Three habitat types used in this study were delineated based on hydrological attributes of the sites invaded by melaleuca. These occasionally inundated (N26.049778˚, W80.441556˚ and N26.052444˚, W80.441278˚), and permanently inundated (N26.161667˚, W80.356097˚ and N26.159667˚, W80.365194˚) sites possessing humid subtropical to tropical climate [8] were located in Broward County, Florida (Figure 1). The hydrological attributes of these habitats are based primarily on the descriptions of Brown et al. 1991) [14] and Kushlan 1991) [15] . Habitats designated as “occasionally inundated” may remain inundated (with less than 0.1 m depth) intermittently for a few hours to several days during or following periods of heavy rain but are neither continuously flooded nor flooded every year, while “permanently inundated” habitats remained inundated year-round [16] where water depth at the research site fluctuated from ca 0.3 - 1.30 m during June of 1997. Soil types in these habitats are dominated by poorly drained organic muck deposited on limestone beds which are generally classified as Histosols [14] . Melaleuca stands are scattered in and around fresh water marshes often associated with the Florida Everglades and the surface water depths in these habitats fluctuate seasonally [15] .

Generally, established melaleuca forests form characteristic domes with predominantly larger trees at the center with progressively younger trees, saplings and seedlings near the edges. Two melaleuca domes (hereafter each referred to as a “site”) in each habitat were divided into three tree size categories (TSC): “small treesize”, at the edge of the dome; “medium treesize”, at the mid-section of the

dome; and “large treesize”, at the center of the dome) based upon the prevalent tree diameter at breast height (DBH) measured at 1.3 meter height from the base. The TSCs were delineated based on their existing position in melaleuca stands. Two permanent plots (10 m × 10 m for the large and medium, and 5 m × 5 m for small tree stands) were established in each section of the two sites (each located ca 1 km apart) within each of the two habitats (Figure 1). These plots’ dimensions were determined based on the approximate average heights of the majority of tees in the areas (at the onset of the experiment) where the plots were to be located. A total of 24 (3 TSC × 2 plots per TSC × 2 sites × 2 habitats) experimental plots were used in this study.

The DBH of melaleuca trees that were likely to contribute to the seed fall within each habitat and TSC were measured and recorded. These measurements were used to describe the tree attributes (density and DBH; Tables [1] [2] ) within each plot in study sites and habitats. Rayachhetry et al. (2001) [17] reported 46% - 84% of melaleuca trees to be reproductive as indicated by the presence of the clusters of seed bearing capsules (fruits).

Two wooden framed boxes, each bearing two seed traps, were randomly placed in each plot. Circular seed traps were constructed from 25-cm diameter (0.098 m²) cylinders that measured ca 15-cm depth and fitted with fine nylon net on one end to catch melaleuca seeds that measured 0.15 - 0.5 mm diameter and 0.6 - 2.0 mm length. One seed-trap was mounted at each of the two corners of two litter trap plot⁻¹ using metallic clips, seed traps were emptied at monthly intervals from July 1997-June 1998. For each habitat type, a total of 48 seed traps were monitored during the study (2 sites × 3 tree size categories × 2 plots × 2 boxes × 2 seed traps). Wooden boxes (0.5 × 0.5 m × 16-cm deep) were raised with legs to a height ranging from 20 - 70 cm above ground level in occasionally inundated sites. Water levels in permanently flooded sites fluctuated from 0.3 to 1.3 m. Therefore, the wooden litter trap boxes were modified to float at least 10 cm above the water surface by mounting a capped 3.8-liter plastic jar under each of the four corners of the supporting frame and tying them loosely to a nearby tree to secure them in place [16] . This allowed seed traps mounted on two sides of the wooden boxes to float above the water lavel.

Melaleuca seeds collected from the two seed traps mounted on each box were bulked during collection to maximize seed number for weighing purposes. Therefore, 24 composite (48 seed traps/2) seed samples were collected habitat⁻¹∙month⁻¹. Each of these composite seed samples were air-dried at room temperature and the foreign objects, such as bark and twig fragments, leaves, insect frass and other dirt particles, were removed. Seeds were either directly counted (if quantities per trap were small) or weighed and sub-sampled, with overall counts extrapolated from the subsamples and the final tallies have been presented as the number of seeds ha⁻¹∙mo⁻¹.

Melaleuca seed quality was measured in terms of viability and germinability percentages based either on 200 seeds or the total number of seeds collected, whichever was greater. The germinability test procedures used in this study were the same as described in Rayachhetry et al. (1998). All tests were performed using sterile petri dishes (5-cm diam.), each containing a sterile pad. The pads were soaked with 2 ml of 0.5% 2,3,5-triphenol tetrazolium chloride (TTC) or sterile deionized water (SDW) that left a thin film of liquid on the surfaces. The seeds were spread onto the soaked sterile pads. The dishes were closed and sealed with Parafilm³ and placed in a dark cabinet drawer at room temperature (25˚C). Seed germinability was assessed on 10 and 14 d after initiation of the testto capture viability and germinability by examining seedmorphology under a dissecting microscope using reflected and/or transmitted light. A seed was considered germinable when an emerging radicle grew out of the seed coat and became visible [10] .

In this study, the independent variables were habitat types and melaleuca TSCs and the dependent variables were the quantities (number of seeds per unit area of the melaleuca stands) and qualities (seed viability and germinability) of fallen melaleuca seeds. However, another set of confounding variables within habitats are melaleuca tree densities and DBH that may influence quantity and quality of fallen melaleuca seeds in the study sites within habitats. Therefore, prior to determining the effects of habitat and TSCs on seed it was necessary to determine the effects of habitat on these confounding variables in the habitat. The tree density and DHBs of plots within each habitat were pooled and tested for normality using Shapiro-Wilkin’s normality test procedures in Sigma Plot [18] . Overall stem density data passed (W-Statistics = 0.936; P = 0.117) the normality tests whereas, the DBH data failed (W-Statistics = 0.885; P = 0.009) this test. Therefore, double-sided logarithmic transformed (natural log-transformed) data for plot stem density and DBH were analyzed using GLM procedure [19] . Main effects and interaction terms that were not significant (P > 0.05) in the model were dropped and the variables with significant (P < 0.05) terms were included in further ad hoc analyses. Multivariate analysis of variance (MANOVA) was applied to assess habitat and TSC effects on confounding variables (density and DBH) as well as the variables of interest (quantity and quality of fallen melaleuca seeds).

The seed viability and germinability percentages were arc-sine transformed [arcsine value = (sqrt (viability or germinability percentage/100))] to stabilize the variance prior to further analyses to determine the effects of habitat and TSC on the quantity and quality (viability and germinability) of fallen melaleuca seeds. We examined possible correlations between the mean tree density and corresponding quantity of fallen melaleuca seeds within habitat applying Spearman’s Correlation Coefficient test procedure on these transformed data. Mean separation for the effects of habitat were performed using Fisher’s protected least significant difference (LSD) while the effects of the TSCs were performed using Duncan’s Multiple Range Test procedure. Although analyses were performed using double transformed data in all but stem density as indicated above, actual values related to the number of seeds and their viability and germinability have been used in the figures and tables presented herein.

The main effects of habitat and TSC on tree (≥1.3 m tall) density were significant while their interaction habitat * TSC was not significant (Table 1). Tree DBH in the study sites was not affected by habitat and habitats * TSC but it was affected by TSCs (Table 1). Mean density and DBH of the trees in the study sites are presented in Table 2. Statistically, stem densities in occasionally inundated habitats were not different across TSC even though there were relatively fewer stems in large tree size categories. On the other hand, stem densities in small and medium TSC in permanently inundated sites were significantly higher than

¹OI = occasionally inundated and PI¹ = permanently inundated. ²Numbers on the column within habitat followed by different letters are significantly different from each other as per Waller-Duncan’s multiple range test at P ≤ 0.05.

those in large TSC (Table 2). Tree stand attributes, such as tree DBH distribution in a forest stand, are important since this parameter is considered as a good predictor of the canopy-held seed reservoir of the serotinous trees (e.g., [17] [20] [21] ). This suggests that the production and releases of melaleuca seeds from trees should increase with an increase in tree DBH. Norghauer and Newbery (2015) [22] studied two tree species, Microberliniabisulcata A. Chev and Tetraberliniabifoliolata (Harms) Hauman in rain forests of Africa and reported a nonlinear asymptotic increase in seed production and subsequent releases in two masting events. In this case, the examination of correlations between the mean tree densities showed no significant correlation (r = −0.126, P = 0.577, N = 22) with the quantity of fallen melaleuca seeds across two habitats. However, viability (r = −0.363, P = 0.0972, n = 22) and germinability: (r = −0.438, P = 0.0412, n = 22) of fallen melaleuca seeds across two habitats indicated a weak to moderately strong correlation with tree densities. This indicated that melaleuca tree density did not effect seed quantity but had some negative effects on their viability and germinability in the studied habitats.

The main effect of habitat and its interaction with TSC did not have significant influence on the quantity of fallen seeds, whereas the main effect of TSC did (Table 3). Therefore, corresponding seed quantity data from occasionally and

permanently inundated habitats were pooled and mean separations were performed by TSCs. The mean number of fallen seeds (m²∙mo⁻¹) and the average number of seeds per tree are presented by TSC (Table 4).

The overall quantity (number of seeds ha⁻¹∙mo⁻¹) of fallen seeds in small and medium TSCs across two habitats were similar but in both these seed quantities were significantly less than the quantities in the large TSCs (Table 4). An average of 8947 seeds fell from a single tree in large TSC compared to 3032 and 4513 seeds in medium and small TSCs, respectively (Table 4). Woodall (1982) [9] studied seed rain pattern in a mature melaleuca tree stand (4600 trees of >5 cm dbh) and reported seed fall of 2260 seeds m²∙wk⁻¹ or ca 9820 seeds m²∙mo⁻¹, this amount approximately matches to the amount we have for our large TSCs. Rayachhetry et al. (2001) [17] conducted a study on melaleuca tree allometry and showed an increase in its canopy-held seed reservoir with the increase in tree sizes. In other systems larger trees in general have also been shown to produce higher number of seeds compared to the trees of smaller sizes (e.g., [23] [24] [25] ). These larger trees are expected to generate higher quantities of monthly seed fall in the forest stands. This assumption is corroborated by the results of a study on scots pine (Pinus sylvestris L.) that showed an increase in the number of seeds fallen with an increase in tree sizes [26] .

The main effects of habitat, TSC, and their interaction terms demonstrated a significant effect on the viability and germinability of fallen melaleuca seeds in our study sites (Table 3). Therefore, the seed viability and germinability attri- butes of fallen melaleuca seeds were analyzed by TSCs within each of the two habitat types.

¹Average number of seeds tree⁻¹∙mo⁻¹ = mean number of fallen seeds m²∙mo⁻¹/average # of trees in OI + PI within corresponding TSCs in Table 2. ²Numbers on the column followed by different letters are significantly different from each other as per Waller-Duncan’s multiple range test at P ≤ 0.05.

The viability and germinability of fallen melaleuca seeds among TSCs are shown in Table 5. The viability and germinability of fallen seeds between small and medium TSCs in occasionally inundated habitats were significantly different; however, these two attributes of large TSC were not different from those in small and medium TSCs. These results indicated no clear cut TSC effects on the quality of fallen melaleuca seeds in occasionally inundated habitat. On the other hand, the viability and germinability of fallen melaleuca seeds in permanently inundated habitat decreased with an increase in the TSCs. Overall, the viability and germinability of fallen melaleuca seeds had inverse relationship with the TSCs. Previously, melaleuca seeds held in older seed-capsules cohorts have been reported to have low viability and germinability due to their age; on the other hand, the larger trees tend to have older cohorts (crop) of seed-capsules [17] . Consequently, seeds contributed by the older capsule cohorts may have contributed to the lower seed viability and germinability in larger tree TSCs. A similar decrease in seed germinability with an increased age of the cone crop has also been reported in canopy-held seeds of a serotinous tree Allocasuarina distylla (Vent.) L. Johnson, and A. nana (Seib. ex Spreng.) L. [27] .

When pooled across TSCs, the quantity (F = 5.94, P = 0.0153), viability (F = 48.55, P < 0.0001), and germinability (F = 42.18, P ≤ 0.0001) of fallen seeds varied significantly between the two habitats. Overall, the occasionally inundated habitat (8961 seeds m²∙mo⁻¹) had significantly higher seed fall compared to those in permanently (6716 seeds m²∙mo⁻¹) inundated habitat (Figure 2(a)). Also, the viability of fallen melaleuca seeds in occasionally inundated habitat (8.84%) was significantly higher than in permanently inundated habitat (5.47%); the seed germinability in both habitats (8.4% in occasionally inundated and 5.07 in permanently inundated) followed trends similar to those observed in their respective viability (Figure 2(b)). These data indicate that a small fraction of fallen seeds also appeared to remain dormant in both habitats as described by Rayachhetry et al. (1998) [10] for canopy-held melaleuca seeds.

When averaged across habitats, the quantity (F = 7.04, P = 0.0010), viability (F = 3.40, P = 0.0346), and germinability (F = 4.65, P = 0.0102) attributes of fallen

seeds were significantly different among the three TSCs. Large TSCs (10446 seeds ha⁻¹∙mo⁻¹) had higher seed fall compared to those in medium (6557 seeds m²∙mo⁻¹) and small (8221 seeds m²∙mo⁻¹) TSCs (Table 4). The viability of fallen melaleuca seeds in small TSC (8.92%) was significantly higher than in medium (7.20%) and large (7.7%) TSCs; the seed germinability in two habitats followed trends similar to those observed in viability (Figure 3).

¹Numbers on the column within habitat followed by different letters are significantly different from each other as per Waller-Duncan’s multiple range test at P ≤ 0.05.

In summary, the quantity and quality of fallen melaleuca seeds was higher in occasionally inundated than in permanently inundated habitats (Table 4 and Table 5) and Figure 2(a) and Figure 2(b)). This trend is consistent with Rayachhetry et al. (1998) [10] who reported less germinability (6.0%) of canopy- held seeds from permanently inundated sites as compared to 7.8% from those in dry (occasionally-inundated) habitats, since canopy-held melaleuca seed reservoirs contribute to the seed fall in the respective studied habitats. On the other hand, melaleuca stands with predominantly smaller TSCs in both habitats had higher viability and germinability of fallen melaleuca seeds. One explanation for higher quality of seeds among the smaller TSCs may be related to their relative location along the periphery of the stand where their flowers were more accessible to pollinators than those of older trees located within the interior. This hypothesis requires further testing but pollens and pollinators have been reported to play important roles in seed qualities among flowering plants and increased pollinator populations often result in better seed set and greater seed quality in flowering plants [24] .