sign in sign up
<<
Home , Alamagan, Farallon de Pajaros, Mariana Islands, Pagan (island), Pandanus tectorius

GENERALIZED GEOLOGIC MAP OF . MARIANA

". '0 '

EXPLANATION ~ Ro,sedreefs ~ PCSf· cOldero lovos ~ POsl-cotdero l0vos ...lhmgnlt eof osh ~ Posl·coidero oshond 'ulf Ie P,r- coiderosuccu$ion .* C,nde, conn .~"'< ".....J Crol er 01 volCono ~ 0 4 3 FOS$,l locollly F'ou,' nOlrd bu l no! eouectee

,~

10 00 500 .0 0 0

". oe o ce'.

e l F IC

o C £ AN

\

\

. T , ~ . " · ""...,. ,

C -4.. e c , " ( ~ I S L .. ~ 0 s

FIG. 1. Generalized geologic map of Pagan, . A Fossil Flora from Pagan, Mariana Islands'

F. R. FOSBERG2 and GILBERT CORWIN3

PYROCLASTIC DEPOSITS on Pagan, northern elevation of 3,136 feet is at the summit ot Mariana Islands, contain numerous im­ . pressions. During geologic investigations in Pagan consists of two active volcanic cen­ 1954, ten collections were made for subse­ ters located within broad circular depressions quent identification and study. Seventeen (calderas) that are connected by a high rugged plant species, all of which now live in the isthmus . It has an area of 18.4 square miles Ma rianas, have been distinguished. and a greatest elevation of 1,890 feet near the The geologi c investigations of Pagan were south end of the isthmus. undertaken by a field party of the Office ofthe The northern caldera is 3Y2 miles in diam­ Engineer, U. S. Army Forces, , staffed eter and has one large central cone, Mt. by members of the U. S. Geological Survey. Pagan, that has an elevation of 1,855 feet. 1. D . Bonham of the field party first noted Relatively recent flows are extensive the plant fossils. Gilbert Corwin, with the aid and form broad plains north, east, and south of M. ]. Terman, also of the field party, and of the volcano; ash is concentrated to the Santiago V. Castro, a resident on the , west (leeward). collected samples and made field studies of The southern caldera is about 1Y2 miles the fossil localities. Fosberg has identified, across. South Volcano within it consists of studied, and described the plant impressions. four coalescing cones, three of which have broad steep-sided craters. Rough lava plains LO CATION AND GEOGRAPHY bound the volcano on the north, east, and Pagan Island is near the center and is the south. To the west the volcano slopes directly largest of the northern Mariana group. It lies to the ocean. between latitudes 18°01' and 18°11' North The isthmus ranges in width from about Y2 and between longitudes 145°41' and 145°49' mile at the north end to 1 mile at the south East, about 280 miles north of and end. Northward along the crest of the isthmus, 1,100 miles south-southeast of Tokyo , the peaks become successively lower. The (Fig. 1). caldera backslopes north of the isthmus are The form a chain dissected by numerous valleys directed away nearly 300 miles long that consists of the from the rim of the northern caldera. summits oflarge volcanoes rising from ocean The volcanic rocks have been divided into depths of as much as 6,000 feet. Uracas two major groups: pre-caldera and post­ () at the north end is one caldera. Both consist of flows and pyroclastic of the most active volcanoes of the western rocks . Pacific. Pagan, Asuncion, Agrigan, and Gu­ The older, pre-caldera group is well ex­ guan have had eruptions since 1900 (Tana­ posed in the caldera walls, along the kadate , 1940). The islands are generally small, ofthe isthmus and southern end ofthe island, relatively high, and rugged. The maximum and in old cliff remnants north and north­ 1 Publication authorized by the D irector, U. S. east of Mt. Pagan (Fig. 1). The lavas and Geologic al Survey. Manu script received March 4, 1957. pyroclastic rocks are products of eruptions by 2 Botanist, U. S. Ge olog ical Survey, Washingron 25, D. C. at least four major volcanoes and a number of 3 Geologist, U. S. Geological Survey, Tokyo. minor vents aligned along or near the axis of

3 4 PACIFIC SCIENCE, Vol. XII, January, 1958

the island . The largest of the old volcanoes dryness as of an extremely porous substratum was probably in the position of the present and of the pioneer nature ofvegetation, which Mt. Pagan. The uppermost unit of the suc­ occupies surfaces of recent volcanic ejecta. cession is composed of a widespread thick In the very few low, wet areas, such as to the tuff-breccia sequence. west and southwest of the Inner Lake, The pre-caldera group is probably of late thickets of broad-leafed are luxuriant Quaternary age. Sea cliffs have been cut in the enough, and places bordering the lake are breccias and older deposits; and reef lime­ somewhat marshy. Luxuriant patches of stones, some of which are now as much as 5 woods also exist in hanging valleys on the feet above present sea level, have since been west side of the south end of the island. A deposited on the wave-cut platforms. Similar mixed scrub forest of low stature forms sea cliffs flanked by terraces and raised reef thickets and patches up to several acres in have been noted on the coasts of extent on plains north and south of Mt. many Pacific islands and are attributed to Pagan - extending up its lower slopes in erosion and deposition during post-glacial places on steep slopes on the west shore of high stands of sea level. Sea levels 5 to 12 the isthmus, and in numerous ravines through­ feet higher than the present one have been out the island . There has been no investiga­ assigned various dates ranging from about tion of the actual composition of this forest, 1,000 to 5,000 years ago. Some that but it contains 15 or more species of trees. may have been introduced by man are rep­ The loose volcanic ash that covers large resented by impressions in samples collected areas, especially on the west side, is largely from the upper breccias and therefore may vegetated by an almost pure stand of sword­ indicate a maximum age for these breccias of grass, Miscanthus floridutus (Labill.) Warb . about 4,000 years-the probable length of This forms a coarse, harsh, brake-like grass­ time since the advent of man in land 1 to 3 m. tall and very dense in places. (Spoehr, 1955a, 1955b). On the steepest slopes and above 250 m. The post-caldera succession is concentrated altitude this grass tends to be shorter and within the calderas. At some places lavas have the clumps more widely spaced. Above 450 flowed over the caldera rims; tuff as much as m. it is sparse to absent. 30 feet thick mantles large portions of the Lava flows may be practically bare, as on caldera blackslopes and the isthmus. the northeast side of Mt. Pagan ; they may Historic records of eruptions are scattered, support scattered clumps of Miscanthus and conflicting, and incomplete. It is likely that trees of Casuarina, as on the east and south­ several major eruptions have taken place east sides ofMt. Pagan and the central upland within the past 200 to 300 years. Marche of the southern part of the island; or, as on (1891: 261) states that a major eruption oc­ many of the flows and lava cliffs to be seen curred in 1872. Extensive lavas northeast and along both sides of the island, they may be southwest of Mr. Pagan and a thick tuff covered by almost pure forests of Casuarina. sequence to the west were probably deposited Casuarina and the fern, N epbrolepis hirsututa .at this time . Eruptions since 1900 have been (Forst. f.) Presl, are among the earliest in­ rnmor. vaders on new lava. Both species were well established on a fresh black aa flow in the VEG ETATION depression at the west base of Mr. Pagan in 1950. This flow has been dated by Tanakadate The present-day vegetation of Pagan gives (1940) as having occurred in 1925. It shows a general impression of semi-aridity . This is no visible weathering. probably not so much a reflection of climatic On plains of ash the vegetation is gen- Fossil Flora from Pagan - FOSBERG AND CORWIN 5 erally grassland with scattered trees or clumps Impressions of trunks were observed of trees. The trees may be , Casuarina, near the base of the upper breccias at several or any of a number of broad-leafed species. places (Fig. 2). In all cases the impressions Many of these areas were under cultivation are nearly horizontal and therefore represent before World War II and are weedy and have fallen branches and trunks. M any are oriented rows of Casuarina and other trees planted by about parallel to radii from Mt. Pagan, sug­ the J apanese as windbreaks. Jatropha gossypi­ gesting that a powerful explosion from the folia 1., a fleshy-stemmed shrub introduced mountain felled,the trees. by the J apanese in the 1930's, has spread and now dominates large areas in the central part THE FOSSIL PLANTS of the island . Clumps oftrees ofvarious kinds Almost the entire collection consists ofleaf mark the sites of houses, still existing or not. impressions, mos t ofthem fragmentary. Some On the gently sloping northwest part of the ofthese have no features preserved that would island is a large plantation. There are make possible even tentative determinations. smaller ones in many part s of the island , both Very careful scrutiny shows that in the entire on plains and on talus cones . are also common in ravine mouths and on steep slopes above the sea. The large plantations are of relatively recent date, but there is no way of knowing the age of smaller clumps of coconuts that are mixed with other vegetation on various parts of the island. Some may very well date from before European visits to the island . FOSSIL LOCALITIES All samples of plant impressions are from the northern half of the island, mostly from tuff of the pre-caldera succession. Impres­ sions were noted in talus blocks at the foot of a high cliff at the south end of the island but no specimens were obtained. One sample (PC-86) was collected from post-caldera tuffs forming the surface of a terrace northeast of the bisected tuff cone; other impressions in post-caldera deposits were observed on the uplands east of the cone. Samples from the pre-caldera succession represent several stratigraphic hori zons (Table 1). M ost collections are from tuffs that under­ FIG. 2. Tuff overlying lava and underlying tuff­ lie or are included in an extensive tuff-breccia breccia ar the west end of the caldera rim. Plant im­ seguence near the top of the succession . One pressions were collected from the tuff below and of the most accessible localities (PC-43) is at interbedded with the dark colored cindery layers in the upper part of the picture. The hole in the cindery layer a sharp bend in the trail at the west end of the was prob ably formed by decay of a tree trunk that had caldera rim. Three samples (PC-58, PC-79, been buried by the ash. The sides of similar hole s PT-7) are from tuffs associated with lavas or elsewhere com mo nly retain impressions of tree bark. A deformed tuff layer near the center of the picture volcanic breccias that underlie the upper may represent the soil hotizon in which the plants breccias. grew. 6 PACIFIC SCIENCE, Vol. XII, January, 1958

TABLE 1 AGE SE QUEN CE OF SAMPLES C ONTAINING P LAN T FOSSILS, C OLLECT ED ON P AGAN I SLAND

NU MBER AGE AND LOCALITY * LITH OLOGY STRATIGR APHIC RELATION

PT- 7 Pre-caldera; Togari Penin sula. Well-indurated, re ddish, fi ne­ Tu ff associated with breccia un­ grained, laminared tuff. derlying th ick flow and cinders of a min or vent; underlain by PC-58 Pre-caldera; Togari Peninsula. Same . thi ck succession of lavas and pyroclastic rocks derived from old volcano to southeast.

PC-7 9 Pre-c aldera; north end ofisland. Po orly consolidated br ownish Tuff overlain by pre -caldera tu ff co ntaining numerous pel­ lavas which are overlain in turn lets. Som e fine-grained yellow by th e upper breccias. Under­ to red dish soil layers asso ciated. lying tu ff breccias mostly cov­ ered by talus .

PC-43 Pre-caldera-base ofupper brec ­ Poorly cons olidated brownish to O verlain by s o-foor thi ck tu ff­ cia sequence; west , west gr ay tu ff and cind ery tuff. Local breccia sequence; underlain by end of caldera rim. coarse cinder beds (Fig. 2). 8 feet of tuffs that are un der­ lain by lavas (Fig. 2).

PC-48 Pre-caldera-base ofup per brec ­ Same as PC-43. Similar to PC-43. cia seq uence ; west coasr Y2 mile so uth of mo nume nt on Ban­ deera Peninsula.

PC-84 Pre-caldera-base ofupper brec­ Brownish, poorly consolidated, O verlain by thin pellet tuff that cia sequence ; east coast, Y2 mile fine tu ff. is overlain by tuff-breccia; un­ N E of bisecte d tuff cone. derlain by gray ash, a yello w­ soil, and lava flow .

PC-85 Pre-caldera-within upper brec­ Coarse, gray, poorly con solidated Ov erlain by tuff-brecc ia; under­ cia sequ ence; same locality as tu ff with numerous large grains, lain by pumice bed, I -foot PC-84. rock fragments, and voids. thick.

ID -2 Pre-caldera- base ofup per brec­ Tuff similar to PC-84. O verlain by thin tu ffs that are cia seq uence; east coast, nort h overl ain by tu ff-breccias; un­ side of Togari Peni nsula. derlain by ash, soil s, and thi ck seq uence of very fine tuff­ breccia (lapiIIi tu ff).

ID -3 Pre-caldera-within up per brec­ Coarse, brownish tuff. N ear middle of tuff-breccia se­ cia sequence; specim ens from quence ; underlain by pumice talus blocks on slope. bed , I-foor thick.

PC-86 Post-caldera-recent tuff form­ Brown , coarse, granular 6-inch O verlain by pellet tuffs and ing surface of terrace Y2 mile tuff·layer that is mo derately in­ locally by and N E of bisecte d tuff cone. durated at surface bu t loose and wash ; underlain by mixed friable a few inches below . loose ash and beach .

• See F igure I. collection less than 20 different species are Margins are rarely well preserved and do not distinguishable. Some species are represented help much in identification . General shape, by many pieces in various states of preserva­ apparent texture , cross section, character of tion . surface, and curvature are sometimes perti­ In most instances the venation ofthe nent. It is obvious that with such material it provides the best distinguishing character. would be hopeless to attempt identification Fossil Flora from Pagan - FOSBERG AND CORWIN 7 unless the probable flora were very restricted. rnent of pinnules and fairly complete pinnae The fact that the deposits are considered rela­ (Fig. 3). tively recent makes reasonable the assump­ This fern is widespread in the western Pa­ tion that the flora was not much different from cific and is found on Pagan today. It occurs the present flora of the Marianas. It must be in shaded places but at times under relatively borne in mind, however, that any determina­ dry conditions. It is surprising that such a tion of material of this sort should be re­ delicate plant should be so well preserved. garded as tentative. In the list that follows, those species which were not in any horizon represented by fairly well-preserved material ? mariannensis Mere. are indicated by question marks. In the lists under the separate sample numbers, species PC-85-3, PC-79-5 not represented in the known present-day PC-79- 4 (a- k) (?) flora are marked by asterisks . Fragmentary impressions (Fig. 4), very Those remains which presented no features difficult to distinguish from those ofPandanus that could suggest any hope for identification tectorius. Leaf veins 10 to 15 per em., equally have been disregarded. Of the remainder the spaced clear to margin, little or no plication better specimens in all samples have been apparent, width up to 4 em. Impressions retained for deposit in the U. S. National showing neither plication, rolled margins, nor Museum. Of these, all have been identified crowding of veins toward margins have gen­ except two. One of these is probably a leaf erally been placed here. Some of these, how­ scar of a compound leaf, which has not been ever, where the venation is not clear, could matched; the other is either a mold of a possibly be Pandanus tectorius. or an impression of a strongly curved or This species is found in the well-explored distorted leaf. parts of the Marianas as far north as Pagan In addition to the specimens collected, molds oftree trunks, probably Pandanus judg­ ing from the abundant transverse leaf scars surrounding the trunks, were noted in several horizons.

In the following list of species the specimen numbers cited immediately below the name of each species are made up of the collector's sample number with an arbitrarily added num­ ber designating material ofsimilar appearance in a sample and letters indicating the various pieces as labelled for deposit in the U. S. National Museum.

POLYPODIACEAE

Pteris quadriaurita Retz .

PC-79-1 (a- f) FI G. 3. Pteris quadriaurit«, fragmentary impression Abundant fragmentary leaf impressions of fro nd show ing portion of rachis, pinnae, with showing very clearly the shape and arrange- pinnules. 8 PACIFIC SCIENCE, Vol. XII, January, 1958

The veins, except the midrib, are of about equal strength (Fig . 5). The tree trunk molds seen probably also belong here . This tree is very common on the Marianas , including Pagan, and is found throughout the tropical Pacific islands, mostly at low altitudes. It is an important component of many forests and is also found scattered in grasslands. Its frequency in this collection

FIG. 4. ? Freycinetia mariannensis, portion of central section of leaf, showing venation. and Agrigan. It has relatives throughout the Indo-Pacific . It grows in woods and ravines and is a liana clim bing on trees and rocks . Pandanus tectorius Park . (P. fragrans Gaud., P. Kafu Mart.) PC-79-3 (a- h) , PC-79-2 (?), PC-86-4 (a- f), PC-43-4 (a- d), PC-43-10, PC-43-11 (a- c) (?), ID2 (F)-6 (?), ID 2 (F)-2 Abundant fragmentary leaf impressions, many ofthem showing venation, many show­ ing characteristic plications, some showing curled margins characteristic of dried leaves (as found under the trees), and several rep­ FIG. 5. Pandanus tectorius, portion of central section resenting the narrow prolonged part near tip . of leaf. Numbers PC-43-11 and ID2 (F)-6 seem to be fragments of the enlarged curved bases of the could be due both to its abundance and to leaves, but this identification is not certain. the hard stiff leaves that lend themselves to Number PC-79-2 seems to be a fragment ofan preservation as impressions. impression of a fruit of this species. This also is not certain. Margins have not been pre­ GRAMINEAE served in good condition in any of the mate­ Miscanthusfloridulus (Labill.) Warb. rial. The veins are generally more crowded especially near the margins, than in Freycinetia, PC-43-6 (a- p), PC-43- 7 (?), PC-43-8, PC­ varying from about 17to 26 per em. depending 43- 9 (a-b), PC-48, PC-79-7 (a- e), PC-79- 8 on the position in the leaf and are closer to­ (a-b) , PC-84-1 (a- d), PC-85-4 (a- d), PC­ gether toward the apex and near the margins. 36-5, ID 2 (F)- l (a- k) , ID 31-1 Fossil Flora from Pagan- FOSBERG AND CORWIN 9

Abundant fragments of leaf impressions, as Pagan , where it dominates large areas of well as a few pieces of stem and leaf-sheath grassland, especially on loose volcanic ash impressions. The leaves are thin and in addi­ . It is a coarse harsh grass as much as 2 tion to a heavy midrib have two orders of to 3 m. tall. It is found from Japan and the lesser veins, with the most prominent ofthese to New Caledonia and east at equal intervals, about 9 to 6 mm. In each throughout Micronesia and (ex­ interval are 3 to 4 very distinct smaller veins. cept ) on volcanic islands. These blades are mostly 1 to 2 em. wide, but several are much wider than that, so that it is PALMAE . hard to match them with available herbarium Cocos nucifera L. material. However, there seems little doubt that they represent this species. The sheaths PC-79-6 show a curved surface, no midrib, and slightly This is the impression of a srrongly plicate irregular veins, about 16 in 5 mm . A mold of single bent section of a seedling leaf (Fig. 7). such a sheath shows a slight amount of com­ The size of the leaf and the amplitude and pression and somewhat of a keel, or angle, as character of the plications match exactly a is commonly observed on the back of sheaths segment of a seedling coconut leaf preserved in this species. The poorly preserved stem in the Bailey Hortorium ofCornell University. fragments are not especially distinctive. This leaf, if similarly bent, can be fitted into The leaves of this species are the most the plications ofthe fossil. Despite the scanti­ abundant fossils in the collection (Fig. 6). In ness of the material this species appears to be some cases they are arranged so regularly as reliably identified. to suggest the pinnae of a palm or the plica­ The coconut palm is now very abundant on tions of a fan palm leaf. The venation, how­ the island, mostly planted. It is generally re­ ever, is very characteristic and is easily garded as being of human introduction on matched in herbarium material. the oceanic islands of the Pacific, although This is the most abundant living .planr on

FIG. 6. Miscanthus jloridulus, part of mass of leaf FIG. 7. Cocos nucifer«, portion ofseedling leaf, show ­ blades, showing venation. ing plication. 10 PACIFIC SCIENCE, Vol. XII, January, 1958 the possibility of the establishment of drift This plant is found throughout the Pacific nuts is not denied . In all probability the islands wherever man has carried it. It has present seedling was from progeny of trees not been found where it could not readily introduced by early human visitors to Pagan , have been originally planted by man . The indicating that the horizon in which it occurs aborigines in many islands are said to utilize may be rather recent. it as a famine food, though it is not very palatable. It has no obvious means of getting ARACEAE around unaided, but once established persists and multiplies rather successfully. Alocasia macrorrhiza (1.) Schott Its presence in this fossil flora is distinctly PC-43-1 (a-f) , 2, 3, 5 surprising. It indicates that the eruption pro­ A very well-preserved mold of a petiole, a ducing the tuff, which preserved the remains, fragment of the upper part of a petiole , vari­ occurred after the earliest visits by man to the ous badly crushed and damaged petiole frag­ island . ments and a stipule, not especially well pre­ LILIACEAE served. Some of these petiole fragments would ? Dianella ensiformis (1.) DC. have been quite unidentifiable by themselves but were recognized when associated with the PC-85-1, PC-85-2 , ID2 (F)-3, 4 other better preserved parts (Fig. 8). Leaf blade impressions and fragments of impressions . Where the outline is preserved these taper more stron gly than in Miscanthus; where venation is evident there is a prominent midrib and fine even venation, the veins of equal thickness, between 3 and 4 per mm . (Fig. 9). This is not known from Pagan as a living plant, though it may well have been over­ looked. It is common elsewhere in the Mari ­ anas in swordgrass vegetation on volcanic soils, especially on old erosion scars and around rocky places. ORCHI DACEAE ? Spathoglottis micronesiaca Schltr. ID 2(F)-7 A reasonably well-preserved impression of a leaf fragment showing strong plication of about the mag nitude of that shown by this species, the width and what remains of the shape, also, being correct. The specimen is 3 cm. wide and has just under 3 plications per cm.(Fig. 10). This species is common in the grassland in the volcanic portions of Guam. It is not known at present on Pagan unless a pink­ flowered plant noted but not collected by FIG. 8. A locasia macrorrhiza, sripule. Bonham is it. Fossil Flora from Pagan - FOSBERG AND CORWIN 11

It probably maybe regarded as one of the few northern elements of the flora of the Marianas ; and the present fossil fairly well disposes of the possibility of it being a recent introduction in the Marianas.

FIG. 10. ? Spathoglottis micronesiaca, portion of leaf blade show ing plication.

HERNANDIACEAE Hernandia sonora 1.

FIG. 9. ? Dianella ensiformis, leaf blades . PT-7-1 A portion of an impression of a leaf blade ULMACEAE with major veins well shown and with some Trema argentea Planch. indication of texture. The appearance of a PC-79-11 firm, stiff texture , palmate venation with veins A partial leaf impression with palmate and somewhat impressed, and the distance be­ net venation so perfectly preserved that there tween the base and the first branches from is little or no doubt as to its identity with the midvein strongly suggest this species.It is Trema argentea. A piece of the opposing part a pantropic lowland or strand tree, common of this impression is preserved on PC-79-1 2 in the Marianas and known from Pagan . (Fig. 11). This plant , which is often and possib ly EUPHORBIACEAE correctly regarded as a variety of Trema orien­ ? M acaranga thompsonii Merr. talis, occurs throughout the Marianas, includ­ ing Pagan , and north through the Volcano PC-79-1 2 and , and with related forms in A poorly preserved impression of part of a continental . leaf showing some major palmate venation. 12 PACIFIC SCIENCE, Vol. XII, January, 1958

MELIACEAE Aglaia mariannensis Merr. PT-7-3 Several imperfect impressions of leaflets, some showing pinnate venation. The veins are rather straight and strongly ascending. The shape of leaflets of this species varies from elliptic to obovate and that of the fossils shows corresponding variation as much as could be expected in the meager material available (Fig. 12a ). This is one of the commonest species of small trees in thickets as well as forests throughout the Marianas, including Pagan . Related species are found elsewhere in Micro­ nesia and westward.

SAPINDACEAE Tristiropsis obtttsangttia Radlk. PC-58-1 (a-e), PT-7-2, 5 FIG. 11. Trema argentea, portion of leaf, sho wing Impressions of leaflets, mostly imperfect, venation. apparently somewh at folded along midribs, rather stiff, some of them showing main vena­ tion of a pinnate character. Some of the more Viewed by itself it would be scarcely iden ­ poorly preserved of these impressions may tifiable, but when placed side by side with a belong to Aglaia, but the oblique bases of leaf of this species the correspondence of the most of them scarcely fit that genus (Fig. 13). venation pattern is striking. Tristiropsis obtusangul« is at present known This is a tree commonly found in the south­ only from Guam and and grows usually ern Marianas, especially in second growth but not always on . and pioneer situations, but more often on limestone than on volcanic soils. It is not known to be living on Pagan at the present TILIACEAE time. Elaeocarpus joga Merr. Melanolepis mttltiglandttlosa (Reinw.) Reich­ PT-7- 4 enb. f. A practically perfect leaf impression, ­ PC-86-2 ovate, showing a short petiole and some vena­ A fragmentary leaf impression showing the tion (Fig. 12b). main veins. The palmate and net venation This species extends throughout the Mari­ pattern checks very well with that of this anas as far north as Pagan. In the southern species, which is common on Pagan as in the Marianas it grows on limestone but in the rest of the Marianas, extending to . north on lava flows. It is a large forest tree. It grows in thickets and secondary scrub Closely related species are found in other forest . parts of Micronesia and the Philippines. Fossil Flora from Pagan - FOSBERG AND CORWIN 13

FIG. 12. A glaia mariannensis, impression on righr, base of large leaflet, th at on left, possibly one of smaller leaflets. Elaeocarp«: joga, impress ion in center, leaf blade and petiole. .

THEACEAE (Fig. 14). It is much smaller than the average for this species but matches well enough ? Eurya nitida Korth. some of the smallest leaves present on avail­ PC-79-10 able herbarium material. The angle of diver­ Two fragmentary leaf impressions, not pos­ gence of the main veins is so wide as to itively identifiable. Onl y bases, with some preclude its belonging to most of the other obscure venation, and rather curiou s minute Micronesian species with palmate venation, surface marking are available. Eurya seems the only possibilities being Thespesia populnea the most likely identity. (1.) Sol. ex Correa, which it does not resemble This is a small-leafed shrub or small tree, in other characteristics and which is rarely so characteristic of volcanic soils and frequently small, and Sida fallax Walp., which has never found isolated Ot in small thickets in sword­ been reported from the Marianas. grass areas. It has been reported from Pagan Premna obtusifolia is a very common tree by Hosokawa as Eurya ladronica H osokawa. throughout the M arianas, includi ng Pagan, growi ng in thickets or secondary grow th as VERBENACEAE well as in original forest. The species is vari­ able and extends over much of the Indo­ Premna obtusifolia R. Br. Pacific region . It has been given various PC-86-1 names, the one most commonly used for A good leaf impression with venation Marianas material being Premna gaudichaudii rather clearly shown but lacking the apex Schauer. 14 PACIFIC SCIENCE, Vol. XII, January, 1958

FIG. 13. Tristiropsis obtusangula , impressio n of leaflet, FIG. 14. Premna obtusifolia, an unusually small leaf showing textu re and venation. blade, with apex missing, showing venation.

Family unknown Sample PC-43 Pandanustectorius, Miscanthusfloridulus, Alo­ Unidentified impression, probably a leaf scar. casia macrorrhiza ID 2(F)-5 It is hard to associate A locasia with sword­ This is a concave impression, more or less grass except where the A locasia might have shield shaped , with 5 (or 6) bundle traces. been in a somewhat shaded ravine bottom, Unidentified mold of a fruit or curved leaf. perhaps with Pandanus trees, and with the PC-86- 3 sides of the ravine covered with swordgrass, This hollow curved impression, lacking possibly extending down almost to the bot­ suggestive details, was not matched with tom of the ravine. On more level ground, if anything. it is not too dry, the Alocasia could have grown in a thicket of Pandanus and possibly PAL EOECOLOGY other trees, closely surrounded by swordgrass. The A locasia suggests the presence or former By listing the species identified from each presence of man, as this plant was in all of the samples it is possible to suggest some­ probability distributed in the Pacific islands thin g of the prob able ecological conditions entirely through human agency. and vegetation types at the time the various depos its represented were laid down . Such Sample PC-48 conclu sions are, of course, extremely tenta ­ Miscanthusfloridulus tive when based on so few species. It is Probably swordgrass vegetatio n. greatly hoped that at some future time more Sample PC-58 extensive collecting may be done in these *Tristiropsis obtusangula beds so that the picture of the past vegetation This is a forest tree, not now found on and conditions on the island may become clearer. * No t now known co be living on Pagan. Fossil Flora from Pagan- FOSBERG AND CORWIN 15

Pagan or in the northern Marianas but com­ Sample ID2 (F) mon on rough limestone on Guam. An asso­ Pandanus tectorius, Miscanthus floridulus, ciation of plants similar to that in which this *Dianella ensiformis, *Spathoglottis micronesiaca species grows on Guam occurs on rough lava Unidentified leaf scar or fruit impression. in the northern Marianas on Is­ The conditions suggested by this small land , which is much like Pagan. Something flora are those of swordgrass , possibly some­ like this association may have existed on what sparse, with either thickets or scattered rough lava in pre-caldera time on Pagan. trees of Pandanus. Sample PC-79 Sample ID31 Pteris quadriaurita, Pandanus tectorius, Frey­ Miscanthus floridulus cinetia mariannensls, Miscanthus floridulus, Cocos Probably swordgrass vegetation. nucifera, Trema argentea, ? *Macaranga tbomp­ From the foregoing lists it will be seen that sonii, ? Eurya nitida there is meager evidence for any comprehen­ From the assemblage of plants represented sive opinions on the past vegetation of Pagan. here the habit could have been the margin of Most of the indications are that at least some a mixed thicket, possibly second growth, of the vegetation was much the same as that bordering on swordgrass. The presence of a found on the island today. Swordgrass' occurs seedling coconut suggests the possibility of in all but the two lowest horizons, and there human disturbance. is little doubt that it dominated the vegeta­ Sample PC-84 tion wherever there was a loose ash substra­ Miscanthurfloridulus tum . Undoubtedly there were scattered trees The vegetation indicated is swordgrass . or clumps of Pandanus and probably mixed thickets and scrub forest of Pandanus and Sample PC-85 other trees. Although the coconut does not ? Freycinetia mariannensis, Miscanthus flori­ appear in most of the lists of species, it may dulus, *Dianella ensiformis very well have been present in all of the later Swordgrass vegetation close to either a horizons, as there is little doubt that it was thicket or a rock cliff, unless what are here introduced by the early Chamorros and no identified as Freycinetia leaves are really Pan­ reason why it would not have persisted danus, in which case it would have occurred whether or not they inhabited the island for as scattered trees or thickets in the swordgrass. any great length of time. An interesting fact is the complete lack of Casuarina in the col­ Sample PC-86 lections. Next to Miscanthus it is the most Pandanus tectorius, Miscanthus floridulus, Me­ abundant component of the modern vegeta­ lanolepis multiglandulosa, Premna obtusifoli« tion of the island. It adds to the scanty evi­ Unidentified mold. dence that Casuarina may be a recent arrival Probably swordgrass with mixed thickets in the Marianas , but.this is still an unsettled or patches of scrub forest of Pandanus and question. broad-leafed trees. The two lowest horizons, with plants com­ Sample PT-7 monly found on rough lava or rough lime­ Hernandia sonora, Aglaiamariannensis, Elaeo­ stone and with one tree not known today on carpusjoga, ? *Tristiropsis obtusangula Pagan , possibly represent a slightly more This association suggests a forest of the mature vegetation than any known at present sort that grows on rough lava on such islands on the island, though it may actually be rep­ as Alamagan at the present day, or on rough resented in the mixed forest type. Something limestone on the islands farther south. like it is prominent on extensive rough lava 16 PACIFIC SCIENCE, Vol. XII, January, 1958

flows on Alamagan, the next island to the TERMAN, and G . W. VIELE. Military Geol­ south of Pagan. The violent explosive erup­ ogy, Pagan Island, Mariana Islands. Ined. tions that formed the calderas and deposited FOSBERG, F. R. 1958. Vascular flora of Pagan the thick tuffs may have completely destroyed Island, Northern M arianas. Pacific Sci. 12(1) : the stands of this type of forest . pp . 17-20. ~o~si l The presence of four species in the M ARCHE, A. 1891. Rapport gene ral sur une collecti ons that are not kn own ro be hVIng mission aux iles Mariannes. Nouvelles Arch. on th e island to day may possibly have little M iss. Sci. Lit. 1: 241-280. Paris. significance, for the living flora has not been SATO, Z. 1941. Ge ology of Babelthuap Island really well collected and those species may ( main island) and the anne xed islets. actually be present now. However, 01( an Amer. Mus. Nat. Hist., Bul. 38(72): 92-104. island with frequent and at times violent vol­ (Translation by R.Endo, 1949, filed in U. canic activity the continued existence of any S. Geol. Survey Library.) bu t the most tenacious species of plants is SPOEHR, A. 1955a. The human background hazardous, and it is more than likely that some of Pacific science. Sci. Monthly 81: 3- 9. tha t were once growing on Pagan may have - -- 1955b. Personal communication. been eliminated. TANAKADAT E, H . 1940. Volcanoes in the Mariana Islands in the Japanese mandated REFERENCES South . Bulletin V olcanologique II (6) : CORWIN, GILBERT, 1. D . BONHAM , M. J. 199-223.