Implications of a Long-Term Mast Seeding Cycle for Climatic

Received: 6 August 2018 Revised: 10 February 2019 Accepted: 25 February 2019 DOI: 10.1111/1440-1703.12014

ORIGINAL ARTICLE

Implications of a long-term mast seeding cycle for climatic entrainment, seedling establishment and persistent monodominance in a Neotropical, ectomycorrhizal canopy tree

Terry W. Henkel1 | Jordan R. Mayor2

1Department of Biological Sciences, Humboldt State University, Arcata, Abstract California In Guyana, we examined temporal variability and ecological consequences of seed- 2Environment and Planning Division, ICF, based reproduction in the ectomycorrhizal, monodominant leguminous canopy tree San Francisco, California Dicymbe corymbosa using an extensive 2003–2017 dataset encompassing two

Correspondence masting events. Annual seed output, predation and nutrient investment were Terry W. Henkel, Department of Biological recorded in primary D. corymbosa forests. Seedling establishment, survival and Sciences, Humboldt State University, growth were monitored. Mast seeding occurred in 2003 and 2016 and low seeding Arcata, CA 95521. Email: [email protected] events occurred every 3–4 years, separated by non-reproductive years. In accordance with predicted climatic entrainment of regional masting, El Niño-intensified Funding information Humboldt State University; National dry seasons preceded each masting event but not low seeding events. The regularly Geographic Society, Grant/Award Numbers: variable seeding pattern by D. corymbosa suggests that innate supra-annual period- 7435-03, 8481-08; National Science icity may initiate D. corymbosa flowering. Internal resource thresholds may also Foundation, Grant/Award Numbers: DEB- 0918591, NSF DEB-1556338 contribute to periodicity, but El Niño-associated intensified dry seasons appear necessary to entrain mass flowering and successful mast seeding across the region. In accordance with a predator satiation hypothesis, seed predation was proportionally higher during low seed years. Competitive gains from the 2016 masting event were evidenced by strong replenishment of declining seedling pools.

KEYWORDS Detarioideae, Dicymbe, El Niño, Fabaceae, Guiana Shield

1 | INTRODUCTION Fabaceae (Henkel, Mayor, & Woolley, 2005; Newbery, Chuyong, & Zimmerman, 2006). Mast seeding (or “masting”) is the supra-annual synchro- Dicymbe corymbosa Spruce ex Benth. (Fabaceae subfam. nous heavy fruiting of a plant (Kelly & Sork, 2002). Mast Detarioideae) is a canopy tree from the central Guiana Shield seeding is exhibited primarily by iteroparous perennial that provides a classic example of mast seeding in a mono- woody plants and is best known from temperate ecosys- dominant tree in lowland tropical moist forest (Henkel, tems (e.g., many Fagaceae, Pinaceae; Crone & Rapp, 2003; Henkel et al., 2005). The species meets the require- 2014; Herrera, Jordano, Guitlan, & Traveset, 1998; ments for “persistent monodominance” as it dominates Kelly & Sork, 2002; Koenig & Knops, 2000; Sork, 1993). >60% of the over-, mid-, and understory individuals in a While a recent masting review failed to elaborate on tropi- stand, and lacks edaphic specialization (Connell & Lowman, cal mast seeding (Pearse, Koenig, & Kelly, 2016), signifi- 1989). Dicymbe-dominated forests cover much of the lower cant examples are known in Asian Dipterocarpaceae elevation terrain in the central Pakaraima Mountains of Guy- (Ashton, 1987; Janzen, 1974) and Neo- and Afrotropical ana (Degagne, Henkel, Steinberg, & Fox, 2009; Fanshawe,

1952). Dicymbe corymbosa exhibits several life history traits extended period encompassing non-reproductive, low seed which, taken together, are unique among tropical tree spe- and mast seeding years?, (b) With comparable datasets cies worldwide. These include: (a) the ectomycorrhizal between mast years and low seed years, is there evidence for (ECM) habit (Henkel, Terborgh, & Vilgalys, 2002), a predator-satiation evolutionary mechanism?, and, (c) What (b) broad edaphic amplitude (Henkel, 2003), (c) cumulative are the impacts of a long inter-mast period on D. corymbosa shoot and root reiterative growth leading to large, multi- seedling populations and persistent monodominance? stemmed, clonal individuals of indeterminate age (Woolley, Henkel, & Sillett, 2008) and (d) episodic, large-scale sexual recruitment through mast seeding (Henkel et al., 2005). This 2 | MATERIALS AND METHODS collusion of evolved traits allows partial escape from density-dependent effects on recruitment, control over 2.1 | Study site above- and belowground resources, and competitive exclu- The study was conducted in monodominant D. corymbosa sion of sympatric non-ECM tree species, ensuring forest along the Upper Potaro River in Guyana's Pakaraima D. corymbosa spatiotemporal dominance (e.g., Henkel et al., Mountains. Study plots were located within a 15-km radius 2005; McGuire et al., 2008; Smith et al., 2017; Woolley of a permanent base camp near the Potaro River (518004.800 et al., 2008). 0 00 – Henkel et al. (2005) detailed the 2003 mast seeding event N; 59 54 40.4 W), with elevations of 700 750 m above sea in the central Pakaraima Mountains of Guyana. This mast level. The site was situated in an undulating valley 20 km featured regional synchrony, high stand-level reproductive east of Mt Ayanganna, and was densely forested with pri- resource investment, and ballistic/barochorous seed dis- marily Dicymbe-dominated and to a lesser extent mixed for- – persal. Density-dependent mortality effects were low, and a ests of the Eschweilera Licania association (Degagne et al., large shade-tolerant seedling cohort was established. 2009; Fanshawe, 1952). In the D. corymbosa- Regional entrainment of this masting event was posited to monodominant stands no other ECM tree species occur be due to an intensified dry season coincident with the (McGuire et al., 2008). Regional synchrony of 2002–2003 El Niño Southern Oscillation (hereafter “El D. corymbosa seed production was assessed along a 37-km Niño”) event. However, there remains uncertainty with transect from the Upper Potaro to the Upper Ireng River regards to climatic initiation and entrainment of (Henkel et al., 2005). Details of climate, geology, soils, for- D. corymbosa seeding events and the ecological importance est structure and the 2003 masting event can be found in of mast and non-mast years in long-term seedling Henkel et al. (2002, 2005) and Henkel (2003). recruitment. Because of uncertainty in the causes of masting, including the roles of resource, biotic and climatic factors in driv- 2.2 | Study plots ing regional entrainment (Crone & Rapp, 2014; Koenig, In 2003, five 50× 50 m2 (0.25-ha) study plots (P1–P5) were Knops, Carmen, & Pearse, 2015; Pearse et al., 2016), we randomly positioned within 2 km of the Potaro base camp in planned to track the reproductive behavior of D. corymbosa larger areas of monodominant D. corymbosa forest (Henkel through a complete masting cycle. This would allow accu- rate determination of the coefficient of variation of seeding et al., 2005). Two additional Upper Potaro 0.25-ha plots – in the species (sensu Kelly, 1994) for comparison to other (P6 P7) were added in 2005, with P6 located 1.5 km south- masting species, assessment of relationships of flowering east of the base camp, and P7 located 5 km southwest of the with regional weather cues, and shed light on the competi- base camp. In 2009, three spatially distant 0.25-ha plots – tive consequences of variable seeding. Following the 2003 (P8 P10) were added along the 37-km Potaro-Ireng transect. mast, data were obtained annually over an unexpectedly P8 was located 15 km east of the base camp, P9 was located long inter-mast period spanning 2004–2015, within which 10 km east of the base camp and P10 was located 9 km east three low seed years occurred. With the super El Niño and of the base camp, all on low, flat ridges. In July 2016, an teleconnected intensified dry season conditions for northern additional spatially distant plot (P11) was established 37 km South America arising in late 2015 and extending into 2016, south of the base camp on a slope with brown sand soil near we predicted that a D. corymbosa mass flowering would the Upper Ireng River. Similar to P1–P5, P6–P11 were occur in February–March of 2016, followed by a mast heavily dominated by D. corymbosa with soils that were uni- seeding event, and designed field work accordingly. In addi- formly acidic and low in extractable P (<12 ppm; Henkel tion to fully documenting the 2016 mast, we specifically et al., 2005). Twenty 3 × 3m2 (9 m2) seed traps were ran- asked: (a) Is there a proximal relationship between weather domly positioned on the ground in each study plot to annu- variables, seeding patterns and resource relations over an ally assess D. corymbosa seedfall. HENKEL AND MAYOR 3

2.3 | Inter-mast monitoring 2004 to 2015 2.5.2 | Seedfall and predation Seeds and fruit valves were counted in each seed trap at the Undamaged seeds and fruit valves were removed and coun- mid-July peak dispersal period and during the end of the dis- ted from each seed trap in P1–P7 at 5-day intervals from persal period in late August. In most years, all traps had no 20 June to 28 July, then in early September after the end of or very few seeds or valves, and these were considered non- seed dispersal. Given the ballistic/barochorous mechanism reproductive years. In 2006, 2009 and 2013, it was evident of seed dispersal in D. corymbosa (described in Henkel that a small scale but synchronous seeding event was occur- et al., 2005), seed and valve removal every 5 days from an ring and seed traps were visited more frequently between individual trap ensured that all seeds and valves present in mid-July to early September. Seeds and valves were counted the trap at the next visit were mature and had fallen in the and post- and pre-dispersal predation was recorded using previous 5 days, and allowed summing of the total inputs methods described below. Reproductive biomass and min- over the entire sampling period (see Zagt, 1997 for the simi- eral nutrient investment for the 2006, 2009 and 2013 low lar congener Dicymbe altsonii). Damaged seeds and fruit seed years were calculated using the mean seed and valve valves were separated and sorted into predation categories dry weights taken from 100 bulk grab samples of composite and counted at each sampling event, based on direct obser- collections made across plots P1–P5 following the methods vation of predators and/or damage symptoms. Pre- and post- described below. In each of the original masting plots P1– dispersal predation categories were: (a) ant feeding on seeds P5, new seedlings established during these low seed years in (primarily harvester ants; post-dispersal), (b) larval bore the 2 × 4.5 m2 (9 m2) survivorship subplots (up to 10 per holes in fruit valves and seeds (putatively Bruchidae, Cole- plot; Henkel et al., 2005) were tagged and survivors counted optera; pre-dispersal; Janzen, 1969), (c) bird-diagnostic tears annually up through July 2017, along with surviving seed- and rips of fruit valves (macaw and parrot; pre-dispersal), lings and saplings from the pre-2003 and 2003 cohorts. For (d) fungal hyphal growth and putrescence of seeds in the logistical reasons, survivorship was not sampled at 96 and absence of animal damage (pre- and post-dispersal) and 132 months. (e) terrestrial vertebrate tooth marks on seeds (post-dispersal). For the spatially distant plots P8–P10, these counts were performed in early and late July and early September. 2.4 | Leaf litterfall The most distant plot, P11, was counted once for all seeds To provide comparative data on D. corymbosa vegetative and fruit valves in early September. Subsequently, seed traps resource allocation and detrital inputs, in June 2004, twenty in all plots were revisited 1 year post-mast on 5 July and 1-m2 wire litter traps were suspended on posts 1.5 m above 30 August. the ground and randomly positioned in each of plots P1–P6. Leaf litter was collected from these traps at approximately 2.5.3 | Reproductive resource investment 3-month intervals up to June 2008. From the individual traps Freshly fallen mature seeds and fruit valves (n = 200 for air-dried, D. corymbosa leaves were separated from heter- each) were randomly collected from various plots in mid- ospecific leaves and dried further for a minimum of 72 h and July, air-dried, further dried with low radiant heat (ca. 40C) weighed. A random sample (n =5)oftheD. corymbosa over campfire coals for a minimum of 72 hr, weighed and leaves were elementally analyzed using methods described mean dry weight was calculated for each component. These below. Sample dry weights and mineral nutrient concentra- means were used to estimate stand level seed and valve dry tions were used to calculate annual leaf litter biomass and − mass by multiplying the mean dry weight by the mean num- minerals on a kg ha 1 basis. ber (m−2) of each respective fruit part, then extrapolated to a kg ha−1 basis. To obtain an estimate of stand-level flower 2.5 | 2016 mast flowering, seedfall, predation, dry mass, ten 1-m2 subplots were randomly positioned on reproductive resource investment and seedling the ground within each plot and previously fallen buds, brac- establishment teoles and perianth parts were collected, dried, weighed and 2.5.1 | Flowering extrapolated as above. Four samples each of seeds, valves, buds, bracteoles and In late February 2016, the proportion of sun-exposed crown perianth parts were randomly selected and dried in air-tight branches bearing flowers was recorded for all D. corymbosa containers with silica beads for a minimum of 36 hr until a canopy trees in P1–P10. Frequency and intensity of pollina- constant weight was obtained. These were subsequently ele- tor visitation during full anthesis was also qualitatively mentally analyzed for N concentration by flash combustion noted. Abscission of flower parts other than ovaries was and gas chromatography (AOAC, 2003), and P, Ca and Mg complete by mid-March. concentration by acid digest and emission spectrometry 4 HENKEL AND MAYOR

(Gavlak, Horneck, Miller, & Kotuby-Amacher, 2003). miners. Monthly OLR anomalies are positively correlated Reproductive mineral investment for the 2016 mast was esti- with reduced cloud cover and increased solar irradiance of mated for each plot by multiplying the mean nutrient con- the forest canopy, and negatively correlated with precipita- centration of each reproductive component by the mean dry tion, in northeastern South America in the months following mass quantity (m−2) of that component extrapolated to a kg strong El Niño events in the central Pacific (Kogan, 2000; ha−1 basis. Lyon & Barnston, 2005). We examined the relationship between summed monthly 2.5.4 | Seedling establishment and survival OLR and precipitation anomalies that were apparent during the period of August to March immediately preceding each New 2016 mast seedlings were tagged and counted in known 1998–2016 seeding event (i.e., three mast and three September 2016 in each of the P1–P5 seedling survivorship low seeding events). In the central Pakaraima Mountains, subplots established in 2003 (Henkel et al., 2005). Total area this period encompasses the beginning of two typically short sampled in these long-term survivorship subplots was dry seasons (August–October and February–March) and 469 m2 (n = 10 9-m2 subplots for P1–P4; n = 31 3-m2 sub- through the period of flower anthesis (late February), polli- plots for P5). Some subplots were lost due to tree falls. Sur- nation, and the beginning of fruit development (early viving 2016 mast seedlings were recensused at 6- and March). The increased canopy irradiance and associated 12-month intervals in 2017. photoassimilative boosts could stimulate reproductive investment during these months (Graham, Mulkey, Kitajima, 2.6 | Seedling growth 2007–2017 Phillips, & Wright, 2003). Anomalies were also summed from the backcasted 2 years prior August to March period Annually within P1–P5 from 2007–2010, and subsequently to account for the possibility of current year flower initia- in 2017, up to six seedlings from each of the pre-2003 seed- tion responding to the previous year's environmental ling and 2003 seedling cohorts were randomly selected cues, a proximal relationship implicated in other masting within each survivorship subplot and their height measured species (Kelly et al., 2013; Newbery, Chuyong, & from ground level to the uppermost apical bud. Seedling Zimmerman, 2006). height values used in subsequent analyses were the average height per subplot. In July 2017, hemispherical photography 2.8 | Statistical analyses was used to quantify percentage canopy openness directly above each survivorship subplot. A north-oriented digital All data were graphically analyzed using JMP 13.1.0 statisti- color photograph was taken at the center of each subplot cal software (SAS Institute Inc., Cary, North Carolina) and using a Canon EOS Rebel T3 camera (Canon Inc., Tokyo, statistically tested in either JMP or the R 3.3.1 Statistical Japan) fitted with a Rokinon 8 mm F3.5 fisheye lens Environment (R Core Team, 2016). Mixed effect regression (Samyang Optics Co., Masan, South Korea) held level at models were fit with the lme4 package (Bates, Mächler, 2 m above the ground (“fisheye” setting). The photo- Bolker, & Walker, 2015). Plot, or subplot within plot, was graphs were analyzed with GapAnalyzer 2.0 software treated as a random effect to accommodate any unique clus- using standard protocols giveninFrazer,Canham,and tered variances associated with environmental conditions. Lertzman (1999). All linear model fits were assessed for violations of normal- ity and heteroscedasticity. Hypothesis testing of explana- 2.7 | Climatic data tory variables used Likelihood Ratio tests with Type III ANOVA Satterthwaite approximation of degrees of free- For the period 1996–2017, satellite-derived outgoing long dom in the lmerTest package (Kuznetsova, Brockhoff, & wave radiation (OLR) monthly anomalies (deviation of out- Christianson, 2017). going re-radiated energy from the long-term monthly mean) Specifically, seedfall during mast and low seed years and precipitation anomalies on a 2.5 lat/long grid resolution were compared using mixed effect models comparing the centered on the Potaro site (5N, 60W) were downloaded seedfall during two quantified mast seeding events to the from the International Research Institute for Climate Predic- three low seed years. The random effect of plot permitted tion (http://iridl.ldeo.columbia.edu/maproom/.Global/). This comparison across years differing in plot number. Seedling time span encompassed an observed 1998 and the quantified height in 2017 was regressed against canopy openness using 2003 and 2016 D. corymbosa mast seeding events. Addi- linear mixed effect models of the log value of seedling tional anecdotal information concerning dry season condi- height with the random effect of plot (P1–P5). Decadal seed- tions following El Niño events was obtained from local ling growth rates of the pre-2003 and the 2003 seeding inhabitants of the region, including indigenous farmers and cohorts were compared for the period of 2007–2017 using HENKEL AND MAYOR 5 the random effects of subplot nested within plot. For both seedling correlation analyses, marginal (fixed) and condi- tional variance (defined as both fixed and random factors) 2 R GLMM values were reported as goodness-of-fit statistics using the MuMIn package (Barton, 2017), as defined in Nakagawa and Schielzeth (2012). Pre-dispersal and post- dispersal predation plot values for combined mast seed years (n = 14) and combined low seed years (n = 27) were compared using a nonparametric Wilcoxon rank sum test. Cumu- lative August–March monthly OLR and precipitation anomalies for each of years 1998–2016 were summed and a logistic test with χ approximation used to predict mast seeding or low seeding years.

3 | RESULTS

3.1 | Reproductive behavior during the inter- mast years FIGURE 2 Post-dispersal seed predation for Dicymbe corymbosa During the years 2004, 2005, 2007, 2008, 2010, 2011 and for low seed years and mast seeding years. Predation was significantly 2012 no seed output was detected for D. corymbosa. These higher during low seed years (n = 27) relative to mast seeding years years were considered non-reproductive. Low seed years (n = 14; p < .0001, nonparametric Wilcoxon rank sum test). Median occurred in 2006, 2009 and 2013 in which seedfall was values are shown in white, quartiles with solid fill, and an outlier in red recorded at low but consistent levels in all traps and plots sampled, with mean (±SE) seeds m−2 of 2.3 ± 0.4 (2006), 3.6 ± 0.4 (2009) and 3.2 ± 0.7 (2013) (Figure 1). The proportion of post-dispersal seed predation during low seed years was significantly higher than that of mast years (p < .001, Figure 2), as was pre-dispersal fruit predation (p < .001, data unavailable for the 2003 mast; Figure 3).

FIGURE 3 Pre-dispersal fruit predation for Dicymbe corymbosa for low seed years and mast seeding years. Predation was significantly higher during low seed years (n = 27) relative to mast seeding years (n = 14; p < .0001, nonparametric Wilcoxon rank sum test). Median values in are shown in white and quartiles with solid fill FIGURE 1 Estimated seed fall in Dicymbe corymbosa plots from 2003 to 2016. Black box plots represent the two observed mast years and gray box plots represent low seed years. Seed fall in 2003 was 3.2 | 2016 flowering measured in P1–5, 2006 in P1–P7, and 2009–2016 in P1–P10. Median values are shown in white, and quartiles with solid fill. Seed fall in the A regionally synchronous D. corymbosa mass flowering two mast years was significantly higher than that of low seed years event occurred in late February 2016, with canopy trees in based on a random effect model comparing the fixed effect of masting full anthesis in all D. corymbosa monodominant stands. and the random effect of plot (p < .001, F = 68.8) Heavy flowering ranged from 75 to 100% of sun exposed 6 HENKEL AND MAYOR crown branches for all canopy trees in each of P1–P10. The the dry dehiscent legumes were flattened-elongate and large flowers at full anthesis were 2–3 cm diameter, with ranged from 10 to 40 cm in length with up to 13 seed cream-colored petals, in masses on exposed crown branches. pockets; mean dry mass was 19.1 g per valve. Mean dry Observed pollinators were bees, with numerous daytime mass of flower parts was 66.9 g m−2. Extrapolated to the visits to anthetic flowers. Within 2–3 days of opening, an stand level, combined dry mass of seeds, fruit valves and individual flower's petals abscised and fell to the forest floor. flower parts of the 2016 mast averaged 3,117 ± 327 kg ha−1 The observed mass flowering in February across all plots across P1–P10 (range 1,444–4,884 kg ha−1; Table S1). Con- was corroborated by the dry masses of abscised flower parts comitant mineral nutrient investment in reproductive parts − ranging from 30.4 to 99.4 g m 2 across P1–P10 in July was: N: 41.2 ± 5.0 kg ha−1; P: 1.6 ± 0.2 kg ha−1; K: 12.5 (Table S1). ± 1.7 kg ha−1; Ca: 5.6 ± 0.7 kg ha−1; and Mg: 3.2 ± 0.4 kg ha−1. The average reproductive resource invest- 3.3 | 2016 seed output, synchrony, predation ment values for dry mass and minerals of the 2003 and 2016 and reproductive investment mast events were generally two-fold greater than those of the low seed years (Table 1). The seedfall mean (±SE) values for D. corymbosa during the 2016 mast for P1–P10 ranged 2.8 ± 0.3 to 17.5 ± 2.3 − 3.4 | Leaf litter relative to reproductive seeds m 2, with a global mean across plots of 8.9 ± 1.5 − resource investment seeds m 2 (Table S1). This global mean approached the − 12.5 ± 1.6 seeds m 2 seen in the 2003 mast and was signifi- Over a 4-year period spanning 2004–2008 D. corymbosa cantly higher than those of the low seed years (Figure 1). average annual leaf litterfall dry mass was estimated at Extrapolations to the stand level yielded a mean estimate of 1,664 kg ha−1 year−1, whereas the average of the 2003 and − 89,999 seeds ha 1, ranging from 27,778 to 174,778 seeds 2016 mast years' reproductive litterfall dry masses was 45% − ha 1 across the 10 plots (Table S1). Plots with lower seed higher at 3,048 kg ha−1 year−1 (Table 1). Average mast year output (e.g., P5, P6) had sustained some recent treefall loss reproductive investment of mineral nutrients ranged from of reproductive crown branches. Regional synchrony in 20 to 40% higher than that of annual leaf litterfall for N, K masting was inferred by the uniform seed production mea- and Ca, and was 69% higher for P, although Mg mast invest- sured in the Upper Potaro plots P1–P7, the Potaro-Ireng ment was 63% lower (Table 1). transect plots P8–P10 9–15 km distant, and 37 km distant plot P11 on the Upper Ireng River (with 6.2 ± 0.4 seeds − 3.5 | Seedling survivorship m 2; Table S1). Subsequently, no seed output was detected in 2017. The calculated population level coefficient of varia- Twelve months after establishment of the 2003 mast seed- tion (CVp; SD across all years divided by the mean seed out- lings, total D. corymbosa seedling density (i.e., pre-2003 − put across all years) for D. corymbosa from 2003 to 2017 and 2003 cohorts) was >8 seedlings m 2, with an across- − was 1.87. plot range of 41,000–92,000 seedlings ha 1 (Henkel et al., Proportional seed predation was low during the 2016 2005). New seedlings were subsequently established at low mast. Pre-dispersal predation by beetle larvae (14.4%) was densities during the low seed years of 2006, 2009 and 2013 greater than that by birds (1.3%). Post-dispersal predation by (Figure 4). Collective 2004–2015 annual mortality rate of several species of ant averaged 10.9 ± 0.8% of seeds on the seedlings across cohorts was 5%; total surviving density had − ground, with fungal damage found in only 3% of seeds. declined to 3.2 seedlings m 2 by June 2015, 1 year prior to Post-dispersal predation by vertebrates was essentially the 2016 mast seeding (Figure 4). Seventy eight percent of absent; only five seeds out of 17,208 that were directly these surviving seedlings were ≥ 13 years of age, of either − examined had small rodent tooth marks, and no evidence the 2003 mast cohort (1.5 seedlings m 2) or the pre-2003 − − was found of ground-turning by peccaries or seed scatter- group (1 seedling m 2). Only 22% (0.7 seedling m 2) of the hoarding by agoutis. Some interplot variation in predation surviving seedlings were from the low seed years of 2006, levels was recorded (Table S1), but even in plots with higher 2009 and 2013. post-dispersal predation (e.g., P3) the amount of seeds that The 2016 mast seeding produced 3.2 new seedlings − escaped predation to germinate and establish seedlings m 2 across P1–P5. Of these 2016 seedlings, 2.9 seedlings − amounted to tens of thousands ha−1. m 2 had survived after 6 months. After 12 months 2.6 Seeds were flattened-orbicular in shape and ranged from seedlings m−2 had survived, restocking the across-cohort 2 to 4 cm diameter across the longest axis. Mean 2016 seed total standing pool to an average of nearly 5.8 seedlings dry mass was 5.2 g per seed, very close to the mean of 5.1 g m−2, with an estimated range of 34,200–68,000 seedlings recorded in the 2003 mast (Henkel et al., 2005). Valves of ha−1 (Figure 4). By this time, 75% of all living seedlings HENKEL AND MAYOR 7

TABLE 1 Dry mass (kg ha−1 yr−1) and combined mineral nutrient investment across all Dicymbe corymbosa plotsa during the 2003 and 2016 masting events and intervening low-seed years, Pakaraima Mountains, Guyana

Year Mast? Reproductive litter N P Mg K Ca 2003 Yes 2,978 32.5 1.6 2.8 9.7 5.3 2006 No 1,172 15.6 0.5 1.3 3.5 2.5 2009 No 1,749 22.3 0.8 1.8 5.5 3.5 2013b No 926 8.9 0.4 0.7 3.8 1.2 2016 Yes 3,118 41.2 1.6 3.2 12.5 5.6 Mast year avg 3,048 36.8 1.6 3.0 11.1 5.4 Low seed year avg 1,461 19.0 0.7 1.5 4.5 3.0 Δ mast vs. low seed year +52% +48% +56% +50% +59% +44% Leaf litter N P Mg K Ca Annual inputsc 1,664 28.5 0.5 4.9 7.6 3.3 Δ mast reproductive litter vs. leaf litter year +45% +23% +69% −63% +32% +39%

Note: Values represent annual tree biomass losses to reproductive litter (seeds, fruit valves, and flower parts) and D. corymbosa leaf litter biomass. No litterfall estimates are available for mast years. aNumber of 0.25-ha Upper Potaro plots: 2003: n = 5; 2006: n = 7; 2009, 2013, 2016: n = 10. bNo flower parts available in 2013, omitted from summary statements in results (otherwise would be the lowest). cAnnual averages calculated post-abscission leaves from 2004–2008 in P1–P6; % N values used in calculations obtained in 2004 only. were derived from the 2003 and 2016 masting events, 3.6 | Seedling growth 2007–2017 15% were from the low seed years, and 10% were from Seedling height was positively correlated with canopy open- the pre-2003 age group. Similarly, 43.1% of the 144 sap- ness in 2017 for both pre-2003 seedlings and 2003 cohort lings (i.e., 1–10 cm diameter individuals) monitored since seedlings (R2c = 0.34 and 0.31, respectively, when plot spe- 2003 were still alive in 2017. cific variance was accounted for; p = .014) (Figure S1). For the 11 year period (2007–2017) average annual seedling growth rates were nearly identical for pre-2003 and 2003 seedlings at 1.04 and 1.05 cm growth year−1 (p < .001), respectively, when accounting for plot and subplot specific variance (Figure S2).

3.7 | El Niño-related climate effects and D. corymbosa seeding The three known D. corymbosa masting events of 1998, 2003 and 2016 were preceded by El Niño-driven sustained positive monthly OLR anomalies in Guyana's Pakaraima Mountains (Figures 5 and S3). Negative precipitation anomalies followed the same pattern (Figure 5). Local inhabitants of the Upper Potaro study region reported dramatic increases in daily sunshine in the 1997–1998 and 2002–2003 dry sea- FIGURE 4 Seedling establishment and survivorship of Dicymbe sons (Henkel et al., 2005), as well as in the 2015–2016 dry – corymbosa in plots P1 P5 following seeding events from 2003 to season reported here, in each case with greatly reduced sur- 2017. Months begin with 0 at time of establishment of the 2003 mast face water flows and numerous accidental forest fires associ- cohort. Mast cohorts were initially measured at time of establishment ated with slash and burn farming. The summed August– and at 6 and 12 months. Survivorship of pre-existing cohorts was measured at 12 month-intervals (apart from 96 and 132 months when March monthly OLR anomalies preceding the 1998, 2003 plots could not be revisited). Recruitment and survivorship of low seed and 2016 mast seedings were significantly higher and pre- year cohorts was measured initially at 12 months after each seeding cipitation anomalies significantly lower than those of most event. By 168 months the total pool was dominated by seedlings from non-mast years (Logistic Fit, p < .001, χ approximation; masting events (25% from 2003, 49% from 2016) Figure 6a,b). The low seeding events of 2006, 2009 and 8 HENKEL AND MAYOR

FIGURE 5 Correspondence between Dicymbe corymbosa seeding events and monthly outgoing long-wave radiation (OLR) and precipitation (PRECIP) anomalies from 1997 to 2017. The OLR anomaly is departure of the monthly OLR from long-term monthly means; negative anomalies imply increased cloudiness and enhanced probability of precipitation; positive anomalies imply decreased cloudiness/precipitation (satellite derived). Similarly, the PRECIP anomaly is departure of the monthly precipitation from the long-term monthly mean. The blue-shaded areas indicate the OLR and PRECIP measurements for months preceding observed flowering and fruiting of D. corymbosa; summed values from these were used to predict masting in Figure 6. The pink-shaded areas indicate the same integrated time period backcasted an additional year for comparison purposes. The 1998 mast was observed and recorded anecdotally but the years immediately following this period represent an area of uncertain seed production. Data are derived from a 2.5 by 2.5 deg. area centered on 5N longitude and 60W latitude, an area encompassing the study area in the Pakaraima Mountains of Guyana (data source: http://iridl.ldeo.columbia.edu/maproom/.Global/)

2013 did not follow strong, prolonged El Niño events or described from other biotically pollinated and abiotically sustained positive monthly OLR anomalies in the preceding dispersed species based on a meta-analysis of 570 seeding August–March period (Figures 5 and S3). For non-seeding variability studies of >6 years duration (i.e., 0.74; Kelly & years, the summed preceding August–March OLR anomaly Sork, 2002). It may be that CV values alone poorly differen- values were generally low. An exception was that the tiate bimodal, supra-annual seeding species from annually sustained August–March positive OLR and negative precipi- seeding species that have high year-to-year variability in tation anomalies in 2009–2010 did not result in a detectable seed production (Norden et al., 2007). seeding event in 2010 (Figures 5 and 6a,b). Backcasting of – the August March period to 2 years prior to reproduction 4.1.1 | Regional entrainment to El Niño had no predictive power for either masting or low seeding teleconnections events (Figure 6c,d). Population-wide synchronization of masting requires entrainment of flowering, typically in response to a broad environ- 4 | DISCUSSION mental cue (Pearse et al., 2016). There is evidence across the Guiana Shield that the intensified dry season conditions of 4.1 | Climate variables, seeding patterns, and reduced cloud cover and increased solar irradiance provide resource relations over a long inter-mast such a cue (Hammond & ter Steege, 1998; ter Steege & Per- interval saud, 1991; Thomas, 2001). The last three D. corymbosa Over 2003–2017, all stands of D. corymbosa exhibited high masting events were preceded by strong Pacific El Niño con- inter-annual variability in seeding, with two synchronous ditions and in western Guyana highly positive OLR anomalies masts and three low seed years occurring in the 15 year persisted for several months leading up to the February– period. The resulting CV of 1.87 is similar to that described March flowering period relative to all other years (Figures 5 for masting, wind-pollinated, predator-dispersed seed plants and 6a,b). These conditions have also been repeatedly corrob- (i.e., 1.77, Kelly & Sork, 2002), but much higher than that orated on the ground by local inhabitants of the Upper Potaro HENKEL AND MAYOR 9

Adams, 2001; Sork, 1993; Wright, Carrasco, Calderón, & Paton, 1999; Wright & Van Schaik, 1994) followed by NSC depletion after masting (Miyazaki, Hiura, Kato, & Funada, 2002; Sala, Hopping, McIntire, Delzon, & Crone, 2012). Additionally, NSC storage between masting events appears minimal (Hoch, Siegwolf, Keel, Korner, & Han, 2013; Ichie et al., 2013) and during masting fruit bearing branches appear carbon autonomous (Hasegawa, Koba, Tayasu, Takeda, & Haga, 2003; Hoch, 2005). Such results further implicate a possible connection between elevated irradiance and carbon gain prior to masting.

4.1.2 | Low seed years Summed OLR or precipitation anomalies did not differenti- ate non-reproductive years from low seed years in the months prior to, or 1 year preceding, flowering (Figure 6). Therefore, apart from some other unknown regularly occurring environmental cue, it is possible that the phenologically synchronous low seed years are internally entrained. Unlike the “alternate year bearing” seen in D. altsonii (Zagt, 1997), D. corymbosa reproduction appears intrinsically separated by 3–4 years (e.g., 2003, 2006, 2009, 2013, 2016, and corroborated by an observed flower bud initiation in January 2019; T. W. Henkel, personal observation). In this scenario, a sufficiently long but predictable flowering periodicity may allow accumulation of mineral nutrients capable of supporting mast fruiting during El Niño events (discussed further below). Admittedly, our data set is of short duration relative to FIGURE 6 Cumulative monthly outgoing long-wave radiation the indeterminate lifespan of D. corymbosa trees (Woolley (OLR) and precipitation (PRECIP) anomalies from differing preceding et al., 2008). Nonetheless, the aforementioned periodicity august–march time periods relative to mast or low seed years. Filled mechanism could also explain the lack of reproduction in squares = mast seed years; open squares = low seed years; filled 2010 despite the intensified dry season conditions in preced- triangle = expected mast seed year; filled circles = no detectable seed ing months (Figures 5 and 6a,b). For instance, a 2010 mast year. The summed august–march OLR values were significantly may have failed because the preceding low seed year higher, and PRECIP significantly lower (logistic fit, p < .001, χ approximation) for the three mast seed years (in bold blue) relative to resulted in extensive resource depletions, with the subse- low seed years as seen in (a) and (b). Backcasting two years did not quent enhanced dry season irradiance insufficient to initiate result in significant prediction of observed mast or low seed years, flowering in such an “off” year (Borchert, 1983; Tucker, (c) and (d) respectively 2003). This idea is akin to the “match-mismatch” hypothesis developed by Cushing (1969, 1990) to describe how climatic variation affects fish recruitment. As a result of the interac- – study region for multiple El Niño events from 1997 2016. It tion of innate periodicity and climatic cycles the increasing is possible that these consecutive months of prolonged high frequency of extreme El Niño events over the next 50 years irradiance provide both the entrainment cue for mass (Cai et al., 2014) could alter both the magnitude and fre- flowering and the non-structural carbohydrates (NSC) neces- quency of resource pulses in D. corymbosa forests as posited sary to “fuel” the enormous carbon investment in masting. for Malaysian dipterocarps (Chen et al., 2018). While NSC data are currently lacking for D. corymbosa,the results of other studies corroborate such a mechanism. Accu- 4.1.3 | Mineral nutrient proximate drivers mulation of NSC immediately prior to a masting event has been observed in tropical and temperate canopy trees Proximate controls over supra-annual seeding include con- (Newbery, Chuyong, & Zimmerman, 2006; Piovesan & straints imposed by resource limitations to flower or 10 HENKEL AND MAYOR fruit/seed production (Pearse et al., 2016). Reproductive compounds of D. corymbosa seeds (Henkel et al., 2005), events can temporarily exhaust internal nutrient supplies and along with current seed predation performed largely by select for alternating patterns in reproduction as seen in insects, it is conceivable that invertebrate predator selection many species of wild and cultivated trees (e.g., Davis, 1957; pressure may have been a driver for a predator satiation Lloyd, 1980; Newbery, Chuyong, & Zimmerman, 2006; mechanism, as recently posited for Bornean dipterocarps Rosecrance, Weinbaum, & Brown, 1998). Dicymbe (Iku et al., 2017). However, prior vertebrate-based selective corymbosa may have resource constraints that favor forces for masting evolution cannot be ruled out (Janzen, extended supra-annual periodicity in flowering which, as 1971, 1974; Waller, 1993). Additionally, chemical defense posited here with an interval of 3–4 years, could allow for of seeds may have been lost after the evolution of masting. multi-year replenishment of resources depleted by seeding All told, there are limitations to the tractability of questions events of any scale. which address how extant community ecology of long-lived Fruit valves and flower parts constituted >70% of the trees may have been shaped by macro-evolutionary pro- total D. corymbosa mast reproductive litter and this material cesses (Johnson & Stinchcombe, 2007). entered the detritus pool. Mineral nutrients contained therein could potentially be lost to leaching. Recovery of these min- 4.3 | Seedlings and monodominance eral nutrients by extensive ectomycorrhizas may therefore be imperative, especially for P which is barely detectable in Stand-wide expanses of D. corymbosa seedlings benefit mineral soils at <12 ppm, and Ca which is mined directly from periodic mast seeding to maintain their high densities from leaf litter (Mayor & Henkel, 2006). The average repro- necessary for persistent monodominance. Over the ductive dry mass inputs from the two masting events was 2003–2015 period prior to the 2016 mast, 83% of pre-2003 ca. 3.0 t−1 ha−1, nearly twice the average annual leaf litter and 75% of 2003 mast seedlings had died, and seedling con- inputs of ca. 1.7 t−1 ha−1. Masting P and Ca investments tributions from the three intervening low seed years were were 69 and 39% greater, respectively, than that lost to minimal. Density-dependent seedling mortality is the rule annual leaf litterfall (Henkel et al., 2005; Table 1). Such for tropical tree species regardless of whether they exhibit large investments suggest that P accumulation during pro- masting or monodominance (Queenborough, Burslem, longed inter-mast periods may be a prerequisite for masting Garwood, & Valencia, 2007). The long-term dataset on oligotrophic tropical soils, as proposed for African reported here for D. corymbosa suggests that without a sub- detarioids (Chuyong, Newbery, & Songwe, 2000; Newbery, sequent masting event the standing seedling and sapling 2005; Newbery, Alexander, & Rother, 1997; Newbery, pool would continue to decline, albeit slowly. Conversely, Chuyong, & Zimmerman, 2006) and the Bornean diptero- the near-term survival of D. corymbosa seedlings in deep carp Dryobalanus aromatica (Ichie & Nakagawa, 2013). In shade, necessary to maintain the standing seedling stock, contrast, temperate masting trees in ECM Fagaceae appear may be dependent on rapid ectomycorrhization – a trait not more limited by N (Han, Kabeya, Iio, Inagaki, & Kakubari, shared with competing species in D. corymbosa forest 2014; Smaill, Clinton, Allen, & Davis, 2011), reflecting (Henkel, 2003; McGuire et al., 2008). global gradients in relative nutrient availability (Mayor The rare transition of an individual D. corymbosa seed- et al., 2015). ling to sapling, pole and ultimately canopy tree must be very slow and dependent on small, transient light gaps resulting 4.2 | Predator satiation and masting from stem attrition of highly reiterated, stand-dominating adult trees (Newbery, Chuyong, Zimmerman, & Praz, 2006; Selection for mast seeding requires an ultimate evolutionary Woolley et al., 2008). The positive correlation between can- driver to favor increased synchrony in supra-annual repro- opy openness and seedling heights in 2017 (Figure S1) production (Pearse et al., 2016). Hypothesized drivers for vided a glimpse of such gap-driven seedling growth. All masting evolution include predator satiation (i.e., seed preda- told, mast seeding may play a fundamental role in tors consume a smaller proportion of large seed crops) and maintaining D. corymbosa dominance, by providing the pollination efficiency (i.e., higher cross-pollination in heavy- periodic pulse of shade-tolerant seedlings necessary for flowering years) (Bogdziewicz, Espelta, Muñoz, Aparicio, & slow, but inexorable, accession to the canopy. Bonal, 2018; Crone & Rapp, 2014; Janzen, 1974). In the case of D. corymbosa, preliminary evidence supports a role ACKNOWLEDGMENTS for predator satiation due to the significantly higher proportional predation during low seed years relative to mast years. Funding was provided to T.W.H. by the National Science Given the absence of a vertebrate-based seed dispersal Foundation (NSF) grants DEB-0918591 and NSF DEB- mechanism in D. corymbosa, the low phytochemical defense 1556338, the National Geographic Society‘s Committee for HENKEL AND MAYOR 11

Research and Exploration grants 7435-03 and 8481-08, the Cushing, D. H. (1969). The regularity of the spawning season of some Linnaean Society of London, and the Humboldt State Uni- fishes. ICES Journal of Marine Science, 33,81–92. https://doi.org/ versity Foundation. Field assistance was provided by 10.1093/icesjms/33.1.81 Cushing, D. H. (1990). Plankton production and year-class strength in C. Andrew, H. Andrew, M. Chin, F. Edmund, L. Edmund, fish populations: An update of the match/mismatch hypothesis. K. Hasebe, P. Henkel, D. Husbands, P. Joseph, V. Joseph Advances in Marine Biology, 26, 249–293. https://doi.org/10.1016/ and L. Woolley. Research permits were granted by the Guy- S0065-2881(08)60202-3 ana Environmental Protection Agency. Carolyn Delevich Davis, L. D. (1957). Flowering and alternate bearing. In Proceedings of provided data on 2016 seedling heights. David Newbery the American Society for Horticultural Science (Vol. 70, provided invaluable insights on an earlier version of the pp. 545–556). Geneva, New York: The Society. manuscript. Three anonymous reviewers improved this work Degagne, R. S., Henkel, T. W., Steinberg, S. J., & Fox, L. (2009). Iden- with constructive editorial critiques. This paper is number tifying Dicymbe corymbosa monodominant forests in Guyana using – 227 in the Smithsonian Institution‘s Biological Diversity of satellite imagery. Biotropica, 41,715. https://doi.org/10.1111/j. 1744-7429.2008.00446.x the Guiana Shield publication series. Fanshawe, D. B. (1952). The vegetation of British Guiana: A preliminary review. Institute paper no. 39. Oxford: Imperial Forestry ORCID Institute. Frazer, G. W., Canham, C. D., & Lertzman, K. P. (1999). Gap light Terry W. Henkel https://orcid.org/0000-0001-9760-8837 analyzer (GLA), Version 2.0: Imaging software to extract canopy structure and gap light transmission indices from true-colour fisheye photographs. New York: Simon Fraser, Burnaby, and Insti- REFERENCES tute of Ecosystem Studies. Gavlak, R., Horneck, D., Miller, R. O., & Kotuby-Amacher, J. (2003). AOAC. (2003). Official methods of analysis of AOAC international Soil, plant andwater reference methods for the western region. (17th ed.). Gaithersburg: AOAC International. WREP − 125 (2nd ed.). Fairbanks: University of Alaska. Ashton, P. S. (1987). Dipterocarp reproductive biology. In H. Leith & Graham, E. A., Mulkey, S. S., Kitajima, K., Phillips, N. G., & M. J. A. Werger (Eds.), Biogeographical and ecological studies Wright, S. J. (2003). Cloud cover limits net CO uptake and growth (pp. 219–240). Amsterdam: Elsevier. 2 Proceedings of the Barton K. (2017) MuMIn: Multi-model inference. R package version of a rainforest tree during tropical rainy seasons. 1.40.0. https://CRAN.R-project.org/package=MuMIn National Academy of Sciences of the United States of America, – Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear 100, 572 576. https://doi.org/10.1073/pnas.0133045100 mixed-effects models using lme4. Journal of Statistical Software, Hammond, D. S., & ter Steege, H. (1998). Propensity for fire in Guia- – 67,1–48. https://doi.org/10.18637/jss.v067.i01 nan rainforests. Conservation Biology, 12, 944 947. https://doi.org/ Bogdziewicz, M., Espelta, J. M., Muñoz, A., Aparicio, J. M., & 10.1046/j.1523-1739.1998.012005944.x Bonal, R. (2018). Effectiveness of predator satiation in masting Han, Q., Kabeya, D., Iio, A., Inagaki, Y., & Kakubari, Y. (2014). oaks is negatively affected by conspecific density. Oecologia, 186, Nitrogen storage dynamics are affected by masting events in Fagus – 983–993. https://doi.org/10.1007/s00442-018-4069-7 crenata. Oecologia, 174, 679 687. https://doi.org/10.1007/s00442- Borchert, R. (1983). Phenology and control of flowering in tropical 013-2824-3 trees. Biotropica, 15,81–89. Hasegawa, S., Koba, K., Tayasu, I., Takeda, H., & Haga, H. (2003). Cai, W., Borlace, S., Lengaigne, M., van Rensch, P., Collins, M., Carbon autonomy of reproductive shoots of Siberian alder (Alnus Vecchi, G., … Jin, F.-F. (2014). Increasing frequency of extreme El hirsuta var. sibirica). Journal of Plant Research, 116, 183–188. Niño events due to greenhouse warming. Nature Climate Change, https://doi.org/10.1007/s10265-003-0085-7 4, 111–116. https://doi.org/10.1038/nclimate2100 Henkel, T. W. (2003). Monodominance in the ectomycorrhizal Chen, Y.-Y., Satake, A., Sun, I.-F., Kosugi, Y., Tani, M., Numata, S., Dicymbe corymbosa (Caesalpiniaceae) from Guyana. Journal of … Wright, S. J. (2018). Species-specific flowering cues among Tropical Ecology, 19, 417–437. https://doi.org/10.1017/ general flowering Shorea species at the Pasoh research Forest, S0266467403003468 Malaysia. Journal of Ecology, 106, 586–598. https://doi.org/10. Henkel, T. W., Terborgh, J., & Vilgalys, R. J. (2002). Ectomycorrhizal 1111/1365-2745.12836 fungi and their leguminous hosts in the Pakaraima Mountains of Chuyong, G. B., Newbery, D. M., & Songwe, N. C. (2000). Litter Guyana. Mycological Research, 106, 515–531. https://doi.org/10. nutrients and retranslocation in a central African rain forest domi- 1017/S0953756202005919 nated by ectomycorrhizal trees. New Phytologist, 148, 493–510. Henkel, T. W., Mayor, J. R., & Woolley, L. P. (2005). Mast fruiting https://doi.org/10.1046/j.1469-8137.2000.00774.x and seedling survival of the ectomycorrhizal, monodominant Connell, J. H., & Lowman, M. D. (1989). Low-diversity tropical rain Dicymbe corymbosa (Caesalpiniaceae) in Guyana. New Phytologist, forests: Some possible mechanisms for their existence. American 167, 543–556. https://doi.org/10.1111/j.1469-8137.2005.01431.x Naturalist, 134,88–119. Herrera, C. M., Jordano, P., Guitlan, J., & Traveset, A. (1998). Annual Crone, E. E., & Rapp, J. M. (2014). Resource depletion, pollen cou- variability in seed production by woody plants and the masting con- pling, and the ecology of mast seeding. Annals of the New York cept: Reassessment of principles and relationship to pollination and Academy of Sciences, 1322,21–34. https://doi.org/10.1111/nyas. seed dispersal. American Naturalist, 152, 576–594. https://doi.org/ 12465 10.1086/286191 12 HENKEL AND MAYOR

Hoch, G. (2005). Fruit-bearing branchlets are carbon autonomous in Lyon, B., & Barnston, A. G. (2005). ENSO and the spatial extent of mature broad-leaved temperate forest trees. Plant, Cell & Environ- interannual precipitation extremes in tropical land areas. Journal of ment, 28, 651–659. https://doi.org/10.1111/j.1365-3040.2004.01311.x Climate, 18, 5095–5109. https://doi.org/10.1175/JCLI3598.1 Hoch, G., Siegwolf, R. T. W., Keel, S. G., Korner, C., & Han, Q. M. Mayor, J. R., & Henkel, T. W. (2006). Do ectomycorrhizas alter leaf- (2013). Fruit production in three masting tree species does not rely litter decomposition in tropical monodominant forests of Guyana? on stored carbon reserves. Oecologia, 171, 653–662. https://doi. New Phytologist, 169, 579–588. https://doi.org/10.1111/j.1469- org/10.1007/s00442-012-2579-2 8137.2005.01607.x Ichie, T., & Nakagawa, M. (2013). Dynamics of mineral nutrient stor- Mayor, J., Bahram, M., Henkel, T., Buegger, F., Pritsch, K., & age for mast reproduction in the tropical emergent tree Tedersoo, L. (2015). Ectomycorrhizal impacts on plant nitrogen Dryobalanops aromatica. Ecological Research, 28, 151–158. nutrition: emerging isotopic patterns, latitudinal variation and hid- https://doi.org/10.1007/s11284-011-0836-1 den mechanisms. Ecology Letters, 18,96–107. https://doi.org/10. Ichie, T., Igarashi, S., Yoshida, S., Kenzo, T., Masaki, T., & Tayasu, I. 1111/ele.12377 (2013). Are stored carbohydrates necessary for seed production in McGuire, K. L., Henkel, T. W., de la Cerda, I. G., Villa, G., temperate deciduous trees? Journal of Ecology, 101, 525–531. Edmund, F., & Andrew, C. (2008). Dual mycorrhizal colonization https://doi.org/10.1111/1365-2745.12038 of forest-dominating tropical trees and the mycorrhizal status of Iku, A., Itioka, T., Kishimoto-Yamada, K., Shimizu-kaya, U., non-dominant tree and liana species. Mycorrhiza, 18, 217–222. Mohammad, F. B., Hossman, M. Y., … Meleng, P. (2017). Increased https://doi.org/10.1007/s00572-008-0170-9 seed predation in the second fruiting event during an exceptionally Miyazaki, Y., Hiura, T., Kato, E., & Funada, R. (2002). Allocation long period of community-level masting in Borneo. Ecological of resources to reproduction in Styrax obassia in a masting year. Research, 32,537–545. https://doi.org/10.1007/s11284-017-1465-0 Annals of Botany, 89,767–772. https://doi.org/10.1093/aob/ Janzen, D. H. (1969). Seed-eaters versus seed size, number, toxicity mcf107 and dispersal. Evolution, 23,1–27 https://www.jstor.org/stable/ Nakagawa, S., & Schielzeth, H. (2012). A general and simple method 2406478 for obtaining R2 from generalized linear mixed-effect models. Janzen, D. H. (1971). Seed predation by animals. Annual Review of Methods in Ecology and Evolution, 4, 133–142. https://doi.org/10. Ecology and Systematics, 2, 465–492. https://doi.org/10.1146/ 1111/j.2041-210x.2012.00261.x annurev.es.02.110171.002341 Newbery, D. M. (2005). Ectomycorrhizas and mast fruiting in trees: Janzen, D. H. (1974). Tropical blackwater rivers, animals, and mast Linked by climate driven resources? New Phytologist, 167, fruiting by the Dipterocarpaceae. Biotropica, 6,69–103. 324–326. https://doi.org/10.1111/j.1469-8137.2005.01468.x Johnson, M. T. J., & Stinchcombe, J. R. (2007). An emerging synthesis Newbery, D. M., Alexander, I. J., & Rother, J. A. (1997). Phosphorus between community ecology and evolutionary biology. Trends in dynamics in a lowland African rain forest: the influence of Ecology & Evolution, 22, 250–257. https://doi.org/10.1016/j.tree. ectomycorrhizal trees. Ecological Monographs, 67,367–409. https:// 2007.01.014 doi.org/10.1890/0012-9615(1997)067[0367:PDIALA]2.0.CO;2 Kelly, D. (1994). The evolutionary ecology of mast seeding. Trends in Newbery, D. M., Chuyong, G. B., & Zimmerman, L. (2006). Mast Ecology & Evolution, 9, 465–470. https://doi.org/10.1016/0169- fruiting of large ectomycorrhizal African rain forest trees: impor- 5347(94)90310-7 tance of dry season intensity, and the resource-limitation hypothe- Kelly, D., & Sork, V. L. (2002). Mast seeding in perennial plants: why, sis. New Phytologist, 170, 561–579. https://doi.org/10.1111/j.1469- how, where? Annual Review of Ecology and Systematics, 33, 8137.2006.01691.x 427–447. https://doi.org/10.1146/annurev.ecolsys.33.020602.095433 Newbery, D. M., Chuyong, G. B., Zimmerman, L., & Praz, C. (2006). Kelly, D., Geldenhuis, A., James, A., Holland, E. P., Plank, M. J., Seedling survival and growth of three ectomycorrhizal Brockie, R. E., … Byrom, A. E. (2013). Of mast and mean: caesalpiniaceous tree species in a Central African rain forest. Jour- Differential-temperature cue makes mast seeding insensitive to cli- nal of Tropical Ecology, 22, 499–511. https://doi.org/10.1017/ mate change. Ecology Letters, 16,90–98. https://doi.org/10.1111/ S0266467406003427 ele.12020 Norden, N., Chave, J., Belbenoit, P., Caubère, A., Châtelet, P., Koenig, W. D., & Knops, J. M. H. (2000). Patterns of annual seed pro- Forget, P.-M., & Thebaud, C. (2007). Mast fruiting is a frequent duction by northern hemisphere trees: A global perspective. Ameri- strategy in woody species of eastern South America. PLoS One, 2, can Naturalist, 155,59–69. https://doi.org/10.1086/303302 e1079. https://doi.org/10.1371/journal.pone.0001079 Koenig, W. D., Knops, J. M. H., Carmen, W. J., & Pearse, I. S. (2015). Pearse, I. S., Koenig, W. D., & Kelly, D. (2016). Mechanisms of mast What drives masting? The phenological synchrony hypothesis. seeding: resources, weather, cues, and selection. New Phytologist, Ecology, 96, 184–192. https://doi.org/10.1890/14-0819.1 212, 546–562. https://doi.org/10.1111/nph.14114 Kogan, F. (2000). Satellite-observed sensitivity of world land ecosys- Piovesan, G., & Adams, J. M. (2001). Masting behavior in beech: tems to El Niño/La Niña. Remote Sensing of Environment, 74, Linking reproduction and climatic variation. Cananadian Journal 445–462. https://doi.org/10.1016/S0034-4257(00)00137-1 of Botany, 79, 1039–1047. https://doi.org/10.1139/b01-089 Kuznetsova, A., Brockhoff, P. B., & Christianson, R. H. B. (2017). Queenborough, S. A., Burslem, D. F. R. P., Garwood, N. C., & lmerTest Package: Tests in linear mixed effect models. Journal of Valencia, R. (2007). Neighborhood and community interactions Statistical Software, 82,1–26. https://doi.org/10.18637/jss.v082.i13 determine the spatial pattern of tropical tree seedling survival. Ecol- Lloyd, D. G. (1980). Sexual strategies in plants. I. A hypothesis of ogy, 88, 2248–2258. https://doi.org/10.1890/06-0737.1 serial adjustment of maternal investment during one reproductive R Core Team. (2016). R: a language and environment for statistical session. New Phytologist, 86,69–79. https://doi.org/10.1111/j. computing. Vienna: R Foundation for Statistical Computing https:// 1469-8137.1980.tb00780.x www.r-project.org/ HENKEL AND MAYOR 13

Rosecrance, R. C., Weinbaum, S. A., & Brown, P. H. (1998). Alternate Woolley, L. P., Henkel, T. W., & Sillett, S. C. (2008). Reiteration in the bearing affects nitrogen, phosphorus, potassium and starch storage monodominant tropical tree Dicymbe corymbosa (Caesalpiniaceae) pools in mature pistachio trees. Annals of Botany, 82, 463–470. and its potential adaptive significance. Biotropica, 40,32–43. https:// https://doi.org/10.1006/anbo.1998.0696 doi.org/10.1111/j.1744-7429.2007.00348.x Sala, A., Hopping, K., McIntire, E. J. B., Delzon, S., & Crone, E. E. Wright, S. J., Carrasco, C., Calderón, O., & Paton, S. (1999). The El (2012). Masting in whitebark pine (Pinus albicaulis) depletes Niño Southern Oscillation, variable fruit production, and famine in stored nutrients. New Phytologist, 196, 189–199. https://doi.org/10. a tropical forest. Ecology, 80, 1632–1647. https://doi.org/10.1890/ 1111/j.1469-8137.2012.04257.x 0012-9658(1999)080[1632:TENOSO]2.0.CO;2 Smaill, S. J., Clinton, P. W., Allen, R. B., & Davis, M. R. (2011). Cli- Wright, S. J., & Van Schaik, C. P. (1994). Light and the phenology of mate cues and resources interact to determine seed production by a tropical trees. American Naturalist, 143, 192–199 http://links.jstor. masting species. Journal of Ecology, 99, 870–877. https://doi.org/ org/sici?sici=00030147%28199401%29143%3A1%3C192% 10.1111/j.1365-2745.2011.01803.x 3ALATPOT%3E2.0.CO%3B2-D. Smith, M. E., Henkel, T. W., Williams, C. G., Aime, M. C., Zagt, R. J. (1997). Pre-dispersal and early post-dispersal demography, Fremier, A. K., & Vilgalys, R. (2017). Investigating niche par- and reproductive litter production, in the tropical tree Dicymbe titioning of ectomycorrhizal fungi in specialized rooting zones of altsonii in Guyana. Journal of Tropical Ecology, 13, 511–526. the monodominant leguminous tree Dicymbe corymbosa. New https://doi.org/10.1017/S0266467400010683 Phytologist, 215, 443–453. https://doi.org/10.1111/nph.14570 Sork, V. L. (1993). Evolutionary ecology of mast-seeding in temperate SUPPORTING INFORMATION and tropical oaks (Quercus spp.). Vegetatio, 107, 133–147. https:// doi.org/10.1007/978-94-011-1749-4_9 Additional supporting information may be found online in ter Steege, H., & Persaud, C. A. (1991). The phenology of Guyanese the Supporting Information section at the end of this article. timber species: A compilation of a century of observations. Vegetatio, 95, 177–198. Thomas, R. S. (2001). Forest productivity and resource availability in How to cite this article: Henkel TW, Mayor JR. lowland tropical forests of Guyana. Tropenbos-Guyana, Series 7. Implications of a long-term mast seeding cycle for Georgetown: Tropenbos-Guyana Programme. climatic entrainment, seedling establishment and Tucker, S. C. (2003). Floral development in legumes. Plant Physiology, persistent monodominance in a Neotropical, 131, 911–926. https://doi.org/10.1104/pp.102.017459 Waller, D. M. (1993). How does mast fruiting get started? Trends in ectomycorrhizal canopy tree. Ecol. Res. 2019; – Ecology & Evolution, 8, 122–123. https://doi.org/10.1016/0169- 472 484. https://doi.org/10.1111/1440-1703.12014 5347(93)90022-H