Herbivores differentially limit the seedling growth and sapling recruitment of two dominant rain forest trees. (J. M. Norghauer & D. M. Newbery).
ESM resource 1 – Map of the 40 gap locations in the 82.5-ha P-plot
In the graphic each open circle represents a canopy gap that was used in the insect herbivore exclusion experiment. The forest area corresponding to the hatched red bar at the westernmost end of the plot was not suitable for sampling because of swampy conditions. The noticeable bare pockets where no gaps were sampled also resulted from mostly swampy or rocky conditions, or a lack of nearby fruiting trees of Microberlinia bisulcata and/or Tetraberlinia bifoliolata.
ESM resource 2 – Starting seedling sizes and light availability in gaps
Table 1. Means (± SE, range in parenthesis) of plant size and light levels of the samples followed in the insect herbivore exclusion experiment in Korup National Park (Cameroon). At the start of the experiment, each tree species’ sample (n) of seedlings was evenly split between unprotected (control) vs. protected (caged) at each of 40 gaps (34 with Microberlinia, 18 with Tetraberlinia, 12 with both spp.). %PPFD is a direct measure of light under overcast conditions; %CO is canopy openness calculated from hemispherical photographs (the 'mid' and 'end' subscripts correspond to measurements made during census 2 and 4, respectively, refer to Fig. 2). Different lower-case letters indicate a significant effect of the species term only following linear mixed model (LMM) analysis (P < 0.05).
Microberlinia bisulcata / Tetraberlinia bifoliolatan / Control / Caged / n / Control / Caged
Census 1 – sizes
Height (cm) / 194 / 21.1 ± 0.33a
(13.5 – 30.3) / 20.9 ± 0.36a
(13.5 – 32.0) / 86 / 19.6 ± 0.44b
(13.7 – 25.6) / 19.8 ± 0.40b
(14.3 – 24.8)
No. of leaves1 / 194 / 5.9 ± 0.18a
(4 – 12) / 5.9 ± 0.16a
(3 – 10) / 86 / 4.2 ± 0.12b
(3 – 8) / 4.1 ± 0.08b
(4 – 6)
Census 2 – light2
%PPFDmid1 / 194 / 5.4 ± 0.38a
(0.81 – 22.2) / 5.6 ± 0.35a
(0.78 – 18.9) / 86 / 5.4 ± 0.48a
(1.6 – 15.5) / 5.5 ± 0.50a
(1.3 – 13.7)
%COmid / 194 / 7.7 ± 0.17a
(4.4 – 11.7) / 7.9 ± 0.20a
(2.9 – 12.9) / 86 / 8.2 ± 0.26a
(5.3 – 10.9) / 8.5 ± 0.33a
(4.9 – 14.1)
Census 4 – light3
%PPFDend1 / 70 + 93 / 4.2 ± 0.59a
(0.34 – 27.9) / 3.7 ± 0.34a
(0.19 – 16.9) / 40 + 41 / 3.8 ± 0.52a
(0.32 – 14.6) / 3.9 ± 0.51a
(0.47 – 12.0)
%COend / 69 + 93 / 6.8 ± 0.15a
(4.0 – 10.8) / 6.8 ± 0.13a
(3.9 – 10.7) / 41 + 41 / 6.7 ± 0.17a
(3.7 – 8.6) / 6.8 ± 0.22a
(3.2 – 9.4)
Notes:
1 Log-transformed in LMMs.
2 Obtained at census 2 for all seedling locations in the original sample (see Fig. 2 and Methods).
3 Survivors only: the sample counts are given as ‘control + caged’ treatments and include ‘replacement’ seedlings (see Methods).
ESM resource 3 – Light availability inside the herbivory treatments
Methods
We compared direct light levels — using the same pair of SKP215 photon sensors — within the herbivore mesh treatments (caged vs. control) in situ based on five collected mini-datasets. First, on a partly cloudy day (11-Apr-2008) absolute PPFD (µmol/m2/sec) was measured simultaneously inside of cages and under the rooftops of their paired controls (n = 15 pairs) at five experimental blocks in the most eastern part of the P-plot. Second, in the morning of an overcast day (11-Nov-2008), absolute PPFD inside and outside of the two mesh treatments, and at the same height above ground, was compared simultaneously only for pairs of ‘caged-control’ seedlings < 2 m apart in 10 randomly selected experimental blocks in the eastern half of the P-plot (n = 20 pairs in total across gaps). Third, on 21-Nov-2008, percent-PPFD transmission of above-canopy radiation reaching inside cages and under rooftops was evaluated for nearly all seedlings at two eastern blocks only, in a balanced sample (n = 10 cages, n = 10 controls). Fourth, on another mostly overcast day (25-Nov-2008), absolute PPFD levels were measured only inside of cages and under control rooftops alternatingly with a single sensor, i.e. not simultaneously, for all gap seedlings (n = 88; 42 cages and 46 controls, including six ‘replacement’ controls) at 10 randomly stratified experimental blocks along the main E-W trail in the P-plot, with mesh ‘age’ scored as “original”, “middle”, or “newer”, corresponding to censuses 1, 2 and 3 respectively (see Fig. 2). For the fifth and final mini-dataset, alternating light readings were again made inside the treatments (18–19-Mar-2009) at 11 newly selected blocks, but measuring instantaneous PPFD outside as well as inside both cages and rooftops at the same height, and also recording the time of measurement — as in the data-set of 21-Nov-2008 — to express these without and within readings as %PPFD transmission through the canopy (for each type of mesh n = 102, 89 and 89, 84 for controls and cages respectively).
Results
These in-situ checks in the forest comparing direct light levels inside the two different mesh structures (caged vs. control) revealed negligible differences between them. For example, on 11-Apr-2008 the mean PPFD was very similar between cages and their paired controls (means ± SE: 120.1 ± 35 and 117.3 ± 31.6 μmol/m2/sec, respectively; paired t-test, d.f. = 14, P = 0.916). In the second dataset from 11-Nov-2008, mean absolute PPFD inside of neighboring cage and control treatment pairs was not significantly different across the 10 blocks (paired t-test, log-transformed values, d.f. = 19, P = 0.825; means ± SE, cages: 14.7 ± 2.3, controls: 17.4 ± 4.3 μmol/m2/sec). Similarly, there was no difference in mean % PPFD transmitted through the mesh material between treatments (i.e. [PPFDinside/ PPFDoutside]*100; paired t-test, P = 0.826; means ± SE for cages and controls, 68.8 ± 3.5% and 69.7 ± 3.6% respectively, n = 20). The small sample tested on 21-Nov-2008 suggested that the percentage of above-canopy radiation reaching seedlings inside the mesh treatments was also similar for cages and controls (respective means ± SE, 4.79 ± 1.00% and 5.09 ± 0.83%; t-test, d.f. = 18, P = 0.820).
From the two larger, more interspersed datasets (from 25-Nov-2008 and 18–19-Mar-2009), very few differences, in terms of light received by seedlings inside the treatments, could be detected between cages and controls. The right-skewed distributions of absolute PPFD
beneath the mesh were very similar between cages and controls (Kolmogorov-Smirnov two-sample test, c = 1.36, d.f. = 2, P = 0.508, n = 42, 46 respectively). Following log-transformation,
mean values (± SE) were similar between cages and controls (unbalanced ANOVA, treatment term, F1, 77 = 0.27, P = 0.607; untransformed means, 22.5 ± 3.1; 20.6 ± 2.8 μmol/m2/sec, respectively). Absolute PPFD also varied significantly among the 10 blocks through the course of the day (block term, F9, 77 = 11.7, P < 0.001). When mesh ‘age’ was first included in the model, the above patterns stayed the same, but the effect of treatment weakened (P = 0.879). This was because light inside the older original mesh was reduced by almost a half of that under previously and recently replaced mesh (age term: F2, 73 = 13.8, P < 0.001, means: “original” 14.9 ± 2.9, n = 37; “middle” 25.4 ± 6.2, n = 11; “newer” mesh age 26.6 ± 3.0 μmol/m2/sec, n = 40).
In the data-set of 18–19-Mar-2009, both PPFD outside of treatment meshes (ambient light) and above the canopy had been recorded, together with PPFD inside the treatment meshes. Considering first this ambient light, expressed in terms of %PPFD transmission through the canopy, the frequency distributions of values were near-identical for cages and controls (Kolmogorov-Smirnov two-sample test, c = 1.48, d.f. = 2, P = 0.478, n = 37, 40 respectively), as it was for %PPFD inside of the treatment meshes (c = 1.14, P = 0.566). Like the recordings of 25-Nov-2008, there was no significant difference between cages and controls in %PPFD inside of mesh treatments (unbalanced ANOVA, log-transformation, treatment term, F1, 65 = 0.97, P = 0.328; untransformed means: cages 3.3 ± 0.44%; controls 2.7 ± 0.34%). Similar results were obtained for ambient %PPFD (separate unbalanced ANOVA, F1, 65 = 1.62, P = 0.207; means 3.7 ± 0.49%; 4.2 ± 0.62%). Finally, and as before, for both of these variables significant differences
in mean %PPFD were detected among the 11 blocks (F10, 65 = 4.05 and F10, 65 = 13.3, respectively, both P < 0.001).
ESM resource 4 – Mesh transmission of light and weathering
Methods
The percent light transmission of unused mesh was measured near the camp-site on six separate pairs of cages and rooftops (without bamboo sticks). This was done under overcast sky conditions (8–9-Nov-2008) at 5-m intervals along six 50-m long transects that covered a wide range of canopy cover (open clearings to closed forest; one pair of treatments per transect, 66 locations in total). Moving from location to location, a single Skye photon sensor was positioned at 30 cm height and an ‘open’ reading was first taken, followed immediately by a reading under a rooftop, or inside a cage (both without sticks). These readings, once initiated, were sequentially made in this order (i.e., not thereafter at random, or alternatingly).
At census 4 we removed the mesh from experimental seedlings (see Fig. 2). Because we were interested mainly in checking changes in mesh transmission between herbivory treatments on fast-growing seedlings, we focused on the used mesh consisting of medium- and large-sized cages and rooftops; and for a subset of these, during collection, we ribbon-labeled whether the treated seedling had lived or died [including replacements]. A photon sensor was positioned at heights of 60, 75 and 100 cm per corresponding cage size, and absolute PPFD light transmission through the mesh treatment was recorded. To obtain percent light transmission, the time to the nearest minute was also recorded to compare with above-canopy radiation being monitored at camp (see Methods). This was done on 13-Nov-2009 and 18-Nov-2009 for 53 large, and 202 medium, cages and rooftops (n = 255 in total), split across three gap-like locations near the camp: simulated large gap (center of clearing), partial shade (edge of clearing), and in a nearby
small gap (tree-fall). To better match conditions in the field, large cages and some large rooftops were measured only in the large gap and its edge, and not in the small gap.
Results
Considering only the subsample with diffuse PAR ³ 1.00 (outside the treatments), which is more realistic of gap-like conditions, the original-sized cage installations made of new mesh material transmitted on average c. 83.8% (SE = 0.61, median = 84%, n = 32) of diffuse light. Similarly, for the control rooftop treatments, the new mesh material allowed 83.4% of light to pass (SE = 0.70, median = 84%, n = 61).
In November 2009, no significant differences in light transmission were found amongst medium-sized treatments (cages: 74.0 ± 4.7% vs. controls: 73.3 ± 4.1%; n = 82, 120), but these values significantly exceeded that transmitted by worn mesh of large-sized treatments, for which in turn, cages transmitted significantly more light than rooftops (66.7 ± 5.7% vs. 69.9 ± 5.3%, n = 41, 12, respectively) (unbalanced ANOVA; cage size x treatment interaction, F2, 249 = 5.3, P = 0.022; means compared using average LSD at 5% significance level). Light transmission was significantly higher where more direct light occurred (center of clearing) than where diffuse light was also present (edge of clearing and tree-fall gap) (block effect of location: F2, 249 = 18.8, P < 0.001). These findings suggested only small treatment differences for transmission of light through worn mesh of different sizes, and that direct light was more transmissible than diffuse light through the used mesh material.
When the PPFD readings for ambient light, and for inside cages and rooftops, were expressed as percent above-canopy radiation, and plotted for all locations on log-log axes (cages and rooftops separately), their regression slopes were 0.9938 and 1.012 respectively. Owing to the large sample sizes these slopes were marginally significantly different (grouped regression on
log-transformed values: F1, 251 = 5.94, P = 0.045). Nevertheless, the reductions in light availability going from ambient conditions to inside treatments, when expressed as the absolute
differences in percentage of light received through the canopy were, on average, not greater for cages than rooftops (means: 7.96% vs. 8.28%, medians: 4.45% vs. 4.86%, n = 123, 132, respectively; Mann-Whitney U-test, d.f. = 1, P = 0.456).
For the subset of control rooftops — no seedlings in cages of medium or large size died — mesh above seedlings that had died let c. 2% more light through than mesh over seedlings that survived the experiment (t-test, t = -1.89, d.f. = 45, P = 0.066; means: 71.7 ± 0.62% and 73.5 ± 0.65%, n = 27, 20 for live and dead respectively). Agreement was found when the difference in %PPFD of above-canopy radiation through the mesh was evaluated instead: rooftops under which seedlings had died showed significantly lower reduction in light than those in which seedlings had survived (t-test, t = -3.63, d.f. = 45, P < 0.001; 6.6 ± 1.00% and 11.1 ± 0.75%, respectively).
ESM resource 5 – Hemispherical photographs
Under cloudy conditions, a single color photograph was taken above each experimental seedling using a digital camera affixed with a fish-eye lens (Nikon Coolpix 950, 990, 995 and FC-E8 converter, respectively). Photographs were taken with the LCD screen oriented to the north on a tripod with a bubble leveling unit (models GT0530 + G11178, respectively; Gitzo SA, Italy) at a height of 1 m, and thus above the original-sized mesh treatments or, as needed, higher for seedlings that grew fast in gaps. The first set of photographs was taken on 17–24-Oct-2008, which corresponded to c.10-11 mo into the experiment, and thus its approximate temporal midpoint (i.e., census 2, see Fig. 2). To quantify variation in the light environment at the experiment’s end, that is c. 22 mo after installation (i.e. census 4, see Fig. 2), a second complete set was taken on 7–9-Nov-2009. All photographs were taken using the same settings, i.e., no flash, centre-weighted with the REGULAR lens (in the last period, however, at a very small subset of locations the FISHEYE1 lens converter setting was used to compare effects of camera setting on later light calculations).