Classic research on elevational gradients in plant-herbivore interactions holds that insect herbivore pressure is stronger under warmer climates of low elevations. However, recent work has questioned this paradigm, arguing that it oversimplifies the ecological complexity in which plant-insect herbivore interactions are embedded. Knowledge of antagonistic networks of plants and herbivores is however crucial for understanding the mechanisms that govern ecosystem functioning. We examined herbivore damage and insect herbivores of eight species of genus Ficus (105 saplings) and plant constitutive defensive traits of two of these species, along a rainforest elevational gradient of Mt. Wilhelm (200 – 2700 m a.s.l.), in tropical Papua New Guinea. We report overall herbivore damage 2.4% of leaf area, ranging from 0.03% in Ficus endochaete at 1700 m a.s.l. to 6.1% in F. hombroniana at 700 m a.s.l. Herbivore damage and herbivore abundances varied significantly with elevation, as well as among the tree species, and between the wet and dry season. Community-wide herbivore damage followed a hump-shaped pattern with the peak between 700 and 1200 m a.s.l. and this pattern corresponded with abundance of herbivores. For two tree species surveyed in detail, we observed decreasing and hump-shaped patterns in herbivory, in general matching the trends found in the set of plant defences measured here. Our results imply that vegetation growing at mid-elevations of the elevational gradient, i.e. at the climatically most favourable elevations where water is abundant, and temperatures still relatively warm, suffers the maximum amount of herbivorous damage which changes seasonally, reflecting the water availability.

We performed the study along the Mt. Wilhelm elevational gradient in the Central Range of Papua New Guinea. The surveyed part of the transect, ca. 25 km long, comprises six study sites spaced regularly at 500 elevational metre intervals, from the lowest study site located within the lowland floodplains of the Ramu River at 200 m a.s.l. (05° 44′ S, 145°20′ E) to the elevational limit of Ficus distribution at 2700 m a.s.l. (05°48′ S, 145°09′ E). Mean annual temperature (measured by data loggers Comet R3120 placed in forest understorey) decreases from 27.4°C at the lowland site to 8.37°C at the tree line at a constant rate of 0.54°C per 100 elevational metres (see Sam et al. 2019 for more detail). Air humidity measured by the same dataloggers ranged between 90 and 100% at all elevations, however the rainfall was not evenly distributed during the year (see Sam et al. 2019 for more detail). Average annual precipitation is 3,288 mm (measured at local meteorological station, data provided by Phil Shearman) in the lowlands, rising to 4,400 mm at 2,700 m a.s.l., with a distinct condensation zone between 2,500 and 2,700 m a.s.l. (Sam et al. in prep).  Mean monthly precipitation along the gradient is 315 mm, and mean monthly precipitation between the two survey (i.e. wet season) periods was 398 mm. The elevational gradient, further description of study sites, climatic and habitat characteristics are published elsewhere (Tvardikova 2013, Sam & Koane 2014, Sam et al. 2015, Sam et al. 2019).

Unfortunately, no tree genus was distributed along the entire elevational gradient. Therefore, we selected species from the widely distributed genus Ficus (Moraceae) as a model system, which allowed us to work from the lowlands up to 2,700 m a.s.l. (05° 48′ S, 145°09′ E, Table 1). Ficus is an exceptionally species-rich genus, with New Guinea as the center of its diversity (Ronsted et al. 2008). Ficus also has high diversity (at least 75 species) and abundance (typically >5% of stems with DBH ≥1 cm) along the Mt. Wilhelm elevational gradient (L. Sam, unpubl. data). Ficus has an upper elevational limit at 2,900 m a.s.l at Mt. Wilhelm, as well as elsewhere (Berg & Corner 2005). Some species of Ficus (e.g. F. hahliana Diels, 1935) have particularly wide elevational ranges (Berg & Corner 2005).

We selected and tagged experimental saplings belonging to 4 – 6 evergreen Ficus species at each elevational study site (Table 2). As far as possible, we selected the locally most common species which had also broad elevational ranges. The selected plant species did not produce any exudates or sugar droplets attracting ants. We selected between 14 and 20 saplings per elevational study site, i.e. 3 – 5 per plant species and elevation or 105 saplings along the gradient (Table 2). For statistical independence, we allowed at least 80 m between any pair of individuals. We visually assessed saplings of the focal species and selected individuals that looked similar, had approximately 500 leaves growing within a well-developed crown 2.5 – 4 m above the ground. The saplings did not have any ant nests and did not have any abnormally high herbivory or fungal damage. Average leaf-sizes of the selected species ranged from 16.31 to 154.10 cm2, and two species (Ficus arfakensis King, 1888 and F. endochaete Summerhayes, 1941) had significantly different sized leaves at some elevations (Table 2).

At the beginning of the experiment, at the end of the dry season between 31-Aug-2014 and 3-Nov-2014 (Table 1), we first collected all arthropods (described below). Then we counted all leaves present at the sapling (to be able to estimate the total leaf area of the sapling) and further we proceeded with collection of leaves for herbivory measurement. To assess herbivore damage, we randomly selected two branches (with ca. 30 leaves each; 57 ± 5.6 (mean ± SE) in total per sapling, i.e. up to ca. 15% of standing foliage) per sapling. We clipped these two branches and collected all leaves from them. This survey period denotes the first point measurement (t = 0) of herbivore damage. A branch clipping (i.e. simulation of an artificial herbivory by a vertebrate herbivore) from a grown sapling should have only short-term and non-significant effect on subsequent herbivory measurement and plant growth in our experiment (Strauss et al. 1996, Seldal et al. 2017). 

Using a 50 x 50 cm2 white backgrounds, we took photographs of all collected and flattened leaves from each sapling (i.e. we took as many photos as needed to photograph all collected leaves). Using Adobe Photoshop CS6 (Adobe Systems Inc., USA). We first outlined the missing edges on the photographed leaves based on the expected shape and whitened or blackened various damages or holes on the leaves. Using various guides, google image searches, and our previous expertise, we carefully distinguished leaf damage caused by chewing and mining herbivores (which was whitened), while we did not consider fungal damage or mechanical leaf damage (which we blackened). We then turned the photos to black and white pictures. We used ImageJ version 1.47 (National Institute of Health, USA) to calculate the remaining leaf area (a, in cm2), the extrapolated leaf area without any herbivore damage (b), and the area lost to herbivory (c = b - a). We then estimated the percentage of leaf–area loss as c/b x 100. We calculated the herbivory loss as loss in % of are per leaf and in cm2 per 100cm2. These two values were highly correlated, so we used only % of herbivory per leaf in analyses and figures.

We conducted a second survey of herbivore damage, on the very same saplings as used in the first survey, approximately six months after the first survey (between 21-Mar-2015 and 11-May-2015, Table 1) at the end of the rainy season of 2014/2015. Similarly, to the first survey, we conducted the arthropod collection first. Then we collected all leaves from each sapling into a bag. We randomly selected 55-60 of leaves (to match the samples size of the first survey), photographed them on the same white background as earlier, and analysed them in the same way as in the first survey.

After each survey, we obtained values for 57 ± 5.6 (mean ± SE, min = 50, max = 95) leaves per sapling. We weighed all scanned and the remaining leaves to calculate the total leaf area for each sapling from the second survey. We used measured mean leaf size per sapling multiplied by the calculated number of leaves on sapling to obtain total leaf area of the sapling the beginning of the experiment (first survey)  The amount of herbivore damage was assessed blindly; viz. the field collector (BK) and the research assistant (AM) handling the leaves did not know that the study aimed to measure leaf damage, preventing any bias in leaf selection.

We measured herbivory on at least three saplings of each plant species at each elevational study site (Table 2) and in average of 438 leaves per plant species and elevation (or 114 ± 11.8 leaves per sapling), which is a number recommended by previous studies (Zhang et al. 2015, Kozlov & Zvereva 2017). By analysis of 11,448 leaves, our study belongs to those larger (Bito et al. 2011, Metcalfe et al. 2014, Kozlov & Zvereva 2017). The number of individual saplings sampled per species and elevational study site seemed to be sufficient as their herbivory between individuals did not differ significantly in majority of cases (Table S1) and the model with tree number as a random effect differed only marginally from model without random effect (AIC = 2.12). We therefore decided to use mean herbivory per species at each elevation and survey in further analyses.

Arthropod survey

Arthropod censuses were performed destructively during both surveys at the same individuals that were used for herbivory measurements, just before we started measurements of herbivorous damage and leaf collections. We slowly lowered the trunk of the sapling above a mosquito net, wrapped it to the mosquito net and sprayed with fast knock-down insecticide (Mortein®). After a while, we shook foliage firmly, opened the net and collected all arthropods (>1mm) and preserved them in vials filled with DNA grade ethanol. In laboratory, arthropods were counted, identified into three feeding groups (i.e. “chewing herbivores”, “predators”, “other arthropods” who have no relationship to herbivorous damage as we measured it and do not act as mesopredators – e.g. adult flies, pupae, adult Lepidoptera etc., sap sucking herbivores). Abundances of chewing herbivorous arthropods was then calculated as number of individuals in feeding guild per m2 of leaf area.

Plant defensive trait measurements

We measured plant defensive traits for two plant species (out of the 8 surveyed for herbivory). For each of these two species at each elevation, we collected leaf discs from five individuals. Where possible, the same individuals were used for herbivory and insect survey and trait measurements, but up to two individuals per plant species and elevation were sampled from different individual. These were usually individuals with smaller amount of leaves than was acceptable for the herbivory and herbivore survey. We collected two 4.5 cm2 leaf discs per leaf from 20 young, but fully expanded leaves for each individual, avoiding the central vein (1 g of dry weight in total on average). The discs were air-dried and stored in silica gel. Half of them were used for the analysis of triterpenes and half of them for measurement of physical traits.

Physical Traits: We measured trichome density and specific leaf area (SLA). Trichome density and SLA are parameters of leaf morphology with a possible impact on leaf-chewing insects. The total number of trichomes per 10 mm2 and their average length was measured on five leaf discs per individual using ImageJ (ver.1.48). Values for dorsal and ventral sides of the discs were combined. Specific leaf area (SLA) was calculated as the area per unit mass of five dried leaf discs collected from five leaves of known diameter for each individual.

Triterpene analysis: Dried powdered leaf tissue samples were ground with methanol in TissueLyser. Then they were centrifugated and aliquot was saved for further use. Terpenoids in aliquot were measured on a Dionex Ultimate 3000 LC system equipped with an Open XRS autosampler and coupled to a Q Exactive Plus Orbitrap mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). A reversed phase Kinetex was used for separating the analytes. For further measurement details, refer to the Methods S1 in Supplementary information.