A Numerical Taxonomic Study of Hawaiian Reef Corals! DENNIS POWERS2

ABSTRACT: Sixty characters were measured and used in multivariate statistical programs to study the systematics of 20 species of Hawaiian corals. Correlation and distance phenograms and a computer-generated, three-dimensional model were used to develop phenetic rankings of species groups at levels corresponding to the taxonomic categories of genus, family, and, provisionally, suborders.

THE SCLERACTINIA is an order of the Coelen­ All observations on soft, live parts were terata. These stony corals have one of the better made either on living corals at the University invertebrate fossil records. Numerous attempts of Hawaii Institute of Marine Biology or from have been made to classify them by various colored photographs taken there. criteria, ranging from gross anatomy to physiol­ Twenty species, referred to as operational ogy. In the literature, the different classifications taxonomic units (O'Tl.I's) (Sokal and Sneath, seem to be almost as numerous as the species 1963), were collected, and 60 characters (listed being classified (Hyman, 1940; Bourradaile and below) were measured for each. Qualitative Potts, 1963; Dana, 1890; Duerden, 1904; Hick­ characters were rated 1.0 if the character was son, 192 4; Fowler, 1885-1889; Ogilvie, 1896; present and 0.0 if the character was lacking. All Robinson, 1923; Vaughan and W ells, 1943). quantitative characters were measured between Consequently, there is no general agreement on 50 and 100 times. The mean was used as the the systematics of the order. The controversies value of the character measured. If there was arise because various authors have chosen differ­ greater variation within an OTU than between ent characters to emphasize in their classifica­ the means of different O'TU's, the character tions. The effects of such emphases can be was discarded as insignificant. Table 1 lists the minimized by measuring a large number of species by the randomly assigned code numbers, kinds of characters and treating them in statisti­ which were used to reduce subjectivity when cal programs that preclude inadvertent weight­ recording the values of the characters. Ideally, ings. This procedure was used to classify 20 the author should neither assign the code num­ species of scleractinian corals and to compare bers nor see the code key until he has finished the results with prior classifications in an effort the analysis. to help stabilize the taxonomy (systematics) of CHARACTERS MEASURED the order. 1. Septa pr esent or absent . 2. If septa present, degre e of septal development. MATERIALS AND METHODS 3. Septa margins: beaded, dentate or smooth . 4. Are septa fenestrate ? All corals studied were collected from the 5. Secondary septa present or absent. waters around the island of Oahu, Hawaii; most 6. T erti ary septa present or absent. were taken from Kaneohe Bay in the summer 7. Qu aternary septa present or absent . 8. Are directive mesenteries pres ent or absent ? of 1967. All species except Cyphastrea ocellina 9. Are septa costate? and Pocillopor« ligttlata were found in samples 10. Corall ite walls pres ent or absent. from several different reefs. Tubastrea anrea was 11. Zooxanthellae present or absent. obtained from a patch reef within the bay and 12. Exot heca present or absent. 13. Endotheca present or absent . from a sunken ledge outside the bay. 14. Peritheca present or absent. 15. Pali present or absent. 1 This work was supported by the U . S. Atomic 16. Snapticulae present or absent. Energy Commission. Manuscript received June 6, 1969. 17. Collines present or absent. 2 Di vision of Biological and Medical Research, Ar­ 18. If septa are fenestrate are they laminar in later gonne N ational Laboratory, Argonne , Ill inois. stages? 180 N umerical Taxonomy of Reef Corals- PowERS 181

TABLE 50. Average bathymetric range. 51. Average height of calyx. CODE NUMBER GENUS AND SPECIES 52. Average height of dentitions or beads. 53. D istance between calices. 1 Pocillopora ligulata 54. Average annu al increase of length (Edmondson, 2 M ontipora oerrilli 1929 ; Tam ura and Hada, 1932; Stephenson and 3 Psammoeora stellata Stephenson, 1933). 4 Porites evermanni 55. Average percent of weight increase (Ref. same 5 Pauona ex planulata as above) . 6 Fun gia patella 56. Glucose-6-phosphate dehydrogenase activity Cyphastrea oee/lina 7 (Powers, Lenhoff, and Leone, 1968) . 8 Pocillopora meandrina 57. 6-Phosphogluconate dehydrogenase activity 9 M ontipora verrucose (P owers, Lenhoff, and Leone, 1968). 10 Porites compressa 11 Fungia seutaria 58. Ratio between length and wid th of calyx. 12 L eptastrea bottae 59. Number of calices per unit area. 13 T ubastrea aurea 60. Ratio of height to width of calyx. 14 Pocillopor« damicornis 15 M ontipora patula 16 Psammoeora oerrilli After the values of the characters were 17 Porites lobata recorded for each OTV, these data were 18 Pavona uarians punched onto IBM cards for processing in a 19 Fun gia [ragilis G.E. 625 digital computer . The group of pro­ 20 Leptastrea purpurea grams used were from NT-SYS Numerical Taxonomy System of Mttltivariate Statistical 19. If peritheca present is it porous or nonporous ? Programs (Rohlf et aI., 1967) . The data were 20. Is stomodaeum smooth or rough? stand ardized by characters. Correlation and dis­ 21. Is sphincter well developed ? stance coeffi cients were computed for each pair 22.N ematocysts per tentacle (many or few ) ? of species, based on the standardized data. Be­ 23. If corallite walls are presen t are they poro us or nonporous ? cause the coefficients measure different aspects 24. If colonial, are the coelentera intercommunicating of phenetic similarity, both methods were used by a basal canal system? (Rohlf and Sokal, 1965; Rohlf et aI., 1967) . 25. Is primary pigmentation carotinoid? After the entir e correlation and all distance 26. Int ra- or extra tentacular budding. 27. Are calices above genera l surfa ce of coral? matrices had been calculated, a clustering 28. Is coral colonial or solitary ? analysis program was used t o do an unweighted 29. Is columella developed ? pair group method of arithmetic averages 30. If columella developed is it low-boss or absent? (VPGMA) to summerize the information con­ 31. If columella developed is it styloid or laminar ? 32. Trabeculae, many or few. tained in the 20 X 20 matrices. The program 33. Trabecular inclination from axis of divergence also generated "phenograms" (Figs. 1 and 2) (Vaughan and Wells, 1943) . and cophenetic value matrices. Another program 34. Epitheca present or absent. was used to compare the correlation and dis­ 35. Septa or peritheca present or absent. tance matrices with their respective cophenetic 36. Is snaptot heca present or absent? 37. Is corallurn cercoid? value matrices (see correlation at bottom of Figs. 38. Is corallum plocoid ? 1 and 2). Another program calculated the 39. Is corallum meandroid ? eigenvalues and eigenvectors of the 60 X 60 40. Is corallum montiporoid ? matrix of correlations among the characters, 41. Is corallum thamnestroid ? using the Jocobi method (Rohlf et aI., 1967) . 42.If montiporoid, is it explanulate? The 20 OTV 's were then projected onto the 43. If plocoid is it glomerate? 44. Mural denticles presen t or absent. first three eigenvectors. 45. Found mostly on wind ward or leeward side of Another program computed the shortest, reef? simply connected network through the series 46. Found in sunny or dark areas ? of 20 poin ts (OTV's) , computed a perspective 47. Relative domin ance on reefs . view of a three-dimensional scatter diagram, 48. Is corallum generally porous or nonporus ? 49. Number dead colonies per 100 ft of reef front ­ and prepared a magnetic tape for a B.L. plotter, age. which was used to prepare Figure 3. 182 PACIFIC SCIENCE, Vol. 24, April 1970

COEFFICIENT OF CORRELATION -0.23 0.03 0.28 0.53 0.78 1.00 r------Pocilloporo Iiguloto - - - Poci Iloporo meondr ino - - - Pocilloporo domicornis ---Montipora verrilli -- - Montiporo patulo -- - Montipora verrucosa - - - Psammocora stellato --- Psammocoro verrill i - - - Pavona expla nulata - - - Pavona var ians --- Porites evermann i --- Porites lobata - - - Porites compressa - - - Fungia patella --- Fungia scutario

L..- -- - Fungia frog il is

-- - Cyphostrea ocell ina --- Leptastrea bottae --- Leptastreo purpureo "------Tubastrea aurea

CORRELATION =0.954 FIG. 1. A computer printed coefficient of correlation phenogram. The correlation between the origina l cor­ re lation matrix and the cophenetic value matrix is below the phenogram .

RESULTS AND CONCLUSIONS CLASSIFICATION ACCORDI NG TO V AUGHAN ANDW ELLS (1 943) Hawaii is in the northern region of the lush coral growth zone. The islands are in the central Cl ass : ANTHOZOA Ehrenberg, 184 3 Pacific and have a coral fauna most like that Order : SCLERACTINIA Bourne, 1900 of Japan and the tropical Pacific islands. A few Suborder: ASTROCOENIIDA Vaughan and Wells, species, however, are common to the Panama­ 1943 nian area. There are only 52 species from Family: ACROPORIDAE Verrill, 1902 1. Montipora uerrilli 14 genera present in the Hawaiian group 2. M ontipOl'a patula (Vaughan and Wells, 1943), and the present 3. Montip Ol'a uerrucosa study has sampled 9 of these genera. Vaughan Family: SERIATOPORIDAE Mi lne- Edwards and and W ell's classification of the Hawaiian corals Haime, 1849 4. Pocillopora ligulata is given below. As will be seen, my data strongly 5. Pocillopora meandrina support this classification. 6. Pocillopora damicoruis Numerical Taxonomy of Reef Corals-PowERS 183

Sub order : FUNGIIDA D un can , 1884 15. Fungia scutaria Famil y: THAMNASTERIIDAE Vaugh an and 16. Fungia [ragilis Wells, 1943 Subord er : FAVIIDA Vaughan and Wells, 1943 7.Psammacora stellata Fam ily: FAVIIDAE Gregory, 1900 8. Psammaeora uerrilli Subfamily: MO NTASTREINAE Vaugh an and Family: AGARICIIDAE Gray, 1847 Wells, 1943 9. Pavona explanulata 17. Leptastrea bonae 10. Pavona uarians 18. Leptastrea purpure« Family: PORITIDAE Gray, 1842 19. Cypbastrea ocellina 11. Porites euermanni 12. Porites lobata Suborder : DENDROPHYLLIIDA Vaugh an and 13. Porites compressa W ells, 1943 Family: FUNGIIDAE D ana, 1848 Family: DENDRO PHYLLIIDAE G ray, 1847 14. Frmgia patella 20. Tubastrea aurea

DISTANCE 2 .16 1.71 1.26 0.81 0.36 0.0 ,....------Pocillopora ligulata ,....------Pocillopora meandrina '------; '------Poci Iloporo domicornis -- - Montiporo verrilli --- Montiporo potulo '------Mont ipora verrucoso --- Porites evermanni --- Por ites. lobato - - - Por ites compresso --- Psommocoro stellato --- Psommocora verr illi -- - Pavona explanulato Pavono vor ians Fung ia patella Fungia scutario Fungio frogilis Cyphastrea ocel li na Leptastrea bottoe Leptastrea purpurea Tubastrea aurea

CORRELATION =0 .959 FIG. 2. A computer printed distance phen ogram . The correlati on betwe en the original distance matr ix and the cophenetic value matrix is below the phenog ram. 184 PACIFIC SCIENCE, Vol. 24, April 1970

15

7 17

8

14

FIG. 3. Perspective 3-D model of species. T he numbered balls are the coded species. Lines connect each species with the next closest species. Vertical lines indicate the position of the species with respect to the base of the model.

The ph enogram in Figure 1 summarizes the of their genera. In addition, Ftmgia tragilis relationships given by the correlation matrix. shows a slight phenetic uniqueness. Cyphastrea The correlation between the phenogram and the is clustered with the two Leptastrea species, original 20 X 20 matrix is 0.954. Therefore, which is not surprising because these genera are the phenogram is an excellent representation of considered to be in the same family (Vaughan the original matrix. Although it resembles one, and Wells, 1943). Furthermore, all the pocil­ this ph enogram should 110t be confused with a loporans are clustered at lower correla tions with phylogenetic tree. Obviously, Figure 1 shows the montiporans. The psammocorans and that members of the same genera are grouped pavonans are definitely clustered together. All fairly tightly together. It is noteworthy, how­ other linkages are at such levels that their ever, that Pocillopora liglliata appears to be phenetic affinities are ignored at this time. If phenetically more different from members of the number of species being compared were this genus than other species are from members greatly increased, the linkage groups with Numerical Taxonomy of Reef Corals-PowERS 185

greater coefficient values probably would not be Leptastrea species. Despite the small number of affected, but the lower linkage groups (clustered genera represented, the groupings shown in at low values) would probably change. There­ Figure 3 are an accurate representation of the fore, the lower linkage groupings should be ac­ data from 60 characters. The suggested larger cepted with caution until a more comprehensive clusters correspond to the suborders of Vaughan study, including a better and broader repre­ and Wells. However, Figure 3 also suggests sentation of genera, can be made. that the clusters above the genus level are not The phenogram in Figure 2 summarizes the spherical, which would account for phenogram data found in the distance matrix. The correla­ irregularities. An expanded study is in progress tion is 0.959, and so the phenogram is highly to test this hypothesis. representative of the matrix. The results are similar to those of the preceding section. Species in each genus and family are clustered SUMMARY togeth er. The distinction between Tubastrea 1. The numerical taxonomic study grouped aurea and the rest of the corals is sharper than members of the same generavand families to­ in the correlation phenogram. The arrangement gether. of the other groups, however, is much less cer­ 2. The groupings provisionally suggest the tain. Undoubtedly, several of the distant clusters subordinal classification adopted by Vaughan are real, but it is difficult to decide which group­ and Wells (19 43). ings are important and which are artifacts re­ 3. These data conform more closely to the sulting from the small number of genera repre ­ classification of scleractinian corals provided sented. by Vaughan and Wells than to others that have The computer-generated, thr ee-dimension been proposed. model (Fig. 3) provides more information on 4. Numerical taxonomic methods are indi­ the phenetic interrelations of the larger clusters cated to be extremely useful adjuncts in sclerae­ of species, that is, those grouped at a distance tinian systematics. value of 1.26 and larger, in Figure 2, and a correlation coefficient of 0.28 and smaller, in Figure 1. The model shows the members of ACKNOWLEDGM ENTS the same genera grouped close together with the exception of Pocillopora liglllata (code I would like to express my sincere apprecia­ number 1), which is far enough away from the tion to Dr. F. James Rohlf for his encourage­ other pocilloporans to suggest that it may be ment, guidance and uncommon patience. In considered as another .genus in the family addition, I would like to thank Dr. C. A. Leone, Seriatoporidae. The position of Pocillopora li­ Dr. R. Koehn , Dr. H. Lenhoff, Dr. Philip glllata in the drawing explains why the Porites Helfrich and the University of Hawaii Insti­ group has lesser affinity to Pocillopora although tute of Marine Biology for encouragement and the group is actually much closer to the mean special assistance and for providing .. diving position of Psammocora, Pavona or Ftt1lgia than equipment. to the mean position of Pocillopora. Conse­ This study was supported in part by Na­ quently, until more evidence from an expanded tional Science Foundation Grant GB-6134, study is available, the slight affinity of Porites U.S. Public Health Service Traineeship Grant to Pocillopora should be considered doubtful, 5-T01-AI-00256, the University of Kansas and due simply to the presence of the unusual Computation Center, and Argonne National member Pocillopora ligulata. Laboratories. Figure 3 reaffirms the uniqueness of Tubastrea. There is a suggestion of weaker LITERATURE CITED affinities among Porites, Psammocora and Pa­ uona. Ftt1lgia either is phenetically unique or BOURRADAILE, 1. A., and F. A. POTTS. 1963. has a weak affinity for Porites. The model re­ The Invertebrata. Cambridge Un iversity affirms the affinity of Cyphastrea to the two Press, London. Pp. 181-187. 186 PACIFIC SCIENCE, Vol. 24, April 1970

DANA, J. D. 1890 . Corals and coral islands. ROBINSON, W . I. 1923. The ancestry of the Dodd Mead, N.Y. Hexacoralla . . . . American Journal of DUERDEN, J. E. 1902. Morphology of the Science, vol. 6, pp. 424-426. Madreporaria. II. Increase of mesenteries, ROHLF, F. JAMES. 1967. Correlated characters etc. Anna ls and Magazine of Natural History , in numerical taxonomy. Systematic Zoology, ser. 7, vol. 10, pp. 96-115. vol. 16, pp. 109-126. EDMONDSON, C. H. 1929. Growth of Hawaiian ROH LF, F. J., and R. R. SOKAL. 1965. Co­ corals. B. P. Bishop Museum Bulletin no. efficients of correlation and distance in nu­ 58, 38 pp . merical taxonomy. Kansas University Science FOWLER, G. H. 1885-1 889. Anatomy of the Bulletin, vol. 45, no. 1, pp . 3-27. Madreporaria. Quarterly Journ al of Micro­ ROHLF, F. J., J. KISHPAUGH, and R. BARTCH ER. scopical Science, vol. 25, pp . 577-597; vol. 1967. Numerical taxonomy system of multi­ 27, pp. 1-16; vol. 28, pp . 1-19, 415-430; variate statistical programs. University of vol. 30, pp. 405- 422. Kansas Computation Center, Lawrence, Kan­ HICKSON, S. J. 1924. An introduction to the sas. study of recent corals. University of Man­ SOKAL, R. R., and P. H. A. SNEATH. 1963. chester Press. Principles of numerical taxonomy. W . H. HYMAN, 1. H . 1940. The invertebrates: Proto­ Freeman and Co., San Francisco. zoa through Ctenop hora. McGraw-Hill, New STEPHENSON, T. A. 'and A. STEPH ENSON . York. Pp. 599-623. 1933. Growth and asexual reproduction in OGILVIE, M. M. 1896. Microscopic and system­ corals. Scientific Reports of the Great Barrier atic study of madreporian types of corals. Reef Expedition, vol. 3, no. 7, pp . 167-217. Philosophical Transactions of the Royal TAM URA, T ., and Y. HADA. 1932. Growth rate Society of London , ser. B, vol. 187, pp . 83­ of reef building corals. Tohoku Imperial 345. University, Scientific Reports, vol. 7, ser. 4, POWERS, D. A., H .M. LENHOFF, and C. A. pp. 433-455. LEONE. 1968. Glucose-6-phosphate dehydro­ VAUGHAN, T . W ., and J. W . W ELLS. 1943. genase and 6-phosphogluconate dehydro­ Revision of the suborders, families and gen­ genase activities in coelenterates. Comparative era of the Scleractinia. Special Papers of the Biochemisty and Physiology, vol. 27, no. 1, Geological Society of America, no. 44, pp. 139-144. 363 pp .