Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
TECHNICAL REPORT NO. 52
August 8, 1985
A GEOSTATISTICAL APPRAISAL OF REGIONAL
GEOCHEMICAL DATA ON MARINE SEDIMENTS
FROM THE SW PACIFIC IN REGARD TO
EXPLORATION FOR DETRITAL, BEDROCK
PHOSPHATIC AND HYDROTHERMAL
MINERAL DEPOSITS
by
R.N. Coward & D.S. Cronan Marine Mineral Resources
Programme, AGRG, Department of Geology
Imperial College, London SW7
Prepared for:
COMMITTEE FOR CO-ORDINATION OF JOINT
PROSPECTING FOR MINERAL RESOURCES IN
SOUTH PACIFIC OFFSHORE AREAS
(CCOP/SOPAC)
As a contribution by
UNDP PROJECT OFFICE
PROJECT RAS/81/102
INVESTIGATION OF MINERAL
POTENTIAL OF THE SOUTH
PACIFIC
- 1 -
ABSTRACT
Multivariate statistical analysis of geochemical data for 17 elements on over
600 sediment samples from the S.W. Pacific has helped to elucidate the major background
controls on the sediment composition and identify anomalies related either to sea floor
mineralisation or mineralisation on nearby land areas. Volcaniclastic, carbonate, phyllosilicate,
authigenic manganiferous and continental sources supply the major sediment components
and control the background sediment chemistry. Superimposed on background
concentrations of elements are a large suite of anomalies reflecting detrital, phosphatic
and hydrothermal mineralisation and which are scattered throughout the whole area.
Detrital anomalies occur in the N. Fiji Basin, off the Solomon Islands, New Caledonia, New
Zealand and Samoa, and possibly in the Tofua Trough and off the Kermadec Islands.
Phosphatic anomalies occur in the N. Fiji and Manus Basins, and within the Solomon
Islands and Vanuatu chains. Hydrothermal anomalies occur off Epi and In the N. Fiji, Manus
and Lau Basins.
- 2 -
INTRODUCTION
Geostatistical analysis of regional geochemical data is a commonly used technique
in mineral exploration programmes on land. However, to date, it has found little application in
marine mineral exploration. Undoubtably the main reason for this is that geochemical data sets
sufficiently large to be amenable to the geostatistical techniques usually employed in regional
geochemical exploration on land have hitherto just not been available from the marine
environment. However, as a result of painstaking collection and analysis of sediment samples in
the SW Pacific area over a number of years, a sufficiently large data set to justify a
geostatistical regional geochemical treatment, approximately 650 samples in all, is now
available and has been used in this work.
It could be argued that marine sediment samples lend themselves better to a regional
geochemical treatment than stream sediment or rock and soil samples on land. The reason for
this is that the mixing processes generally operative in the marine environment will tend to
broaden point source geochemical anomalies and thus, while reducing their amplitude, will
widen their extent. A wider spaced sample distribution should thus be acceptable than in on
land regional geochemical exploration programmes.
The sample base used in this work is derived from several sources. The nucleus
is the data set of Cronan and Thompson(1978) the samples being re-analysed by inductively
coupled plasma spectrometry for a larger number of elements. This has been supplemented by
data on samples collected in the Tonga-Kermadec Ridge, Lau Basin, Harve Trough and off
northern New Zealand areas, which have been described by Cronan et al.(1984). Finally,
about 200 additional previously unanalysed sediment samples from the CCOP/SOPAC
collections and taken from throughout the CCOP/SOPAC area, have been analysed for the
first time. This total data set Is one of the largest, i f not the largest, ever employed for marine
mineral exploration purposes.
The data described in this work have been employed mainly in attempting to outline
possible areas of detrital and phosphatic minerals on the sea floor in the S.W. Pacific, together with possible mineralisation in bedrock on adjacent islands. However, some of the data are
also of relevance in exploration for hydrothermal mineralisation, and this subject i s also
addressed by way of supplementing an earlier coverage of hydrothermal mineralisation within
the region (Cronan,1983).
The statistical treatments employed in this work go far beyond the univariate
statistical analyses applied to the individual data sets previously, and considerably supplement
- 3 -
and enhance those first attempts to outline the main controls on the composition of the
sediments. Nevertheless, some of the areas found to be anomalous in the present work have also been recognised as being anomalous using univariate statistics (Cronan & Thonpson,1978;
Cronan, 1983, Cronan et al.,1984; Hodkinson et al., in press, Cronan, in press). The main
aims of the present work therefore are i) to recognise the "old" univariate (single element) anomalies as multivariate (more than one element) anomalies also, and to maximise their
contrast, and ii) discover "new" anomalies in the data set. For the purpose of this study an
anomaly is defined as a population of samples which is geochemically distinct from the
background population that is normal for the prevailing sediment type in any given area. Just
because an element concentration is high is not considered sufficient justification for the term
anomaly to be applied. To be anomalous a population must be remote (seperable) from the
gaussian distribution of background values. This approach strengthens anomaly selection
based simply on individual element enrichments.
ANALYTICAL METHODS
The sediment samples were air dried over silica gel until no further moisture was
absorbed, and then hand or machine ground to a fine powder using agate vessels. Chemical
analysis of the samples was performed by dissolving the sediments in a mixture of hydrofluoric
nitric and perchloric acids, evaporating to dryness, taking up the residues in 1M HCI and
injecting the solutions into an ARL 34000 inductively coupled argon plasma emission
spectrometer and reading for Li, K, Be, Mg, Ca, Sr, Al, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn and
P. Accuracy and precision were better than+ 10% based on the routine use of duplicates and
reference materials.
CORRELATION ANALYSIS
A correlation matrix is the starting point for all the multivariate statistical techniques used in the present study. In a regional survey of this nature with a large sample spacing,
one would expect to sample only a few anomalous mineralised samples. Therefore the most
that one can expect from a correlation matrix is an expression of the background relationships between variables. There are several methods of conditioning the correlation matrix to best
represent such background relationships. These are data transformation and outlier correction
methods.
Data Transformation - The ideal transformed data values in multivariate statistical analysis of geochemical data will have normal gaussian distributions. In most geological
- 4 -
situations many trace element frequency distributions are positively skewed. In attempts to normalise data in exploration geochemistry, the log transform is usually applied to push
positively skewed distributions of elements towards the ideal gaussian distributions. However,
applying a log transform sometimes results in the production of negatively skewed distributions.
To avoid this happening it is possible to perform a Box - Cox (1964) power transformation
which optimises the normality of sample populations. The transformation is of the form:
1 X = (X - 1) / ^
X = transformed element value
X = untransformed element value
^ =power.
This transformation was applied to the SW Pacific data set.
Outlier Correction - It is well known in geostatistics that outlying samples adversely
affect the calculation of correlation coefficients between elements. This is because
background variable means (important parameters in the calculation of covariance) are
strongly influenced by atypical observations. The problem of outliers can be overcome by
either univariately or multivariatly trimming the data set, or by the calculation of robust
correlation coefficients. Univariate and multivariate trimming both involve the physical removal
of the outlying samples from the data set prior to the calculation of the correlation matrix.
These are reinstated before the calculation of any interpretative parameters and identification
of anomalies. However, rather than remove the outlying samples from the data set, it would be
better i f the means used in the calculation of the covariance were robust, that is if the effect
of outlying samples were downweighted in the calculation of the means. Several different
weighting techniques are available. The one used in the present work is one where the
influence of outliers away from the group mean is reduced by assigning weights inversely
proportional to Mahalanobis Distance, which is given by:
- - D2=(X-X)S-1(X-X)t
D = Mahalanobis Distance
X = Sample composition
X = Vector of variable means
S-1= Covariance matrix
t = Transform of matrix
The robust correlation matrix for the data discussed in this work is given in Table 1.
- 5 -
PRINCIPAL COMPONENTS ANALYSIS
Principal Components Analysis is a multivariate statistical technique capable of
extracting covarying groups of elements from a suitably conditioned correlation matrix. The starting point for the analysis in this work is the robust correlation matrix for Box-Cox
transformed data values (Table 1). An iterative program (BPCA2) was used on the Imperial College computer system to generate the rotated loadings matrix. Varimax rotation was found
to be necessary to resolve geologically interpretable components. MINITAB was subsequently used to compute principal components scores using standardised data in the simple matrix
algebra.
The rotated loadings matrix provides groups of elements that are mutually correlated.
These are illustrated along with their corresponding eigenvalues in Table 2. Important elements
from the five components are shown below with decreasing correlations from left to right.
1) (V, Mg, Fe, Ti, Co, Ai)
2) (Sr,Ca)
3) (Li,K, Be, Ti, Al)
4) (Mn, Cu, P, Zn, Co, Ni)
5) (Ni,Cr)
What has to be done in an exploration geochemical interpretation of such data is to explain the
covariance of the elements by variation in the amount of a particular mineral phase or phases
in the sediment. In order to accomplish this, residual probability plots of principal components
scores, score maps, sediment lithology data, partition analysis and bulk chemical data have
been employed. The interpretations based on these treatments outline the main geological and
geochemical controls on the element associations in the majority of the samples. However,
other minor processes cannot he excluded.
PCA 1 (V, Mg, Fe, Ti, Co, Al). Fig. 1
Using Inflexion points on the probability plot of principal component scores it is
possible to recognise four populations of samples, A - D, in a component that can be
interpreted as largely representing the amount of clastic volcanic material in the sediments.
The populations represent different sediment types:-
A = Clays from the South Penrhyn Basin
B = Clays from the Tokelau and S. Fiji Basins and selected volcanic sands
- 6 -
C = A gradational population from volcanic sands to carbonate oozes
D = Carbonate rich deposits
These populations are plotted as different symbols on the score map (Fig. 2). negative scores represent increasing components scores.
Increasing
Population A consists solely of clays from the South Penrhyn Basin, while population
B contains days from the South Fiji and North Tokelau Basins. Also Included in population B
are volcanic sands from the western flank of the northern Tonga Ridge, the Southern Tonga
Ridge, the central Kermadec Ridge near the Kermadec Islands, off the New Georgia Group in
the Solomon Islands, offshore northern Samoa, in the channel between San Jorge and Santa
Isabel in the Solomon Islands, and in the Sigatoka river mouth, Fiji. Scores for populations C
and D can be interpreted In terms of the variation in amount of volcanic material in the
sediments. Samples with low component scores (high positive scores) are "clean" carbonate
deposits with no volcanic component. Samples at the negative end of population C are
volcanic sands. Petrographic work has shown that phases responsible for the element
groupings in this component are basic/intermediate glass with accessory orthopyroxene,
clinopyroxene, olivine, plagioclase and opaque iron oxides.
P.C.A. 2 (Sr,Ca) Fig. 3.
Three populations of samples are recognisable in Fig. 3, A - C, and these are plotted
on the score map, Fig. 4. Increasing negative scores represent Increasing component scores.
A = Carbonate sands and oozes
B = Mixed population
C = Volcanic sands and clays.
This component can be interpreted as representing the amount of biogenic carbonate in the
sediments.
P.C.A.3 (Li, K, Be, Ti, Al) Fig. 5.
The dominant process explaining the element associations on PCA 3 can be regarded as the amount of phyllosilicate minerals (clay minerals) in the sediments. Examination of the
probability plot (Fig. 5) reveals two major populations, B and C, incorporating most samples,
together with a minor population A. Increasing positive scores represent increasing component
scores, and the populations include (Fig. 6):-
- 7 -
A = Carbonate sediments from predominantly pelagic areas
B = Impure carbonate oozes largely from the North Fiji Basin and
volcaniclastic sediments from the Tonga-Kermadec Ridge and
Solomons/Vanuatu chain
C = Basin clays and muds largely from the New Hebrides, Manus,
Penrhyn and S. Fiji Basins and on the Pacific Plate, and
volcaniclastics from offshore Samoa and Raukumara Plain silts.
The basinal clays and muds score high on this component due to their large fine grained
phyllosilicate fraction rich in AI and K. Source enrichments in all high loading component
variables or anomalously high values in one or two of the component elements can be held to
explain isolated samples in the Vanuatu/Solomons chain plotting in the upper population.
Raukumara Plain silts and offshore Samoan volcaniclastics have small phyllosilicate fractions
yet score high on PCA 3. In the case of Raukumara silts this is thought to be due to their
containing a continental derived fraction enriched in ail the high loading variables (Li,K, Be,
Al,Ti) relative to the dominant basaltic/andesitic composition of the bulk of the volcani-
clastic material in the region. By contrast, Samoan volcaniclastics score high due to their
containing very high concentrations of Ti (1-2%) which more than compensates for moderate
concentration of the remaining high loading variables in the calculation of the scores.
P.C.A. 4 (Mn, Cu, P, Zn, Co, Ni) (Fig.7).
The dominant mineral phase explaining the grouping of elements on this principal
component is likely to be manganese oxide and associated absorbed/coprecipitated ions.
Three populations of samples are recognisable on the probability plot of scores (Fig. 7), B
containing most samples and anomalous populations A and C. These populations are plotted on
the score map (Fig. 8). Increasing positive scores represent increasing component scores; and
the populations are:-
A = Anomalously low scoring carbonate sediments and Raukumara Plain silts.
B = Sediments ranging from carbonate oozes to volcaniclastic bearing clays
and muds.
C = Anomalously high scoring basinal clays/muds from various basins throughout
the region, and on the Pacific Plate. Also included are manganese rich
carbonate oozes from the N. Fiji Basin in which the manganese is thought
to have a hydrothermal component (Cronan,1983).
The main controlling process on manganese oxide concentration in the sediments
- 8 -
will be the rate of deposition of major diluting phases such as carbonates and volcaniclastics.
Consequently, sediments near or below the CCD and away from volcaniclastic influencer score higher on PCA 4 due to their low rate of sedimentation allowing hydrogenous manganese
oxides to accumulate in abundance. Only where a non-hydrogenous source of manganese is available, such as in hydrothermally active areas, do more rapidly depositing sediments score
high on PCA 4.
The association of P on this component is, at first sight, rather strange as the principal
association of P should be with phosphatic deposits, and these are normally most abundant on
seamounts in the study area (Cullen, in press). However, phosphatic fish debris are, like
manganese oxides, a sometimes important component of slowly accumulating pelagic
sediments. Furthermore, possible reworking of phosphatic debris on seamounts and their
transport down slope could contribute to the concentration of P in basinal sediments in the
vicinity.
P.C.A. 5 (Ni, Cr) (Fig. 9).
PCA 5 is the smallest of those delineated in this work and is interpreted as being a
mixed component partly reflecting local supply of lithogenous sediments of anomalous
composition. Examination of the probability plot of scores reveals three populations of samples
A - C. Increasing positive scores represent
increasing component scorer. The populations are :-
These are plotted on the score map Fig. 10.
A = Anomalously low scoring samples consisting mostly of carbonate
sands and pure carbonate oozes.
B = Population B contains most samples and the scores are normally
distributed.
C = Anomalously high scoring samples consisting of clays from S. Penrhyn
Basin and volcaniclastics from offshore Samoa together with isolated
samples in the Solomon Islands.
Penrhyn Basin clays in population C probably score anomalously high because of
Ni adsorption by manganese oxides. Samoan volcaniclastics score high not just because they
are enriched in Cr and Ni, but because they also have relatively high Mg concentrations.
Magnesium is an element that loads moderately on PCA 5 and thus enhances the scores.
The variations in the sample scores in population B are probably assignable to a number
of processes. These include Cr and Ni incorporation into volcaniclastics off New Zealand, Cr
- 9 -
and Ni in sediments derived from the weathering of nickeliferous laterites on New Caledonia
and possibly elsewhere, Cr and Ni enrichment in volcaniclastic sands within the Solomon Islands
chain (e.g. channel between San Jorge and Santa Isabel), and variable adsorption of Ni by
manganese oxides in basinal samples from throughout the area.
Pure carbonate deposits score anomalously low because of their near total lack of
Cr and Ni containing mineral phases.
SUB-SET STATISTICS
Principal Components Analysis reveals three major components explaining background
element variation in the Southwest Pacific. These have been interpreted as being a reflection
of volcaniclastic, biogenic and phyllosilicate components in the sediments. However,
inspection of the annotated PCA score probability plots allows a sub-division of the populations
into carbonate sand, carbonate ooze, volcaniclastic, phyllosilicate rich, continental
and mixed populations.
It has been recognised by several workers that division of terrestrial multielement
regional geochemical data into geological or physiographic sub groups may improve its
interpretation (Howorth,1973; Chapman,1976). It is evident therefore that a division of the
present data into the identified sediment populations may improve the reliability and contrast of
anomalies. Such a division has been done largely on the basis of the inflexion points on the
PCA probability plots, and sub-set statistics calculated. The population sub-sets are plotted as
different symbol shapes in Fig. 11, and summary robust statistics for the population sub-sets
are shown in Table 3.
RIDGE REGRESSION ANALYSIS
Linear or multilinear regression is a popular mineral exploration technique on land.
Ridge regression is a recently developed form of regression analysis which has a considerable
application for exploration purposes. it has been developed at Imperial College for land based
exploration programmes by Turner(1980) and Davis(1982), but has never previously been used
in the marine environment. The present study therefore comprises the first application of ridge
regression analysis to a marine data-set.
Regression analysis is a multivariate statistical technique whereby one predicts the
concentration of a dependent variable using several independent predictor variables. An
- 10 -
assumption of the ordinary multiple linear regression technique is that there is no correlation
between predictor elements. However, in geological situations, one often finds correlation
(multicolinearity) between predictors. In these cases ordinary multiple linear regression
Coefficients are numerically too large and/or of the wrong sign. Ridge regression analysis
produces more stable coefficients. To perform the analysis, an interactive program (RIDGE) written by R.J. Howorth, was used on the Imperial College computer system.
The aims of ridge regression analysis in the context of the present work are to develop
regression equations which predict the concentration of elements at sample sites based on the
operation of one or more of the background processes resolved using Principal Components
Analysis. The resulting residual values (difference between predicted and actual element
concentrations) will outline anomalous samples/regions in the study area. Ridge regression
equations have been developed for Cr, Ni, Mn, Fe, Cu, Zn and P.
The program (RIDGE) extracts background relationships between variables from a
correlation matrix. Therefore, the data should be conditioned to reduce the effects of outlying
samples and enhance background before calculating coefficients. Program structure dictated
that multivariate trimming and iterative editing of anomalous residuals using line printer residual
probability plots be used.
Fig. 12 shows a map of the study area with all the anomalous residual sites marked
based on ridge regression analysis of population sub-sets. As expected, use of population
sub-sets gave improved anomaly recognition over use of the total data set. Groups of anomalous
samples are boxed. However, no ridge regression analysis was performed on carbonate sands
because the sample population was too small to give meaningful results, and its content of
potentially economic minerals is likely to be very low.
In order to show actual anomalous residual values at sample sites, the selected areas
boxed in Fig. 12 are enlarged and represented as seperate figures (Figs 13 - 24) with residual histogrammes. All the sample sites with histogramme representations are anomalous. The bars
on the histogrammes show all the residual values for the sample site expressed as a percentage of their respective thresholds. The horizontal dotted lines are positive (+ 100%)
and negative (- 100%) thresholds. Residual values which break through the thresholds are
anomalous.
- 11 -
REGIONAL GROUPINGS OF ANOMALOUS RESIDUAL PLOTS FOR SUB-SETTED DATA
Area A, North Fiji Basin - Ellice Basin Borders (Fig. 13).
This area borders Area C to the south (Fig. 15) where many more anomalous samples
have been Identifled. The causes of the anomalies may be similar. Being far from land, localised detrital influences on the chemistry of the sediments can be discounted, and the
anomalies must be ascribable to non-detrital processes. The phosphorus anomaly in sample
302 Is associated with a local shallowing and is probably due to phosphate formation. The anomalies of Fe, Cu and Zn, a well known hydrothermal association, although not present
together in the samples, might reflect hydrothermal activity in the area. Such an interpretation
has been placed on similar anomalies in Area C, just to the south. However, independent evidence of hydrothermal activity in the area is currently not available, although von
Stackelberg et al (in press) have reported hydrothermal crusts from several stations in the same
general region (ca 14' 30'S, 177'E).
Area B, Manus Basin (Fig. 14).
Manus Basin sediments exhibit a number of anomalies, but manganese anomalies are
predominant. Cronan(1983) pointed to the presence of manganese anomalies in two Manus
Basin cores, which were considered to define a line of possible hydrothermal manganese
enrichment closely related to the position of a transform fault. The additional data presented in
Fig. 14 strongly support this supposition, and define the position of the likely hydrothermal
contribution more closely. Phosphorus anomalies in the east of the area can probably be
related to sea floor phosphorite enrichments, while the scattered zinc and iron anomalies
may be hydrothermal in origin, but cannot be adequately explained without further work.
However, it Is worth pointing out that the main Zn anomaly at the NE end of the New Britain is
close to the hydrothermal activity at Matupi Harbour (Ferguson and Lambert, 1972) and this
would lend support for it being hydrothermal in origin. A hydrothermal source of phosphorus (cf
Froelich et al,1982) can thus not be ruled out here.
Area C, North-West Fiji Basin (Fig. 15).
Area C falls to the NW of Fiji in an area where hydrothermal enrichments in the
sediments were postulated by Cronan(1983). Subsequent work in the area by von Stackelberg
et al (in press) has confirmed the occurrence of hydrothermal activity there by the finding of
- 12 -
hydrothermal crusts and sediments, and sulphides impregnating basalt. A range of elements of
hydrothermal affinities exhibit anomalies in many of the samples from Area C, iron, copper and
zinc being the most widespread. Coupled with these anomalies in one sample, but isolated in
others, are Ni anomalies indicative of a likely hydrogenous as well as hydrothermal component
to the sediments. Sporadic phosphorus anomalies could be related to phosphate occurrences,
although none appear to occur on seamounts where phosphates would be most likely to occur.
Reworking and downslope transport are possible mechanisms to redistribute phosphorus in
sediments (Summerhayes,1972). A small Cr anomaly In sample 401 could probably be related
to the inclusion of volcaniclastic material or its alteration products in the sediments.
Area D, Channel between Santa Isabel Island and San Jorge island, Solomon Islands (Fig. 16).
The salient feature of the sediments in area D are the large Ni anomalies that they
exhibit, reaching a maximum of 1623%. These can probably be readily explained in terms of
the close proximity of nickeliferous laterites on the adjacent land. The iron anomalies can
probably be traced to a similar source. While telling us nothing about mineralisation that we did
not already know about, the presence of such large Ni anomalies in the sediments in the
vicinity of known Ni mineralisation provides a good independent test of the validity of the
statistical methods used.
Also In these sediments are high background levels of Cr, Including one sample that
contains more than half a percent Cr. This can no doubt be related to known Cr mineralisation
in the north of San Jorge Island and Cr bearing heavy mineral sands on the beaches,
Area E., off New Caledonia (Fig. 17).
Like in Area D, the main anomaly in sediments of Area E is one of nickel, probably
derived from the nickeliferous laterites of New Caledonia. In three of the four anomalous
samples there is a Cr anomaly also, probably, In view of its strong ultrabasic association, of
similar affinities to the Ni.
Area F, Northern Lau Basin (Fig. 18).
Area F contains samples from the northern Lau Basin and the Tofua Trough. In the
Lau Basin, Cronan(1983) postulated the likely occurence of submarine hydrothermal activity on
the basis of univariate geochemical anomaly identification. What were considered t o be
hydrothermally metal enriched sediments were subsequently reported there by Cronan et
al(1984). More recently, the findings of von Stackelberg et al (in press) have confirmed these
- 13 -
conclusions by the recovery of a range of hydrothermal deposits in the Lau Basin, including
disseminated sulphides.
The Lau Basin anomalies identified in the present work are predominantly zinc ones
and may be assignable to the hydrothermal activity documented above. However, in one of the samples with anomalous zinc, Ni is anomalous also, and alone In one other sample,
Suggesting a hydrogenous component to the sediments as well as a hydrothermal one.
The Tofua Trough anomalies are primarily of copper, with one of Zn. None are large. Such anomalies are surprising in view of the absence of any known hydrothermal activity in
the area, and may be related to copper and/or copper-zinc mineralisation in the
volcaniclastics which make up the bulk of the sediment in the Tofua Trough (Cronan et al,
1984). Crawford (pers. comm. 1985) has reported disseminated chalcopyrite in volcanics
dredged from the vicinity of Epi, Vanuatu, (see Area H below) and similar disseminations may
be present in the volcanics from which the volcaniclastic sediments in the Tofua Trough are
derived. However, submarine hydrothermal metal enrichments in the area cannot be
entirely ruled out at the present stage of knowledge.
Area G, Solomon Islands, New Georgia Group (Fig. 19).
Sediments from off the New Georgia Group exhibit a number of anomalies, of which
those of nickel are the most prominent. This is interesting, as no nickeliferous laterites are
shown on the Mineral Occurrences Map of the Solomon Islands as being present in the New
Georgia Group. The largest Ni anomaly is in sample 1537, taken in the channel between
New Georgia and Kolombangara. Possibly unrecorded Ni mineralisation exists in the near
coastal area of one or other of these islands. Other prominent Ni anomalies occur in sediments
between Kolombangara and Vella Lavella, to the NW of Vella Lavella, and south of Vangunu.
Scattered Cu and Zn anomalies could be related to known Cu, Zn minerallsation within the
Group, while small phosphorus anomalies are probably related to phosphate bearing material.
Area H, Epi, Vanuatu (Fig,20).
Sediments associated with the submarine volcanoes off Epi and the clearly discernable
hydrothermal influences on them have been described by Exon and Cronan(1983). The data
presented in Fig. 20 highlight the anomalous iron contents of these sediments, which have
been subjected to further work on the 1984 S.P.Lee Tripartite programme cruises (Greene
et al, in prep). One sample exhibits a P anomaly, probably assignable to phosphate bearing
material.
- 14 -
Further analyses of samples from off Epi which were collected during the 1984 Tripartite
programme (not included in the statistical analysis) have confirmed the hydrothermal iron
enrichments in them (Greene et al, in prep.). Maximum iron contents reach about 20% in the
sediments.
enrichment over the average Cu contents of the sediments off Epi of only 0.02%. The
possibility that this may represent distal exhalative copper sulphide mineralisation associated
with the submarine volcanic activity cannot be ruled out. However, Crawford (pers. comm 1985) has reported disseminated chalcopyrite in the volcanic rocks dredged off Epi, and the
copper enrichment in the sediments reported here may result from the inclusion of the
degradation products of some of this material in the sediments described here.
Area K, Southern Harve Trough - Kermadec Ridge (Fig. 21).
One sample containing 11% Fe exhibits almost 0.1% Cu, a considerable
A more or less east-west transect to the south of the Havre Trough exhibits zinc and
zinc-iron anomalies in the sediments. The eastern end of the transect is close to the volcanically
active Kermadec Islands. A zinc anomaly also occurs in sediments to the east of Esperance
Rock. Based on bulk sediment chemistry, Cronan et al(1984) could find no evidence of hydro-
thermal inputs in the Havre Trough. If such a situation prevails in Area K, the anomalies
recorded above most probably reflect the westward transport of volcaniclastic material from
the Kermadec Islands.
Area L, Raukumara Plain (Fig. 22).
Being so close t o New Zealand, Raukumara Plain sediments will be considerably
affected by continental runoff. In the principal components analysis described
earlier in this work, these sediments group together as a distinct population.
Because the precise nature of the continental influence on the deposits cannot
be established wi th certainty without further work, i t i s difficult t o assign the
manganese and nickel anomalies in them t o any particular source.
Area N, Samoa (Fig.23).
Off Samoa are considerable Ti enrichments in the sediments ranging up to 1.96% Ti (Fig.23). These samples are mainly in deep water (over 2000 m) and range down
to abyssal depths of over 4000m. They are variable lithology , but most have a visibily
- 15 -
identifiable volcanic component almost certainly derived from the adjacent islands, and
the Ti enrichments are probably associated with this phase.
Two of the deep water Ti enriched samples from off the western end of the Group
exhibit large Ni anomalies (Fig. 23). These samples are particularly enriched in volcaniclastic
material and the Ni anomalies may be associated with this phase. However, elsewhere
in the region, such as off the Solomons and New Caledonia, similar anomalies appear
to be Influenced by known nickeliferous laterites. A relationship between the Ni anomalies
off Samoa and possible nickeliferous lateritisation on land is worth exploring further.
Area O, S.E. of Fiji (Fig. 24).
South-east of Fiji are a number of sediments exhibiting manganese anomalies.
The origin of these anomalies is difficult to explain because the area in question is not
one of known or even suspected sea floor volcanic or hydrothermal activity. Furthermore,
known manganese mineralisation on land in Fiji is largely confined to the western and
northern parts of the island. Further work is needed to establish the source of these
anomalies.
- 16 -
SUMMARY AND CONCLUSIONS
Multivariate statistical techniques have been applied to a large regional geochemical
data set of over 600 samples from the SW Pacific which have been analysed for 17 elements.
The aim has been to identify chemically anomalous sediments reflecting either sea floor
mineralisation and/or minerallsation on adjacent land areas.
Principal Components Analysis of a robust correlation matrix for Box-Cox transformed
data has revealed five groups of elements which between them account for more than
70% of the variance of the data. The element groups can be interpreted in terms of
geological background processes and are as follows :-
PCA 1 (V, Mg, Fe, Ti, Co, Al) - Scores represent the amount of volcaniclastic
material in the sediments.
PCA 2 (Sr, Ca) - Scores represent the amount of biogenic calcium carbonate
in the sediments.
PCA 3 (Li, K, Be, Ti ,Al) - Scores represent the amount of phyllosilicate (clay
minerals) in the sediments.
PCA 4 (Mn, Cu, P, Zn, Co, Ni) - Scores represent the amount of manganese
oxide and associated coprecipitated/absorbed elements in the sediments.
PCA 5 (Ni,Cr) - Scores represent mixed processes of partly localised element
supply.
The total data were subsetted on the basis of the Principal Components Analysis
Into carbonate sands, carbonate oozes, volcaniclastic rich sediments, phyllosilicate, rich
sediments, continental sediments and a mixed sediment population.
Subsetted data have been subjected to Ridge Regression Analysis which is a
new multivariate technique which takes into account multicolinearity. The aim of Ridge
Regression on the SW Pacific data set has been to develop equations which predict the
concentration of a dependant element based on the operation of one or more of the
background processes resolved using Principal Components Analysis. Probability plots
- 17 -
of residuals can be used to locate anomalies. Several such anomalies and anomalous areas have been identified and described individually, and are as follows :-
a) Fe, Cu, Zn and P anomalies in the northern part of the North Fiji Basin possibly reflecting hydrothermal activity and phosphatic sedimentation.
b) Mn, Fe, Zn and P anomalies in the Manus Basin possibly reflecting hydrothermal activity and phosphatic sedimentation.
c) Fe, Cu, Zn, Ni, P and Cr anomalies in the North West Fiji Basin possibly reflecting hydrothermal activity, phosphatic sedimentation and volcaniclastic input.
d) High background levels of Cr together with Ni and Fe anomalies in the
channel between Santa Isabel and San Jorge Islands, Solomon islands, possibly reflecting
detrital chromite mineralisation in the sediments and/or onshore Ni and Cr mineralisation.
e) Ni anomalies off New Caledonia possibly reflecting derivation from the
nickeliferous laterites onshore.
f) Zn and subsidiary Ni anomalies in the northern Lau Basin possibly reflecting
hydrothermal activity with some hydrogenous input.
g) Cu and Zn anomalies in the Tofua Trough, Tonga, possibly reflecting
volcaniclastic or hydrothermal inputs.
h) Ni, Cu, Zn, Fe and P anomalies around the New Georgia Group, Solomon
Islands, possibly reflecting mixed onshore mineralisation, and phosphatic sedimentation.
i) Fe anomalies off Epi, Vanuatu, related to hydrothermal activity together with
Cu enrichment possibly related to the hydrothermal activity or sedimentation of Cu bearing
volcaniclastics.
j) Zn and Fe anomalies in the southern Harve Trough possibly reflecting
volcaniclastic inputs from the Kermadec Islands.
k) Mn and Ni anomalies on the Raukumara Plain off New Zealand, possibly
reflecting a continental input there.
l) Ti enrichments and Ni anomalies off Samoa possibly reflecting volcaniclastic
inputs and/or onshore mineralisation.
m) Mn anomalies off S.E.Fiji of uncertain orogin.
- 18 -
REFERENCES
Champman,(1976) To be supplied.
Crawford, T.(1985) Univ of Tsamanla, pers comm.
Cronan, D.S.(1983) Metalliferous sediments of the CCOP/SOPAC region of the
southwestern Pacific, with particular reference to geochemical exploration
for the deposits. CCOP/SOPAC Tech. Bull 4. Suva.
Cronan. D.S. (in press) Regional geochemistry of sediments from the S.W. Pacific
in (Cronan, D.S. ed) Sedimentation and Mineral Deposits in the Southwestern
Pacific Ocean. Academic Press. London.
Cronan, D.S. & Thompson, B.(1978) Regional geochemical reconnaissance survey
for submarine metalliferous sediments in the S.W. Pacific. Trans. Instn. Min.
Metall. B87 p 87-89.
Cronan, D.S., Moorby, S.A., Glasby, G.P., Knedler, K., Thompson, J. & Hodkinson,
R.(1984) Hydrothermal and volcaniclastic sedimentation on the Tonga-Kermadec
Ridge and in its adjacent marginal basins. In (Kokelaar B.P. & Howells, M.F.
eds) Marginal Basin Geology. Geol.Soc.Lond.Spec.Publ 16, p 137-149.
Cullen, D.J.(in press) Submarine phosphatic sediments of the S.W. Pacific. In (Cronan, D.
ed) Sedimentation and Mineral Deposits in the Southwestern Pacific Ocean,
Academic Press, London.
Davis (1982) To be suppiled.
Exon, N.F. & Cronan, D.S.(1983) Submarine hydrothermal iron deposits off Epi, Vanuatu.
Mar.Geol. 52, p M43-52.
Ferguson, J. & Lambert, I.B.(1972) Volcanic exhalations and metal enrichments at Matupi
Harbour, New Britain, T.P.N.G. Econ.Geol. 67 p 25-37.
- 19 -
Froelich, M., Bender, M.L., & Heath, G.R.(1982) Phosphorus accumulation rates In
metalliferous sediments on the East Pacific Rise, Earth Planet Sci. Lett., 34, p
351-359.
Greene, G. et (in prep) Joint Cruise Rept., S.P. Lee 1984 Vanuatu - Solomons Cruise.
Hodkinson, R, et al (in prep) To be supplied.
Howorth,R.(1973) To be supplied.
Summerhayes,C.P.(1972) To be supplied.
Turner, (1980) To be supplied.
von Stackelberg, U and the shipboard scientific party (in press) Hydrothermal sulfide
deposits in Back-Arc spreading centers in the southwest Pacific.