Introduction
Tropical rainforests are known for their exceptional biodiversity, often concentrated in their canopies (Ødegaard Reference Ødegaard2000). Despite the vital role canopy habitats play in sustaining a significant portion of global biodiversity (Basset et al. Reference Basset, Cizek, Cuénoud, Didham, Novotny, Ødegaard, Roslin, Tishechkin, Schmidl, Winchester, Roubik, Aberlenc, Bail, Barrios, Bridle, Castaño-Meneses, Corbara, Curletti, Duarte Da Rocha and Leponce2015, de Souza Amorim et al. Reference de Souza Amorim, Brown, Boscolo, Ale-Rocha, Alvarez-Garcia, Balbi, Barbosa, Capellari, De Carvalho, Couri, Dios, Fachin, Ferro, Flores, Frare, Gudin, Hauser, Lamas, Lindsay and Rafael2022), they receive relatively little attention compared to tropical forest understories. Nevertheless, the multilayered structure (i.e. vertical and horizontal spaces) of tropical forests supports rich flora and fauna, along with their interactions, which require our closer attention (Erwin Reference Erwin and Levin2001).
Arthropods, especially herbivorous insects, can cause significant damage to plants (Cyr and Pace Reference Cyr and Pace1993, Novotny et al. Reference Novotny, Miller, Baje, Balagawi, Basset, Cizek, Craft, Dem, Drew, Hulcr, Leps, Lewis, Pokon, Stewart, Allan Samuelson and Weiblen2010), impacting their competition and community composition (Coley Reference Coley1983). This damage is often most pronounced in the forest understory, where limited light, intense competition for resources, and reduced leaf area make plants more vulnerable to herbivory (Coley and Barone Reference Coley and Barone1996).
The interaction between plants and insect herbivores (hereafter, ‘herbivores’) is a central theme in ecology (Maraia et al. Reference Maraia, Charles-Dominique, Tomlinson, Staver, Jorge, Gélin, Jancuchova-Laskova, Sam, Hattas, Freiberga and Sam2024; Metcalfe et al. Reference Metcalfe, Asner, Martin, Espejo, Huasco, Amézquita, Carranza-Jimenez, Cabrera, Baca, Sinca, Quispe, Taype, Mora, Dávila, Solórzano, Vilca, Román, Bustios, Revilla and Malhi2014; Sam et al. Reference Sam, Koane, Sam, Mrazova, Segar, Volf, Moos, Simek, Sisol and Novotny2020, Reference Sam, Mrazova, Houska Tahadlova, Kollross, Maraia, Moreira and Abdala-Roberts2024; Shao et al. Reference Shao, Zhang and Yang2021). Although insect–plant interactions have been widely studied along latitudinal (Dyer et al. Reference Dyer, Singer, Lill, Stireman, Gentry, Marquis, Ricklefs, Greeney, Wagner, Morais, Diniz, Kursar and Coley2007; Anstett et al. Reference Anstett, Naujokaitis-Lewis and Johnson2014; Moles et al. Reference Moles, Bonser, Poore, Wallis and Foley2011) and altitudinal gradients (Sam et al. Reference Sam, Koane and Novotny2015; Zvereva et al. Reference Zvereva, Zverev and Kozlov2022), relatively little is known about their dynamics across vertical forest gradients (Kristensen et al. Reference Kristensen, Rousk and Metcalfe2020). Vertical gradients within forests are steeper, involving multilayered foliage with distinct traits over shorter distances (Nakamura et al. Reference Nakamura, Kitching, Cao, Creedy, Fayle, Freiberg, Hewitt, Itioka, Koh, Ma, Malhi, Mitchell, Novotny, Ozanne, Song, Wang and Ashton2017). Vertical forest gradients experience climatic variations like temperature and humidity (Haack et al. Reference Haack, Borges, Grimm-Seyfarth, Schlegel, Wirth, Bernhard, Brunk, Henle and Pereira2022; Stiegel et al. Reference Stiegel, Entling and Mantilla-Contreras2017), which can result in stratum-specific biotic interactions.
Vertical forest layers support distinct arthropod communities, including herbivorous insects and predatory arthropods play key roles in shaping forest dynamics (Basset et al. Reference Basset, Novotny, Miller and Kitching2003). Herbivorous insects, such as caterpillars (Lepidoptera larvae), beetles (Coleoptera), and sap-sucking hemipterans, feed on plant tissues, contributing to foliage loss and affecting plant growth and regeneration (Novotny et al. Reference Novotny, Miller, Baje, Balagawi, Basset, Cizek, Craft, Dem, Drew, Hulcr, Leps, Lewis, Pokon, Stewart, Allan Samuelson and Weiblen2010). In contrast, predatory arthropods, including spiders (Araneae), predatory ants (Hymenoptera), and mantises (Mantodea), regulate herbivore populations through top-down control, mitigating excessive herbivory and maintaining ecological balance (Roslin et al. Reference Roslin, Hardwick, Novotny, Petry, Andrew, Asmus, Barrio, Basset, Boesing, Bonebrake, Cameron, Dáttilo, Donoso, Drozd, Gray, Hik, Hill, Hopkins, Huang and Slade2017; Sanders and Platner Reference Sanders and Platner2007; Leles et al. Reference Leles, Xiao, Pasion, Nakamura and Tomlinson2017). The composition and density of these guilds vary along the vertical gradient, with herbivores typically concentrated in the lower strata where leaves are more palatable (Coley and Barone Reference Coley and Barone1996; Lowman Reference Lowman1985; Piper et al. Reference Piper, Altmann and Lusk2018), while predators are more abundant in the upper canopy, where they exploit a diverse prey base and benefit from increased temperature and reduced humidity (Basset et al. Reference Basset, Cizek, Cuénoud, Didham, Novotny, Ødegaard, Roslin, Tishechkin, Schmidl, Winchester, Roubik, Aberlenc, Bail, Barrios, Bridle, Castaño-Meneses, Corbara, Curletti, Duarte Da Rocha and Leponce2015; Stork et al. Reference Stork and Blackburn1993). Understanding how these feeding guilds interact across vertical stratification is crucial for predicting the cascading effects of herbivory on forest health and ecosystem stability.
Despite vertical forest gradients having gained some attention, challenges persist in accessing them, particularly in wet tropical regions (Barker and Pinard Reference Barker and Pinard2001; Barker and Sutton Reference Barker and Sutton1997). Most studies on vertical forest gradients have been conducted in temperate forests (Leidinger et al. Reference Leidinger, Seibold, Weisser, Lange, Schall, Türke and Gossner2019; Schowalter et al. Reference Schowalter, Zhang and Progar2005; Šigut et al. Reference Šigut, Šigutová, Šipoš, Pyszko, Kotásková and Drozd2018; Stiegel et al. Reference Stiegel, Entling and Mantilla-Contreras2017; Thomas et al. Reference Thomas, Sztaba and Smith2010) and in tropical dry forests (Neves et al. Reference Neves, Araújo, Espírito-Santo, Fagundes, Fernandes, Sanchez-Azofeifa and Quesada2010, Reference Neves, Silva, Espírito-Santo and Fernandes2014; Silva et al. Reference Silva, Leal, Espírito-Santo and Morais2017). Additionally, these investigations typically focus on documenting the abundance and species richness of various arthropod groups (Basset et al. Reference Basset, Cizek, Cuénoud, Didham, Novotny, Ødegaard, Roslin, Tishechkin, Schmidl, Winchester, Roubik, Aberlenc, Bail, Barrios, Bridle, Castaño-Meneses, Corbara, Curletti, Duarte Da Rocha and Leponce2015; de Souza Amorim et al. Reference de Souza Amorim, Brown, Boscolo, Ale-Rocha, Alvarez-Garcia, Balbi, Barbosa, Capellari, De Carvalho, Couri, Dios, Fachin, Ferro, Flores, Frare, Gudin, Hauser, Lamas, Lindsay and Rafael2022), neglecting research on biotic interactions between herbivores and the plants (Metcalfe et al. Reference Metcalfe, Asner, Martin, Espejo, Huasco, Amézquita, Carranza-Jimenez, Cabrera, Baca, Sinca, Quispe, Taype, Mora, Dávila, Solórzano, Vilca, Román, Bustios, Revilla and Malhi2014). Thus, research combining abiotic and biotic factors along the vertical forest gradient is crucial for providing insights into the evolutionary dynamics of plant–animal interactions, such as herbivory, which, in turn, has implications for forest health (Jaworski and Hilszczański Reference Jaworski and Hilszczański2013).
Humidity and temperature are known to affect forest functioning crucially (Lin et al. Reference Lin, Chen, Zhang, Fu and Fan2017). Plant transpiration and heat generation depend largely on soil moisture and atmospheric vapour pressure deficit, both influenced by air humidity (Damour et al. Reference Damour, Simonneau, Cochard and Urban2010). In arid conditions, plants may close stomata to reduce water loss, leading to increased leaf temperatures (Lin et al. Reference Lin, Chen, Zhang, Fu and Fan2017). This effect is especially pronounced from the forest understory to the canopy (Nakamura et al. Reference Nakamura, Kitching, Cao, Creedy, Fayle, Freiberg, Hewitt, Itioka, Koh, Ma, Malhi, Mitchell, Novotny, Ozanne, Song, Wang and Ashton2017). Such variability influences herbivores and arthropods, affecting their body size, density, diversity, diet preferences, distribution, and phenology (Bauerfeind and Fischer Reference Bauerfeind and Fischer2013; Clissold and Simpson Reference Clissold and Simpson2015).
Moreover, there are steep vertical gradients in temperature and humidity within tropical forests, strongly influencing plants and arthropod physiology. In lower forest strata, cooler temperatures and higher humidity favour the persistence of softer, more palatable leaves, supporting dense herbivore populations. As one ascends through the forest layers, temperature generally increases and humidity decreases, exposing organisms to harsher, drier conditions (Nakamura et al. Reference Nakamura, Kitching, Cao, Creedy, Fayle, Freiberg, Hewitt, Itioka, Koh, Ma, Malhi, Mitchell, Novotny, Ozanne, Song, Wang and Ashton2017; Haack et al. Reference Haack, Borges, Grimm-Seyfarth, Schlegel, Wirth, Bernhard, Brunk, Henle and Pereira2022). Elevated temperatures at the canopy level can accelerate arthropod metabolic rates, growth, and development but may also impose thermal stress, favouring larger-bodied species that can better regulate internal temperatures (Pincebourde et al. Reference Pincebourde, Dillon and Woods2021; Laird-Hopkins et al. Reference Laird-Hopkins, Ashe-Jepson, Basset, Cobo, Eberhardt, Freiberga, Hellon, Hitchcock, Kleckova, Linke, Lamarre, McFarlane, Savage, Turner, Zamora, Sam and Bladon2023). Similarly, reduced humidity can intensify leaf toughness by promoting higher dry matter content, altering leaf nutritional quality, and subsequently influencing herbivory patterns (Damour et al. Reference Damour, Simonneau, Cochard and Urban2010, Lin et al. Reference Lin, Chen, Zhang, Fu and Fan2017). Thus, vertical microclimate gradients play a central role in structuring plant defences and arthropod community dynamics, with cascading effects on herbivory across forest strata.
Despite the limited seasonality of many tropical forests, even slight seasonal rainfall variability can yield significant differences (Allen et al. Reference Allen, Dupuy, Gei, Hulshof, Medvigy, Pizano, Salgado-Negret, Smith, Trierweiler, Van Bloem, Waring, Xu and Powers2017). Particularly, in wetter tropical regions, species remain evergreen (Coley and Barone Reference Coley and Barone1996). However, woody tropical rainforest plants in families, including Sapotaceae, Meliaceae, Fabaceae, Cannabaceae, Malvaceae, and Euphorbiaceae, may shed leaves during brief dry periods, which can impact the population dynamics of herbivores (Morante-Filho et al. Reference Morante-Filho, Arroyo-Rodríguez, Lohbeck, Tscharntke and Faria2016). Conversely, the beginning of the wet season, particularly when young leaves emerge, provides essential resources for many herbivores (Barone Reference Barone2000; Lowman Reference Lowman1985). Unfortunately, there has been little focus on studying seasonal herbivory patterns in weakly seasonal tropical rainforests, and research on herbivory damage accumulation and its change over time is rare (Coley and Barone Reference Coley and Barone1996; Quiroz-Pacheco et al. Reference Quiroz-Pacheco, Mora, Boege, Domínguez and del-Val2020). Typically, studies measure standing herbivory at a given time (Houska Tahadlova et al. Reference Houska Tahadlova, Mottl, Jorge, Koane, Novotny and Sam2023; Kozlov and Zvereva Reference Kozlov, Zvereva, Cánovas, Lüttge and Matyssek2017; Neves et al. Reference Neves, Silva, Espírito-Santo and Fernandes2014; Zhu et al. Reference Zhu, Sun, He, Cai, Zhu and Ji2024), with changes over time seldom measured (but see Frost Reference Frost2022; Lowman Reference Lowman1985; Weissflog et al. Reference Weissflog, Markesteijn, Lewis, Comita and Engelbrecht2018).
Plant traits and species diversity also influence insect herbivory and arthropod community dynamics (Castagneyrol et al. Reference Castagneyrol, Giffard, Péré and Jactel2013, Reference Castagneyrol, Giffard, Valdés-Correcher and Hampe2019; Ebeling et al. Reference Ebeling, Hines, Hertzog, Lange, Meyer, Simons and Weisser2018; Maraia et al. Reference Maraia, Charles-Dominique, Tomlinson, Staver, Jorge, Gélin, Jancuchova-Laskova, Sam, Hattas, Freiberga and Sam2024; Martini et al. Reference Martini, Sun and Chen2022). Increased plant species richness and certain leaf traits diversity foster diverse arthropod communities, benefiting arthropod assemblages (Ebeling et al. Reference Ebeling, Hines, Hertzog, Lange, Meyer, Simons and Weisser2018; Harrison et al. Reference Harrison, Philbin, Gompert, Forister, Hernandez-Espinoza, Sullivan, Wallace, Beltran, Dodson, Francis, Schlageter, Shelef, Yoon and Forister2018). Leaf traits such as leaf toughness (LT), leaf dry matter content (LDMC), trichome density (TD), and specific leaf area (SLA) influence herbivory by shaping arthropod community composition and damage extent (Azevedo-Schmidt and Currano Reference Azevedo-Schmidt and Currano2024; Barbour et al. Reference Barbour, Rodriguez-Cabal, Wu, Julkunen-Tiitto, Ritland, Miscampbell, Jules and Crutsinger2015; Matsuki and MacLean Reference Matsuki and MacLean1994). LT acts as a barrier against herbivores, especially leaf-chewing insects (Caldwell et al. Reference Caldwell, Read and Sanson2016; Malishev and Sanson Reference Malishev and Sanson2015), correlating positively with LDMC, which increases with tree height (Kenzo et al. Reference Kenzo, Mohamad and Ichie2022). Leaves in the forest canopy, exposed to higher light intensity, have higher LDMC, making them tougher and more resistant to insects (Caldwell et al. Reference Caldwell, Read and Sanson2016). TD also increases towards the canopy (Ichie et al. Reference Ichie, Inoue, Takahashi, Kamiya and Kenzo2016), acting as a physical barrier and reducing leaf palatability. Similarly, SLA serves as a strategy for water conservation and protection against herbivores. Plants with low SLA typically thrive on infertile soils, while those with high SLA are more prevalent on fertile soils (Hodgson et al. Reference Hodgson, Montserrat-Martí, Charles, Jones, Wilson, Shipley, Sharafi, Cerabolini, Cornelissen, Band, Bogard, Castro-Díez, Guerrero-Campo, Palmer, Pérez-Rontomé, Carter, Hynd, Romo-Díez, de Torres Espuny and Pla2011; McIntyre Reference McIntyre2008). In drier environments, plants tend to develop thicker and more resistant leaves, while in humid environments, especially in the understory or shade, leaves are thinner and softer (Pinho et al. Reference Pinho, Tabarelli, Engelbrecht, Sfair and Melo2019), making them more palatable to herbivores. However, these leaf traits can vary between plant species, among individuals, across vertical forest strata, as well as across microclimatic gradients and seasonality, undoubtedly influencing herbivore population dynamics and arthropod community structure within forest ecosystems (Bröcher et al. Reference Bröcher, Ebeling, Hertzog, Roscher, Weisser and Meyer2023; Castagneyrol et al. Reference Castagneyrol, Giffard, Péré and Jactel2013; Jia et al. Reference Jia, Yang, Castagneyrol, Yang, Yin, He, Yang, Zhu and Hao2023; Whitfeld et al. Reference Whitfeld, Novotny, Miller, Hrcek, Klimes and Weiblen2012).
In this study, we investigate how microclimate, seasonality, and leaf traits influence insect herbivory (hereafter, ‘herbivory’), as well as the body size and density of arthropods, across the vertical forest gradient in a weakly seasonal tropical rainforest in Papua New Guinea. We test the following five hypotheses:
H1 Herbivory will be higher in the lower forest strata due to the prevalence of thinner, softer leaves with higher SLA (Coley and Barone Reference Coley and Barone1996) and lower TD (Ichie et al. Reference Ichie, Inoue, Takahashi, Kamiya and Kenzo2016).
H2 Herbivory rates will increase during the wet season, due to a higher abundance of newly flushed, less-defended leaves and increased damage to mature leaves (Barone Reference Barone2000; Coley and Barone Reference Coley and Barone1996).
H3 Arthropod density will be greater in the upper canopy compared to lower forest strata (Basset et al. Reference Basset, Aberlenc, Barrios, Curletti, Bérenger, Vesco, Causse, Haug, Hennion, Lesobre, Marquès and O’Meara2001).
H4 Arthropod body size will be larger in the canopy, where larger body size facilitates thermoregulation under harsher abiotic conditions (Coley and Barone Reference Coley and Barone1996; Laird-Hopkins et al. Reference Laird-Hopkins, Ashe-Jepson, Basset, Cobo, Eberhardt, Freiberga, Hellon, Hitchcock, Kleckova, Linke, Lamarre, McFarlane, Savage, Turner, Zamora, Sam and Bladon2023; Pincebourde et al. Reference Pincebourde, Dillon and Woods2021).
H5 Due to the leaf economics spectrum (Wright et al. Reference Wright, Reich, Westoby, Ackerly, Baruch, Bongers, Cavender-Bares, Chapin, Cornelissen, Diemer and Villar2004), we expect lower leaf quality, lower SLA, and higher toughness in canopy leaves compared to understory leaves. These traits will negatively influence chewing herbivory rates (Stiegel et al. Reference Stiegel, Entling and Mantilla-Contreras2017).
Methods
Study site and plant species selection
The study was conducted in the Baitabag lowland rainforest, Madang Province, Papua New Guinea (38 m above sea level, 5°08’ 18.21” S, 145°46’ 24.61” E) (Figure 1). The climate is humid, with an annual rainfall of 3,558 mm, a moderate dry season from July to September, and an average air temperature of 26.5°C (McAlpine et al. Reference McAlpine, Keig and Falls1983). The study site is part of the Kau Wildlife Management Area, covering 365 hectares of pristine rainforest. The terrain is hilly, with a nearby narrow stream. Soil is silty clay. In 2019, the New Guinea Binatang Research Center established a 1-hectare long-term botanical plot equipped with a canopy crane. The Kakoba Research Canopy Crane is 45 m tall, covering a 50 m radius, suitable for monitoring of plant growth and canopy research.
Figure 1. Design of the study: Black trees illustrate an example of vertical sampling of seven branches along one replicate of one tree species. This design was used because some trees did not have branches available at all heights (A). Photo of Kakoba canopy crane with its height indicated (B). Map of Papua New Guinea with the orange dot indicating the location of the study site (Baitabag, Madang Province) (C).
Seven plant species (Table 1) were selected within a 1-hectare plot based on their common occurrence in both the understory and as canopy trees. The accessibility of their crowns from the crane was considered for higher strata. Several individuals were selected from each species so that three individual branches considered for the study could be selected at each of the following heights: 1 m, 5 m, 10 m, 15 m, 20 m, 25 m, and 30 m (i.e. 7 plant species × 3 branches × 7 heights = up to 147 samples). Note that while all branches at 1 and 5 m height represent separate individuals, several branches could have been collected from a single individual at higher strata. The structure of the data was considered in the models. However, this design could not be fully applied to Donella lanceolata, which lacked branches at heights of 15 and 20 m; Dysoxylum arborescens, which lacked branches at 25 and 30 m; and Erythrospermum candidum and Gymnacranthera farquhariana var. paniculata, which lacked branches at 30 m. As a result, a total of 129 samples were assessed. Furthermore, two branches were lost during the experiment, preventing full replication of the herbivory measurements in time, so the total number of sampled branches was reduced to 127 for herbivory. Whole saplings were used for samples representing 1 m height, while branches were selected for all higher heights. For simplicity, we refer to both branches and saplings as ‘branches’. In higher strata, branches were typically collected from a single individual tree to cover all forest layers of the plot, as the emergent trees reached slightly above 30 m. All branches were selected only if they had more than 30 leaves and were at least 3 m apart. The Canopy Research Crane was used to sample branches that were 5 m above ground. Branches 1 m above ground were sampled from the ground. Prior to selection, branches were carefully examined to ensure no ant nests, unusually high levels of herbivory, or fungal damage were present.
Table 1. List of tree species, the number of branches sampled, and the mean number of leaves per branch estimated during three different sampling seasons, including the mean leaf area of the leaves (m2) of these branches. The three surveys were conducted at the beginning of the dry season (June 2021), the beginning of the wet season (November 2021), and at the end of the wet season (March 2022). Sampling heights ranged from 1 m to 30 m, with increments of 5m (i.e. 1 m, 5 m, 10 m, 15 m, 20 m, 25 m, and 30 m), except for the 4 m increment between 1 m and 5 m
Herbivory measurement in the field and laboratory
The experiment was conducted during three surveys between June 2021 and March 2022. We conducted the first survey at the beginning of the dry season in June 2021, the second at the start of the rainy season in November 2021, and the third at the end of the rainy season in March 2022. Initially, individual branches with over 30 leaves (Table 1) were selected. Thirty leaves per branch were marked with permanent ink markers. The leaves were distributed on three different twigs per branch, using coloured strings for easy identification during subsequent surveys. Leaf markings were regularly checked and renewed if affected by weather, especially rainfall and moisture on the leaves. Digital cameras (Nikon camera® and Olympus Tough TG-6 red camera®) were used for the first and second surveys, photographing each marked leaf against a white cardboard sheet (50 × 50 cm) with a ruler taped at each corner for scale. In the second survey, new leaves and shoots from marked branches were included. In total, 185 new leaves appeared. In the final survey, all marked and newly grown leaves were collected and scanned (Brother scanner, PDF format, 600 dpi). For species with compound leaves, namely, Pometia pinnata and Dysoxylum arborescens, each pinnule attached to the petiole was marked and considered to be an individual leaf.
Herbivory analysis
Assessment of herbivory caused by chewing herbivores followed a previously established protocol at the level of three twigs per branch (Sam et al. Reference Sam, Koane, Sam, Mrazova, Segar, Volf, Moos, Simek, Sisol and Novotny2020). We outlined missing leaf edges using expected shapes, converted photos to black and white using Photoshop® software (Adobe Photoshop CS6, Adobe Systems Inc.), and calculated consumed and remaining leaf areas with ImageJ® software (version 1.47, National Institutes of Health). We summed the consumed and remaining leaf areas from the 10 leaves per twig. Then, we expressed the proportion of leaf area loss per twig per m² of foliage (i.e. obtained three measurements per branch). Due to inaccuracies caused by photographing the leaves on the trees from the crane gondola (e.g. due to shadows, angles, etc.), the damage per twig might have slightly decreased between re-measurements over time. For such twigs, we considered the change in herbivory damage to be 0.
Leaf trait measurements
At the end of the experiment (March 2022), we collected both the initially marked leaves and newly sprouted young leaves growing above the originally marked leaves, on the branches marked with coloured strings. After scanning the leaves for herbivory analysis, we randomly selected five leaves from each sample (i.e. including mature and young leaves). From each of these, 10 circular leaf discs (15 mm in diameter) were obtained using a cork borer, excluding the primary vein. The individual fresh leaf discs were then weighed. Subsequently, we oven-dried leaf discs at 60°C for 72 hours and then re-weighed them to determine dry mass. Using these dried leaf discs, we analysed SLA, LDMC, LT, and TD (Cornelissen et al. Reference Cornelissen, Lavorel, Garnier, Díaz, Buchmann, Gurvich, Reich, Steege, Morgan, Heijden, Pausas and Poorter2003). TD was generally very low, and we observed no pattern of vertical forest stratification; hence, it was excluded from the analyses. SLA was calculated by dividing the area of each leaf disc (mm2) by its dry mass (mg). LDMC was determined by dividing leaf dry mass by fresh weight. For TD (number of trichomes per cm²), we used three leaf discs. Each leaf disc was analysed under a Dino-Lite Digital Microscope, Version 2018Q3©, to count the leaf TD in a random 1 mm² area on both the abaxial and adaxial surfaces, avoiding the midrib. We measured LT using an FL50 penetrometer with a 1.5-mm-wide tip (Stepper Motor Powered Test Stand, TVO 500N55S, Sauter GmbH) to puncture the leaf.
Arthropod sampling
Just prior to each herbivory damage sampling event (to avoid disturbing the foliage), we used a beating method to collect arthropods from individual branches (Montgomery et al. Reference Montgomery, Belitz, Guralnick and Tingley2021). The branches were shaken vigorously five times to dislodge any arthropods onto the 1.5 × 1.5 m sheet (Ribeiro et al. Reference Ribeiro, Borges, Gaspar, Melo, Serrano, Amaral, Aguiar, André and Quartau2005). We collected all arthropods longer than 1 mm and preserved them in 8 ml vials containing absolute ethanol. We also searched the foliage for any remaining arthropods and took note of the individuals that escaped during the beating. Later, body sizes of all arthropods were measured under a stereo microscope (in mm), identified to order (Araneae, Hymenoptera, Coleoptera, Hemiptera, Blattodea, Orthoptera, and Lepidoptera larvae), and categorized into two feeding guilds: predators and herbivores (hereafter, ‘herbivores’). The remaining individuals were grouped into a third guild as ‘others’ (i.e. those not predating and those not causing chewing damage to the leaf, which means sapsuckers, gallers, parasitoids, fungivores, wood-borers, arthropods not feeding during the adult stage, etc.). Although this third group doesn’t affect the plants directly, these insects are prey of predatory arthropods, playing an important role in arthropod community dynamics. We therefore consider them only in some analyses, and we do not link them directly to leaf damage. Arthropod density was calculated as the number of individuals per m² of total leaf area of individual branches, for all arthropods as well as for the three guilds. The total leaf area of branches was calculated as the median of the estimates of the number of leaves per branch, multiplied by the mean leaf area (for given species and strata) obtained from the dry scans.
Microclimatic data
Microclimatic data (air temperature and humidity) were recorded hourly over the whole length of the experiment from June 2021 to March 2022 using data loggers (COMET S3120E, Comet Logger). Positioned on the research crane, six data loggers were placed at intervals of 0 m, 7 m, 14 m, 21 m, 28 m, and 35 m along the forest strata. These loggers covered the entire expanse of the tree canopy, including up to the emergent treetops. However, several data loggers failed and their locations did not match the forest strata heights surveyed by us, and we therefore use the data only for discussion, not for any rigid analyses (Figure S3).
Statistical analyses
All statistical analyses were performed in statistical software R version 4.2.2 (R Development Core Team 2022). We first built generalized linear mixed models, following a Gaussian distribution with log link function, using the package ‘glmmTMB’ (Brooks et al. Reference Brooks, Kristensen, Benthem, van Magnusson, Berg, Nielsen, Skaug, Mächler and Bolker2017) to test the effect of forest stratum and season on the total arthropod density (i.e. number of individuals per m² of foliage) and the densities of three feeding guilds (herbivores, predators, and other arthropods) or the two most abundant arthropod orders. The models contained forest strata (as a continuous variable), season (factor of 3 levels), and their interaction as fixed predictors, with plant species (factor of 7 levels), and tree individual (factor of 66 levels) as random effects. Next, we used the most parsimonious model to predict the densities and their standard errors using the function predict from the package merTools (Knowles et al. Reference Knowles, Frederick, Whitworth and Knowles2016). Body sizes of all individual arthropods were analysed in a similar way, but the models included additional fixed factor guild (3 levels) and additional random factor branch (129 levels).
To analyse the impact of vertical strata and time on herbivory damage (proportion of leaf area lost per m² of foliage), we conducted generalized linear mixed models as we did for densities but with a beta error distribution for herbivory damage. In models for herbivory damage, we used ‘branch’ as a random factor (factor of 127 levels, data for 2 branches from which we had no collection of herbivory damage) in addition to all factors used in the models for densities.
Finally, we constructed a set of candidate models integrating leaf traits (SLA, LDMC, TD, and LT) as continuous variables. The full model further contained forest stratum (continuous variable), season (factor of 3 levels) as fixed predictors, plant species (factor of 7 levels), and tree individual (factor of 66 levels) as random effects. We then used the dredge function in the package MuMIn for stepwise selection. As above, we used a beta distribution for herbivory damage models and a Gaussian with a log link function for the density of arthropods.
Results
Herbivory damage along vertical gradient during the wet and dry seasons
In total, we surveyed herbivory damage caused by chewing arthropods on 9,425 leaves from 127 sampled branches. This represents 1,135 data points instead of 1,143 (127 branches × 3 surveys × 3 twigs), as we were unable to resample several branches in some months. Mean herbivory damage (across three twigs per branch) was significantly affected by forest strata, by the month of collection (season), and by their interaction (Table 2).
Table 2. Comparisons of generalized linear mixed models testing for the effects of stratum and time (and their interactions) on chewing herbivory damage. Results of the analysis of deviance are based on the delta AICc (corrected Akaike Information Criterion). The most parsimonious models are underlined and indicated in bold font. Branch, tree, and plant species were used as random factors in all models
In all surveys, the herbivory damage was highest in the understory and decreased non-linearly towards the canopy (Figure 2). As we measured the herbivory on the leaves remaining on the branches between June 2021 and March 2022, leaf damage increased over time. In June 2021, at the beginning of the dry season, the mean standing herbivory (i.e. the cumulative leaf damage recorded at a given point in time) was 5.28 ± 0.07% (SE) across all strata (Figure 2). During the dry season, from June 2021 to November 2021, herbivory increased to 7.13 ± 0.09% (SE). Then, during the wet season, from November 2021 to March 2022, it increased to 11.18 ± 1.41% (SE) (Figure 2).
Figure 2. Herbivore damage (% per m2 of leaf area) across vertical stratification during three sampling seasons distinguished by colours. Each coloured dot represents the damage on one twig in each season, within one of the 127 sampled branches. Fitted lines represent the fit from the best model from Table 2, and dashed lines represent 95% confidence intervals of the predicted model.
Effect of forest strata and season on density and body size of arthropods
Overall, we collected 3,445 arthropod individuals from 407.07 m² of foliage. Arthropod density was affected by stratum, season, and their interaction (Table 3, Figure 3). At the beginning of the dry season in June/July 2021 (Figure 3), arthropod densities followed a C shape and were higher in the canopy and in understory and lowest at mid-heights. At the beginning of the wet season in November 2021, the arthropod densities increased significantly in the canopy but remained unchanged in the forest understory (Figure 3). At the end of the wet season, in March 2022, the densities in the canopy returned to the state at the beginning of the dry season, and the understory now had slightly increased abundances.
Table 3. Comparisons of generalized linear mixed models testing for effects of stratum and time and their interactions on densities of all arthropods, chewing herbivore arthropods, and predatory arthropods, spiders, and beetles. Results of the analysis of deviance based on the delta AICc (corrected Akaike Information Criterion). The most parsimonious models are indicated in boldface and underlined. Tree and plant species were used as a random factor in all models
Figure 3. Each coloured dot represents arthropod density in one branch in a given season. Fitted lines represent the fit from the best model from Table 3, and dashed lines represent 95% confidence intervals of the predicted model.
Most of the arthropods we collected were adult Araneae (1,320), Hymenoptera (i.e. ants; 1,138), Blattodea (221), Coleoptera (210), Hemiptera (171), Orthoptera (155), and larval Lepidoptera (100). All other orders were rare, with fewer than 38 individuals. Based on the initial identifications to orders and other taxa (e.g. spiders, ants), most of the collected arthropods were identified as predators (2,705 individuals, 11.67 ± SD 0.89 individuals per m2 of foliage). On the other hand, we found only 595 chewing herbivores (2.44 ± SD 0.24 individuals per m2 of foliage). While the density of the herbivorous arthropods was affected only by strata and the season, predatory arthropods were affected by strata and the season as well as by the interaction between these two factors (Figure 4, Table 3). Thus, the trends described above for all arthropods seemed to be driven by the predatory arthropods (which were overall more abundant than chewing herbivores) but were different for herbivorous arthropods (Figure 4). Coleoptera and Araneae were the two most abundant orders representing chewing herbivores and predators, respectively, and they both were affected by time of the collection, stratum, and interaction between these factors (Table 3, Figure S1).
Figure 4. Densities of chewing herbivore arthropods and predatory arthropods (i.e. the number of all arthropods longer than 1mm per 1m2 of foliage) across all strata throughout the experiment. Fitted lines represent the fit from the best model from Table 3, and dashed lines represent 95% confidence intervals of the predicted model.
We treated the plant species as a random factor in our models because they naturally differed in the densities of the arthropods they hosted. For example, Donella lenceolata hosted markedly higher densities of arthropods (29.27 ± SD 2.14) than any other plant species, and Pometia pinnata (5.92 ± SD 2.09) hosted the lowest densities (Table S1).
The body size of arthropods found on the host plants differed between strata, season, and the feeding guild (Table 4). The arthropods were generally larger in the canopy and smallest in the understory (Figure 5, Figure S2). Very small arthropods (ca. 1–2 mm) were missing in the upper canopy layers, while the largest arthropods (i.e. > 15 mm) were found only in the uppermost canopy layers (Figure 5). Overall, body sizes of herbivorous arthropods were more variable and bigger than predatory arthropods (Figure 5, Figure S2).
Table 4. Comparisons of multi-predictor models analysing factors (i.e. time, stratum, guild, and their interactions) affecting the body size of arthropods. Results of the analysis of deviance based on the delta AICc (corrected Akaike Information Criterion). The most parsimonious model is indicated in boldface and underlined
Figure 5. Relationship between the size (measured as body length) and forest stratum (height in m). Distribution of body sizes of arthropods identified as chewing herbivores (Herb) and predators (Pred) at the different forest strata. Arthropods without a trophic relationship (others) are not shown in the graph. The line indicates the median value for each group and stratum.
Variation in leaf traits along vertical forest stratification
Nearly all leaf traits measured and across all tree species were correlated with vertical forest gradient, although these correlations were weak. LDMC (R2 = 0.51) and LT (R2 = 0.08) increased with higher forest strata, whereas SLA (R2 = 0.61) decreased as height increased (Figure 6). Stepwise selection-built models for arthropod densities considered forest stratum (P < 0.001, Est. = 6.17, Z = 7.164) to be the best predictor of arthropod density together with LT (P < 0.001, Est. = −0.410, Z = −3.483), SLA (P < 0.001, Est. = −0.543, Z = −3.495), and LDMC (P = 0.007, Est. = −0.329, Z = −2.711). Stepwise selection-built model for herbivory damage considered forest strata (P < 0.001, Est. = −0.214, Z = -3.322), SLA (P < 0.001, Est. = 0.024, Z = 3.711), and LDMC (P = 0.023, Est. = 0.013, Z = 2.275).
Figure 6. Measurement of leaf traits, namely, specific leaf area (SLA), leaf dry matter content (LDMC), and leaf toughness (LT) along the vertical forest gradient. SLA decreased with the increasing height of the forest, while LT and LDMC increased with the height.
Discussion
Insect herbivory across forest stratification during the wet and dry seasons
In line with our first hypothesis, herbivory damage decreased with increasing forest height. During the experiment (June 2021–March 2022), herbivore damage accumulated more in the lower strata compared to the higher strata and doubled by the end of the rainy season. Additionally, herbivory occurred mostly during the rainy season. These results are consistent with some, but not all, studies from other tropical forests. For example, a study in the semi-deciduous tropical forest of Barro Colorado Island in Panama found higher herbivory in the forest understory compared to the canopy (Barone Reference Barone2000). Similarly, Frost (Reference Frost2022) observed a comparable pattern, particularly the damage caused by chewing herbivores, in a Neotropical forest at the Amazon Conservatory of Tropical Studies Field Station and Canopy Walkway near the Sucusari River in Peru. Further, Lowman (Reference Lowman1985) reported that leaves close to the ground level (0–3 m) were more heavily grazed by insect herbivores than leaves farther from the ground in Australian subtropical rainforests. This trend was also noted in temperate regions (Stiegel et al. Reference Stiegel, Entling and Mantilla-Contreras2017, Thomas et al. Reference Thomas, Sztaba and Smith2010). In contrast, a study in the tropical dry forest of Mata Seca State Park, Brazil, found higher herbivory in the canopy than in the understory (Neves et al. Reference Neves, Silva, Espírito-Santo and Fernandes2014). Similarly, Basset (Reference Basset1991) assessed insect herbivory on the subtropical tree species Argyrodendron actinophyllum in a rainforest near Brisbane, Australia, and found that herbivory was higher in the middle vertical layer of tree crowns exposed to sunlight (15–25 m), mostly on the young leaves.
Thus, there are three possible explanations for higher herbivory damage in the understory compared to the canopy. First, understory leaves are typically larger, thinner, and softer, with higher chlorophyll content due to less sunlight, aiding in resource conservation (Salgado-Luarte and Gianoli Reference Salgado-Luarte and Gianoli2012). These traits make them more susceptible to herbivore damage because understory plants often prioritize growth over defence, resulting in leaves more palatable to insects (Barone Reference Barone2000, Ernest Reference Ernest1989, Dudt and Shure Reference Dudt and Shure1994). Second, the emergence of young leaves during the rainy season offers herbivores a chance to feed on nutritious, less-defended leaves, potentially leading to higher herbivory. Finally, the soil at our study site might affect the herbivory of the trees that grow in it. It has been shown that the vertical pattern of herbivory potentially changes with the soil type, with herbivory being higher in the canopy than in the understory on loam soils (Shao et al. Reference Shao, Zhang and Yang2021).
In contrast, forest canopies or higher strata are exposed to extreme conditions like increased temperature and lower humidity due to direct sunlight (Stiegel et al. Reference Stiegel, Entling and Mantilla-Contreras2017). These conditions result in increased concentrations of defensive compounds such as carbon and phenolics and lower nutrient levels, particularly nitrogen. Consequently, canopy leaves become smaller, tougher, thicker, and less nutritious, rendering them less palatable to herbivores (Massad et al. Reference Massad, Richards, Philbin, Yamaguchi, Kato, Jeffrey, Oliveira, Ochsenrider, de Moraes, Tepe, Cebrian-Torrejon, Sandivo and Dyer2022). However, other studies have reported higher herbivory in the canopy, suggesting that this could result from leaf flushing during the rainy season. This is likely due to the synchronization of insect activity with the availability of young, highly nutritious plant tissues (Basset Reference Basset1991; Neves et al. Reference Neves, Silva, Espírito-Santo and Fernandes2014), which nonetheless aligns with our explanation regarding the sprouting of young leaves during the rainy season.
Densities of arthropods and body sizes across vertical forest stratification during the wet and dry seasons
Our data partially supported the hypothesis that arthropod densities are highest in the uppermost stratum compared to the lowest. However, we found similar densities in both the understory and the highest stratum, with the uppermost and lowest strata having the highest densities and mid-strata the lowest at the beginning of the dry season (June 2021). Arthropod densities were low during the dry season but increased at the start of the rainy season (November 2021), peaking especially in the uppermost stratum and remaining similar in the lowest stratum as in June 2021, at the end of the rainy season (March 2022). Vertical patterns of predatory arthropod density were similar to trends for all arthropods. Their abundances significantly increased in the highest stratum at the start of the rainy season before returning to dry season levels by the end of the rainy season. Herbivore density followed a different distribution pattern, with the highest density observed in the uppermost stratum at the beginning of the rainy season. Arthropod body sizes were largest in the uppermost stratum and smallest near the ground at 1 m (Figure 5), with over 2,700 predators and only 595 herbivores collected. Both guilds had larger body sizes in the uppermost stratum, but herbivores were generally larger than predators. Seasonality and forest strata influenced both the density and body size of arthropod communities, though guild did not affect body sizes; arthropods belonging to different guilds had different body sizes.
The highest herbivory was observed in the lowest stratum. Surprisingly, this pattern did not correlate with the density of herbivorous arthropods, as the highest density was observed in the canopy. This indicates that although larger herbivorous arthropods were present in the higher strata, their overall abundance was lower. This suggests that larger herbivorous arthropods in the higher strata might be permanent residents rather than transient. Our result aligns with Basset et al. (Reference Basset, Aberlenc, Barrios, Curletti, Bérenger, Vesco, Causse, Haug, Hennion, Lesobre, Marquès and O’Meara2001), who studied arthropod diel activity in the tropical rainforest of Gabon. They found high herbivore turnover between day and night in the upper canopy, with no significant influx of herbivores from lower foliage at night, suggesting that upper canopy herbivores are likely resident and well-adapted to those environmental conditions. Alternatively, the higher density of arthropods in the uppermost stratum could be influenced by plant species (Grimbacher and Stork Reference Grimbacher and Stork2007). For instance, we observed the highest density of arthropods on D. lanceolata and the lowest on P. pinnata, suggesting that certain arthropod groups prefer specific plant species as hosts (Novotny et al. Reference Novotny, Miller, Baje, Balagawi, Basset, Cizek, Craft, Dem, Drew, Hulcr, Leps, Lewis, Pokon, Stewart, Allan Samuelson and Weiblen2010; Dyer et al. Reference Dyer, Singer, Lill, Stireman, Gentry, Marquis, Ricklefs, Greeney, Wagner, Morais, Diniz, Kursar and Coley2007). Barone (Reference Barone2000) found a similar pattern in two Panamanian tree species, Quararibea asterolepis and Alseis blackiana, concluding that the role of adult trees as herbivore sources for saplings depends on the importance of new leaves to the tree species’ herbivores, a factor varying between species. Additionally, Novotny et al. (Reference Novotny, Basset, Miller, Drozd and Cizek2002) demonstrated that most herbivores feed on several closely related congeneric plant species, potentially leading to niche partitioning of food resources.
Although herbivores were larger than predators in the uppermost stratum, their abundance was lower. Predators, particularly spiders (Araneae) and predatory ants (Hymenoptera), such as Oecophylla smaragdina and Crematogaster polita, were abundant in our samples and could have played a regulatory role in controlling herbivores in the higher strata (Zvereva et al. Reference Zvereva, Paolucci and Kozlov2020). Moreover, the larger body sizes of arthropods may be attributed to coping with competition between individuals and among different arthropod groups (Stork and Blackburn Reference Stork and Blackburn1993), responding to evolutionary pressures (Shine Reference Shine1989), and withstanding intense microclimatic conditions (Stiegel et al. Reference Stiegel, Entling and Mantilla-Contreras2017). For instance, arthropods with larger body sizes may have a competitive advantage over those with smaller body sizes, thereby outcompeting them for food resources (Basset and Angelis Reference Basset and Angelis2007) as well as mating opportunities (Alcock Reference Alcock2013). Moreover, being ectotherms, arthropods with larger body sizes may have a higher advantage over those with smaller body sizes in regulating their temperature (Laird-Hopkins et al. Reference Laird-Hopkins, Ashe-Jepson, Basset, Cobo, Eberhardt, Freiberga, Hellon, Hitchcock, Kleckova, Linke, Lamarre, McFarlane, Savage, Turner, Zamora, Sam and Bladon2023; Pincebourde et al. Reference Pincebourde, Dillon and Woods2021). Consequently, their ability to thermoregulate their body temperature enables them to occupy niches in the higher canopies more effectively than those with smaller body sizes.
As confirmed in this study, we observed that smaller arthropods occupied the lowest stratum, just 1 m above the ground. Smaller herbivores may prefer this niche because leaves in the higher strata tend to have lower nutrient content and greater toughness, which can negatively impact their high metabolism and poor digestion. Consequently, they may focus on young leaves, which are more abundant in the understory (Čížek Reference Čížek2005), especially during the rainy season when young leaves emerge (Coley and Barone Reference Coley and Barone1996). Another reason for our results might be methodological: the tiny arthropods might have disproportionately escaped detection in the canopy compared to the understory, due to the logistical difficulties of sampling the canopy from the crane gondola.
Effect of vertical forest stratification and temperature on leaf traits
Leaf physical traits are crucial for defending plants against herbivores by providing direct structural defences, making it harder for insects to consume leaves and affecting herbivore behaviour and preferences (War et al. Reference War, Paulraj, Ahmad, Buhroo, Hussain, Ignacimuthu and Sharma2012). We examined four physical traits, namely, LT, TD, LDMC, and SLA, as predictors of arthropod density and insect herbivory along a vertical forest gradient. All traits except for TD correlated with forest stratification. Trichomes were rarely observed in the focal plant species, and their density showed no pattern along the vertical forest gradient. However, the relationships between the other leaf traits mentioned above were highly correlated with vertical forest stratification and exhibited steep gradients. Forest stratum significantly influenced these leaf traits. The results of the stepwise selection (dredge function) of the models (Table S2) were used to explain the density of arthropods and herbivory damage along the vertical forest stratification. We allowed for non-linear patterns along vertical strata and used SLA, LT, and LDMC as explanatory factors, with individual trees and tree species as random factors. We showed that LDMC, SLA, and LT were the main predictors of arthropod density, while LDMC and SLA were the best predictors for herbivory damage. These findings align with previous studies highlighting the importance of forest strata in influencing leaf traits (Castagneyrol et al. Reference Castagneyrol, Giffard, Valdés-Correcher and Hampe2019; Kenzo et al. Reference Kenzo, Mohamad and Ichie2022; Stiegel et al. Reference Stiegel, Entling and Mantilla-Contreras2017).
Moreover, local microclimate could potentially influence these leaf traits. In our study, we observed that temperature increased with height and fluctuated between day and night, but humidity did not follow the expected trend, likely due to condensation affecting the data loggers’ measurements. However, this did not significantly affect arthropod density in the canopy, which agrees with the findings of Basset et al. (Reference Basset, Aberlenc, Barrios, Curletti, Bérenger, Vesco, Causse, Haug, Hennion, Lesobre, Marquès and O’Meara2001). In weakly seasonal tropical rainforests like those in Papua New Guinea, the effects of temperature and humidity are likely mediated by consistently wet conditions, high plant diversity (Cámara-Leret et al. Reference Cámara-Leret, Frodin, Adema, Anderson, Appelhans, Argent, Guerrero, Ashton, Baker, Barfod, Barrington, Borosova, Bramley, Briggs, Buerki, Cahen, Callmander, Cheek, Chen and van Welzen2020), and intense biotic interactions (Coley and Barone Reference Coley and Barone1996). These complex and interacting factors underscore the importance of studying microclimate–trait relationships in such systems, where traditional assumptions about abiotic drivers may not hold.
In conclusion, this study examines how vertical forest stratification, seasonality, physical leaf traits, and microclimate influence insect herbivory, arthropod densities, and their body sizes. Vertical stratification, seasonality, and temperature were direct determinants, while leaf defence traits had indirect effects on arthropod dynamics and insect herbivory in Papua New Guinea’s weakly seasonal tropical rainforest. Increased temperatures across the vertical gradient correlated with shifts in leaf traits. Specifically, the higher temperature in the higher strata correlated positively with increased LDMC and LT and negatively with SLA, potentially affecting herbivore behaviour and arthropod dynamics. By examining multiple vertical strata, we identified detailed patterns not apparent when comparing only the understory and canopy across wet and dry seasons. Our results confirm that insect herbivory decreases with increasing forest height and is higher in the lower strata and at the end of the rainy season and that larger arthropods are found in the upper strata, with higher densities in both the highest and lowest strata. We also noted that predatory arthropods might play important roles in regulating herbivore damage, especially in the higher strata. However, we did not measure the predation rate, which needs to be investigated for a definitive answer. Future research should combine chemical and physical leaf defence traits to assess herbivory along the vertical forest gradient over wet and dry seasons to reveal insect–plant interactions in weakly seasonal tropical rainforests, like those at our study site. Nonetheless, this study supports the hypothesis that insect herbivory, arthropod densities, and body sizes are influenced by vertical stratification, seasonality, and temperature, with indirect effects of leaf traits needing further exploration.
Our research demonstrates that the high canopy of primary forests supports dense populations of arthropods, particularly during the wet season. Disturbances to the canopy, often caused by selective logging in tropical regions, can negatively impact both the number of species and the individuals that are able to survive (Pangau-Adam et al. Reference Pangau-Adam, Slowik and Waltert2021; Willottf Reference Willottf1999). Additionally, we observed significant seasonal fluctuations in arthropod populations, which may be further influenced by the severity of El Niño oscillations or other climate changes.
The canopy crane, located in collaboration with the local landowner community, has provided financial support, job opportunities, and a chance for local involvement in sampling efforts. This initiative has also enhanced their understanding of key ecological processes, which in turn has fostered a stronger commitment to protecting the forest and maintaining it as a site for ongoing research, thereby helping to prevent further logging activities.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0266467425100163.
Acknowledgements
We are grateful to the New Guinea Binatang Research Center for providing logistics and accommodation for this research at its biological field station. We also extend our sincere thanks to the local field assistants of the Kau Wildlife Management Area for their support to H.M. during the experiment.
Financial support
This work was supported by the European Research Council – Starting Grant (BABE 805189) (awarded to K.S.), and the analysis and writing were funded by the Czech Science Foundation project 22-17593M (also awarded to K.S.).
Competing interests
The authors declare that they have no conflict of interest.
