Characterisation of oceanic sediments and basalts from Hole 833B, New Hebrides Island Arc (Vanuatu) from downhole measurements. Alain Rabaute, Louis Briqueu, Peter K. Harvey, Henri Mercadier To cite this version: Alain Rabaute, Louis Briqueu, Peter K. Harvey, Henri Mercadier. Characterisation of oceanic sediments and basalts from Hole 833B, New Hebrides Island Arc (Vanuatu) from downhole measurements.. Scientific Drilling, Copernicus Publications, 1997, 6 (2), pp.91-101. <hal- 00642165> HAL Id: hal-00642165 http://hal.upmc.fr/hal-00642165 Submitted on 17 Nov 2011 HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destin´ ee au d´ epˆ ot et ` a la diffusion de documents scientifiques de niveau recherche, publi´ es ou non, ´ emanant des ´ etablissements d’enseignement et de recherche fran¸cais ou ´ etrangers, des laboratoires publics ou priv´ es.
13
Embed
Characterisation of oceanic sediments and basalts from ... · Geochemical tool string The Schlumberger Geochemical Logging Tool (GLT) comprises four separate tools using three separate
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Characterisation of oceanic sediments and basalts from
Hole 833B, New Hebrides Island Arc (Vanuatu) from
downhole measurements.
Alain Rabaute, Louis Briqueu, Peter K. Harvey, Henri Mercadier
To cite this version:
Alain Rabaute, Louis Briqueu, Peter K. Harvey, Henri Mercadier. Characterisation of oceanicsediments and basalts from Hole 833B, New Hebrides Island Arc (Vanuatu) from downholemeasurements.. Scientific Drilling, Copernicus Publications, 1997, 6 (2), pp.91-101. <hal-00642165>
HAL Id: hal-00642165
http://hal.upmc.fr/hal-00642165
Submitted on 17 Nov 2011
HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, estdestinee au depot et a la diffusion de documentsscientifiques de niveau recherche, publies ou non,emanant des etablissements d’enseignement et derecherche francais ou etrangers, des laboratoirespublics ou prives.
Characterisation of oceanic sediments and basalts from Hole 833B, New HebridesIsland Arc (Vanuatu) from downhole measurements.
Alain Rabaute 1, Louis Briqueu 1, Peter K. Harvey 2 and Henri Mercadier 1
1 Laboratoire de Géochimie Isotopique, URA-CNRS 1763, ISTEEM, CC 066, Université Montpellier 2, Place E. Bataillon. 34095 Montpellier Cedex 05.France.
2 Borehole Research, Department of Geology, University of Leicester, LE1 7RH, UK.
Scientific Drilling
Alain Rabaute 2
Abstract. During Ocean Drilling Program Leg 134, amixed sequence of volcaniclastic and carbonate sedimentswas cored together with basaltic breccias and a single basaltsill in Hole 833B. In this study, data from Schlumberger’sGeochemical Logging Tool, together with physical propertylogs, characterise part of the section and establish a completelog-derived lithostratigraphy. The major transition betweenthe sedimentary rocks and a basaltic sill at 830 mbsf (metresbelow sea floor) is clearly seen as well as the distinctionbetween the silicate-rich sediments and the carbonate-richpart of the section. The transition between the volcanicsandstone interbedded with basaltic and sed-lithic breccia ofUnit III, and the volcanic sandstone of Unit IV is clearlymarked by the physical property logs. The accuracy of thelogging data was evaluated and discussed with an improvedlog versus core correlation. The cluster analysis of thecomplete data set shows a correlation with the data providedby the mechanical caliper device, which in turn indicates thequality of the borehole.
Introduction
Ocean Drilling Program (ODP) Leg 134 was devoted to thestudy of the New Hebrides island arc zone. The NewHebrides island arc has two characteristics. First, inopposition to the general westward subduction in this part ofthe Pacific Ocean, it is the result of an eastward subductionof the Australia-India plate beneath the North Fiji basin.Second, the arc presents an important structural feature: anancient oceanic ridge, the d'Entrecasteaux Zone (DEZ), whichis colliding with the central part of the arc (Figure 1) [5, 12,7, 13]. ODP Leg 134 showed that the DEZ is a complexstructure which consists of an Eocene volcanic arc in itssouthern part [9, 6, 29], and a slice of oceanic crustconsisting of transitional back-arc basalts in its northern part[21, 28]. The collision started around 3 Ma and presentlycauses the uplifting of the plate which supports the arc [4,33, 34, 35, 36, 23]. The two eastern and western chains ofislands (Figure 1), still rising, are interpreted as the result ofthe collision [8, 37]. The intra-arc Aoba basin, which islocated between these two chains [14, 15, 11], preserves the
entire volcanic history of the arc since Miocene time. SeeMacFarlane et al. (1988) for a review of the geology of theNew Hebrides Island Arc. ODP Leg 134 had three main objectives: a) to study thenature of the d'Entrecasteaux Zone, b) to study the effects ofits collision on the accretionary complex, and c) to study thetemporal evolution of the magmatism of the arc in order toevaluate the effects of the postulated polarity inversion of thesubduction during the opening of the North Fiji basin [10,3]. During this leg, wireline logging was carried out at foursites: 829, 830 (the accretionary complex), 831(Bougainville guyot) and 833 (North Aoba basin) (Figure 1).At each site geochemical, resistivity and lithodensity toolswere run. Considering the incomplete core recovery in thelogged interval, the aim of this study is to provide for Hole833B a continuous lithostratigraphy on the basis of the logsdata analysis. This will be compared to the lithostratigraphyderived from the cores and will help in interpreting theunrecovered intervals. The physical properties logs will adduseful information concerning rock density, resistivity orporosity, that may allow to see changes in lithologiespreviously seen as homogeneous on the cores. In order toachieve these objectives, we first have checked the reliabilityof the geochemical logs with a improved comparisonbetween some cores XRF analyses and logs data. We thenperformed a lithostratigraphic classification of the physicalproperties and the geochemical data sets, using somestatistical techniques (clustering, principal componentanalysis).
Hole 833B
Hole 833B has a total depth of 1001.1 metres. The boreholepasses through four main lithologic units (Figure 2) [31].From top to bottom, the first unit is composed of volcanicashes mixed with sediments and volcanic silts. The second isa basaltic breccia with interbedded volcanic sandstone. Thethird is carbonate-rich and includes foraminiferal sediments,limestone and nannofossil chalk. The last unit is a basalticsill with some intercalated sediments. Hole 833B is located
Scientific Drilling
Alain Rabaute 3
in a typical basin depositional environment, the Aoba basin.The lithological sequence here presents a regularstratification and a simple tectonic history. As geochemical measurements were only recorded in thishole over the interval: 550 mbsf to 850 mbsf, we will focus
our study on that interval. Core recovery was about 80% forthe considered interval. A full geological interpretationrequires a full sequence which can be obtained from theanalysis of the available logging data.
Geochemical tool string
The Schlumberger Geochemical Logging Tool (GLT)comprises four separate tools using three separate modes ofgamma-ray spectroscopy. The natural gamma-ray tool
(NGT), measures the spectrum of natural gamma radiation ofthe formation. The concentrations of potassium (K, inelemental wt%), thorium (Th, in ppm) and uranium (U, inppm) are then derived from the spectrum [20] as well as themeasure of total gamma radiation (SGR). The dual porositycompensated neutron tool (CNT) is used with the activationaluminium clay tool (AACT). The concentration ofaluminium (in elemental wt%) in the formation is derivedfrom these measurements [26]. At the bottom of the tool
Figure 1. a) Regional bathymetry (in meters) of the southwest Pacific(after Kroenke et al., 1983). b) Detailed bathymetry (in meters) of thecentral New Hebrides island arc, showing location of Sites 827-833. CDB= Central d’Entrecasteaux Basin.
4000
6 0 0 00 1000
km
160 °E 170 °
N e wC a le donia
V a nua tu
AU S T R ALIA
km
2000
North Loyalty Basin
West TorresMassif
Santa Cruz Islands
Torres Islands
Banks Islands
M atthewIsland H unter
Island
New Caledonia
4 0 0 0
40
00
15˚S
20˚
2000
4000
6000
2000
2000
4000
2000
6000
4000 d’Entrecasteaux Zone
4000
4000
6000
6000
2000
30˚
10˚S
2000
165˚E 170˚
2000
2000
4000
South Loyalty Basin
South New Hebrides Trench
166˚E
Mere Lava Island
Maewo Island
North FijiBasin
Pentecost Island
Ambrym Island
Epi Island
Aoba Island
Malakula Island
Espiritu Santo Island
Vanua Lava Island
Santa Maria Island
Banks Islands
North Loyalty Basin
South AobaBasin
Bougainville Guyot
North AobaBasin
831
Southd’EntrecasteauxChain
828
168˚
Vanuatu
833832
167˚
827
829
8302000
4000
4000
3000
3000
5000
4000
4000 2000
1000
2000
1000
2000
1000
West TorresMassif
3000
4000
5000
60001000
1000
2000
2000
15˚
14˚S
16˚
Northd’EntrecasteauxRidge
d’Entrecasteaux Zone
Scientific Drilling
Alain Rabaute 4
string, the gamma spectroscopy tool (GST) carries a pulsed14 MeV neutron source coupled with a gamma rayscintillation detector to measure the energy spectrum ofgamma rays generated by the neutron-capture reactions. The
weight percent of the elements silicon, iron, calcium, sulfur,titanium and gadolinium can be derived from this spectrum,combined with elements derived from the NGT and AACT.The GST also measures chlorine and hydrogen which arepresent in high concentrations in the borehole fluid [18].Measurements are made every 0.1564 m (6 inches).
Physical properties
Physical properties are also considered in this study in orderto evaluate correlations between the geochemical data and thephysical properties of the drilled formation. The bulk density(RHOB) and the photoelectric factor (PEF), are measured bythe high temperature litho-density tool (HLDT), which usesa radioactive 137Cs source and two detectors. This sourceemits 662 keV gamma radiation to which the formationresponds by the Compton effect, which gives the bulkdensity of the formation (RHOB, in g/cm3) and an estimateof the photoelectric factor (PEF, in barns/electrons). Threemeasurements of electrical resistivity representing threedifferent depths of investigation and vertical resolution are
provided by the dual induction tool: IDPH, deep induction,IMPH, medium induction, and SFLU, shallow sphericallyfocused resistivity [27, 25]. Measurements are made every0.1564 m (6 inches).
Logs quality
Comparison of core and log data.
In order to evaluate the accuracy of the well-logging data, X-ray fluorescence (XRF) analysis was carried out on 12samples from the sedimentary sequence (645 to 690 mbsf)and 11 samples from the basaltic flow (830 to 866 mbsf).Analyses were performed on fusion beads using a lithiumtetraborate + lanthanium oxide flux to give theconcentrations of the same six major oxides: CaO, SiO2,Al2O3, Fe2O3, K2O and TiO2, with those measured by theGLT string. Replicate analyses of rock standards show thatthe major element data are precise within 0.5 to 2.5%, andare considered accurate to ~ 1% for Si, Ti, Fe, Ca, and K,and between 3% and 5% for Al. The results are presented inTable 1. Laboratory core samples are often difficult to locate relativeto log depths because of incomplete core recovery. In anattempt to minimize this problem, we used the method ofAgrinier & Agrinier (1993) [1]: after recovering a core, onlyits upper depth and its bottom depth, the number of samplesand their individual thicknesses are known for "sure". Fromthese data, and only with the assumption that the relativepositions of samples in the core are preserved, we are able toevaluate the most probable sample depth interval for anysample of the core. Agrinier & Agrinier show that theprobability distribution of depth is given by Euler's Bêtadistribution. This method however assumes that thelithology between the samples is represented by thelithology of the samples. That is not always true, as theunrecovered parts may indicate a change in the lithology, andthe model detailed below allows us to overcome thisproblem. We then have for each sample the most probableinterval of depth expressed with a mean and a standarddeviation. The corresponding logging interval (CLI) chosenin the further study will be the 95% confidence interval,expressed as the mean depth plus and minus two times thestandard deviation. The results are given in Table 2. All thelogging values, for each major element measured by theGLT, within that interval are then averaged to find for eachsample the corresponding geochemical loggingcompositions. In order to understand better and to improve the correlation,a simple model has been made by choosing three averagingintervals on the basis of the initial corresponding logginginterval (CLI) found with the method of Agrinier & Agrinier(1993, see above). Model A is CLI ± 20 cm, Model B isCLI ± 50 cm, and Model C is CLI ± 100 cm. For eachmajor element considered, logging values within each newinterval are averaged. The mean value obtained for eachsample are compared to the XRF value and the discrepancy isexpressed in relative uncertainty percent. The aim is, whenmodeling interval increases, to see if there is a decrease inthe discrepancy, expressed as the difference between
laboratory analyses and log values (Figure 4a), or thescattering of these data (Figure 4b).
Results and interpretation.
A first comparison of the core and loggeochemical data is shown in Figure3, where the open circles stand forsamples taken from the volcano-clastic/sedimentary sequence (UnitIV), while the full triangles representthe basaltic sill (Unit V). There is noobvious correlation between the well-log data and X-ray measurements oncore except for Fe2O3, K2O andpossibly TiO2. For Fe2O3, except forone basalt sample that clearly liesabove the 1:1 line, all the others liereasonably close. Almost all thesediment samples are under the linebut quite close. The potassium alsoshows a reasonable correlation,though one basalt sample lies wellabove the line. In particular, we cannote that this is theonly case where the coremeasurements for the sedimentssamples show a greater range than thelogging data. SiO2 shows points wellgrouped around the 1:1 line, with theexception of two points, one aboveand one below the line. However, thevariation in composition is still too
great in comparison to the core measurements. The basalticsill has been described as homogeneous [31], with a averageSiO2 content of 52 wt%, as shown by the coremeasurements, while the logging data lies between 47 wt%
Table 1. Major elements compositions of volcanoclastic and carbonate sediments (samples 64R to 66R) and the basaltic sill (samples 81R to 85R)from Site 833. * were analyzed at Laboratoire de Géochimie Isotopique, Université de Montpellier II, France. + are from Shipboard ScientificParty, (1992), Site 833. ° are from Hasenaka, et al. (1994). Total iron is expressed as Fe2O3
T, or as FeO* and is calculated as follows: FeO* =(Fe2O3
T from XRF analyses * 10) / 9.
Sample Calculated Standard Corresponding logged interval
Core/Section Interval (cm) mean depth deviation (σ) top bottom
Table 2. Estimated depth and its standard deviation (σ) for each sample analyzed. The 95%confidence interval (mean ± 2σ) will be considered as the corresponding logged interval (CLI).
Scientific Drilling
Alain Rabaute 6
and 68 wt%. The same observation can be made fortitanium.
Figure 4a displays, for Model A, B and C (see above), therelative uncertainty percent between XRF data and theircorresponding logged intervals. Open circles stand forsamples taken from Lithological Unit IV, and full trianglesrepresent the basaltic sill (Lithological Unit V). Shadedsymbols are the arithmetic means. Figure 4b is thescattering of the previous data for each model used. As seenbefore in Figure 3, titanium, calcium and aluminium showthe greatest variations. Silica , iron and potassium give thebest results. Except for calcium, the sedimentary sequenceshows greater scattering than the basaltic sill. This is due tothe wide number of different lithologies present in Unit IV,whereas the basaltic sill is known to be an homogeneouslow K basalt. Considering only the basalt data, oneinteresting output is that the modeling does not improve thecorrelation as the mean does not move between each model,except for silica (Figure 4a). Although the scatteringdecreases for silica, titanium and calcium, and can beconsidered as constant for iron, aluminium and potassium(Figure 4b), an overall uncertainty of 10 to 15% remains forall elements (Figure 4a) except calcium, whose behavior is
discussed below. The modeling does not affect thesedimentary sequence data as well, except for silica andtitanium. In this case however, the mean uncertainties aregreater: 10% for silica, 25 to 30% for aluminium, iron and
potassium, and 35 to 40% for titanium and calcium (Figure4a). The scattering decreases for most of the considered majorelements (Ti, Si, Fe, Al), but increases surprisingly for Ca,and slightly for K.The constant uncertainty of 10 to 15% for all elements cannot be interpreted as the result of a poor calibration of thetool string during acquisition. If it were, the same relativepercentage error should be expected in both sediments and thebasaltic flow. Considering the calcium in our model, oneexplication of its behavior is that it can be expressed as CaOor CaCO3. A decision about which, or what combination ofthese molecules, Ca has to be quoted, is made duringprocessing. This decision is based on shipboard XRF andCaCO3 analyses. Creating here an error in the elementdetermination can lead to greater than necessary variability inthe Ca data. One aspect of this problem is illustrated inFigure 9, where a selection of shipboard CaCO3 analyses are
0
20
40
60
80
100SiO
2
0
20
40
60
80
100TiO
2
0
20
40
60
80
100Al
2O
3
Rel
ativ
e di
scre
panc
y be
twee
n X
RF
and
logg
ing
data
0
20
40
60
80
100
FeO*
0
20
40
60
80
100
CaO
the different steps of resampling of the logging data
Model A Model CModel B0
20
40
60
80
100
K2O
Model A Model CModel B
0
6
12
18
0 6 12 18
CaO
20
40
60
80
20 40 60 80
SiO2
0
1
2
3
0 1 2 3
K2O
log
(wt%
)
0
0.5
1
1.5
0 0.5 1 1.5
TiO2
0
6
12
18
0 6 12 18
FeO*
0
15
30
45
0 15 30 45
Al2O
3
core (wt%)
Figure 3. Comparison of core and geochemical logging data. The X-axisis the measure of each oxide obtained by X-ray fluorescence analysis oncores. Standard deviation is one percent of the measure. The Y-axis is themean oxide concentration and standard deviation for the well-loggingdata interval corresponding to each sample. Full triangles = basalticsequence, and open circles = sedimentary sequence.
Figure 4. a) Relative discrepancy percent between XRF and calculatedlogging data. Each calculated logging datum is the average over aprobable Corresponding Logged Interval (CLI), that corresponds to theprobable depth interval found for the XRF sample with the method ofAgrinier & Agrinier, ±20 cm for Model A, ±50 cm for Model B, and ±100cm for Model C. The relative discrepancy percent is calculated asfollows: {( |[element]xrf - [element]calc logging value| ) / [element]xrf } *100. Full triangles = basaltic sequence, and open circles = sedimentarysequence. The shaded symbols represents the mean of the discrepanciesin each sequence.
Scientific Drilling
Alain Rabaute 7
plotted together with some XRF CaO data from Table 1(CaCO3 calculated = CaO xrf * 1,78463).
Derivation of lithostratigraphy
Principles.
In order to develop a coherent lithostratigraphy using allgeochemical and physical property data, two descriptivemultivariate statistical techniques have been employed. Thesimplest of these techniques is principal component analysis(PCA), which attempts to simplify a data set while keepingthe original information. This provides a representation ofthe data that allows easier visualization of the relationshipsamong the various parameters. In particular, highlycorrelated variables become ‘attached’ to individualcomponents and in this way the apparent dimensionality onthe system is reduced. A product of the principal componentsanalysis is a new set of coordinates, the component scores,which define the original data points with respect to the newcomponent axes. These can be plotted and manipulated in thesame manner as the original variables. The second approach we have used is a classificationtechnique which allows the data to be grouped together into
discrete groups such that the data points within each groupare more similar to one another than they are to those in anyother group. This technique, iterative non-hierarchical clusteranalysis (INCA), requires no a priori information about thedistribution of the data and as such can be used to provide anessentially unbiased classification. The classification used inthis study assumes minimum euclidian distances betweensamples. The principles of this method are described byLebart et al. (1984, [19]) and Spath (1980, [32]). Shepherd etal. (1987, [28]), Harvey et al. (1989, [16]) and Pelling et al.(1991, [24]) have applied it to geochemical problems. Atypical INCA analysis starts with the choice of ‘seed points’,which will be the first temporary centroids of the furtherclusters. These seed points can be determined randomly, but
Figure 4. b) Evolution of the scattering of the logging data around thetrue value represented by the XRF data while averaging intervalincreases (Model A to C). The data scattering is expressed as thestandard deviation of the discrepancies in the sedimentary sequence(shaded circles), and in the basaltic sequence (shaded triangles).
6
7
8
9
10
11
4
5
6
7
8
9
Sedimentarysequence
Basaltic sill
SiO2
24
26
28
9
10
TiO2
30
32
34
36
38
10
11
Al2O
3
Sta
nd
ard
dev
iatio
n ~
Dat
a sc
atte
rin
g
16
17
18
19
6
7
8
FeO*
20
21
22
23
24
18
20
22
24
26
CaO
18
19
20
13
14
15
K2O
Model A B C
4000
5000
6000
7000
8000
9000
10000
2 4 6 8 10 12Sum
of
dist
ance
s be
twee
n gr
oup
cent
roid
s
Number of groups5
Figure 5. Diagram showing the number of INCA groups versus the sumof the distances between groups centroids. The arrow shows the arbitrarychoice previous INCA analysis which corresponds to the point when thecurve flattens out.
Scientific Drilling
Alain Rabaute 8
we chose the method of Ball & Hall (1967, [2]), whichconsists in taking, as initial seed points, a few points in thedata set that are already separated by a minimum distance.This approach allows faster determination of the differentclusters. Then, the data points are allocated to the nearestseed point, in terms of simple euclidian distance, in our casebetween the data point and the seed point (now called clustercentroid). When all points have been assigned, new centroidsare calculated for each cluster as the average of all samples
allocated to thatcluster. The datapoints are allocatedagain to the nearestcentroid. Thisprocess of re-allocation andcalculation of newcentroids continuesuntil the systemconverges (nofurther relocation) ora set number ofiterations has beencompleted. Thenumber of groupscan be specifiedbefore the analysisor determined duringthe latter by theclassifyingalgorithm. For eachdata set, such adetermination mustbe done before thecluster analysis, inorder to know theappropriate numberof groups thatallows the bestdiscrimination of thedata. The algorithmcalculates, for agiven data set, thenumber of iterationsrequired forconvergence and thesum of the distancesbetween groupcentroids, and doesthis analysis for anincreasing numberof clusters (forexample, from twoto six clusters). Byplotting the sums ofthe distancesbetween centroidsversus the numberof groups, the mostappropriate numberof groups is chosenwhere the curve
flattens out (see Figure 5). For example, models with fivegroups gave the best results for correlation with thegeological data in Figure 6.
Results and interpretation.
The results of the principal component and cluster analysesof the whole logged interval are summarised in Figure 6.
Comp. 1 Comp. 2 Comp. 3 Groups
-5.0 5.0 -4.5 4.5 -7.5 7.5 1 2 3 4 5
Caliper
820.4
790.8
761.1
731.5
701.8
642.5
612.8
583.2
553.5
672.1
850.1
Depth(mbsf)
Figure 6. Summary of the statistical analysis (PCA + INCA) of the entire logging data set for Hole833B. The three left hand columns show the variation in component score with depth for the threedominant principal components. The fourth column summarises an INCA analysis based on thecomponent scores. The caliper curve, which indicates the shape of the borehole, is put over thefourth column to emphasize the correlation between Group 3 and bad hole data.
Scientific Drilling
Alain Rabaute 9
The principal components analysis was based on the majorelements, Ca, Si, Fe, Ti, K and Al, together with the
physical properties: bulk density, shallow resistivity andphotoelectric factor. The principal component scores for thethree dominant components were subsequently subjected toINCA analysis. Figure 6 shows the variation in thecomponent scores and INCA classification with depth. Thecaliper curve, which indicates the shape of he borehole, hasbeen put over the groups column. The INCA analysissuggested a five group classification. In Figure 6, the lengthof the bar in the column ‘Groups’, representing each datapoint or set of data points allows discrimination between thedifferent groups. The pattern used for each group iscompletely arbitrary. The groups are numbered in order ofincreasing length of its respective bar. A strong transition at830 mbsf divides the log in two parts: Above this transition,Groups 1, 2 and 3 are present, while below this depth onlyGroups 4 and 5 are found. The sharp segregation into twodifferent associations of groups is very clear and correlates
precisely with the contact between the volcano-sedimentarysequence above and the top of the basaltic sill below. The results of the principal components analysis issummarised in Table 3. The eigen-values are listed for the
first four components; the first three, which represent almost70% of the total variation in the data set, are discussedbelow. The first component is seen to reflect the sametransition between the associations of Groups 1, 2, 3 andGroups 4 and 5. It is characterised by high loadings on thephysical properties (RHOB, SFLU and PEF) and FeO*.High values of these variables in the basalt result incorresponding high positive scores for component 1 whichcan be seen clearly in Figure 6. Component 2, which isdominated by a high negative loading on CaO, and a highpositive loading on SiO2, determines the difference betweenGroup 1 and Group 2. As such, component 2 distinguishesbetween the carbonate-rich layers (giving negative scores forGroup 1) from the silica-rich layers (positive scores forGroup 2). The third component correlates with Group 3,which appears at irregular intervals along the log. Thiscomponent is dominated by high positive loadings for TiO2,K2O and Al2O3. High scores on this component are clearlycorrelated with caliper peaks that show bad hole conditions.This coincidence leads to two possible interpretations: thedata are suspect because of the collapse of the boreholewalls, and the statistical model classifies everything as onesingle group that should not be taken into account in furtherlithological interpretation. On another hand, this increase inthe borehole diameter can be due to the presence of a specificlithology (sand layer, shale), and Group 3 is a marker of thisparticular lithology. In any case, this demonstrates that thequality of the borehole is a supplementary constraint on thevalidity of the well-logging data.Figure 8 summarises the results of two successive INCAanalyses over the sedimentary section of the logged interval(550 to 830 mbsf): The geochemical data together with thetotal amount of natural gamma radiation (SGR) and the Thand U contents separated the section into four discriminatinggroups, labeled from 1 to 4 in Figure 8. The second INCAanalysis found four other groups, labeled from 5 to 8, on thebasis of four physical properties (medium and deepresistivity (IMPH, IDPH), photoelectric factor (PEF) andbulk density (RHOB)). Figure 7 displays the copheneticcorrelation between the groups, and Table 4a and b showsthe group centroids composition, for both analyses.Lithological units inferred from the logs interpretation is putnext to the lithologies seen on the cores.
Table 4 allows to find thegeneral chemical or physicalcharacteristics of the lithologiesrepresented by the groups:
- Groups 1 and 3 areclose together considering thevariables used in the analyses(Figure 7), and are then goingto correspond to very similarlithologies. Both groups showrelative high silica and lowcalcium content, and relativelow gamma radiation. Theyprobably correspond to thesiliciclastic sediments (volcanic
sandstone, basaltic breccia). Group 1 shows higher contentin iron and titanium than Group 3, marking then somemixture between basaltic clasts (Fe- and Ti-rich) or breccia
eigen-values % of contributionto the totalvariance
Table 3. Results of the Principal Component Analysis of the complete dataset shown in Figure 6: eigen-values for the first four eigen-vectors,percentage of contribution to the total variance, and cumulative percent.Principal component loadings matrix for 3 components. The first threeeigen-vectors account for 69% of the total variance, which is acceptablein our case.
Figure 7. Cophenetic correlation trees for the INCA analyses using the geochemical data set (tree a) and thephysical properties data set. The cophenetic correlation represents the degree of distortion of the dendrogram.
Scientific Drilling
Alain Rabaute 10
and volcanic sediments, whereas Group 3 will mostly standfor more homogeneous volcanic sandstone.
- Group 2 shows high content in calcium andrepresent the carbonate-rich layers. It is relatively U-rich,suggesting some organic carbon-rich levels or the presenceof heavy minerals.
- Group 4 shows the highest amount of gamma-radiations, and the highest content in K, Ti and Th. Thisgroup can indicate the clay-rich intervals present in thesequence.
- Group 5 is very distinctly defined (Figure 7), andpresents by far the highest resistivity, and the highestdensity. It can indicate the presence of clasts, hard rock orvery well lithified sediments layers.
- Group 6 and 7 are close together and differ mostlyby their resistivity, Group 7 defining a relatively moreresistive lithology than Group 6.
- Group 8 show the lowest density, photoelectricfactor and resistivity values, and appears very occasionallythroughout the sequence. It can indicate unconsolidated shale,sand or clay-rich layers. On the basis of theresults of both analyses,the logged interval canbe separated into threeLog-Units. Log-Unit I: it is foundbetween 550 and 580mbsf and, ascharacterised by Groups5 and 1, is more resistiveand denser, and showsrelative high contents iniron and titanium and issiliceous. Furthermore,Group 5 refers to distinctlithology (Figure 7). Itcorresponds fairly well toUnit III, characterised byig- and sed-lithic breccia,mixed volcanic
sandstone/siltstone. Log-Unit II: it isdefined between 580 and755 mbsf, and can besubdivided into 4 sub-units that are slightly different.
- Sub-Unit A: defined between 580 and 623 mbsf, ischaracterised by Groups 1, 3 and 7. This defines a relativelyresistive lithology, dominantly siliceous, which correspondfairly well to the volcanic lithified sandstone and siltstonefound on the cores at this depth. The presence of Group 6 atsome intervals suggests some unconsolidated zones. Eachstatistical analysis put the end of this sub-unit at twodifferent depths, and this end is thus defined as a transitionzone between 623 and 627 mbsf. Here, we see that bothGroups 1 and 3 are still present, but Group 6 has replacedGroup 7. Although still being siliceous, the numerousdifferent layers intercalated there lead to a less consolidatedzone, thus marked by a lower resistivity.
- Sub-Unit B: defined between 627 and 666 mbsf, ischaracterised by Groups 2, 3 and 6. The scattered occurrencesof Groups 2 and 3 suggests again several intercalated layers
Lithological unitsfromlogs from cores1 2 3 4
Chemical propertiesGroups
Physical properties
5 6 7 8
750
800
550
600
650
700
Log-
Uni
t II
Log-
Uni
t III
Uni
t III
Uni
t IV
Log-
Sub
unit
ALo
g-S
ubun
it B
Log-
Sub
unit
CLo
g-S
ubun
it D
Depth(mbsf) re
cove
ry
Log-
Uni
t I
Figure 8. Summary of the two separate INCA analyses performed on partof the sedimentary sequence (550 to 830 mbsf) of Hole 833B. The first twocolumns summarise the groups resulting from the analyses (1 to 4 for thegeochemical data, and 5 to 8 for the physical properties data). The thirdcolumn displays the lithological units inferred from the logs, as well as thelithologies and units seen on the cores (see Figure 66, Shipboard Scientificparty, 1991, Site 833). The symbols used are the same as in Figure 2. Theextreme right-hand column gives an idea of the core recovery of thissection.
0 10 20 30 40
650
660
670
680
690
CaCO3 (wt%)
Dep
th (
mbs
f)
XRF analyses (see Table 1)Shipboard CaCO3 analyses
Figure 9. Diagram showing selectedShipboard CaCO3 analyses plottedtogether with calculated CaCO3 datafrom XRF CaO measurements (seeTable 1; CaCO3 calculated = CaO xrf *1,78463).
Scientific Drilling
Alain Rabaute 11
of different nature, leading to a global lower resistivity(Group 6). This sub-unit signs the appearance of Group 2,which correlates very well with the appearance of thelimestone/mixed-sedimentary. Group 3 is still present andmarks some intercalated volcanic sandstone layers. The endof this sub-unit has also been figured as a transition zone(666 to 676 mbsf), only characterised by Groups 3, 6, and 8,that suggests a shale or clay-rich layer, which is consistentwith the silty claystone mixed with clayey volcanic silt and
nannofossil chalk seen on the cores.- Sub-Unit C: defined between 676 and 700 mbsf, is
characterised by Groups 1, 3 and 6. We have here the samesequence as in Sub-Unit A, but less consolidated as lessresistive. Again, this correlates with the lithology seen onthe cores, which is more or less the same as from 580 to623 mbsf.
- Sub-Unit D: characterised by the scatteredoccurrence of all Groups 1 to 4, suggesting a sequence madeby thin intercalations of volcano-clastic, carbonate-rich andclay-rich layers, more or less well lithified as indicated bythe average resistivity throughout this sub-unit (Groups 6and 7). The occurrence of Group 5 at 726 and 737 mbsfsuggests some more consolidated fine-grained intervals thatwere seen in Cores 70R and 71R (see [31]). The base of thissub-unit is also marked by a transition zone (755 to 762mbsf) characterised by Groups 3, 4 and 6 and thus interpretedas interbedded volcano-clastic siltstone/sandstone with clay-richer intervals. Log-Unit III: it is found between 762 and 830 mbsf andcharacterised mostly by Groups 2 and 6. This can beinterpreted as a carbonate-rich sequence, dominated bylimestone or lithified calcareous chalk or sandstone. Theoccurrence of Group 3 at 790 mbsf and Group 4 at 808 and821 mbsf suggests respectively a volcano-clastic silica-richinterval and some clay-rich layers (claystone, clayeylimestone) at the end of the section. The presence of Group 7at 782 mbsf and between 815 and 830 mbsf indicates betterlithified and more homogeneous intervals.
Summary and Conclusion.
The logs data from Hole 833B have been carefully analysedto establish a detailed and continuous lithostratigraphy in an
attempt to overcome the occasionally poor recovery.Following an evaluation of the quality of the geochemicaldata used in the interpretation, multivariate statisticaltechniques were used to infer a log-based lithostratigraphywhich shows excellent agreement with the core observations.It has been demonstrated that, in case of geochemicallogging data, the decisions made during the processing of theyields in order to obtain oxide weight percent data will becrucial for the quality of the final data. The uncertainties
calculated are not leading to awrong interpretation of the data inour case, but reducing these errorscould help in the presence of thinbeds, mixed layers orheterogeneous environments. Thelogs analysis has permitted someextra subdivisions (Figure 8)based on changes in physicaland/or chemical properties, andclarifies the continuity of all coreobservations through nonrecovered intervals. The log-basedlithostratigraphy indicates well-and poor-consolidated intervals,the occurrences of carbonate-richand clay-rich lithologies, andshows that a unit considered as awhole on the cores can be
separated in several consistent sub-units which arecharacterised by different chemical compositions, or differenttextures and fabrics. Principal component analysis allowed agood discrimination between the carbonate-rich and silica-rich sediments, and showed the importance of the quality ofthe borehole in the data obtained (Figure 6).
Acknowledgments.
The authors gratefully thank the CNRS for supporting thiswork by a financial contribution of the “GéosciencesMarines” public research program.
References.
1. Agrinier, P. & Agrinier, B. (1994) A propos de la connaissance de laprofondeur à laquelle vos échantillons sont collectés dans les forages.C. R. Acad. Sci. Paris, série II, vol. 318, pp 1615-1622.
2. Ball, G.H. & Hall, D.J. (1967). A clustering technique for summarizingmultivariate data. Behavioral Science. 12. pp 153-155.
3. Barsdell, M., Smith, I.E.M., and Sporli, K.B. (1982). The origin ofreversed geochemical zoning in the Northern New HebridesVolcanic Arc. Contrib. Mineral. Petrol., 81: pp 148-155.
4. Chung, W.Y. & Kanamori, H. (1978) A mechanical model for platedeformation associated with aseismic ridge subduction in the New-Hebrides arc. Tectonophysics.Vol. 50. pp 29-40.
5. Collot, J.Y., Daniel, J., and Burne, R.V. (1985). Recent tectonicsassociated with the subduction/collision of the d'Entrecasteaux zone inthe central New Hebrides. Tectonophysics, Vol. 112. pp 325-356.
6. Collot, J.Y., Pelletier, B., Boulin, J., Daniel, J., Eissen, J.P., Fisher, M.A.,Greene, H.G., Lallemand, S. and Monzier, M., (1989). Premiersrésultats des plongées de la campagne SUBPSOI dans la zone decollision des rides d'Entrecasteaux et de l'arc insulaire des NouvellesHébrides. C. R. Acad. Sci., série II, vol. 309, pp 1947-1954.
7. Collot, J.Y. & Fisher, M.A. (1991). The collision zone between theNorth d'Entrecasteaux Ridge and the New Hebrides Island arc. Part.1: Seabeam morphology and shallow structure. J. Geophys. Res., Vol.96. pp 4447-4478.
8. Collot, J.Y. & Fisher, M.A. (1992) The d’Entrecasteaux Zone-NewHebrides island arc collision zone: an overview. In Greene, H.G.,
Table 4. Centroids compositions, corresponding to the average composition of the group, for INCAanalyses of geochemical data and physical properties data sets shown in Figure 8.
Scientific Drilling
Alain Rabaute 12
Collot, J.Y., Stokking, L.B., et al., Proc. ODP, Init. Repts., 134. pp 19-31.
9. Dubois, J., Deplus, C., Diament, M., Daniel, J., and Collot, J.Y., (1988).Subduction of the Bougainville seamount (Vanuatu): mechanical andgeodynamical implications. Tectonophysics, Vol. 149. pp 111-119.
10. Falvey, D.A., (1975) Arc reversals and a tectonic model for the NorthFiji Basin. Austal. Soc. Explor. Geophys. Bull., Vol. 6 pp 47-49.
11. Falvey, D.A., and Greene, H.G., (1988). Origin and evolution of thesedimentary basins of the New Hebrides Arc. In Grene, H.G., andWrong, F.L. (Eds) Geology and offshore Resources of Pacific IslandsArcs-Vanuatu region. Circum-Pac. Counc. Energy and Miner.Resour., Earth Sci. Ser., vol. 8, pp 177-200.
12. Fisher, M. A., (1986). Tectonic processes at the collision of thed'Entrecasteaux zone and the New Hebrides Island Arc. J. Geophys.Res., vol. 91. pp 10470-10486.
13. Fisher, M.A., Collot, J.Y. & Geist, E.L. (1991) The collision zonebetween the North d'Entrecasteaux Ridge and the New HebridesIsland Arc. Part. 2. structure from multi channel seismic data. J.Geophys. Res., vol. 96, pp 4479-4495.
14. Greene, H. G. and Johnson, D.P., (1988) Geology of the Central Basinregion of the New Hebrides Arc inferred from single channelseismic-region. Circum-Pac. Counc. Energy and Miner. Resour.,Earth Sci. Ser., vol. 8, pp 177-200.
15. Greene, H. G., Macfarlane, A., Johnson, D.P., and Crawford A.J.,(1988). Structure and tectonics of the central New Hebrides Arc. InGrene, H.G., and Wrong, F.L. (Eds) Geology and offshore Resourcesof Pacific Islands Arcs-Vanuatu region. Circum-Pac. Counc. Energyand Miner. Resour., Earth Sci. Ser., vol. 8, pp 177-200.
16. Harvey, P.K. and Lovell, M.A. (1989) Basaltic lithostratigraphy ofOcean Drilling Program Hole 504B. Nuclear Geophysics, vol. 3, N°2,pp 87-96.
17. Hasenaka, T., Briqueu, B., Coltorti, M., Baker, P., and Fujinawa, A.(1994) Magmatic evolution of the North-Aoba intra-arc basin: Sites832 and 833. In Greene, H.G., Collot, J.Y., Stokking, L.B., et al., Proc.ODP, Sci. Results, 134: College Station TX (Ocean Drilling Program),375-392.
18. Hertzog, R.C., Colson, L., Seeman, B., O’Brien, M., Scott, H.,McKeon, D., Grau, J., Ellis, D., Schweitzer, J., and Herron, M. (1989)Geochemical logging with spectrometry tools. SPE FormationEvaluation. vol 4(2), pp 153-162.
19. Lebart, L., Morineau, A. and Warwick, K.M. (1984) Multivariatedescriptive statistical analysis. John Wiley & Sons. New York.
20. Lock, G.A. & Hoyer, W.A. (1971) Natural gamma-ray spectrallogging. The Log Analyst. 12(5), pp 3-9.
21. Macfarlane, A., Carney, J.N., Crawford A.J., and Greene, H. G.,(1988). Vanuatu - a review of onshore geology. In Grene, H.G., andWrong, F.L. (Eds) Geology and offshore Resources of Pacific IslandsArcs-Vanuatu region. Circum-Pac. Counc. Energy and Miner.Resour., Earth Sci. Ser., vol. 8 pp 1.
22. Maillet, P.M., Monzier, M., Selo, M., and Storzer, D., (1983). Thed'Entrecasteaux zone (southern Pacific). A petrological andgeochronological reappraisal. Mar. Geol., vol. 53 pp 179-197.
23. Pelletier, B. & Louat, R. (1989) C.R.Acad.Sci.Paris. 308, serie II. pp123-130.
24. Pelling, R., Harvey, P.K., Lovell, M.A. and Goldberg, D. (1991)Statistical analysis of geochemical logging tool data from Hole 735B,Atlantis fracture zone, Southwest Indian Ocean. In Proc. ODP,Scientific Results, vol. 118, pp. 271-283.
27. Serra, O. (1984) Fundamentals of Well Log Interpretation: TheAcquisition of Logging Data: Elsevier. Amsterdam.
28. Shepherd, A., Harvey, P.K. & Leake, R.C. (1987) The geochemistryof residual soils as an aid to geological mapping; a statisticalapproach. Journal of geochemical exploration.. Vol. 29, pp 317-331.
29. Shipboard Scientific Party, 1992. Site 828. In Collot, J.Y., Greene,H.G., Stokking, L.B., et al., 1992. Proc. ODP, Init. Repts., 134:College Station, TX (Ocean Drilling Program), pp 139-177.
30. Shipboard Scientific Party, 1992. Site 831. In Collot, J.Y., Greene,H.G., Stokking, L.B., et al., 1992. Proc. ODP, Init. Repts., 134:College Station, TX (Ocean Drilling Program), pp 317-386.
31. Shipboard Scientific Party, 1992. Site 833. In Collot, J.Y., Greene,H.G., Stokking, L.B., et al., 1992. Proc. ODP, Init. Repts., 134:College Station, TX (Ocean Drilling Program), pp 479-557.
32. Spath, H. (1980) Cluster analysis algorithms for data reduction andclassification of objects. Ellis Horwood, Chichester.
33. Taylor F.W., Issacks, B.L., Jouanic, C., Bloom, A.L., and Dubois, J.(1980). Coseismic and Quaternary vertical tectonic movements, Santoand malekula Islands, northern New Hebrides islands arc. J Geophys.Res., vol. 85, pp 5367-5381.
34. Taylor, F.W., Jouanic, C., and Bloom, A.L., (1985). Quaternary upliftof the Torres Islands, northern New Hebrides frontal arc: comparison
with Santo and Malekula Islands, central New Hebrides frontal arc. J.Geol., vol. 93, pp 419-438.
35. Taylor, F.W., Frohlich, C., Lecolle, J., and Strecker, M. (1987)Analysis of partially emerged corals and reef terraces in the centralVanuatu arc: comparison of contemporany coseismic and nonseismicwith Quaternary vertical movement. J Geophys. Res., vol. 92, pp4905-4933.
36. Taylor, F.W., Edwards, R.L., Wasserburg, G.J., and Frohlich, C.(1990) Seismic recurrence intervals and timing of aseismic subductioninferred from emerged corals and reefs of the central Vanuatu (NewHebrides) frontal arc. J Geophys. Res., vol. 95, pp 393-408.
37. Taylor, F.W. (1992) Quaternary vertical tectonics of the centralNew-Hebrides island arc. In Greene, H.G., Collot, J.Y., Stokking,L.B., at al., Proc. ODP, Init. Repts., vol. 134. pp 33-42.