Quantitative image analysys: Application to planktonic foraminiferal paleoecology and evolution

QUANTITATIVE

IMAGE ANALYSYS

FORAMINIFERAL

: APPLICATION

PALEOECOLOGY

TO PLANKTONIC

AND EVOLUTION

by NANCY HEALY-WILLIAMS *

ABSTRACT

RI~SUMI~

A complete understanding of the taxonomic and paleoenvironmental information contained in the morphology of foraminifera depends heavily on quantitative methods to describe phenotypic and ecophenotypic variation. Automated image analysis systems now make it possible to rapidly and precisely describe the two dimensional shape of almost any microfossil. Our results using Fourier shape analysis with Fourier series in closed form demonstrate the feasibility of using these shape descriptors as a quantitative biometric tool. Analysis of the planktonic foraminiferal species, Neogloboquadrina pachyderma and Globorptalia inflata, indicate that the test outlines of these species reflect various physical parameters of the water column.

Une compr6hension compl6te des informations taxonomiques et pal6oenvironnementales fournies par la morphologie des foraminif~res d~pend en grande partie des m~thodes quantitatives pour d6crire la variation ph~notypique et ~coph~notypique. Maintenant, les syst~mes d'analyse automatique d'image permettent de d6crire rapidement et pr6cis6ment la forme en deux dimensions de presque chaque microfossile. Nos r6sultats, utilisant l'anaiyse de forme selon Fourier, avec les s6ries de Fourier en forme ferm6e, d~montrent la possibilit6 d'utiliser ces descripteurs de forme comme un instrument biom~trique et quantitatif. L'analyse des foraminif~res planctoniques Neogloboquadrina pachyderma et Globorotalia inflata indique que les plans de test de ces esl~ces refl~tent divers param&res physiques de la colonne d'eau.

FORAMINIFERAL MORPHOMETRICS - BACKGROUND The morphological characteristics of fossil foraminiferal tests are the basis for making phylogenetic, taxonomic, as well as paleoecological inferences about fossil species. The tests of foraminifera are notoriously plastic, exhibiting a wide range of morphological traits, some of which are an expression of varying amounts of optimal and stressful environmental conditions (Kennet, 1976). Early on, Parker (1962) recognized the difficulties in differentiating phenotypic from ecophenotypic variation in fossil

foraminifera. By increasing our knowledge of the factors which influence the present day morphology of living and recent species, it may be possible to quantitatively model the variation in that species through time. Recent attempts to use biometric analysis to understand the morphological variability observed in Recent and Quaternary fossil foraminifera involved exacting linear measurements of test dimensions (Malmgren and Kennett, 1976, 1977, 1981). These authors made

* Department of Geology, University of South Carolina, Columbia, SC 29208, USA.

Geobios, M~m. special 8

p. 425-432, 2 fig., 1 tabl.

Lyon, 1984

- - 426 --

significant improvements in the applicability of biometrics to foraminiferal populations through the use of multivariate statistical analysis of biometric, faunal and oceanographic data. Even in these quantitative approaches, however, difficulties arise in accurately describing the overall shape of many species using univariate or a small number of predetermined multivariate measurements. The number of linear test measurements becomes increasingly large in order to quantify the important characteristics which determine the overall test shape, i.e., chamber number, chamber size, chamber arrangement, sutural characteristics, rate of chamber expansion, kummerform/ normal form, etc. Whether these dimensions have morphological or environmental significance is not known and, as a result, dimensions are often considered which do not always provide important morphological or environmental information.

We believe that Fourier shape analysis via Fourier series in closed form is an objective and unambiguous biometric method for determining which shape components are important in morphological variations within planktonic foraminiferal populations (HealyWilliams and Williams, 1981 ; Healy-Williams, in press;Scott, 1975, 1976, 1980, 1983). Scott (1974, 1980) discussed the importance of using quantitative biometric techniques to describe the variability of foraminiferal tests. The primary goal of such biometric studies should be to better understand the factors controlling or influencing morphological variability. Recent developments in the Fourier shape technique (Ehrlich and Weinberg, 1970;Full and Ehrlich, 1982) in addition to the continuing advances in automated image analysis, make this an attractive method for studying morphological variation in foraminifera and other fossil groups.

FOURIER SHAPE ANALYSIS

Fourier shape analysis provides a rapid, objective and quantitative means of resolving the important characteristics of test shape with no a priori designation of what constitutes an important morphological parameter. One of the major advantages of Fourier shape analysis over most univariate and multivariate biometric techniques is that this approach is not based on the ratios of a limited number of axes which are assumed to be of equal taxonomic importance in all species or subspecies. Nor does Fourier shape analysis require that the investigator define homologous points prior to morphometric analysis which may thereby force preconceived opinions regarding the form onto the shape analysis. Another advantage is that the variability of the entire data set is preserved for analysis and not reduced to mean shape. By viewing this variability in frequency distribution diagrams, variation in population structure can be discerned. Closed from Fourier series is a mathematical expression of cosines used to describe any generally convex two dimensional outline to a specified degree of precision. A Fourier series segments the shape into components of fixed geometry (figure eight, trefoil, quadrefoil, etc.) which when combined can produce the empirical outline. The shape of the specimen is represented by a series of harmonics in the form :$5 R ( O ) = RO

+

n~l

Rncos(nO-~I~n)

where Rn represents the amplitude of the <>harmonic, and the remainder concerns the phase angle of that harmonic. Each harmonic amplitude represents the relative contribution to the empiric shape of a characteristic shape component. For example, the amplitude of the second harmonic represents the contribution to the empiric shape of a <~figure eight >>and is thus a measure of elongateness. The third represents a trefoil, the fourth a quadrefoil, etc. In general, the <~harmonic amplitude represents the shape contribution of a ~~ clover. The observed shape is thus partitioned into a series where gross shape (elongateness, triangularity, etc.) is measured by the harmonic amplitudes of the lower harmonic orders, and increasingly fine-scale surface sculpture is measured at higher orders. In our studies we have primarily used harmonic amplitude values to delineate shape changes. In order to reconstruct the original shape, both phase angle (the angular relationship of each harmonic) and amplitude value are required. In all of the species we have analyzed to date, we have found that phase angles are relatively invariant within a singular taxon. For example, analysis of phase angle for the third harmonic in specimens of Globorotalia truncatulinoides revealed that the scatter about the mean was + 5 ~ to 10~ which is insignificant in comparison to the range in amplitude values for the same specimens, 0.02 to

-- 427 -0.195. Similar results have been obtained for other harmonics and other species. These results indicate that, while the phase angle of a given harmonic does not vary, the important morphological information is contained within the harmonic amplitude values for all species we have studied to date. Phase may play a role in some taxa but this has yet to be determined. Another important aspect of this consistent harmonic orientation is that the same shape component is described for each specimen thereby insuring homology. In our image analysis system, the periphery of each test is digitized into 200-1000 edge points and submitted to the Fourier program which : a) converts the cartesian coordinates to polar coordinates ; b) calculates the center of gravity of the specimen ; c) relocates the grain origin at the center of gravity ; d) calculates the Fourier coefficients (amplitudes and phase angles) for the first twenty-four harmonics using a Fourier series in closed form (Ehrlich and Weinberg, 1970 ; Full and Ehrlich, 1982). These harmonics are the shape descriptors which represent specific contributions to the total shape (harmonic amplitude). Each harmonic can be used as a morphologic variable, and the test shape can be described as precisely as desired, depending on the number of harmonics measured. The amplitude spectrum for a collection of foram tests comprises the primary data field used for ensuing analysis. Each foraminiferal test in a sample contributes an amplitude value for every term of the series ; the twenty-four terms calculated produce an N x 24 data matrix (N = the number of tests) for any one sample. Accordingly, for each term of the series, shape frequency diagrams can be calculated by plotting amplitude values as a function of frequency of occurrence. Thus, as each test can be described by twenty- four harmonic amplitude values, each sample can be described by twenty-four shape frequency

distributions. With the data in this form, samples may be statistically compared harmonic by harmonic. The overall shape of these frequency distributions carries information regarding the shape nature of the sample. In our work, 50-100 individual foraminiferal tests per sample are mounted onto standard 60 square micropaleo slides with a water soluble mounting medium. We have determined that an orientation error of • 10~ does not significantly affect the measurements (Pharr & alii, submitted). We have also found that 50-100 specimens from a quantitative split were sufficient to derive statistically significant shape data. The automated particle shape analysis system at the University of South Carolina (Fico, 1980) interfaces a video digitizer with a microprocessor enabling the automated digitization of any shape outline. Once the foraminifera are correctly mounted, approximately 200 specimens can be digitized per hour. A television camera is mounted onto a standard reflected light binocular microscope. The analog signal from the video scanner is digitized (by a Cat 100S video digitizer) into a grided array of points to which each is assigned a gray scale value. The edge-finding algorithm of the microprocessor uses a ~t thresholding, procedure based on a near-white shade of gray in order to locate the foraminiferal boundary. Communication between the microprocessor/digitizer and the operator is through a standard ASCII keyboard connected to a CRT and a 10 inch black and white display monitor. The display monitor shows the original analog signal from the camera plus the digital representation of the image. The operator can signal the microprocessor to accept or reject the digitized shape. Once the operator accepts the digitization, the microprocessor passes the data to diskettes which are then submitted ot the shape analysis programs described above.

APPLICATION OF FOURIER SHAPE ANALYSIS TO FORAMINIFERAL MORPHOMETRICS The test of any morphometric technique is in the results obtained from that technique. During the last two years we have had great success in using Fourier series to delineate : 1) shape families in fossil speci-

mens of planktonic foraminifera and 2) shape changes in living planktonic material. Briefly described herein are results from two of these studies.

SHAPE CHANGES IN FOSSIL FORAMINIFERAL POPULATIONS

Neogloboquadrina pachyderma (EHRENBERG) has long been one of the most important foraminiferal species in paleoceanographic studies of Quaternary and late Cenozoic marine sediments. Numerous stu-

dies of this species have documented that coiling direction and chamber number in the final whorl (4, 4 , 5) have important relationship with high latitude water mass properties (Ericson, 1959 ; Kennett, 1968

--

428

--

-

429

-

Herb, 1968 ; Malmgren and Kennett, 1972 ; Vella, 1974 ; among others). This variability also poses problems in the taxonomy of the species, however. In an attempt to objectively quantify this morphologic variability, we performed Fourier shape analysis on both sinistral and dextral forms with 4, 4 ~ a n d 5 chambers in the final whorl and regarded these forms as phenotypic variations of the same species (Bandy, 1972 ; Kennett, 1968 ; Malmgren and Kennett, 1972 ; Srinivasan and Kennett, 1975). Approximately 70 N. pachyderrna per sample were analyzed from 15 core top samples (N > 1000 total) located in the southern Indian Ocean (38~ - 53~ All specimens were removed from split portions of a 125~m-180~m size fraction. The purpose of our study was twofold : 1) to test the ability of Fourier series to delineate the

-

-

major morphotypes of N. pachyderma ; 2) to quantify these morphotypes and relate them, if possible, to hydrographic parameters of the overlying water masses. Fourier analysis was able to precisely determine that four harmonics are required to describe the shape variability in N. pachyderma and that the significant shape components are related to the chamber arrangement (Figure I). As shown in Figure I, specimens with high harmonic 3 values tend to be 4 chambered but with a triangular shape usually resulting from an inflated final chamber. Typical quadrate specimens have high harmonic 4 values. With increasing shape complexity, the higher order harmonics, 5 and 6, are able to resolve the specimens which have 4 ~ and 4 to 5 chambers respectively.

TABLE 1

COILING (% S i n i s t r a l )

~

=

SALINITY

DISSOLVED

0 M/~

o2rfl,/n

HARMONIC 3

-0.52

+0.44

+0.49

-0.44

HARMONIC 4

+O.O1

+0.10

-0.04

+0.04

HARMONIC 5

-0.72

+0.68

+0.61

-0.57

HARMONIC 6

-0.74

+0.72

+0.59

-0.58

level of significance

Fig. 1 - -

TEMPERATURE

o = 0.01

Scanning electron micrographs of representative shapes of Neogloboquadrina pachyderma as described by harmonics 3 (3a-e) ; ,4 (4a-d) ; 5 (Sa-e) ; 6 ( 6 a - 0 . M a g n i f i c a t i o n : x 400.

Photos au microscope ~lectronique h balayage de formes representatives de Neogloboquadrina pachyderma d6erites par les harmoniques. Grossissement : x 400.

-

MOCNESS MEAN

o

08 .oF

HARMONIC 2

.,o

.q,

.),z

430

-

-

-

G. inf/ofo

38

MEAN HARMONIC 3

.q4

o5

MEAN HARMONIC4

.o,6 .oT

07

h

WATER ,

TEMPERATURE

~,2 ,

~4 ,

~6

*C ,

la

,

I -

25

E

50 i t

I00

125

150

J

175

MOCNESS MEAN .09

HARMONIC

.lO

JI

.12

;~

62

MEAN .1~

G. i n f / o t a

HARMONIC3

.06

.07

.08

MEAN HARMONIC4 .05

.04

.05

WATER TEMPERATURE

,/

.6.. 8.. i.o. i z

0 IOO.

200/

300. /

G,

)

\

J

/

/ /

500'

iI

\

//

600. 700' 800' 900'

,> /

,.4

II

/

/

400,

)C

/

/,/, i

)

i

I000.

Fig. 2 -

Fourier harmonic amplitude values for specimens of Globorotalia inflata from MOC 38 (top) and MOC 62 (bottom) with temperature profiles of the water column. (t I >>indicates inferred temperature value of G. inflata from ~ 180 analysis (Fairbanks, Lament Doherty Geological Observatory). Valeurs de l'amplitude harmonique de Fourier pour les sp6cimens de Globorotalia inflata de MOC 38 (haut) et MOC 62 (bas) avec des profils de la temp6rature de la colonne d'eau. << I >>indique la valeur de ta temp6rature de G. inftata d6duite de l'analyse de 180 (Falrbanks, Lament Doherty Geological Observatory).

The preliminary results of this study indicate that the integration of several shape components is required to fully understand the morphological changes in N. pachyderma and the relatioship between the different forms. A linear correlation analysis (Table 1) indicates that the shapes defined by harmonics 3 and 4 are not related to environmental parameters and possibly represent genotypic information. However,

harmonics 5 and 6 both have correlations with the hydrographic parameters, indicating that the change to 4"/= or 5 chambers may be an ecophenotypic response. We believe that by quantifying these shape components it will be possible to enhance the utilization of N. pachyderma in paleoenvironmental studies of high latitude sediments.

LIVING PLANKTON MATERIAL Analysis of living plankton material provides us with information concerning the effect of biotic and abiotic factors on test morphology and greatly enhances biometric studies of fossil foraminifera. Material from the MOCNESS tows (multiple opening/closing net environmental sensing system, Wiebe & alii, 1976) is ideal for such studies because of the discrete sampling interval afforded by the system. Specimens of

Globorotalia inflata were analyzed from two tow sites in the North Atlantic : 1) MOC 38, slope water ; 2) MOC 62, Cold Core ring. The specimens for this study were provided by R. Fairbanks (LDGO) who also performed the isotopic analyses on the specimens. The maximum entropy method (Full & alii, 1981) determined that harmonics 2 (elongateness), 3

--

(triangularity) and 4 (quadrateness) contained the most shape information. For MOC 38 tows from 0150m (Figure 2), mean values for harmonic 2, 3 and 4 all indicate a sharp change in shape at the level of thermocline. The majority of specimens beltJw 100m are less elongate, less quadrate and more triangular. For tow array MOC 62, (from 150 to 925m), where the thermocline is less well developed harmonic 2 values also reflect a gradual shape change (Figure 2). Harmonic 3 has no significant trends with depth and harmonic 4 reflects only a minor decrease in quadrateness with depth. In both tow arrays G. inflata shows

431

--

no distinct trend in size with tow depth. The observed decreases in harmonic 2 values for living G. inflata from the North Atlantic appear to be proportional to the strength of the thermocline. The results from the first Fourier biometric analyses of living plankton indicate that shape changes are occurring in the water column and that these changes appear to be sensitive to the temperature values within the water column. We believe that the results from this species and other living species may increase our understanding of planktonic foraminiferal morphology and their application to paleoceanographic problems.

VI - - SUMMARY As stated previously, the primary goal of biometric studies should be to (1) better understand the factors controlling or influencing morphologic variability and (2) enhance our ability to study the paieoecology of foraminiferal populations as they reflect paleoceanographic and paleoclimatic changes. We believe that Fourier shape analysis provides an objective and unambiguous biometric method for determining which shape components are important in morphol9gical variation within planktonic foraminiferal populations. We make no claim that Fourier shape analysis is the ultimate biometric tool but only a means of obtaining rapid and precise morphometric

data. Afterall, every investigator does not bring to a study the same questions and therefore, methodologies will need to differ. Fourier descriptors can be used successfully to provide a deeper understanding into the nature of foraminiferal populations and their response to ecologic variables. This success has arisen not only from Fourier shape analysis via Fourier series in closed form but also from a rapid image acquisition and analysis system described in this paper. The application of image analysis to foraminiferal morphology is an important step towards understanding the ecology and paleoecology of these organisms.

Acknowlegements

The author thanks Richard Fairbanks of Lamont Doherty Geological Observatory for access to the MOCNESS specimens and to his unpublished isotope data. Thanks also go to Robert Ehrlich and Douglas Williams for their helpful discussion of the research and the application of image analysis

to foraminiferal biometrics. Thanks also go to Jean-Robert Hedouin for his assistance with the French translations. This research was supported by National Science Foundation grant OCE-8110167.

REFERENCES BANDY O.L. (1972) - Origin and development of Globorotalia (Turborotalia) pachyderma (EHRENBERG).Micropaleontology, 18 (3), 294-318. EHRLICH R. & WEINBERGB. (1970) - An exact method for characterization of grain shape. J. Sedim. PetroL, 40 (1), 475-484.

ERICSON D.B. (1959) - Coiling direction of Globigerina pachyderma as a climatic index. Science, 130 (3369), 219220. FICO C. (1980) - Automated particle shape analysis : Development of a microprocessor controlled image analysis system. Unpublished Masters Thesis, University of South Carolina.

- - 432 - -

FULL W.E & EHRLICH R. (1982) - Some approaches for location of centroids of quartz grain outlines to increase homology between Fourier amplitude spectra. Mathematical Geology, 14, 43-55. HEALY-WILLIAMS N. & WILLIAMS D.F. (1981) - Fourier analysis of test shape of planktonic foraminifera. Nature, 289, 485-487.

lineage ; DSDP Site 284, South-west Pacific. Paleobiology, 7 (2), 230-240. PARKER F.L. (1962) - Planktonic foraminiferal species in Pacific sediments. Micropaleontology, 8, 219-254. PHARR E.B. Jr., WILLIAMS D.F. & HEALY-WILLIAMS N Diachronous niche expansion and morphological adaptation by Globorotalia truncatulinoides in the late Quaternary. Science, (submitted).

HEALY-WILLIAMS N. (1983) - Fourier shape analysis of Globorotalia truncatulinoides from late Quaternary sediments in the southern Indian Ocean. Marine Micropaleontology, (in press).

SCOTT G.H. (1974) - Biometry of the forarniniferal shell. in : R.H. Hedley and C.G. Adams (ed.), ForaminOrera, vol. 1, 55-152.

HERB R. (1968) - Recem planktonic foraminifera from sediments of the Drake Passage, Southern Ocean. Eclog. Geol. Heir., 6, 467-480.

SCOTT G.H. (1975) - Variation in Gioborotalia miozea (Foraminifera) from the New Zealand Neogene. New Zealand J. GeoL Geoph., 18, 865-880.

KENNETT J.P. (1968) - Latitudinal variation in Globigerina pachyderma (EHRENBERG) in surface sediments of the southwest Pacific Ocean. Micropaleontology, 14, 305318.

SCOTT G.H. (1976) - Variation in Globorotaliapuncticulata sphericomiozea from New Zealand (Foraminifera, Neogene). New Zealand J. GeoL Geoph., 19, 855-869.

KENNETT J.P. (1976) - Phenotypic variation in some Recent and late Cenozoic planktonic forantinifera. In: R.H. Hedley and C.G. Adams (ed.) ForaminQ'era, vol. 2, Academic Press, London, 111-170. MALMGREN B.A. & KENNETT J.P. (1972) - Biometric analysis of phenotypic variation : Globigerina pachyderma (EHRENBERG) in the South Pacific Ocean. Micropaleontoiogy, 18, 241-248. MALMGREN B.A. & KENNETT J.P. (1976) - Biometric analysis of phenotypic variation in Recent Globigerina bulloides d'ORBtGNY in the southern Indian Ocean. Marine Micropaleontology, 1, 3-25. MALMGREN B.A. & KENNETT J.P. (1977) - Biometric differentiation between Globigerina bulloides andGlobigerinafaiconensis in the southern Indian Ocean. J. Forarain(feral Res., 7 (2), 130-148. MALMGRI-N B.A. & KENNETT J.P. (1981) - Phyletic gradualism in a late Cenozoic planktonic foraminiferal

SCOTT G.H. (1980) - The value o f outline processing in the biometry and systematics of fossils. Paleontology, 23, 757-768. SCOTT G.H. (1983) - Biostratigraphy and histories of Upper Miocene-Pliocene Globorotalia, South Atlantic and Southwest Pacific. Marine Micropaleontology, 7, 369383. SRINIVASAN M.S. & KENNETT J.P. (1975) - Paleoceanographically controlled ultrastructural variation in Neogioboquadrina pachyderma (EHRENBERG) at DSDP Site 284, South Pacific. In ; Andrews J.E., Parkham G. & alii, Initial Reports Deep Sea DrtTling Project, vol. 30, Washington, U.S. Gov. Print. Off., 709-721. VELLA P. (1974) - Coiling ratios of Neogioboquadrina pachyderma (EHRENBERG):Variations in different soze fractions. GeoL Soc. America, Bull. 85, 1421-1424. WlEBE P.H., BURT K.H., BOYD S.H. & MORTON A.W. (1976) - A multiple opening/closing net and environmental system for sampling zooplankton. J. Marine Res., 34 (3) : 313-326.