Masting in a Temperate Tree: Evidence For

Austral Ecology (2012) 37, 175–182

Masting in a temperate tree: Evidence for

environmental prediction?aec_2260 175..182

K. C. BURNS School of Biological Sciences,Victoria University of Wellington, PO Box 600,Wellington 6140, New Zealand (Email: [email protected])

Abstract Many plant species produce large fruit crops in some years and then produce few or no fruits in others. Synchronous, inter-annual variation in plant reproduction is known as ‘masting’ and its adaptive significance has yet to be fully resolved. For 8 consecutive years, I quantified every fruit produced by 22 females of a New Zealand tree species (Dysoxylum spectabile), which has an unusual habit of taking a full calendar year to mature fruits after flowering. Fruit production varied strongly among years and was tightly synchronized among trees. Annual variability in fruit production declined with total reproductive output, indicating trees with lower fecundity exhibited a stronger tendency to mast. Although unrelated to temperature, annual fruit production was positively related to precipitation during annual periods of fruit development, and negatively related to fruit production in the previous year. Seedlings had higher rates of survivorship in a wet, high-seed year than in a dry, low-seed year, suggesting that seedlings might be drought sensitive. Therefore, D. spectabile produced large fruit crops during periods of high rainfall prior to fruit maturation, which may enhance survivorship of drought-intolerant seeds. Results were inconsistent with several hypotheses that are widely believed to be the most likely explanations for masting. Instead, results were consistent with the environmental prediction hypothesis, suggesting that this hypothesis may be more important than previously appreciated.

Key words: Dysoxylum spectabile, fruit, New Zealand, plant reproduction, precipitation.

INTRODUCTION been derived to explain the adaptive significance of masting. For example, if reproductive fitness acceler- Many plant species reproduce heavily in some years ates with reproductive effort, then annual reproductive and forgo reproduction altogether in others. Synchro- variability could arise from an evolutionary economy nous, inter-annual variation in reproductive effort is of scale. One mechanism promoting such an economy commonly known as ‘masting’ and the phenomenon of scale is predator satiation (Silvertown 1980; Kelly occurs in perennial plant populations across the globe et al. 2000; Kitzberger et al. 2007; Espelta et al. 2008). (Herrera et al. 1998; Koenig & Knops 2000; Koenig The predator satiation hypothesis argues that inter- et al. 2003; Liebhold et al. 2004; Norden et al. 2007). annual variation in seed set starves seed predators in However, the adaptive significance and evolutionary low-seed years and then overwhelms seed predators precursors of mast-fruiting have yet to be fully resolved during high-seed years, when too many seeds are pro- (see Kelly 1994; Kelly & Sork 2002). duced to be completely harvested. A second mecha- The resource-matching hypothesis provides the nism promoting an economy of scale is the pollination most parsimonious explanation for masting (Norton & efficiency hypothesis, which states that masting would Kelly 1988; Sork et al. 1993; Monks & Kelly 2006). It be selected for if the probability of successful pollina- states that plants allocate a constant fraction of assimi- tion increases with synchronized reproductive effort lated carbon each year to reproduction. Years with (Nilsson & Wästljung 1987; Smith et al. 1990; Kelly benign climatic conditions that promote high rates of et al. 2001). Explanatory hypotheses based on econo- carbon assimilation therefore experience greater seed mies of scale are traditionally the most widely accepted production. In years with poor climatic conditions, explanations for the evolutionary origins of annual plants fix less carbon and produce fewer seeds. Deter- reproductive variability (Kelly 1994; Keeley & Bond ministic processes are unimportant under the 1999). However, deterministic processes other than resource-matching hypothesis and inter-annual varia- those based on an economy of scale might also select tion in seed set results from climatic variation directly. for annual reproductive variability. Alternatively, annual reproductive variability may be The environmental prediction hypothesis postulates adaptive and three non-exclusive hypotheses have that plants mast to capitalize on benign environmental conditions for seedling establishment during particular Accepted for publication March 2011. years (Janzen 1971; see Silvertown 1980; Smith et al.

1990). Some authors list the environmental prediction the dynamics of conifer-broadleaf forests throughout New hypothesis under the heading of economy of scale (e.g. Zealand (Dawson 1988; Wardle 2002). Kelly 1994). However, the term ‘economy of scale’ is Dysoxylum spectabile (Forst. f.) Hook. f. (Meliaceae) is a derived from economics, where it refers to a direct bird-pollinated and bird-dispersed tree species that grows to relationship between manufacturing costs and produc- 5–10 m tall and is endemic to New Zealand. Flowers are produced in clusters on drooping panicles attached to tree tion (i.e. when cost declines as a function of production trunks in the Austral winter (July) and are pollinated mostly itself, rather than some external variable). If applied by two bird species (tui, Prosthemadera novaeseelandiae and accurately to mast-fruiting, it should refer to an waxeyes, Zosterops lateralis) in Otari-Wilton’s Bush. Fruits are increase in fitness that results directly from an increase initially round, green and approximately 3 cm in diameter. At in production (e.g. reduced accessory costs, increased maturity, they dehisce to expose six to eight arillate seeds pollination efficiency or enhanced attractiveness of fruit enclosed in orange, fleshy fruit pulp. It takes an unusually crops to frugivores).Therefore, keeping in line with the long time for fruits to mature; they are retained on parent original use of the term, I will treat the environmental plants for a year or more after flowering, leading to temporal prediction hypothesis as distinct from an economy of overlap between flowers and fruits. In some years fruits can scale. The environmental prediction hypothesis has also be aborted just prior to dehiscence. Although the breed- been tested infrequently (Williamson & Ickes 2002; ing system of D. spectabile has yet to be resolved precisely, available data indicate that it is ‘leaky dioecious’, meaning Piovesan & Adams 2005) and results are equivocal that most individuals are either male or female, but some (Piovesan & Adams 2001; Koenig et al. 2010). Support males occasionally set very small amounts of fruit (Braggins has also come mostly from fire-dominated ecosystems, et al. 1999; Gardner 2007). where species use fire as a cue for mast-fruiting (Kelly Like many isolated islands, New Zealand lacks native 1994; Keeley & Bond 1999). mammals (aside from two species of bat). However, since the Here, I investigate annual reproductive variability in arrival of humans approximately 700 years ago, many her- a temperate tree species to test the environmental pre- bivorous mammals have been introduced to New Zealand diction hypothesis. I quantified every fruit produced that have deleterious effects on native plant communities. by 22 female trees from 2002 to 2009 in an old-growth Arguably the most destructive is the brush-tail possum (Tr i- forest on the North Island of New Zealand. Data were chosurus vulpecula), which has a particular preference for the leaves, flowers and fruits of D. spectabile (Nugent et al. 2002; used to answer four questions: (i) Is annual fruit pro- Buddenhagen & Ogden 2003). Prior to this study in 1993, a duction synchronized among individuals in the popu- major possum control programme was initiated in Otari- lation? (ii) Is among-individual variation in annual Wilton’s Bush (Otari Native Botanic Garden and Wilton’s variation in annual fruit production related to tree Bush Reserve 2007). Between July 1994 and June 1996, 50 diameter or total fecundity? (iii) Is annual variation in poisoned bait stations were established at even spatial inter- fruit production related to temperature and precipita- vals across the reserve, which initially destroyed over 1000 tion during the year-long period of fruit maturation, or possums. Only five possums were eliminated after a second fruit production in the previous fruiting season? (iv) more intensive poisoning programme in November 1998 (see Might seedling survivorship differ between years char- O’Leary 1999). Periodic poisoning has been conducted ever acterized by high and low fruit production? since and possum numbers have been maintained at nearly undetectable levels.

Field observations METHODS A stratified, random sampling design was used to select 60 trees along a 3-km trail that traversed the length of the Study site and species reserve. Five trees were randomly selected from within 12 sampling localities that were evenly distributed across the All data were collected in Otari-Wilton’s Bush, which is length of the trail. Each tree was permanently marked with a located on the southern tip of the North Island of New uniquely numbered metal tag and its diameter at breast Zealand (41°14′S, 174°45′E). Otari-Wilton’s Bush is a height was measured at the end of the observation period. forest reserve comprised of 75 ha of old-growth and regen- Sampling included a range of tree sizes but avoided repro- erating conifer-broadleaf forest (see Burns & Dawson 2005, ductively immature individuals (<10 cm diameter at breast Burns & Beaumont 2009 for detailed site descriptions). height and <5 m tall). The parent material is greywacke; steeper slopes are com- The sex of each tree could not be accurately ascertained prised of a stony colluvium and valley bottoms are covered when they were initially marked, so all trees were treated in silty loam. Mean annual temperatures average 12.8°C identically throughout the observation period. However, total and range from an average daily minimum of 6.3° in July to fruit set tallies gave a good indication of the sex of each tree, an average daily maximum of 20.6° in February. Annual as only ‘female’ trees produce significant amounts of fruit. I rainfall averages 1250 mm. Rainfall occurs year-round; specifically defined female trees as those that produced 10 or however most (~ 75%) falls in winter (May–September). more fruits throughout the study period and only female Frosts and snow are uncommon. Naturally occurring fires trees were included in individual-based statistical analyses are exceedingly rare and fire plays an insignificant role in (questions one and two in the Introduction). doi:10.1111/j.1442-9993.2011.02260.x © 2011 The Author Journal compilation © 2011 Ecological Society of Australia TREE MASTING 177

Most studies on the annual reproductive variability of survivorship. Seed mortality was defined by the total disap- individual trees have documented the fruiting behaviour of pearance of a seedling that previously germinated approximately 30 trees (mean = 32.5, range 5–120 individu- successfully. als, Koenig et al. 2003). Because previous work suggests that D. spectabile is dioecious (Braggins et al. 1999; Gardner 2007), I initially marked 60 trees in the hope of including 30 Statistical analyses trees of each sex to approximate the average sample size of most studies. However, at the close of observations, 22 trees To test whether annual variation in fruit production was were ‘female’. Therefore, the sample size of all individual- synchronized among trees, Spearman rank correlations based statistical analyses was n = 22 because of a male-biased were used to compare annual fruit production by each population sex ratio. female tree to summed fruit production by all other females Visual searches for ripe fruits were made with binoculars in the population. A non-parametric test was used because just prior to fruit dehiscence in July–August for eight con- most trees failed to set fruit in many years. This resulted in secutive fruiting seasons (2002–2009). Exact counts of fruits a large number of zeros in the dataset, which made trans- produced by each marked tree were facilitated by their large forming the annual distribution of fruit production to size and conspicuous location on tree trunks (i.e. cauliflory). conform to parametric assumptions impossible. A large Several trees were rooted to the forest floor by multiple number of significant, positive correlations and a high stems. For these trees, diameter and fruit count measure- average rank correlation coefficient would indicate synchro- ments were made separately for each stem and then summed. nous population-level fruit production. Mean annual syn- All branches produced by multi-stemmed individuals could chrony (rp) was also calculated as the mean of all pairwise be unambiguously assigned to a single individual. Pearson correlation coeffecients between individuals, fol- Analyses of the climatic correlates of population-level fruit lowing Koenig et al. (2003). production were conducted using precipitation and tempera- Two analytical approaches were used to test whether ture data obtained from New Zealand’s National Climate masting intensity of each female tree was associated with its Database (CliFlo 2010). Climatic observations were made size or fecundity. First, the coefficient of variation in annual daily throughout the study period at a location approximately fruit production was calculated for each female tree by divid- 6 km northwest of the study site. Because D. spectabile fruits ing the standard deviation of annual fruit production across take approximately one calendar year to mature, total daily the observational period by its average annual fruit produc- precipitation was summed from 1 August, when flowering tion (CVi). Multiple regression was then used to test whether usually terminated, to 31 July of the following year when fruit the CVi for each female was related to its diameter or total dehiscence commenced. Maximum daily temperatures were fecundity, which was defined as the sum of all fruits that it averaged similarly. produced throughout the study period. Data were natural logarithm transformed to conform to assumptions. Second, the scaling relationship between average annual fruit produc- Seed germination experiment tion by each female tree and its standard deviation in annual fruit production was evaluated. Because the standard devia- To evaluate whether rates of seed germination and seedling tion of any randomly sampled population increases passively survivorship might differ between high and low years of seed with its mean, if annual variation in fruit production varies production, I conducted a seed germination experiment passively with total fecundity, this relationship will be isomet- during the last 2 years of observations. Most female trees set ric (i.e. slope = 1, intercept = 0, see Taylor 1961). However, large amounts of fruit in 2007 and failed to set fruit in 2008. statistical differences in the slope and intercept of this rela- Therefore, the first cohort of seeds used in the experiment tionship from isometry (one and zero, respectively) would came from a high-seed year and the second came from a indicate annual variation in fruit production varies dispro- low-seed year. In both years, 100 seeds were collected hap- portionally with total fecundity. I therefore regressed average hazardly from 10 to 15 trees, from which 60 seeds were annual fruit production against the standard deviation of randomly selected, cleaned of pulp and stored in moistened average annual fruit production and obtained estimates of blotting paper placed within small plastic containers at 5°C the slope and intercept parameters using maximum likeli- for 3 months. Afterwards they were buried approximately hood in SPSS (2002), in addition to their 95% confidence 1 cm below the soil surface at 3-m intervals on alternating intervals (CI). sides of the entrance of same trail used in observations. Seeds Finally, I used multiple regression to test whether were sown within segments of 10 cm diameter plastic piping population-level fruit production could be predicted by tem- that were hammered into the forest floor so that their lip perature and precipitation during fruit maturation, or by projected 2 cm above the forest floor, to facilitate relocating total fruit production in the previous year. Total annual fruit seeds and seedlings in subsequent observations. A single seed production, summed among all 60 individuals in the popu- was sown in each plastic container in the spring of 2007 and lation, was used as the dependent variable. Total annual an additional 60 containers were placed immediately adja- precipitation and average maximum daily temperature were cent to these containers in spring 2008. These containers used as independent variables, in addition to summed fruit housed the second cohort of seeds, which were treated production among all individuals in the previous year of identically to the 2007 cohort. After transplantation, all observations. The first year of observations (2002) was seeds were monitored at 4-month intervals to record seed omitted from this analysis because fruit production in the germination (which was defined by the presence of two com- previous year was unknown.Therefore, the sample size of this pletely developed, undamaged cotyledons) and seedling analysis was n = 7.

Fig. 2. Standard deviation in annual fruit production versus average annual fruit production in 22 Dysoxylum spectabile trees. The slope of the observed relationship is significantly less than one and the intercept is significantly greater than zero, indicating trees with lower total fruit production Fig. 1. Fruit set by 22 Dysoxylum spectabile trees from 2002 display stronger masting cycles. to 2009. Individual trees are listed as rows, years are listed as columns and symbols are proportional to the total number of significantly positive (i.e. P < 0.05). Average individual fruits produced by each tree each year following Koenig et al. synchrony (the mean of all pairwise correlation coeffi- (2003). The far right column lists the coefficient of variation Ϯ in annual fruit production for each tree across the study cients, rp) was 0.614 0.269 (range =-0.21, 0.98). period (CVi) and the bottom row lists the total number of Annual reproductive variability declined with total fruits produced by all 22 trees each year.The last value in the fruit production.The overall multiple regression model last column is the population-level coefficient of variation in relating CVi to tree diameter and total fruit production annual fruit production (CVp). was significant (r2 = 0.357, P = 0.006). However, only total fruit production contributed to the model (b=-0.672, P = 0.002), tree diameter was unimpor- RESULTS tant (b=0.276, P = 0.148). Total fecundity was also unrelated to tree diameter (b=0.084, P = 0.180).The Of the 60 randomly selected trees that were marked for relationship between the standard deviation in annual observation, 38 produced fewer than 10 fruits through- fruit production and average annual fruit production out the study period, suggesting they were male.There- confirmed this result (Fig. 2). The standard deviation fore, the population appeared to be male-biased, with increased with the mean (r2 = 0.947, P < 0.001). approximately one-third of the population (22 trees) However, the intercept was greater than zero (0.933, being female. Female trees produced an average of 95% CI = 0.514–1.352) and the slope was less than one 129.3 Ϯ 236.7 SD fruits per year (Fig. 1).Totalaverage (0.855, 95% CI = 0.763–0.947), indicating less fecund annual fruit production across the population ranged trees exhibit higher annual reproductive variability than from 0.3 fruits per individual in 2008 to 350.0 fruits per more fecund trees. individual in 2009.The average coefficient of variation Annual fruit production was positively related to among trees (mean CVi) was 139, while the population- precipitation during annual periods of fruit develop- level coefficient of variation (CVp) was 104. ment and negatively related to fruit production in the Fruit production was highly synchronized among previous year (Fig. 3). The overall multiple regression female trees. Annual fruit production in 77% of model relating annual fruit production to annual pre- female trees (17 out of 22) was positively correlated cipitation, mean maximum air temperature and the (P < 0.05) with total annual fruit production by previous year’s fruit production was significant the rest of the female population (average correlation (r2 = 0.822, P = 0.044). Both precipitation (b=0.677, coefficient = 0.770 Ϯ 0.180, range = 0.250–0.980). P = 0.039) and previous reproductive effort (b= Out of the total 231 pairwise comparisons among indi- -0.650, P = 0.034) contributed to the model. Tem- viduals, none was significantly negative and 107 were perature was unimportant (b=0.177, P = 0.429). doi:10.1111/j.1442-9993.2011.02260.x © 2011 The Author Journal compilation © 2011 Ecological Society of Australia TREE MASTING 179

Fig. 3. Relationships between annual fruit production, fruit production in the previous year (left) and total annual precipitation during fruit development (right) in a population of Dysoxylum spectabile from 2003 to 2009. In both instances, the effect of the other independent variable was removed; residuals of the relationship between annual fruit production and total annual precipitation is shown at left and residuals of the relationship between annual fruit production and annual fruit production is shown at right.

did not differ statistically between years (c2 = 1.9, P > 0.100). Approximately 80% of 2007 seedlings were still alive 1 year after planting and 60% were still alive 2 years after planting, while none of the 2008 cohort survived for longer than 4 months. Seedling survivorship in the ﬁrst year after planting violated the assumption of equal survivorship between cohorts (c2 = 34.0, P < 0.001).

DISCUSSION

Approximately two-thirds of the study population produced little or no fruit throughout the study period. Fig. 4. Results of a seed germination and seedling survi- This suggests that they were functionally male and the vorship experiment conducted with seeds produced during a mast year (white symbols) and a non-mast year (black remainder of the population was female, which sup- symbols). Symbols represent total per cent survivorship at ports previous work indicating that D. spectabile is 4-month intervals and numbers beneath symbols refer to the dioecious (Braggins et al. 1999; Gardner 2007). Nev- total number of living seedlings. Over 80% of the 60 seeds ertheless, many apparently ‘male’ trees occasionally that were sown in the mast year (2007) germinated, of which produced small numbers of fruit (<10). Every effort 50% were alive 2 years later. Over 60% of the 60 seeds that was made to insure accurate fruit counts, but D. were sown in the non-mast year (2008) germinated, all of which were dead 4 months later. spectabile trees often grow quite close to one another with interspersed canopies, so the observation of fruits on apparently ‘male’ trees could have resulted from Seed germination and seedling establishment were small sampling errors. Alternatively, incomplete sexual higher in the high-seed year (2007) than in the low- segregation, whereby some male trees occasionally seed year (2008) (Fig. 4). Over 70% of seeds planted produce viable female ﬂower parts has been docu- in spring 2007 (43 out of 60 seeds) germinated, while mented experimentally in other plant species just over 50% of seeds planted in spring 2008 (31 out (Humeau et al. 1999; Venkatasamy et al. 2007). This of 60 seeds) germinated. Assuming an equal probabil- phenomenon is known as ‘leaky dioecy’ and appears to ity of germination between cohorts, germination rates be particularly common on isolated islands.

The timing of fruit production was tightly synchro- cal rainforests (Dawson 1988).While D. spectabile trees nized among female trees. Synchronous fruit produc- can grow to substantial heights, there are many other tion is a hallmark of mast-fruiting plant populations tree species that grow much taller. So fully mature D. and often operates over large spatial scales (Koenig & spectabile trees often occur beneath taller, canopy- Knops 1998; Schauber et al. 2002; Liebhold et al. emergent species such as Beilschmiedia tawa, Knightia 2004; Lamontagne & Boutin 2007; Masaki et al. 2008; excelsa and Dacrydium cupressinum, which can shade Sanguinetti & Kitzberger 2008). On the other hand, them significantly and presumably lower the rate at the coefficient of variation in annual fruit set declined which they fix carbon.Trees in New Zealand forests can linearly with total fruit production. Therefore, more also harbour high liana loads, which might compete fecund trees exhibited less pronounced annual repro- with larger trees for light and subsequently lower energy ductive variability. This result was not biased by the levels available for reproduction.Alternatively, the weak two least fecund trees, which produced 54 and 92 relationship between fecundity and tree size may result fruits, respectively. With these two trees omitted from from sampling errors,as measurements of tree diameter analyses, CVi still declined with total fruit production are only coarse estimates of overall tree size (see Phillips (r2 = 0.189, P = 0.032), and the standard deviation of et al. 2002). fruit production in individual trees failed to scale iso- Annual fruit production in female trees was nega- metrically with total fruit production (r2 = 0.947, tively related to fruit production in the previous year, P < 0.00, intercept = 0.901, 0.317–1.484 95% CI, indicating that trees ‘rest’ following high-seed years. slope = 0.862, 0.738–0.986 95% CI). This pattern is common in many mast-fruiting species Explanations for mast-fruiting based on economies and suggests that trees alternate between growth and of scale, such as the predator satiation and pollination reproduction between years (see Koenig & Knops efficiency hypotheses, predict that reproductive fitness 1998). However, ‘switching’ between growth and accelerates with synchronized reproductive effort in reproduction could not be investigated per se because plant populations. If reproductive fitness in D. spect- D. spectabile does not produce annual growth rings (K. abile varied according to an economy of scale annual C. Burns, 2006, pers. obs.), so annual growth rates reproductive variability should increase with fecundity. could not be compared with annual rates of reproduc- However, the opposite pattern was observed. Annual tion. Nevertheless, lower fruit production in years fol- reproductive variability declined with total reproduc- lowing high-seed years falsifies the resource-matching tive effort, suggesting that other processes are at work. hypothesis, which postulates that inter-annual variation Several native birds prey upon D. spectabile seeds in fruit production results from inter-annual variation (e.g. kaka Nestor meridionalis), but I saw no evidence of in climatic conditions alone (see Monks & Kelly 2006). avian seed depredation throughout the observational High-seed years were correlated with precipitation, period. The seeds and fruits of D. spectabile are heavily indicating that fruits matured immediately following predated by introduced mammals, particularly the periods of higher rainfall. Most mast-fruiting species in brush-tailed possum (T.vulpecula). Given the extensive New Zealand set fruit following high summer tem- effort to control possum numbers in Otari-Wilton’s peratures (Schauber et al. 2002), which suggests that Bush, it seems unlikely that results are biased by the trigger for the production of large fruit crops in D. possum predation. It is doubtful that possums have spectabile might be unusual. Studies in other localities influenced the evolution of mast-fruiting in D. spect- have also documented that large fruit crops are asso- abile, given the short period the two species have been ciated with precipitation (e.g. Piovesan & Adams in contact. I also saw no evidence of seed predation 2001). However, in these instances drought triggers by arthropods in any of the seeds collected for the large fruit crops. Higher rainfall during the build-up to germination experiment. Native birds such as tui (P. seed release raises the possibility that D. spectabile novaeseelandiae) and waxeyes (Z. lateralis) interact might adjust reproductive effort according to the soil mutualistically with D. spectabile as both pollinators conditions that seeds are likely to encounter following and seed dispersers. But given that mass flowering and dispersal. Forests with closed canopies and deep soil fruiting is likely to satiate pollinators and frugivores, horizons can store large amounts of water after heavy mutualists are likely to select against annual reproduc- rain, and sustained periods of drought are often nec- tive variability (Kelly & Sork 2002). essary to lower soil water potential (see Dungan et al. Both total fecundity and annual reproductive vari- 2007). Given the time lag between the entry and exit ability were unrelated to tree diameter. This is some- of precipitation in forested ecosystems, tree species what surprising because larger trees in the forest canopy with water-demanding seeds and seedlings could be at that are exposed to full sunlight should have higher rates a selective advantage if they released their seeds imme- of carbon fixation than smaller trees in the forest under- diately after wetter periods. If D. spectabile seeds are story. One explanation for this result stems from the sensitive to drought, high-seed years following periods vertical complexity of broadleaf-Podocarp forests in of elevated rainfall could help trees ‘anticipate’ periods New Zealand, which are as structurally diverse as tropi- that are beneficial for recruitment. doi:10.1111/j.1442-9993.2011.02260.x © 2011 The Author Journal compilation © 2011 Ecological Society of Australia TREE MASTING 181

Several lines of evidence suggest that D. spectabile ment of white spruce trees in western North America seeds are indeed sensitive to drought. First, Burrows increases substantially when high-seed years coincide (1999) investigated the germination behaviour of D. with fire (Peters et al. 2005, see also Boutin et al. 2006). spectabile seeds experimentally and found that they Others have suggested that the environmental predic- failed to germinate after they were separated from fruit tion hypothesis may not always be restricted to fire- pulp and left to air dry for several days. Therefore, prone ecosystems. For example, dipterocarps in successful germination would seem to require moist Southeast Asian rainforests typically set fruit after conditions immediately following seed dispersal. ENSO events, which often damage or kill adult trees Second, results from experimental seed plantings thus providing forest gaps for seedling recruitment in this study showed that rates of seedling survivor- (Williamson & Ickes 2002). A similar situation may ship were higher during the high-seed year, which occur with D. spectabile. I saw no evidence that adults experienced higher rainfall coincidentally with seed were sensitive to drought, but their seeds and seedlings sowing (188.9 mm from September to November in appear to be drought sensitive, suggesting that timing the 2007 vs. 133.6 mm in 2008). Germination rates high fruit production to coincide with periods of heavy were statistically similar between years, perhaps indi- rainfall may increase the probability of recruitment in cating that drought is not an important factor deter- the absence of marked adult mortality. mining seed mortality. However, both cohorts of seeds Overall, results were inconsistent with the resource- were stored in cool, moist conditions in the laboratory matching hypothesis and various hypotheses based on in an effort to break seed dormancy prior to seed economies of scale, but were consistent with the envi- sowing. In hindsight, planting seeds immediately would ronmental prediction hypothesis. Like most investiga- have yielded useful information on the link between tions of the evolution of mast-fruiting however, data germination and precipitation. The experiment was provide only correlative evidence for the processes also unreplicated among years. Direct support for a responsible for annual reproductive variability (cf. positive effect of precipitation on long-term trends in Lazaro et al. 2006). Additional experiments on tree recruitment would require many more years of precipitation-driven recruitment patterns and longer- trials wherein seeds were planted immediately follow- term observations of fruit abortion are needed to test ing collection. the environmental prediction hypothesis more Finally, high-seed years typically occurred in alter- thoroughly. Support for this hypothesis also hinges on nate years with the exception of 2006, when the whole identifying a physiological mechanism that would fruit crop appeared to be aborted. Most trees ‘rested’ enable females to detect periods of drought during the in 2005, suggesting they had sufficient energetic build-up to seed release, which is currently unknown. resources for high fruit production in 2006. Both sexes Nevertheless, results clearly indicate that the environ- flowered profusely and female trees produced large mental prediction hypothesis is a viable explanation for crops of unripe fruits, yet annual precipitation was low. the evolution of elevated annual reproductive variabil- Just prior to fruit maturation most trees appeared to ity outside fire-prone ecosystems. withdraw water and nutrients from fruit capsules, which became unusually brittle and discoloured.When ACKNOWLEDGEMENTS fruits dehisced, they only split-open partially and the fleshy arils within were desiccated and blackened. Thanks to Otari-Wilton’s Bush staff for advice and Other mast-fruiting species are known to abort fruits permission to conduct the study and to two anony- during adverse weather conditions (Espelta et al. mous referees who provided constructive comments 2008). A similar circumstance may occur with D. spec- on an earlier draft. Funding was provided by Victoria tabile, which may abort fruit crops in drier years to University of Wellington. avoid drought-induced seed and seedling mortality. Results were generally consistent with the environmental prediction hypothesis, which postulates that REFERENCES mast-fruiting has evolved to promote recruitment Boutin S., Wauters L.A., McAdam A. G., Humphries M. M., Tosi during years that are favourable for seed and seedling G. & Dhondt A. A. (2006) Anticipatory reproduction and survivorship. Previous support for the environmental population growth in seed predators. Science 314, 1928–30. prediction hypothesis comes mostly from fire-prone Braggins J. E., Large M. F. & Mabberley D. J. (1999) Sexual ecosystems and is believed to be less important than arrangements in kohekohe (Dysoxylum spectabile, the predator satiation and pollination facilitation Meliaceae). Telopea 8, 315–24. hypotheses (Kelly 1994; Koenig et al. 2010). For Buddenhagen C. E. & Ogden J. (2003) Growth and survival of Dysoxylum spectabile (Meliaceae) seedlings in canopy gaps. example, semelparous bamboos typically mast-flower N.Z. J. Bot. 41, 179–83. after adult plants grow for long periods and build-up Burns K. C. & Beaumont S. (2009) Scale-dependent trait cor- large fuel loads, which when burned might create space relations in a temperate tree community. Austral Ecol. 34, for seedlings (Keeley & Bond 1999). Similarly, recruit- 670–7.

Burns K. C. & Dawson J.W. (2005) Patterns in the diversity and Liebhold A., Sork V. L., Peltonen M. et al. (2004) Within- distribution of epiphytes and vines in a New Zealand forest. population spatial synchrony in mast seeding of North Austral Ecol. 30, 891–9. American oaks. Oikos 104, 156–64. Burrows C. J. (1999) Germination behaviour of seeds of the New Masaki T., Oka T., Osumi K. & Suzuki W. (2008) Geographical Zealand woody species Beilschmiedia tawa, Dysoxylum spect- variation in climatic cues for mast seeding of Fagus crenata. abile, Griselinia lucida, and Weinmannia racemosa. N.Z. J. Bot. Popul. Ecol. 50, 357–66. 37, 95–105. Monks A. & Kelly D. (2006) Testing the resource-matching CliFlo (2010) National Institute of Water and Atmospheric hypothesis in the mast seeding tree Nothofagus truncata Research, New Zealand. [Cited 2009.] Available from URL: (Fagaceae). Austral Ecol. 31, 366–75. http://cliflo.niwa.co.nz/ Nilsson S. G. & Wästljung U. (1987) Seed predation and cross- Dawson J. W. (1988) Forest Vines to Snow Tussocks: The Story of pollination in mast-seeding beech (Fagus sylvatica) patches. New Zealand Plants. Victoria University Press, Wellington. Ecology 68, 260–5. Dungan R. J., Turnbull M. H. & Kelly D. (2007) The carbon Norden N., Chave J., Belbenoit P. et al. (2007) Mast fruiting is a costs for host trees of a phloem-feeding herbivore. J. Ecol. frequent strategy in woody species of Eastern South 95, 603–13. America. PLoS ONE 2, e1079. Espelta J. M., Cortés P., Molowny-Horas R., Sánchez-Humanes Norton D. A. & Kelly D. (1988) Mast seeding over 33 years by B. & Retana J. (2008) Masting mediated by summer Dacrydium cuppressinum Lamb. (rimu) (Podocarpaceae) in drought reduces acorn predation in Mediterranean oak New Zealand: the importance of economics of scale. Funct. forests. Ecology 89, 805–17. Ecol. 2, 399–408. Gardner R. O. (2007) Kohekohe (Dysoxylum spectabile)inflower Nugent G., Whitford J., Innes J. & Prime K. (2002) Rapid and fruit. Auckl. Bot. Soc. J. 62, 95–6. recovery of kohekohe (Dysoxylum spectabile) following Herrera C. M., Jordano P., Guitian J. & Traveset A. (1998) possum control. N.Z. J. Ecol. 26, 73–9. Annual variability in seed production by woody plants and O’Leary T. (1999) Revegetation and Forest Management: Otari the masting concept: reassessment of principles and rela- Wilton’s Bush. Wellington City Council, Wellington. tionship to pollination and seed dispersal. Am. Nat. 152, Otari Native Botanic Garden and Wilton’s Bush Reserve (2007) 576–94. Unpubl. report. [Cited 2009.] Available from URL: http:// Humeau L., Pailler T. & Thompson J. D. (1999) Cryptic dioecy www.wellington.govt.nz/plans/policies/otari/pdfs/02-back. and leaky dioecy in endemic species of Dombeya (Sterculi- pdf: aceae) on La Réunion. Am. J. Bot. 86, 1437–47. Peters V. S., Macdonald S. E. & Dale M. R. T. (2005) The Janzen D. H. (1971) Seed predation by animals. Annu. Rev. Ecol. interaction between masting and fire is key to white spruce Evol. Syst. 2, 465–92. regeneration. Ecology 86, 1744–50. Keeley J. E. & Bond W.J. (1999) Mast flowering and Semelparity Phillips O. L., Malhi Y., Vinceti B. et al. (2002) Changes in in bamboos: the bamboo fire cycle hypothesis. Am. Nat. growth of tropical forests: evaluating potential. Ecol. Appl. 154, 383–91. 12, 576–87. Kelly D. (1994) The evolutionary ecology of mast seeding. Trends Piovesan G. & Adams J. M. (2001) Masting behaviour in beech: Ecol. Evol. 9, 465–70. linking reproduction and climatic variation. Can. J. Bot. 79, Kelly D. & Sork V. L. (2002) Mast seeding in perennial plants: 1039–47. why, how, where? Annu. Rev. Ecol. Syst. 33, 427–47. Piovesan G. & Adams J. M. (2005) The evolutionary ecology Kelly D., Harrison A., Lee W., Payton I., Wilson P. & Schauber of masting: does the environmental prediction hypothesis E. (2000) Predator satiation and extreme mast seeding also have a role in mesic temperate forests? Ecol. Res. 20, in 11 species of Chionochloa (Poaceae). Oikos 90, 477– 739–43. 88. Sanguinetti J. & Kitzberger T. (2008) Patterns and mechanisms Kelly D., Hart D. E. & Allen R. B. (2001) Evaluating the wind of masting in the large-seeded southern hemisphere conifer pollination benefits of mast seeding. Ecology 82, 117–26. Araucaria araucana. Austral Ecol. 33, 78–87. Kitzberger T., Chaneton E. J. & Caccia F. (2007) Indirect effects Schauber E. M., Kelly D., Turchin P. et al. (2002) Masting by of prey swamping: differential seed predation during a eighteen New Zealand plant species: the role of temperature bamboo masting event. Ecology 88, 2541–54. as a synchronisation cue. Ecology 83, 1214–25. Koenig W. D. & Knops J. M. H. (1998) Scale of mast- seeding Silvertown J. (1980) The evolutionary ecology of the mast and tree-ring growth. Nature 396, 225–6. seeding in trees. Biol. J. Linn. Soc. 14, 235–50. Koenig W. D. & Knops J. M. H. (2000) Patterns of annual seed Smith C. C., Hamrick J. L. & Kramer C. L. (1990) The advantage production by Northern Hemisphere trees: a global of mast years for wind pollination. Am. Nat. 136, 154–66. perspective. Am. Nat. 155, 59–69. Sork V. L., Bramble J. & Sexton O. (1993) Ecology of mast- Koenig W. D., Kelly D., Sork V. L. et al. (2003) Dissecting fruiting in three species of Missouri oaks, Quercus alba, components of population-level variation in seed produc- Quercus rubra & Quercus velutina (Fagaceae). Ecology 74, tion and the evolution of masting behavior. Oikos 102, 581– 528–41. 91. SPSS (2002) Release 1151 for Windows. SPSS Inc, Chicago. Koenig W.D., Knops J. M. H. & Carmen W.J. (2010) Testing the Taylor L. R. (1961) Aggregation, variance and the mean. Nature environmental prediction hypothesis for mast-seeding in 189, 732–5. California oaks. Can. J. For. Res. 40, 2115–22. Venkatasamy S., Khittoo G. & Keeley S. (2007) Leaky dioecy in Lamontagne J. M. & Boutin S. (2007) Local-scale synchrony Diospyros (Ebenaceae) endemic to the Island of Mauritius. and variability in mast seed production patterns of Picea Plant Ecol. 189, 139–46. glauca. J. Ecol. 95, 991–1000. Wardle P. (2002) Vegetation of New Zealand. Blackburn Press, Lazaro A., Traveset A. & Mendez M. (2006) Masting in Buxus Caldwell. balearica: assessing fruiting patterns and processes at a large Williamson G. B. & Ickes K. (2002) Mast fruiting and ENSO spatial scale. Oikos 115, 229–40. cycles – does the cue betray a cause? Oikos 97, 459–61. doi:10.1111/j.1442-9993.2011.02260.x © 2011 The Author Journal compilation © 2011 Ecological Society of Australia