed (valuesmore than three times the standard deviationfrom the mean value, normal or logarithmic de-

pending on the parameter distribution) and re-moved from the data set to preclude undue in£u-ence on the clustering, leading to unrealisticgroupings [12]. Finally, values were standardisedso that parameters with large values and/or vari-ability again did not predominate.

3. Results

First, we use ArcView to display the interpo-lated geographic distributions of the measuredmagnetic parameters and ratios (Figs. 2, 4, 6and 8) and then show the results of multivariatestatistical analysis of the non-concentration-de-pendent, sediment magnetic properties.

3.1. Magnetic susceptibility

Magnetic susceptibility provides an indicationof how magnetic a sample is:

Susceptibility ðMlfÞ ¼ Mf þ Mp þ Ma þ Md

where Mf is the strong, positive susceptibility con-tributed by ferrimagnets (like magnetite), Mp andMa the weak, positive susceptibility carried by para-magnets (like clay minerals) and high-coercivityminerals (like haematite and goethite), respec-tively, and Md the weak, negative susceptibilityof diamagnets (like calcium carbonate). Suscepti-bility thus often re£ects the concentration of fer-rimagnets within a sample. Magnetic susceptibili-ty values for our sampled range of potentialsources mostly vary from 0.01 to 3.3U1036 m3

kg31, with one extreme value of V23U1036 m3

kg31, measured on a granite from Devon Island(Fig. 1). Highest values are associated with theigneous rock samples (basalts and granites) andlow values with the African soils, Triassic andDevonian sandstones and carbonate-rich sedi-ments dominated by bacterial magnetite. In com-parison, magnetic susceptibility values for theNorth Atlantic surface sediments were found torange from 0.01 to 8U1036 m3 kg31 (Fig. 2). Thesusceptibility data identify a major contrast inpresent-day North Atlantic sediments and sour-ces. The low-latitude zone, extending in a trans-Atlantic belt west and northwest from Africa

Fig. 2. Geographic distribution of susceptibility (Mlf , U1036

m3 kg31) in present-day North Atlantic deep-sea sediments(sample values interpolated to 300 km radius of sample site,using ArcView); the locations of the PSA samples are alsoshown.

EPSL 6791 5-9-03


(from 0 to V30‡N and to V60‡W), displays theuniformly lowest susceptibility values(6 0.5U1036 m3 kg31). This pattern persistseven when adjustment of the susceptibility datais made for carbonate content. In contrast, themiddle to higher latitudes exhibit much higherand more variable susceptibility values (up toV8U1036 m3 kg31). High values (1^6U1036

m3 kg31) are found all around Iceland, possiblyassociated with: (a) the underlying, magnetite-richbasalt of the mid-ocean ridge and the Iceland^Faeroe ridge (Fig. 1) ; (b) movement of sedimentin the South Iceland Basin by Iceland^ScotlandOver£ow Water (ISOW), a component ofNADW; and (c) possible deposition of sub-aer-ially erupted ash (Fig. 1). Additional high-suscep-tibility locations include sites o¡ the east Green-land coast, coinciding with the location of majoriceberg-calving glaciers. Intermediate values occurin the Labrador Sea, Ba⁄n Bay and to the southof Greenland and are also probably associatedwith iceberg rafting. Maximum rates of icebergmelting o¡ eastern Canada occur presently inthe Labrador Current at 54^50‡N [18]. It is alsopossible that there has been some subsequent re-distribution of IRD by turbidite activity or deep-ocean currents (see below). Within the mid- tohigh-latitude belt of high susceptibility values,there are localised regions of slightly lower values,including: o¡ Newfoundland; south of Spitsber-

gen; the eastern Nordic Seas; and southwest ofIreland. In contrast with the South Atlantic [12],no relationship appears to exist between waterdepth and magnetic susceptibility; shallow pelagicsites show no obvious link with lower susceptibil-ity values. The North Atlantic seems dominatedmagnetically by distinctive terrigenous inputsrather than by productivity variations.

These terrigenous inputs become more clearlyde¢ned upon inspection of the frequency-depen-dent susceptibility, which displays an almost in-verse relationship to susceptibility. Values of fre-

Fig. 5. Bar chart showing HIRM (%) for PSA samples.

Fig. 4. Geographic distribution of frequency-dependent sus-ceptibility (Mfd, %) in present-day North Atlantic deep-seasediments (sample values interpolated to 300 km radius ofsample site, using ArcView).

Fig. 3. Bar chart showing frequency-dependent susceptibility(Mfd, %) for PSA samples.

EPSL 6791 5-9-03


quency-dependent susceptibility for the PSAsrange from 0.1 to 10% (Fig. 3), with maximumvalues associated with the Niger and Moroccansoil samples and minimum values with acid igne-ous rocks from Ba⁄n, Devon and north BylotIslands and the unweathered Illinoian (USA)loess. Values for the North Atlantic sedimentsrange from 0 to V14% (Fig. 4). Again, the low-latitude belt is distinctive and conspicuous, withtwo ‘plumes’ of sediment with the lowest suscep-tibility and the highest Mfd values ^ to the westand northwest of the African continent, extendingto 50‡W and 55‡N. Values of Mfd as high asV14% indicate the dominant presence of ultra-¢ne, SP ferrimagnetic grains, i.e. 6V30 nm inmagnetite [19,20]. These two swathes of present-day sediment are thus characterised by low con-centrations of dominantly SP ferrimagnets. Incontrast, the regions of high and intermediate sus-ceptibility around Iceland, Ba⁄n Bay, the Labra-dor Sea and along the coast of Greenland areassociated with low Mfd values (6 4%), indicatingthe dominance of coarser SD to MD ferrimag-nets. The low-susceptibility region in the easternNordic Seas is associated with moderate fre-quency dependence of V4^8%.

3.2. Magnetic remanences

Remanence ratios can be used to characterisethe magnetic stability of samples (e.g. [21]). Therelative abundance of the high-coercivity, mag-netically ‘hard’ minerals, haematite and goethite,is indicated by the HIRM (the remanence ac-quired in ¢elds between 0.3 and 1 T, as a percent-age of the total remanence). HIRM values for thePSAs range from 0 to 76%. The highest HIRMsare exhibited by the red sandstones, the lowest bythe granitic and basaltic samples (Fig. 5). Valuesfor the North Atlantic surface sediments rangefrom 0.3 to 27% (Fig. 6). The region of highestHIRM% values coincides with the low-suscepti-bility/high-frequency-dependence belt between 0and 30‡N to the west of Africa (Fig. 6). Well-drained, oxic soils in humid to sub-humid tropicalregions form haematite and goethite [22]. Haema-tite/goethite input to lower-latitude North Atlan-tic sediments has thus been associated with aeo-

lian transport of such soils from the sub-Saharan/Sahel region of Africa (e.g. [2,23,24]). In additionto this dust-dominated belt, high HIRM values(V10%) are also found close to the eastern coastof North America, between 40 and 50‡N. Whilstthese values might re£ect some distal dust input, itseems more likely that they result from (glacial)erosion of the redbeds in the St. Lawrence region(e.g. [2,25]) transported southwards by the DeepWestern Boundary Current [18]. Smaller, localisedregions of high HIRM values also occur o¡ thecoast of eastern Greenland (at V70‡N) and to thenorth and northeast of the UK. The high easternGreenland values may result from the transportof IRD from Triassic redbeds in the FlemingFjord region (Fig. 5; [26]). Redbeds of Devonianage are found in the central peninsula of northernSpitsbergen (Fig. 5) ; the intermediate HIRM val-ues (6^8%) found in the eastern Nordic Seas maysimilarly be the result of transport of haematite-rich IRD. Other regions, including Ba⁄n Bay, theLabrador Sea and the coast of Greenland southof V65‡N, are characterised by lower HIRM val-ues (6 6%), indicating domination by ferrimag-nets.

Another remanence ratio, identifying the pro-portion of remanence acquired at low applied¢elds (here, IRM20 mT, normalised to the rema-nence acquired at 100 mT), can act as an indica-tor of ferrimagnetic grain size (domain state).

Fig. 6. Geographic distribution of HIRM (%) in present-dayNorth Atlantic deep-sea sediments (sample values interpo-lated to 300 km radius of sample site, using ArcView).

EPSL 6791 5-9-03


High values of this parameter indicate higheramounts of ferrimagnets that are easy to magne-tise, i.e. either coarse MD or viscous grains on theSD/SP border. Values of this low-¢eld remanenceratio for the PSA samples (Fig. 7) range frommaxima of V0.5^0.7 for the Niger and Moroccansoils and the basalt samples, to a minimum of 0.1for the US loess samples and the purest (carbon-ate+) bacterial magnetite sample. For the NorthAtlantic sediments (Fig. 8), IRM20 mT/IRM100 mT

values range from 0.01 to 0.56. Highest values are

found between the equator and 40‡N, extendingto V50‡W. As this corresponds to the ‘dust belt’region of low susceptibility, high frequency depen-dence, high HIRM and intermediate to highMARM/Mlf , it is likely that the high IRM20 mT/IRM100 mT ratio indicates here the presence of vis-cous SD/SP grains. The region surrounding Ice-land is characterised by intermediate values, mostlikely associated with the basaltic ridge material(Fig. 7). Lowest values are found in Ba⁄n Bayand in parts of the Labrador Sea and close to thesouthern coast of Greenland. Given the high sus-

Fig. 10. Geographic distribution of MARM/Mlf ratios inpresent-day North Atlantic deep-sea sediments (sample valuesinterpolated to 300 km radius of sample site, using Arc-View).

Fig. 9. Bar chart showing MARM/Mlf ratios for PSA samples.

Fig. 8. Geographic distribution of low ¢eld remanence(IRM20 mT/IRM100 mT) in present-day North Atlantic deep-seasediments (sample values interpolated to 300 km radius ofsample site, using ArcView).

Fig. 7. Bar chart showing the proportion of remanence ac-quired at low applied ¢elds (IRM20 mT/IRM100 mT) for PSAsamples.

EPSL 6791 5-9-03


ceptibility, low frequency dependence and lowHIRM of these samples, these low-IRM20 mT/IRM100 mT areas probably re£ect the presence oflithogenic SD/PSD ferrimagnets. There is also aregion of low IRM20 mT/IRM100 mT values o¡ theeastern North American coast. These low valuescoincide with low Mlf , Mfd, and HIRM values andhigh MARM/Mlf values, suggesting the presence ofSD ferrimagnets of bacterial magnetite origin (seebelow).

Finally, the susceptibility of ARM (MARM) ishighest for strongly interacting SD ferrimagnets(such as the ferrite chains produced intracellularly

by magnetotactic bacteria) and decreases rapidlywith increasing grain size (e.g. [19,27]). Normal-ising MARM with susceptibility (Mlf ), to removemagnetic concentration e¡ects, values for thePSAs range from 0.2 to 204 (Fig. 9), with thecarbonate sample containing intact bacterial mag-netite displaying the highest value and the redsandstones and igneous rocks the lowest. TheMARM/Mlf values for the North Atlantic surfacesediments are very similar in range to the mea-sured PSAs, from 0.5 to 254 (Fig. 10). MaximumMARM/Mlf values are found in restricted areas closeto the southeastern coast of North America

Fig. 11. Summary of the non-linear mapping for the nine-cluster solution.

EPSL 6791 5-9-03


(V30‡N), south of Newfoundland, the Africancoast and west of the UK. The highest values(s 50) occur in association with low susceptibili-ty, HIRM and IRM20 mT/IRM100 mT values, indi-cating little detrital input of source materials suchas either dust or IRD. We infer the dominantpresence of intact, SD, bacterial magnetite chains(e.g. [28^30]) formed in situ in these areas. Lessextreme values (20^50) o¡ the southeastern coastof North America may be linked to the presenceof broken bacterial magnetite chains. Values inthe range 20^50 are also found for sediments tothe west of Africa, together with high Mfd andIRM20 mT/IRM100 mT ratios, indicating the pres-ence of SD, SD/SP and SP ferrimagnets. Suchgrain size assemblages are characteristic of soil-formed ferrimagnets [27,31]. Minimum MARM/Mlfvalues are observed along the eastern coast ofGreenland, around Iceland, Ba⁄n Bay and theLabrador Sea, indicating for each of these regionsthe contribution of coarser, MD-like ferrimag-nets. Intermediate values of MARM/Mlf (10^20) areassociated with the ‘dust belt’ regions of low sus-ceptibility and high Mfd and HIRM to the westand northwest of the African continent. A regionof intermediate values also surrounds Spitsbergen.

4. Discussion

Geographic presentation of the individual mag-netic properties of the North Atlantic surfacesediments reveals evidence of areal di¡erentiationand distinctive terrigenous (dominantly) and au-thigenic magnetic sources. To integrate the mag-netic data and make objective identi¢cation ofsediment groupings and transport pathways, mul-tivariate analysis is required. Only parameters in-dependent of magnetic mineral concentrationwere used in the fuzzy cluster analysis appliedhere. Compared with the sediments, many of thepotential source samples display magnetic param-eter values which fall as outliers, and so cannot beincluded as end-members. Mixing of sources isthus indicated; mathematical unmixing of sedi-ments in terms of sources is the subject of furtheranalysis and will be reported in a later paper.

The statistical indicators of cluster perfor-

mance, highest F and lowest H, suggest that the‘best’ clustering solution for the magnetic data setfor the North Atlantic present-day sediments isachieved with nine clusters. The parameter meansfor each of these nine clusters are shown in Ta-ble 1. Fig. 11 shows the results of the non-linearmapping (NLM) of the data. Making no pre-sumptions regarding the presence or number ofclusters, the NLM displays a two-dimensionalprojection of the multidimensional data cloud,with minimal distortion of the interdata distances[14]. Six of the nine clusters are well separated (i.e.display large interdata distance): clusters 1, 2, 4,5, 8 and 9. The three remaining clusters, 3, 6 and7, are well-de¢ned, but separated by smaller dis-tances, and fall in the central part of the NLM.Samples dominated magnetically by ¢ne-grainedferrimagnets (cluster 2) plot in the upper left sec-tion of the NLM. Further to the right, samplesbecome increasingly dominated by high-coercivitybehaviour (cluster 5 to cluster 9). In contrast,sediments dominated by coarse ferrimagnets (clus-ters 1, 4, 7, 8) plot in the lower section of theNLM. These are further split into clusters by var-iations in their concentration of high-coercivityminerals. Samples plotting towards the middle

Fig. 12. Geographic distribution of the magnetic nine-clustersolution, based on 276 sediment samples (i.e. excluding thosewith missing magnetic values). Twenty-two samples were un-classi¢ed, i.e. 8% of the sample set. The symbol size repre-sents the degree of a⁄nity with each cluster. The solid linerepresents the approximate path of the NADW.

EPSL 6791 5-9-03


of the NLM (clusters 3, 6, 7) interface betweenthe coarse- and ¢ne-grained clusters, most likelyindicating a mixture of sources. Fig. 12 shows thespatial distribution of all the cluster results. Noinformation on the spatial distribution of samplesis included in the fuzzy c-means cluster analysis.Hence, the distinctive spatial groupings that resultare notable.

4.1. Cluster 2

Samples with strong a⁄nities to cluster 2 arefound close to the coast in two main regions:(a) o¡ the southeast coast of the southern USA(V30‡N) and (b) in a narrow transect extendingsoutheast from the Gulf of St. Lawrence (Fig. 13).Additional, single samples belonging to this clus-ter are distributed throughout the North Atlantic(e.g. o¡ the African coast at V16‡N; south ofIceland; southeast of Greenland; and east of

Spitsbergen). The cluster means indicate the dom-inance of SD ferrimagnets, with the highest MARM/Mlf values (39), low HIRM (3.4%) and low Mfd

(3%). Given that susceptibility values are alsolow in these regions, little detrital input is indi-cated but rather the predominance of bacterialmagnetite, formed in situ. Intact magnetosomechains are characterised by high MARM and lowMfd values (e.g. [29,32]). The slightly higherHIRM of this cluster in comparison to our mea-sured bacterial magnetite representatives may re-sult from a contribution of either windblown dust[2] or glacially derived haematite, transportedfrom the St. Lawrence redbeds region by theDeep Western Boundary Current. It is likelythat bacterial magnetite formation occurs in sur-face sediments right across the North Atlantic atthe present day but is only identi¢ed as the majormagnetic component when detrital inputs areminimal.

Table 1

Cluster1

Cluster2

Cluster3

Cluster4

Cluster5

Cluster6

Cluster7

Cluster8

Cluster9

HIRM (%) 2.97 3.45 4.64 4.32 5.58 5.68 6.23 8.26 10.61MARM/Mlf 6.73 39.04 8.58 2.78 25.43 16.69 9.26 5.24 15.87IRM20 mT/IRM100 mT 0.11 0.13 0.22 0.16 0.26 0.18 0.11 0.10 0.35Mfd (%) 2.10 2.97 2.48 0.78 7.64 4.13 3.59 1.53 7.63

Fig. 13. Location of samples belonging to cluster 2 (dominat-ed by bacterial magnetite). The symbol size represents the de-gree of a⁄nity with the cluster.

Fig. 14. Location of samples belonging to clusters 5 and 9(dominated by wind-blown dust). The symbol size representsthe degree of a⁄nity with the cluster.

EPSL 6791 5-9-03


4.2. Clusters 5 and 9

Samples belonging to cluster 5 and cluster 9 arefound south of 60‡N in two bands: the ¢rst ex-tending northwest and the second west from Afri-ca. Both clusters have distinctive and similar mag-netic properties (Fig. 14). They have the highestMfd means (V8%). Cluster 9 also has the highestHIRM mean (11%). The potential source materi-als measured here, i.e. topsoils from Morocco,Tunisia and Niger, match strongly with these sedi-ment magnetic properties and have strong statis-tical a⁄nity (s 0.8) with these clusters. The onlyexception is that the sediments have slightly high-er MARM/Mlf values; a minor magnetic contributionof V2^10% intact bacterial magnetosomes couldaccount for this di¡erence.

4.3. Clusters 1, 4, 7 and 8

Samples belonging to cluster 1 are mostly re-stricted to the northwestern North Atlantic (Fig.15), being located in Ba⁄n Bay and the north-western Labrador Sea. Additional cluster 1 sam-ples occur o¡ the coast of east Greenland atV70‡N, southwest of Iceland and one samplenorth of Spitsbergen. Mfd and MARM/Mlf are bothlow (Table 1) and HIRM% is the lowest for anyof the clusters. These data indicate the dominance

of MD ferrimagnets. Large numbers of cluster 1samples are located close to present-day calvingglaciers. The geology surrounding these sites isdominated by (titano)magnetite-rich, Precambrianacid igneous rocks and Tertiary basalts. Ice-raft-ing is thus the most likely sediment transportmechanism for these sites. For more proximalsites, debris £ows and turbidites may also be sig-ni¢cant sediment sources (e.g. the nearshorenorthwestern Labrador Sea [33]). IRD in northLabrador Sea samples most likely derives fromicebergs released around Ba⁄n Bay, where thedominant geology is Precambrian granite andgneiss. The geographic distribution of these clus-ter 1 samples closely matches the IRD source(Ba⁄n Bay) and deposition area (Labrador Sea)mapped lithologically by Bond et al. [7] on thebasis of the presence of detrital carbonate(s 10%). However, cluster 1 samples extend fur-ther eastwards than inferred by these authors.

IRD sources for the cluster 1 samples o¡ eastGreenland mostly comprise granodioritic gneissesand granite, with Triassic redbeds further north.Our potential source rock sample (granitic gneissfrom north Bylot Island) provides an obvioussource of coarse, MD ferrimagnets, being charac-terised by high susceptibility (3U1036 m3 kg31),minimal Mfd (6 1%), zero HIRM% (saturated by300 mT), and a high IRM20 mT/IRM100 mT value of0.36. As these magnetic values are more extremethan the cluster 1 means, some subsidiary admix-ture of other sources is likely (e.g. bacterial mag-netite (V3%) to raise the MARM/Mlf and red sand-stone (V4%) to raise the HIRM). A mix of 40%granitic gneiss and 60% Icelandic ash also produ-ces values comparable to, but slightly higher than,the cluster 1 means.

The cluster 1 samples south of Iceland lie alongthe eastern £ank of the mid-Atlantic ridge andprobably result from bottom water transport byISOW. Whilst they have highest a⁄nity with clus-ter 1, they also display some a⁄nity with cluster7, which appears to have a signi¢cant contribu-tion from Icelandic ash as a source material.

Cluster 4 samples are mainly restricted to thenorthwestern North Atlantic, o¡ the southernGreenland coast and in the Labrador Sea (Fig.15). This cluster has the lowest Mfd (0.8%) and

Fig. 15. Location of samples belonging to clusters 1, 4, 7and 8 (dominated by IRD). The symbol size represents thedegree of a⁄nity with the cluster.

EPSL 6791 5-9-03


MARM/Mlf (2.8) values, indicating the dominance ofcoarse MD ferrimagnets. Both the geographic ex-tent and cluster means of cluster 4 are similar tocluster 1; a similar, IRD contribution is indicatedfor most of these samples. Samples at 40‡N, alongthe US coast, may appear too proximal to resultfrom IRD, but the International Ice Patrol re-ports modern iceberg sightings as far south as38‡N [34].

Cluster 7 samples are found in similar regionsto clusters 1 and 4, again suggesting an IRDsource. However, an additional cluster 7 grouplies south of Iceland, along the mid-Atlantic ridge.Cluster 7 has a higher contribution from high-co-ercivity minerals (HIRM=6.2%) and SP ferri-magnets (Mfd = 3.6%). Low values of MARM/Mlf (9)and IRM20 mT/IRM100 mT (0.1) suggest the pres-ence of PSD-type grains. The cluster means cor-respond well with our Icelandic ash samples(although these have slightly higher Mfd values).It is possible, therefore, that this group re£ectstransport and deposition by the bottom watercurrents of the ISOW [6].

Cluster 8 is restricted to two small regions, o¡Newfoundland/Nova Scotia, and south of Spits-bergen. The HIRM values (8%) indicate signi¢-cant contribution by high-coercivity minerals. Ad-ditionally, however, Mfd, MARM/Mlf and IRM20 mT/IRM100 mT values are all low (1.5%, 5 and 0.10,respectively) indicating coarse ferrimagnets, oflikely IRD origin. The Newfoundland/Nova Sco-tia region of high HIRM values may be associ-ated with the glacially derived redbed sediments,redeposited by turbidity or bottom currents, o¡-shore from the St. Lawrence region [2,18]. Thehigh HIRMs north of Norway may be relatedto ice-rafting from Devonian redbeds on Spitsber-gen.

In summary, most of the samples belonging tothe IRD clusters substantiate Bond et al.’s [7]conclusion that most (tracer-bearing) ice circu-lates in the cooler waters north and west of thesubpolar front. Whereas Bond et al. [7] inferredthe presence of IRD from south and west Green-land on the basis of low tracer percentages, thecluster data presented here identify IRD fromthese source regions both to the south of Green-land and in the Labrador Sea.

4.4. Cluster 3

Parameter means for cluster 3 indicate MD fer-rimagnets (low MARM/Mlf and Mfd and slightly high-er IRM20 mT/IRM100 mT and HIRM values thanthe cluster 1 samples). Notably, most samplesare located around Iceland (Fig. 16), coincidingwith Bond et al.’s [7] Icelandic glass lithologicalzone. There are no iceberg releases from Icelandat the present day; Bond et al. [7] suggest that thisregion results from volcanic eruptions onto drift-ing ice. A discrete cluster 3 sample group lies o¡-shore from the east Greenland Scoresby Sund re-gion (Precambrian granites and Tertiary basalts)and may represent IRD.

Further cluster 3 samples occur around theAzores, Canaries/Madeira area, re£ecting erosionof (titano)magnetite-rich, mid-ocean ridge andvolcanics [35].

4.5. Cluster 6

The geographic distribution of cluster 6 is wide-spread (Fig. 17). Notably, however, most samplesextending from the eastern Nordic Seas havestrong memberships to this cluster. These samplescoincide with the low-susceptibility/low-IRD re-gion of the Greenland Sea previously noted byPirrung et al. [4] and also resemble the IRD

Fig. 16. Location of samples belonging to cluster 3 (mid-ocean ridge and volcanic material). The symbol size repre-sents the degree of a⁄nity with the cluster.

EPSL 6791 5-9-03


area mapped (on the basis of s 10% haematite-stained grains) by Bond et al. [7]. Smaller groupsof samples occur o¡ the Floridan coast, northwestof Africa, south of Iceland, north of Spitsbergenand along the eastern coast of Greenland. Thiscluster has a moderate MARM/Mlf mean value(17), Mfd of 4%, HIRM of 5.7% and IRM20 mT/IRM100 mT of 0.18. It appears to relate to areaswhere detrital inputs (from haematite-bearingIRD or dust) are relatively low at present. Somecontribution from bacterial magnetite is alsolikely, given the low detrital supply.

4.6. Summary

Taking into account the cluster means, and theavailable source magnetic data, cluster 2 appearsto re£ect authigenic formation of bacterial mag-netite, with no other signi¢cant detrital input.Clusters 5 and 9 represent aeolian dust. Clusters1, 4, 7 and 8 are characterised by coarse-grainedferrimagnets. From comparison with present-dayobservations and iceberg model data [10,36], themajority of samples classi¢ed in these four clus-ters appear to contain IRD. Studies of glacialNorth Atlantic sediments have similarly shownthat much of the IRD is associated with highconcentrations of coarse-grained (MD and PSD)ferrimagnets with low coercivities and a ‘soft’IRM signal [23,37^39]. Cluster 3 samples are

also characterised by coarse ferrimagnets but theirspatial distribution suggests they are more likelyto be derived from mid-ocean ridge material orvolcanic ash. The remaining cluster, 6, is geo-graphically widespread but with a number of sam-ples con¢ned to the eastern half of the NordicSeas; this cluster probably receives low inputs ofhaematite-bearing dust and/or IRD, coupled withsome in situ formation of bacterial magnetite.

This interpretation of distinct sediment sourcesis supported by the NLM results (Fig. 11). Great-est interdata distance is apparent between theIRD-dominated sediments (clusters 1, 4, 7 and8) and the bacterial magnetite (cluster 2) anddust clusters (5 and 9). Within the IRD clusters,8 is separated from 1 and 4. Cluster 3 (mid-oceanridge/volcanic) is less well separated from 7 (IRDand mid-ocean ridge mixed). Cluster 6 interfacesbetween clusters 3 and 7 and the dust and bacte-rial magnetite groupings. Subsequent runs of thecluster analysis were made, with the bacterialmagnetite and the two dust sample groups re-moved, to assess if any further discrimination be-tween the remaining clusters is possible. The bestcluster solution in this case was six, with the spa-tial distribution and the magnetic parametermeans matching closely those of the original anal-ysis. This indicates both that no further discrim-ination is possible statistically and that the origi-nal nine-cluster solution is robust.

In summary, magnetic characterisation ofpresent-day samples from across the North Atlan-tic, together with similar data for a range of po-tential sediment source areas, identi¢es distinctand statistically robust spatial groupings of sedi-ment. These groupings in turn re£ect mostly de-trital inputs to the sediments, principally of IRDand windblown dust. These sediment transportpathways can be validated by making direct com-parison of the sediment magnetic properties withsource rocks and soils, and with iceberg trajectoryand observation data. The spatial distribution ofthe IRD-dominated groupings suggest that deepwater currents are less signi¢cant than previouslyproposed [6,40] in controlling present-day sedi-ment mineralogy and distribution. Kissel et al.[6] identi¢ed little magnetic variation in sedimentcores obtained along the trajectory of the

Fig. 17. Location of samples belonging to cluster 6. Thesymbol size represents the degree of a⁄nity with the cluster.

EPSL 6791 5-9-03


NADW, and on this basis proposed that suchconstancy could be explained by dominant bot-tom water transport of material from the Nordicbasaltic province. However, the larger number ofsediment samples analysed here presents a di¡er-ent picture, with several di¡erent statistical group-ings occurring along this pathway (Fig. 12). Thisapparent contradiction may be explainable in thatbottom water transport and deposition does occur^ but is overshadowed magnetically wherever oth-er detrital inputs (such as IRD) are supplied. Suchan interpretation is independently supported byPrins et al.’s recent [41], particle size-based studywhich also identi¢es signi¢cant IRD inputs toareas previously thought dominated by ISOW£ow.

In similar vein, in situ authigenesis of bacterialmagnetite probably occurs over much of theocean £oor but is also only identi¢ed where de-trital inputs are low.

5. Conclusions

Magnetic measurements of present-day, deep-sea sediments from across the North Atlanticidentify distinct spatial patterns of sediment sour-ces and transport pathways. Much of the sedi-mentary magnetic signal appears to be controlledby detrital inputs, especially of IRD and wind-blown dust. Sediments identi¢ed as having signi¢-cant IRD inputs, which presently occur as farsouth as 40‡N in the western North Atlantic,can be split statistically into four di¡erent clus-ters, re£ecting di¡erent sources and/or mixing ofsources. These IRD distributions can be com-pared explicitly with observational and icebergtrajectory data, and lithological tracer-based stud-ies. They provide a present-day basis for compar-ison with glacial-stage IRD patterns and modes ofocean circulation (Watkins et al., in preparation).Previous studies have concluded that the magneticsignal of North Atlantic sediments along theNADW trajectory shows little variation, andthat transport and deposition by bottom wateris the dominant process. However, the largerdata set shown here identi¢es signi¢cant inputby IRD for discrete areas along this pathway.

In terms of windblown dust, the magnetic datadelineate the source and spread of dust ; statisticalanalysis identi¢es two, distinct westerly- andnorthwesterly-orientated dust ‘plumes’. Again,these modern dust patterns can be used for com-parison with previous glacial-stage distributions.

Authigenesis of magnetic material is only evi-dent in those locations presently supplied withlittle detrital input. It is likely that bacterial mag-netite is being formed across much of the NorthAtlantic at present but is overwhelmed magneti-cally by other, allochthonous inputs.

Finally, whilst magnetic discrimination of sedi-ment sources and pathways seems e¡ective for themajority of present-day samples, one statisticalcluster (cluster 6) is notable by its rather widegeographic spread. It is likely that these sedimentsreceive relatively small amounts of a mix of po-tential source materials at the present day.

Acknowledgements

This work was carried out under a Natural En-vironment Research Council PhD studentship (04/99/ES/62), jointly supervised by Dr Bigg at theUniversity of East Anglia, and building on earlierdiscussions with Dr Robinson, Manchester Uni-versity. We thank the following for access todeep-sea sediment samples: Dr Hardy, CanadianGeological Survey; Prof. Sarnthein, University ofKiel; Dr Rothwell, University of Southampton;Dr Smith, Scripps Institute of Oceanography; DrBroda and Dr Roosen, Woods Hole Oceano-graphic Institute; Drs de Vernal and Bilodeau,University of Quebec at Montreal ; Dr Matthies-sen (Dr Pirrung, now at Jena), AWI-Bremer-haven; Prof. Andrews, INSTAAR, University ofColorado; and Dr Moros, University of Bergen.We thank the following for access to potentialsource samples: Drs Lavoie, Jackson and Frisch,Canadian Geological Survey; Dr Nakrem, Uni-versity of Oslo; Dr Mcgarvie, Open University;Prof. Kent, Lamont-Doherty Earth Observatory;Dr Grimley, Illinois State Geological Survey; andDr Hounslow, Lancaster University. Dr Dekkers(Utrecht) kindly made available the fuzzy clustersoftware.[VC]

EPSL 6791 5-9-03


Appendix. Magnetic instrumentation

Low-¢eld, room-temperature and frequency-de-pendent susceptibility were measured using a dualfrequency (0.46 kHz ^ Mlf and 4.6 kHz ^ Mhf )Bartington MS2B sensor, with frequency depen-dence calculated as:

Mfdð%Þ ¼Mlf3Mhf

MlfU100

To identify if the geographic distribution ofsusceptibility is in£uenced by carbonate content,a subset (60%) of susceptibility values were cor-rected for carbonate content, using the formula:MlfU(13CO3 content (wt fraction)). Anhystereticremanent magnetisation was imparted to samplesusing a Molspin AF demagnetiser, by placingthem in a slowly decreasing alternating ¢eldfrom a peak of 80 mT. A superimposed DC ¢eldof 0.08 mT was applied parallel to the AC ¢eldand the resulting ARM was measured using aMolspin magnetometer. The ARM is expressedas a mass-speci¢c susceptibility of ARM (Marm,1036 m3 kg31) by normalising with the DC ¢eld:

Marm ¼ ARMf

where ARM is the mass-speci¢c ARM (1034 Am2

kg31) and f is the steady biasing ¢eld applied (inthis case, 0.08 mTw63.66 Am31). ARMs werethen demagnetised, before isothermal remanentmagnetisations were measured (resulting from in-crementally applied ¢elds of 20, 100, 300 and 1000mT), using a Molspin magnetometer. Fields from20 to 300 mT were imparted using a pulse mag-netiser and the 1 T (‘saturation’) ¢eld was im-parted using a Highmoor DC electromagnet.The proportion of high ¢eld remanence acquiredbetween 300 mT and 1 T, the HIRM%, was cal-culated as:

%HIRM ¼ SIRM3IRM300 mT

SIRMU100

References

[1] W.L. Balsam, F.W. McCoy, Atlantic sediments: glacial/

interglacial comparisons, Paleoceanography 2 (1987) 531^542.

[2] W.L. Balsam, B.L. Otto-Bliesner, B.C. Deaton, Modernand last glacial maximum eolian sedimentation patterns inthe Atlantic Ocean interpreted from sediment iron oxidecontent, Paleoceanography 10 (1995) 493^507.

[3] M. Cremer, F. Grousset, J.C. Faugeres, J. Duprat, E.Gonthier, Sediment £ux patterns in the northeastern At-lantic ^ variability since the last interglacial, Mar. Geol.104 (1992) 31^53.

[4] M. Pirrung, D. Futterer, H. Grobe, J. Matthiessen, F.Niessen, Magnetic susceptibility and ice-rafted debris insurface sediments of the Nordic Seas: implications forIsotope Stage 3 oscillations, Geo-Mar. Lett. 22 (2002)1^11.

[5] F. Niessen, D. Weiel, Distribution of magnetic suscepti-bility on the Eurasian shelf and continental slope ^ impli-cations for source areas of magnetic minerals, Rep. PolarRes. 212 (1996) 81^88.

[6] C. Kissel, C. Laj, L. Labeyrie, T. Dokken, A. Voelker, D.Blamart, Rapid climatic variations during marine isotopicstage 3: magnetic analysis of sediments from Nordic Seasand North Atlantic, Earth Planet. Sci. Lett. 171 (1999)489^502.

[7] G. Bond, B. Kromer, J. Beer, R. Muscheler, M.N.. Evans,W. Showers, S. Ho¡mann, R. Lotti-Bond, I. Hajdas, G.Bonani, Persistent solar in£uence on North Atlantic cli-mate during the Holocene, Science 294 (2001) 2130^2136.

[8] K. Matsumoto, An iceberg drift and decay model to com-pute the ice-rafted debris and iceberg meltwater £ux: Ap-plication to the interglacial North Atlantic, Paleoceanog-raphy 11 (1996) 729^742.

[9] J.A. Andrews, Icebergs and iceberg rafted detritus (IRD)in the North Atlantic: Facts and assumptions, Oceanog-raphy 13 (2000) 100^108.

[10] G.R. Bigg, M.R. Wadley, D.P. Stevens, J.A. Johnson,Prediction of iceberg trajectories for the North Atlanticand Arctic Oceans, Geophys. Res. Lett. 23 (1996) 587^3590.

[11] A.M. Schmidt, T. vonDobeneck, U. Bleil, Magnetic char-acterisation of Holocene sedimentation in the South At-lantic, Paleoceanography 14 (1999) 465^481.

[12] M. Hanesch, R. Scholger, M.J. Dekkers, The applicationsof fuzzy c-means cluster analysis and non-linear mappingto a soil data set for the detection of polluted sites, Phys.Chem. Earth 26 (2001) 885^891.

[13] P.P. Kruiver, Y.S. Kok, M.J. Dekkers, C.G. Langereis, C.Laj, A pseudo-Thellier relative palaeointensity record, androck magnetic and geochemical parameters in relation toclimate during the last 276 kyr in the Azores region, Geo-phys. J. Int. 136 (1999) 757^770.

[14] S.P. Vriend, P.F.M. van Gaans, J. Middelburg, A. Nus,The application of fuzzy c-means cluster analysis andnon-linear mapping to geochemical datasets: examplesfrom Portugal, Appl. Geochem. 3 (1988) 213^224.

[15] C.J. Bezdek, R. Ehrlich, W. Full, FCM: the fuzzy

EPSL 6791 5-9-03


c-means clustering algorithm, Comput. Geosci. 10 (1984)191^203.

[16] J.W. Sammon, A nonlinear mapping data structure anal-ysis, IEEE Trans. Comput. C18 (1969) 401^409.

[17] M.J. Dekkers, C.G. Langereis, S.P. Vriend, P.J.M. vanSantvoort, G.J. de Lange, Fuzzy c-means cluster analysisof early diagenetic e¡ects on natural remanent magnet-isation acquisition in a 1.1 Myr piston core from theCentral Mediterranean, Phys. Earth Planet. Inter. 85(1994) 155^171.

[18] D.J.W. Piper, K.I. Skene, Latest Pleistocene ice-raftingevents on the Scotian Margin (eastern Canada) and theirrelationship to Heinrich events, Paleoceanography 13(1998) 205^214.

[19] B.A. Maher, Magnetic properties of some synthetic sub-micron magnetites, Geophys. J. 94 (1988) 83^96.

[20] J.A. Dearing, R.J.L. Dann, K. Hay, J.A. Lees, P.J. Love-land, B.A. Maher, K. O’Grady, Frequency-dependent sus-ceptibility measurements of environmental materials,Geophys. J. Int. 124 (1996) 228^240.

[21] B.A. Maher, R. Thompson, M.W. Hounslow, Introduc-tion, in: B.A. Maher, R. Thompson (Eds.), QuaternaryClimates, Environments and Magnetism, Cambridge Uni-versity Press, Cambridge, 1998, pp. 1^48.

[22] U. Schwertmann, R.M. Taylor, Iron oxides, in: J.B. Dix-on, S.B. Weed (Eds.), Minerals in Soil Environments, SoilScience Society of America, Madison, WI, 1987, pp.

[23] S.G. Robinson, The Late Pleistocene palaeoclimatic rec-ord of North Atlantic deep-sea sediments revealed bymineral magnetic measurements, Phys. Earth Planet. In-ter. 42 (1986) 22^47.

[24] B.A. Maher, P.F. Dennis, Evidence against dust-mediatedcontrol of glacial-interglacial changes in atmosphericCO2, Nature 411 (2001) 176^180.

[25] G.C. Bond, R. Lotti, Iceberg discharges into the NorthAtlantic on millennial time scales during the last glacia-tion, Science 267 (1995) 1005^1010.

[26] P.E. Olsen, D.V. Kent, Long-period Milankovitch cyclesfrom the Late Triassic and Early Jurassic of eastern NorthAmerica and their implications for the calibration of theEarly Mesozoic time-scale and the long-term behaviour ofthe planets, Phil. Trans. R. Soc. London A357 (1999)1761^1786.

[27] Oº . Oº zdemir, S.K. Banerjee, A preliminary magnetic studyof soil samples from west-central Minnesota, Earth Plan-et. Sci. Lett. 59 (1982) 393^403.

[28] N. Petersen, T. von Dobeneck, H. Vali, Fossil bacterialmagnetite in deep-sea sediments from the South AtlanticOcean, Nature 320 (1986) 611^615.

[29] P. Hesse, J.F. Stolz, Bacterial magnetite and the Quater-

nary record, in: B.A. Maher, R. Thompson (Eds.), Qua-ternary Climates, Environments and Magnetism, Cam-bridge University Press, Cambridge, 1998, pp. 163^199.

[30] D.F. McNeill, Biogenic magnetite from surface Holocenecarbonate sediments, Great Bahama Bank, J. Geophys.Res. 95 (1990) 4363^4371.

[31] B.A. Maher, R.M. Taylor, Formation of ultra¢ne-grainedmagnetite in soils, Nature 336 (1988) 368^370.

[32] B. Moskowitz, R.B. Frankel, D.A. Bazylinski, Rock mag-netic criteria for the detection of biogenic magnetite,Earth Planet. Sci. Lett. 120 (1993) 283^300.

[33] J.S. Stoner, J.T. Andrews, The North Atlantic as a Qua-ternary magnetic archive, in: B.A. Maher, R. Thompson(Eds.), Quaternary Climates, Environments and Magnet-ism, Cambridge University Press, Cambridge, 1999, pp.49^80.

[34] NSIDC, International Ice Patrol (IIP) Iceberg SightingsDatabase, National Snow and Ice Data Center/WorldData Center for Glaciology, Boulder, CO, 1995, digitalmedia.

[35] S.G. Robinson, J.T.S. Sahota, F. Old¢eld, Early diagen-esis in North Atlantic abyssal plain sediments character-ized by rock-magnetic and geochemical indices, Mar.Geol. 163 (2000) 77^107.

[36] G.R. Bigg, M.R. Wadley, D.P. Stevens, J.A. Johnson,Modelling the dynamics and thermodynamics of icebergs,Cold Reg. Sci. Technol. 26 (1997) 113^135.

[37] J. Bloemendal, J.W. King, F.R. Hall, S.-J. Doh, Rockmagnetism of Late Neogene and Pleistocene deep-seasediments: Relationship to sediment source, diageneticprocesses and sediment lithology, J. Geophys. Res. 97(1992) 4361^4375.

[38] F.E. Grousset, L. Labeyrie, J.A. Sinko, M. Cremer, G.Bond, J. Duprat, E. Cortijo, S. Huon, Patterns of ice-rafted detritus in the glacial North Atlantic (40^55N), Pa-leoceanography 8 (1993) 175^192.

[39] S.G. Robinson, M.A. Maslin, I.N. McCave, Magneticsusceptibility variations in Upper Pleistocene deep-seasediments of the NE Atlantic: implications for ice raftingand paleocirculation at the last glacial maximum, Pale-oceanography 10 (1995) 221^250.

[40] T.M. Dokken, E. Jansen, Rapid changes in the mecha-nism of ocean convection during the last glacial period,Nature 401 (1999) 458^461.

[41] M.A. Prins, L.M. Bouwer, C.J. Beets, S.R. Troelstra, G.J.Weltje, R.W. Kruk, A. Kuijpers, P.Z. Vroon, Ocean cir-culation and iceberg discharge in the glacial North Atlan-tic: Inferences from unmixing of sediment size distribu-tions, Geology 30 (2002) 555^558.

EPSL 6791 5-9-03


Documents

Magnetic characterisation of present-day deep-sea sediments and … · 2009-09-14 · netic properties may allow identi¢cation of the sources and transport pathways of detrital mag-neticgrainsor,intheabsenceofsigni¢cantdetri-tal