100 PRINCIPES [Vor..4l

Primipes, 1l(2\, 1997, pp. 100-ll I FactorsAffecting the Distributionof a ThreatenedMadagascar Palm Species Dypsisdecaryi

Jonrsol RersrneRsoN'eNn JoHN A. Srreunsn,Jn. Departmentof EcoLogyand EuoLutionaryBiology LI-42, Uniuersityof Connecticut, Storrs,CT 06269-3012

ABSTRACT fruit mesocarp is eaten. Seeds are heavily col- lected and exported for tropical horticulture The complex interactions of human disturbance, environ- use. mental requirements, seed dispersal agents, seed predators, The main purpose of this study was to deter- and herbivores appear to have an important role in the limited mine the effects of variation in microenvironmen- distribution of a threatened palm, Dypsn decaryi, in southeast tal conditions,biotic interactions,and humandis- Madagascar. Human activities limit the growth and the recruit- turbances on the distribution and oerformanceof ment of new individuals throughout much of its range. Seed dispersal agents tend to deposit seeds near the parent Dypsis d,ecaryi.Specifically. this study examined: where seed predation is heavier than elsewhere. The success (I) the importance of variation in light and water of dispersed seeds depends on the conditions where seeds are availability on germination, (2) the role of seed deposited. In nature, dispersed seeds deposited in sites with dispersers,seed predators, and herbivoreson spe- intermediate light and moderate moisture levels have a better cies distribution patterns, and (3) the effect chance of escaping predators, germinating, and becoming of established. human activities on germination and seedling establishment.We focus on the early stagesof the life cycle (seed,seedlings) because these repre- Dypsisdecaryi (Jum.) Beentje & J. Dransf. (Neo- sentcritical phasesin the life historyofplants and d,ypsisd,ecaryl Jum.) (Family ,Subfamily may provide a key to understandingobserved dis- Arecoideae, Tribe ),2a rare, threatened tribution patterns. palm species,is restrictedin southeasternMada- gascar to a nalrow zone (about 20 km wide) Methods betweenthe humid rain forest of the southeastand the dry spiny forest of the southwest(see Fig. 1) Field Sites.The field site was located in the (Du Puy et al. 1992, Ratsirarson L993b, Eboroke AndohahelaNature Reservein southeasternMad- 1994). D. decaryi appearsto function ecologically agascar, in the third of the three noncontiguous "keystone as a species"(Ratsirarson and Silander parcels (25'01'S, 46"09'E). This 500-ha *.-" 1996; seealso Gilbert 1980,Terborgh 1986). It is was establishedin 1939 to protect one of the few locally dominantand showsasynchronous, contin- flourishingpopulations of Dypsisdecaryi (Fig. 3). uous flower production throughout the year, pro- Today, this is the only substantial population viding crucial food resourcesfor pollinators during throughout the range of the specieswhere signif- certain periods ofthe year. Dypsisd,ecaryi can also icant regeneration is occurring and a complete be considereda keystonespecies anthropologi- range of age and size classescan be found (Rat- cally. Local people rely on it for many resources sirarson f993b). Severalsmall, isolated popula- (Ratsirarson I993a\. Leaves are harvested for tions of Dlpsis can also be found along ephemeral thatching (Fig. 2); inflorescencebracts are col- watercoursesin the secondparcel of the reserve, lectedfor makingbaskets, hats, mats, etc., and the wheresome successful recruitment also occurs. In existingpopulations elsewhere, seedling and juve- nile stagesare rare or absent. rPresent address: Depafiement des Eaux et Forets ESSA, Univ- Granitic bedrock is the most abundant parent ersite d'Antananarivo, B.P. 3044, Antananarivo, Madagascar. 'All material the six genera, including Dypsis in the Dypsidinae sub- at the study site and the soils are shallow tribe, have recently been changed to Dypsis (Dransfield and oxisolswith a low pH (Jenkins1987, Nicoll and Beentie 1995). Langrand 1989). The emergent, discontinuous r9e7l RATSIRARSONAND SILANDER:DYPSIS DECARY]

for approximately l0 weeks. Seedswere sown in test beds of locally available, loamy soils (approx- imately equal mixtures of silt, clay, and sand). Germination (emerging leaves) was recorded weekly and followed over five months with results , tallied at 30,60, 90,I2O, and 150 daysaftersow- ing. The percentage of seeds germinating three months after sowing and the time required (in days) for 507oof seedsto germinatein each &eat- ment were recorded. Three levels of seed pretreatmentwere chosen: a control (no soaking)(G1), seedssoaked in water for 24 hours (G2), for 72 hours (G3), or soakedin boiling water for five minutes (Ga). We selected these pretreatmentsbecause they stimulated ger- mination in other palms (Nagao and Sakai 1979, Doughty et al. 1986, Broschat and Donselman lgBB). Three different levels of shading were pro- vided: no shading (S1), shading at 20 cm (heavy shading) (S2), and shading at 40 cm (moderate shading) (S3) abovethe ground. Grassculms were cut and arrangedhorizontally (approximatelyB cm thick) on a woodenframework to provide the shad- ing, which is a standard,local plantation practice. Equipment to measurelight levels in these treat- ments was not available. Three different levels of 0 100 200 300 400 km watering were chosen: no supplementalwatering (W1), watering once a day (W2), and watering 1. Distribution of Dlpsis bcani in Madagascar.Patterned twice a day (W3). Watering was continued area indicates the current natural distribution of the palm' throughout the experiment. In total, this consti- tuted a 4 X 3 X 3 factorial design (36 treatments). canopy is dominated by Dypsis d'ecaryi (Fig. a). Forty seeds were used in each treatment. Each The shrub layer subcanopyis dominated by spe- treatment was replicated twice. A three-wayanal- cies adaptedto the seasonallydry local conditions ysis ofvariance statistical design was used to ana- and include members of the Apocynaceae, lyze the data. Euphorbiaceae,Didieraceae, and Rubiaceae Fruit and SeedDispersal. The predicted seed- (Nicoll and Langrand 1989, Eboroke 7994; Rat- ling recruitment (i.e., dispersal)profile of the palm sirarson,unpublished data). was determined using a seedling recruitment Light and Water Aooilfrility. Citical threshold model, in order to understand the effective dis- light levels and water conditions may determine persal patterns in Dypsis (Ribbens et al. 1994). whether seeds germinate, and seedlings or juve- Two transects of 100 m each were established niles survive (Uhl and Dransfield I9B7). A ger- through the palm population inside the reserve. mination experiment was initiated in June 1992, The number of the palm seedling recruits in each using a nursery plantation near the field site. The successive2 X 2 m quadrat along the transects effects of three different treatments(seed presoak- was counted. Seedling recruitment was predicted ing, shading, and watering regime) on seed ger- as a function of adult size (height) and distance mination were examined.This information can be from potential parents, using a maximum likeli- used not only to understandthe optimal condition hood statistical analysis. This approach deter- for seed germination and the subsequenteffects mines the best match between the observedand on the distribution of the palm, but also to the predicted seedling distributions (see Ribbens enhanceex situ conservationofthe species(Chaz- et aI. 1994 for details). The height of the adult don l9B8). Mature fruits were collected randomly individuals, which is the best predictor of repro- from the natural range ofthe speciesand air dried ductive output (Ratsirarson 1993b), within 20 m PRINCIPES [Vor. 41

Table l. Effect of seedpretreatment on germination o/.

Percentageol seeds Number of days to germinating after 90 days 50% seeds germination Seed treatments (1\/: 18) (,v: 18) No. presoaking(Gl) Average 29.3 rI0.22 SD 22.3 24.62 Range o-72.5 60-170 Presoaking in water for 24 hours Average 21.27 128.33 (c2) SD 2t-B 36.86 Range o42.5 60-200 Presoaking in water for 72 hours Average 38.9 99.55 (G3) SD t? ? 27.3 Range 7.5-42.5 55-160 Presoakingin boiling water for 5 Average 0.00 0.00 minutes(C4) SD 0.00 0.00 Range 0.0-0.0 0.r).0 x Statistically signilicant differences among values are indicated with different letters using Duncan's multiple range test at the

from the transect was recorded. and the distance decaryi, open without Dypsisdecaryi, coveredwith between the center of the quadrat and the adult Dypsis decaryi, and covered without Dypsis individuals was taken. The mapping of adult decaryi. A pile of 50 seedswas set on the ground palms relative to the seedling transect quadrats at each site in J:une1992. Each treatment was was accomplishedby triangulation from compass replicatedl0 times. Seedswere censusedalter 7, bearings and observed distances along the tran- \4, 27, 32, and 50 days. Based on our earlier secr. studies,any removedseeds were most likely eaten SeedPredation and,Herbiuory. Some of the Dvo- by nocturnal mouse lemurs (see also below). srs seedson the ground *"r. "on.n-ed by eitirlr scolytidbeetles (l,anurgus sp.) or noctumalmouse Results lemurs (Microcebusmurinus, Miller). The beetles bore through the bony endocarpand lay their eggs Light and WaterAuailability. The time needed inside the seeds (Fig. 5). The larvae develop for 50Vo of seeds to germinate varied from 55 to inside while feeding on the endosperm and the 200 days (Tables 1-3). The percentageof germi- seed embryo. The emerging females leave after nating seedsafter 90 days was greater (407o) for matingwithin the seedsand searchfor new seeds seedspresoaked in water for 72 hours than for the on which to oviposit(see Ratsirarson I993b). The other treatments (Table I). Soaking in boiling percentageof seedsinfested by thesebeetles was water was lethal. The percentageof germinating countedin July 1992 inside successiveI X 1 m seedsafter 90 days and the time required for 50Vo plots, laid out along a transect through a stand of of the seeds to germinate were greater in treat- 30 parentplants. ments without shading (Table 2). Watering twice Mouse lemurs crack open the dry seeds,and a day increased the percentage of germinating eat the endosperm(Fig. 6). An experimentalstudy seeds after 90 days, and decreasedthe time on seed predation was initiated to determine required for 507o of the seedsto germinate(Table whether the escape of the seedsfrom lemur pre- 3). Open sunlight and supplementalwatering are dation was habitat-dependent.Seed removal was thus important factors promoting seed germina- studied in a two factorial design with the treat- tion. Presoaking,shading, or wateringtreatments ments being beneathor away (i.e., adult individ- alone had a highly significant effect on the time uals at least 10 m away) from parent , and needed for germination as well as the percentage in an opening (gap) or covered (non-gap) site. A seed germination (Table 4). The interaction gap is defined as a cleared area at least 5 m in between shading and watering showedsignificant diameter. All non-gap sites were more than 20 m effects on seed germination after 90 days, but no away from a gap area. The resulting combination effect on the time needed for germination (Table of factors are referred to as: open with Dypsis 4). For seedsthat were not pretreated(Gl treat- r9e7l RATSIRARSONAND SILANDER:DYPSIS DECARYI

Table2. Effect of shading on germination of Dypsis decaryi'

Percentage of seeds Number of days to germinating after 90 days 5070 seed germinations : (1{ : Seed treatments (,V 18) 18)

No shading(Sl) Average 55.0 a* 89.4 a SD r5.9 3I.34 Range ro< at( 55-170 Lightshading t40 cm from the Average 2t.t b 120.06 b ground) (S3) SD 17.0 26.46 Range 0.0-57.5 B0-200 Heavy shading (20 cm from the Average l3.B b 128.67 b ground) (S2) SD 10.5 23.92 Range 0-35 97-tB0 range test at the x Statistically significant differences among values are indicated with different letters using Duncan's multiple 1Vo level. ment), germination was acceleratedwhen planted including ringtailed lemurs (l'emur catta), mouse in full sunlight with supplementalwatering (Table Iemurs (Microcebusmurinus), black parrots (Cor- s). acopsisaasa), and wild pigs (Potamocheruslarao'' Fruit and SeedDispersal. The local distribution tru) (Ratsirarson, unpublished data), but in any of Dypsisdecaryi adults was significantly aggre- casethe viable seedsare apparentlynot movedfar gated (Ratsirarson1993b). The predicted seedling from the parent plant. recruits tend to be concentratedbeneath the par- The behavior of animal dispersal agents was ent plants (Fig. 7). A significant correlation was observedduring the peak fruit production- During seen between the observed and the expected the fruiting season, ring-tailed lemurs (l'emur recruitment in each quadrat along the transect (r catta) and mouse lemurs (Microcebusrnurinus) : 0.577; p < 0.01, see Ratsirarson1993b for consumethe edible parts of the fruit (mesocarp), details). Mean dispersal distance was calculated and drop the unharmed seeds near the parent by distributing expected recruits around a stan- plant. Lemurs are territorial with restricted home dardized parent plant and determining the average ranges.The territory of ring-tailed lemurs (l'emur distance from parent to recruit (see Ribbens et al. catta) in the Beza Mahafaly reserye,for example, 1994 for methods).The mean dispersal distance was reported to be limited between 400 and 900 calculated was quite small, 2.78 m. Fruits are m2 (approximatelya radius of 10-16 m) (Ratsirar- small in size (meandiameter : 1'41 + 0.12 cm, son 1987). The limited territories in this important N : 150) and can secondarilybe moved by grav- dispersal agent may severelyrestrict the effective ity, water, and/or by a variety of vertebrates dispersal of the seeds. Black parrots (Coracopsis

Table3. Effict of wateringon germinationo/Dypsis decaryi'

Percentageof germinating Number of days 507o seeds after 90 days of seeds germinate = Seed treatments (1V lB) (1f: 18)

No watering(Wl) Average 22.5 ax l2B.2A a SD 22.4 36.13 Range 0-82.s 55-200 Watering once a day (W2) Average 29.9 a 105.33 b SD 24.6 22.9I Range 0-72.5 ))-l4J Watering twice a day (W3) Average 37.5 a 104.5 b SD 2t.2 30.71 Range 7.5-72.5 6N0 at the * Statisticallv significant differences among values are indicated with different letters using Duncant multiple range test )"/a leveL PRINCIPES lVoL.41

Table 4. ANOVA: ffict of presoaking, shading, Table 5. Effect of shading and watering on the and watering on Dypsis decaryi germination afi.er gerrnination o/ Dypsis decaryi seeds(non-pretreat- 90 days, and on the time required. (in d,ays)for ed,. 507o of seed"sto germinate. Percentageof seeds P values geminating after 90 days (1\/ : 6) Gemin- Treatment [Mean + SD (Range)] ation Gemin- after ation No shading(Sl) 55.00+ u.53 (30-72.5) Treatmenl df 90 days time Moderate shading (S2) 15.41+ 9.00(5-25) D Heavy shading (S3) 17.50+ 10.72(0-30) b Presoaking 2 0.000 0.006 + Shading 2 0.000 0.000 No watering (Wl) 23.33 26.25(0-72.5) a Watering 2 0.001 0.0r3 Watering once a day + Presoaking X Shading 4 0.832 0.827 (w2) 29.59 2L8I (5-65) Presoaking X Watering 4 0.319 0.523 Watering twice a day + Shading X Watering 4 0.02I 0.603 (w3) 35.00 2r.r5 (12.5-72.5) b Presoaking X Shading X x Statistically significant differences among values are rn- Watering 0.106 0.723 dicated with different letters using Dur"uni multiple range test at the SVolevel.

aasa) consumethe edible parts of the fruits, and also tend to drop the undamagedseeds around the distance from the parent (Fig. B). No other insects parent plant. Wild pigs (Potamocheruslaraatus) were obseryedto feed on Dlpsis d.ecaryiseeds. ingest Dypsi"sfruits and deposit the seedsin their Seeds are also preyed upon by the noctumal feces, often more than 30 m from the closestpar- mouse lemurs (Microcebusmurinus). Their tracks ent plant. However,the seedspassing through the and feces were obvious on remaining seeds.Live pig's digestive systemfail to germinate (07oof the traps were used to capture these lemurs using seeds planted from wild pig's feces germinated seedsas bait. Three mouselemurs were caught in after 10 months; Ratsirarson,unpublished 'Seed data). 46 traps that were set out beginnirlg in June 1992 Predation and,Herbiaory. The averageper- and followed for 2.5 months. centage of seeds infested by lanurgus sp. (Scol- Seed disappearancefrom the seed removal tri- ytidae) beetles inside the l-m2 plots was 92Vo at als was followed over 50 days. Seed removal by 1 m from focal plants. Infestation declined with mouse lemurs differed among treatments.A chi-

2.0

o 1E, o

o 1n

Erl = o, ^- 0, u.5 @

Distancefrom adutt(m)

7. Predicted recruitment profile. The predicted density of recruits per m2 is shown as distributed around a standardizedparenl plant of3 m height. te97l RATSIRARSONAND SILANDER:DYPSIS DECARYI

100

80 s !60 q)c, at

dl .^ 6ev c, 20

0123456 Distancefrom the ParentPlant (m)

B. Seed infestation by scolytid beetles as a function of distance from the parent plant. Points represent the averagepercentage of seeds infested by scolltid beetles inside successiveI X I m plots, at varying distances{rom parent plants. Enor bars are tl SE.1{: 30. square (X') goodnessof fit analysis showeda sig- sisdecaryi cell. Seedremoval underneaththe can- nificant association between seed removal and opy was significantly greater than in the gap area treatment at the 27th (X' : 9.12, df : 3, p < from the 27th day of census onward. 0.05), 32nd (X' : 12.11, df : 3, p ( 0.01), and The effect of insect herbivory on non-seed 50th (1' : 16.32,df : 3,p < 0.01) censusdays stages also appears to be significart in Dypsis (Fig. 9). This significant associationappears to be decaryi. Seedlingsand juveniles were susceptible due to the effect of the closed canopy with no D1p- to rhinoceros beetle (Oryctespynhus, Scarabeidae)

g €40 o o aa E30 0) Ezo z

20 50 60 Numberof days g. Seed removal by mouse lemurs (Microcebu,smurinw). Points represent the number of remaining seeds at different census periods. The four treatments were: seeds placed under a closed canopy, without Dypsn present (filled squares (l); seeds placed ir ur op"r gap, with Dypsis nearby (filled diamonds O); seeds placed under a closed canopy, with Dlpsls (open squares n) and seeds placed in an open gap, without Dypsis(open diamonds 0). Error bars are I SE. PRINCIPES lVoL.4l attack. The beetlesdig into the plant, consumethe absenceof Dypsisd,ecaryi from the nearby wet for- sap, and destroy the apical meristem. Approxi- est communities. Interspecific interactions (e.g., mately 95Vo of.jwenile, and 207o of seedlingmor- competition for light resources)were not explicitiy talities are caused by rhinoceros beetles (Ratsi- examined in this study but may play important rarson 1993o, 6). roles in the distribution of this palm. Hutnan Disturbance. Human activity is one of The low mean dispersal distance (about 3 m) in the most important factors controlling the cunent Dypsis decaryi probably reflects (l) the limited viability and the distribution of Dypsis decaryi availability of dispersal agents(seeds tend to pas- (Ratsirarson I993a). Local people cut and fre- sively fall below parent) or (2) the very limited. quently bum much of the area outside the pro- effective range of dispersal agents(Hubbell 1979, tected reserve for agricultural practices (mainly Howe and Smallwood 1982, Howe 1990). Obser- tobacco and maize plantations). Few or no seed- vations on the main dispersal agents indicate lings and juveniles are to be found outside pro- eilher narrowly restricted territories or home tected areas (Ratsirarson1993b). Although adult ranges(i.e., Lcmur catta and Microcebusmurinus), individuals are resistantto fire (Fig. 10), seedlings or a tendency to feed on the fruit mesocarpand and juveniles are very sensitive (Ratsirarson drop the seedsbelow local foraging sites, near the 1993a,6). The absenceof youngerstages (seed- parent plants. lings and juveniles) in the unprotectedpopulations In the past longer distance dispersal may have limits natural regeneration. In addition, Dypsis been accomplishedby now extinct megafaunadis- decaryi does provide important needs for local persal agents,including the giant lemurs (Archae- people. Leavesand seedsof most individuals out- olemur spp.) and the elephant birds (Aepyorni.s side the reserve and some individuals inside the spp.). Bone and egg fragments oI Archaeolemur reserveare heavily harvested(Ratsirarson I993a, spp. and Aepyornis spp., respectively, have been Ratsirarsonet al. 1996). reported from a cave (Andrahomana),within the current distribution of Dypsisdecaryi in the south- Discussion east of Madagascar(Walker 1967). Based on the cranial and dental morphology of Archaeolemur Poorestgermination occurred when seedswere spp., they appear broadly convergent with Papio neither pretreated nor given supplementalwater- baboons, and were likely to have preferred a ing, which is most likely the conditions of the wooded habitat, had a frugivorous diet, and pos- seeds in their natural habitat. Soil moisture and sessed a greater home range than the present light resources seem to be important keys to lemur species(Martin and Wright 1967, Tattersall understanding the distribution of Dypsis decaryi. 1973, Dewar l9B4). Aepyornisspp. were reported In semi-arid zones,like the site occupied by DW- to be mainly vegetarian (Wetmore 1967). These sis decaryi, a moderatelevel of shading may pro- elephant birds were likely generalist herbivores vide that critical moisture level for germination and frugivores (Dewar 1984). Cracraft (1974) that otherwisewould be low in full sunlight. Par- interpreted the robustness of Aepyomis as evi- tial shading under some tree or shrub cover with dence of forest habitat preferences.These extinct some moisture supply appears to be the optimal megafaunaspecies (Aepyorni"s spp. and Archaeole- environmental situation for Dypsisdecaryi esrab- mur spp.) and,Dlpsi,s d,ecaryilikely occurred in lishment. Seedsin full sunlight may desiccaterap- the sameecological communities,and the fruits of idly and fail to germinate. This is perhaps why Dypsi.smay have been a significant componentof individuals are largely excluded from the drought- their diet. The larger home rangesor territories of stressedconditions of the drier spiny forest region. these megafaunamay have provided a greaterdis- In the easternedge of the spiny forest zone,Dlpsis persal range for the palm seedsin the past, and a decaryi is found only along ephemeral water- wider distribution than is currently observed. courses.Locally, individual palms tend to be con- Most of the seeds fall underneath the parent centrated either on rocky slopes in which water plant where seed predation is high. Seed dispers- seepageis sometimesavailable or along temporaxy ers play an important role in the escapefrom seed watercourses.In the adjacent rain forest, Dypsls predation (Janzen 1971, 1972). The removal of decaryi seedsmay germinatereadily but seedlings seeds from the vicinity of the parent may thus may be poor competitorsunder the continuouslow increase the probability of recruitment since seed light levels found there. This may explain the predators concentrate their activities under the 19971 RATSIRARSONAND SILANDER:DYPSIS DECARYI parent plant (Howe et al. 1985). Palms appear to and fail to germinateowhile seedsdeposited below be more susceptible to seed predation than other a closed canopy are less likely to survive given flowering plants, not only because of the size of the lower light level and the higher predation lev- the seed resource,but also apparently becauseof els by mouse lemurs. The trade-off between dis- more limited dispersal (Vandermeeret al. 1979, persal and predation depends mainly on the Tomlinson 1990). In Euterpe globosa, Janzen microhabitat of the sites where the seeds are (1971)reported that when the densitiesofthe sco- deposited which in turn determinesthe distribu- lytid beetle Cocotrypescarpophagus were at their tion of the palm. In addition, Oryctespyrrhus maximum,the beetlesmay destroyall the seeds beetle herbivory appears to have an impo;tant produced.Willson and Janzen(1972) also found effect on early stages (seedling and juvenile) of that more than B07o of Scheelearostrata seeds development of Dypsisdecaryi populations. Rhi- near parent plants were destroyedby a speciesof nocerosbeetles are reportedto be a commoncause bruchid beetle. of mortality in many palm species. For example, Seed removal by noctumal lemurs appears to one of the main threatslo the Borassusaethiopium be greater underneath a closed canopy than else- palm population in Senegalhas been reported to where. This result may be related to the arboreal be the large-scaleattack by rhinocerosbeetles behavior of Miuocebtu murinw, which may be (Oryctesspp.) (Sambouet al. 1992). Risk of adult better protectedfrom potential predatorsunder- Dlpsis decarylmortality to rhinocerosbeetle attack neathclosed canopies (Richard and Dewarl99l). appears to be very low (much less than |Vo) in Removal of seedsleft in a gap area was very low. contrast (Ratsirarson,unpublished data). Human Mouse lemurs (Microcebu murinus) are reported activities also severely limit the current distribu- to have a wide rangein diet (Richardand Dewar tion of this palm. Intensive leaf harvest (up to 1991),from plant parts (e.g..fruits. seeds,flowers, 1007oofleal numberin each adult individual) and leaves)and insects (moths,mantises, spiders) to the heavy seed collection (up to IO07o of seed small vertebrates(birds. chameleon, mice) (Petter producedper individual) significantly limit growth et al. 1977). Seeds mav be an important dietary and the natural regeneration of Dypsis decaryi componentof these lemurs during certain periods (Ratsirarsonet al. 1996). of the year. More than SVcof Drysis decaryi seeds In sum, the limited natural distribution of Dyp- were seen destroyedbt lemurs (as evidencedby sis decaryi in southeastMadagascar seems to be seed fragmentsremaining) rrithin 1 m of the par- due to the complex interaction of environmental ent plant (Ratsirarson.unpublished data). The requirements,human disturbance,and the effects lower seedremoval rate obsen'edunder D. decrtryi of seeddispersers, seed predators, and herbivores. canopies,compared to other closed-canopycon- The widespreadhuman destructionof natural hab- ditions, undoubtedlyreflects the fact that seeds itat (cutting and burning) associatedwith the local present naturally underneath the parent plants intense harvesting of palm resources (leaves, were not removedprior to the experiment.This seeds)Iimits the growth as well as the recruitment inflated the local densitv of seed in the D. decaryi of new individuals. Although adult individuals are canopies treatment. Here lemurs were probably resistant to the effect of fire, early stagesof the removingexperimental as nell as residentseeds. life cycle are very sensitive. Fruit and seed dis- There was a clear sun'ival advantageto dis- persal are limited and seedlingrecruitment is con- persal of Dypsis decart-i arvav from the parent centrated underneath the parent plant. The cur- plant. However, the advantageof dispersal is rent main animal dispersalagents (lemurs, parrots) relatedto the habitat in which the seedis depos- deposit seeds near their foraging sites close to ited (Howeet al. l9B5). The optimal environment adult palms. Seed mortality by seed predators for seedsurvival doesnot appearto be the optimal (scolytid beetle, mouse lemurs) is higher under- environmentfor germination.Seeds dispersed and neath parent plants and under closed canopythan depositedin a completelyopen area may desiccate elsewhere.Thus, seedsdispersed away from adult

2. Dypsis decaryi leaves used in thatching local village homes. View is east toward the wet, montane forest zone.3. Population of Dypsis decaryi in the third parcel of Andohahela Resene. 4. Emergent canopy of Dypsjs decaryi over the spiny shrub understory in Andohahela Reserve. View is west toward the arid zone. r0B PRINCIPES fVttt..'ll I09 leeTl RATSIITARSO\;.\ND SILANDIitI:D\ PSISDF'(jAR\ I

,!{t f ; fr. .J {\

OIr..theright are see(ls (racke(l open with an holes bv scolytid beetles (ktrturgus sp.) in DtTrsi.sderrrrri seetls. 5. Oviposition (r' 'flre is approxinratel) '!0 t rrr itr tliameter' nrouse letnrtr u,lult lr,uu.gu, sp. inrlicated i,u th. rto*. Muiu.. t".,1 sir -l nurinLs is the rlaitr Microcebus lerel of trLrnk of rhis inrliridual. Adults tarr tolerale a ntotlerate Repeate4 burning at this site-has left charcoal.,,r., on the repeatetlfires. seetllingsantl .iurenils cantlot' PRINCIPES [Vor. 4l

Dypsi,sd,ecaryi may have a higher chance of sur- Hown, H. F. 1990. Seed dispersal by birds and mammais: vival. Subsequentto dispersal, successfulgermi- implications for seedling demography.1n; K. S. Bawa and M. Hadley (eds.). Reproductive ecology of tropical forest nation and recruitment depend on the microhabi- plants, pp. l9l-218. Man and The BiosphereSeries, vol. tat where seedsare deposited. 7, UNESCO,Paris. - ANDJ. Suerlwoon. 1982. Ecologyof seeddispersal. Acknowledgments Annual Review of Ecology and Systematics13:2Ol-228. *r E. W. Scnupp, ,cl,[n L. C. Wtsrlry. 1985. Early We acknowledgethe D.E.F. (Direction des consequencesof seed dispersal for a neotropical tree Eaux et For6ts-Ministdre des Eaux et For0ts) in (Virola surinanensls). Ecology 66: 7BI-7 91. HuBBELL,S. P. f979. Tree dispersion, abundance,and diver- '. Madagascarfor granting permissionto work in the sity in a tropical dry forerst. Science 201: 1299-1309. protected reserve of Andohahela. Special thanks JANZEN,D. H. 1971. Seed predation by animals. Annual are due to S. O'Connor,A. Rajerison, A. Raman- Review of Ecology and Systematics2: 465492. anjatovo,J. V. G. Rasolonirina,and J. Ratsirarson 1972. Escape in space by Sterculia apelata seeds in Rican decid- for their invaluable help and assistancein Mada- from the btg Dysdereusfasciatus a Costa uous forest. Ecoiogy 53: 350-361. gascar. We thank R. L. Chazdon, R. Dewar, K. JEtlrrns, M. D. 1987. Madagascar,an environment profile. Holsinger, R. Kobe, T. Murray, S. W. Pacala, E. IUCN/UNEP/WWF. Gland, Switzerland. Ribbens, and A. F. Richard for their useful sug- MaRrn, P. S. ,cllrl H. E. Wnrcur, JR. 1967. Pleistocene gestionsand commentson the manuscript.J. R. extinctions. Yale University PressoNew Haven, Connect- rcut. acknowledges the encouragement and help in Neceo, M. A. er,rl W. S. Seret. 1979. E{Iect of growth reg- many different ways from R. Dewar, V. Rakoto- ulators on seedgermination of Archontophoenixalexandre. zafy, A. F. Richard, and M. Schwartz. Hoticultural Science14: 182-183. "Bio- This project was financially supportedby NIcoLL,M. E. enn O. LANGRAND.1989. Madagascar:Revue diversity Support Programo'(WWF-Grant number de la Consemationet des Aires Prot6g6es.WWF. Cland, Switzerland. 7517). Additional support was provided by The Prrran. J. J., R. Ar-ercNlec,eNn Y. RuuprnR. 1977. Faune W.W.F. U.S. and The University of Connecticut de Madagascar Mammifdres-Lemuriens, vol. 44. ResearchFoundation. ORSTOM, Paris. RATSIRARSoN,I. 1987. Contribution d 1'etudecomparative de LnrRerunr Cnnl l'eco-ethologiede lcmur catta dans deux habitats differ- ents de 1a reseroe sp6ciale de Bezd Mahafaly. Memoire de fin d'6tudes. Universit6 d'Antananarivo, Madagascar. BRoscHAr, T. K. ANDH. DoNSELMAN.1988. Palm seed stor- 1993a. Importance 6conomique et menace d'un age and germination studies. Principes 32:3-12. palmier en danger de Madagascar: Neod,ypsisd,ecaryr. Cneznou, R. L. l9BB. Conservation-consciouscollectinqr Akon'ny Ala 2:24-25. concerns and guidelines. Principes 32: 1,3-17. 1993b. Population biology and conserryationof an Cnecnarr, J. 1974. Phylogeny and evolution of ratite birds. Madagascartriangle palm: Neod,ypsisd,ecaryr. Ibis 116: 494-521. endangered of Connecticut, Storrs, Con- Drwen, R. E. 1984. Recent extinctions in Madagascar:the Ph. D. Thesis. University Ioss o{ subfossil fauna. In: P. Martin and R. G. Klein necticut. - ANDJ. A. Strenlrn, Jn. 1996. Reproductivebiology (eds.). Quatemary extinctions: a prehistoric revolution, pp. 574.-593. University of Arizona Press, Phoenix, Ari- of a threatened Madagascar triangle palm: Neodypsis zona. decaryi, Jumelle. Biotropica 28: 737-7 45. DoucHrv, S. C., E. N. O'RouRKE,E. P. Bannros, exl R. P. AND A. F. RrcueRl. 1996. Conseroation Mos/ERS. 1986. Germination induction of pygmy date and managementof a threatened Madagascarpalm spe- palm seed. Principes 30: 8t€7. cies: Dypsls decaryi, fumelle. Conservation Biology l0: DRANSFIELD,J. exl H. BEENTJE.1995. PalmsofMadagascar. 40-52. Royal Botanical Carden Kew and The International Palm RTBBENS,E., J. A. SILANDER,Jn., ,tttl S. W. Pecer-e..1994. Society. Recruitment in forests: calibrating models to predict pat- Du Puy. B.. D. Du Puy" ANDV. RANDRTANASoI-o.1992. Some terns of tree seedling dispersion. Ecology 75(6): 1794.- aspectsof the paims of Madagascar,and their cultivation 1806. at the parc de Tsimbazaza,Antananarivo. Principes 36: RtcueRo,A. F. ANDR. Drwen. 1991. Lemurecology. Annual 84-93. Reviewof Ecologyand Systematics22:145-175. EBoRoKE,S. f994. Approche phytosociologiquede la Par- SAMBou,B., J. E. Lawrssoru,erw A. S. BARFOD.1992. Bor- celle 3 de la RNI d'Andohahela. Travaux de Recherches. assu aethiopum, a threatened multiple purpose palm in Denatement Sciences Naturelles -E.E.S.S. Universit6 Senegal.Principes 36: l4B-155. de Toliary, Madagascar. Terrnnselr, l. 1973. Cranial anatomyof Archaeolemurines. Grr-nrnr, L. E. 1980. Food web organizationand conservation AmericanMuseum of Natural History 52: l-1I0. of neotropical diversity. 1n.'M. E. Soul6 and B. A. Wilcox TrnroRcu, J. f986. Keystoneplant resourcesin the Tropical (eds.). Consenation biology: an evolutionary-ecological forest. 12.' M. E. Soul6 (ed.). Conservation biology: the perspective,pp. 1I-34. Sinauer, Sunderland, Massachu- science of scarcity and diversity, pp. 33G-344. Sinauer, setts.USA. Sunderland, Massachusetm. 19971 PALM LITERATURE Ill

Tour-rnson,P. B. 1990. The structuralbiology of palms.Clar- WALKER,L. C. 1967. Locomotoradaptations in recent and endonPress. Oxford. fossil Madagascarlemurs. Ph. D. Thesis. University of UHL.N. ANDJ. Dnensnrr-1. 1987. GeneraPalmarum: a clas- London. sification of palms based on the Work of Harold E. Moore WETMoRE,A. 1967. Re-creatingMadagascar's giant extinct Jr. Allen Press,Lawrence, Kansas. bird. National GeographyMagazine 132: 488493. VANDERMEER,J., J. Srour, erunS. RIsH. 1979. Seeddispersal WTLLSoN,M. F. ANDD. H. JANZEN.1972. Predationon Schee- of a common CostaRican rainforest palm (Weffiageorgii). lea palm seedsby bruchid beetles: seed density and dis- Tropical Ecology 2O: 17-26. tance from the parent palm. Ecology 53:954-959.

Principes,4l(2), 1997, p. fff below'; a sixth symposium took place in 1996 . Since 1990, periodic information on sago palm developmentshas also been furnished by a sago PALMLITERATURE newslettef . 1. Tan, K. (ed.)Sago-76: Papers of the First Inter- SunuelR,tsA.NDHU,S. AND S. Slootnn (eds.) national Sago Symposium.330 pp. Kemajuan Fifth International Sago Symposium.Hat Yai, Kanji, Kuala Lumpur, Malaysia,1977; Stanton, Songkhia,Thailand, 27-29 Jantary 1994. Acta W.R. and M. Flach (eds.)Sogo;The Equatorial Horticuhurae 389. June 1995. 278 pp. Dfl. Suamp as a Natural Resource.Proceedings of 88.00 softcover. the SecondIntemational Sago Symposium.244 pp. Martinus Nijhoff, The Hague, Netherlands, Eleven of the 19 papers in this fifth symposium 1980 (see review in Prinripes 27(l):49-50, proceedings are devoted to sago starch as an 1983); Yamada,N. and K. Kainuma (eds.)Ihe industrial raw material and to secondary palm Third International SagoSymposium: Sago-'B 5. products. Ethanol, protein-enriched sago starch, 233 pp. Tropical Agriculture ResearchCenter, and glucose syrups are examples of starch prod- Yatabe, Tsukuba, Ibaraki, Japan, 1986; Ng ucts discussed. Sago Ieaf pulp for papermaking Thai-Tsiung, Tie Yiu-Liong and Kueh Hong- and sagopith residue as a green manurerepresent Siong (eds.) Toward,sGreater Adaancement of the latter product group. Various aspectsof sago '90s. the Sago Indu.stry in the Proceedingsof palm cultivation are dealt with in six papers on the Fourth International SagoSympositm.225 topics such as research needs, agroforestry,and pp. Ministry of Agriculture, Sarawak,Malaysia, seedling production. Two presentations covered 1991. economics:a cost-benefitanalysis of sagoproduc- 2. The Sixth International Sago Symposiumwas tion and a study of starch productivity. convenedDecember 9-I2,1996 in Pekanbaru, This excellent volume representsanother mile- Indonesia. The theme of the symposium was stone in the ongoingattempt to transformthe sago "Sago: The Future Sourceof Food and Feed." palm (Metroxylonsagu) into a major commercial Proceedingswill be published. species.Prospects ofachieving that goalare prom- 3. Sago Cornmunicationis published three times ising. per year by Tsukuba Sago Fund, 79I-27 To enhance the utility of this review, informa- Inaoka, Tsukuba, Ibaraki 305 Japan. tion on the four preceding international sagosym- posia, held between 1976 and 1990, is given DsNNrsJoHNsott