Harungana madagascariensis (Hypericaceae) is a key phorophyte for native epiphytes and lianas during ecological restoration: case study on an oceanic island

Study sites

Mauritius is a tropical volcanic island located around 900 km east of Madagascar. Centred on 20°15′S and 57°35′E, it is part of the Mascarene Archipelago and falls within the Madagascar and Indian Ocean Islands biodiversity hotspot (Myers et al., 2000). The island is about 1,865 km², with a maximum elevation of 828 m. Mean annual rainfall ranges from 800 mm along the leeward coast to 4,000 mm in the central uplands, with an average annual temperature of 22 °C (Staub, Stevens & Waylen, 2014). Since human colonisation in 1638, extensive habitat transformation has reduced the island’s native vegetation to ∼82 km² or 4.4% of its original extent (Hammond et al., 2015), which nevertheless continues to support a high proportion of native species (Florens et al., 2012). This native vegetation, however, survives as highly fragmented patches (Florens, 2013) that are increasingly dominated by alien woody species, particularly in the understorey (Florens et al., 2016). Although started in the first half of the 20th century (Vaughan & Wiehe, 1941; Baider & Florens, 2025), attempts to restore native vegetation communities by controlling invasive alien plants were upscaled from around the mid-1980s (Strahm, 1993; Jones, 2008), and as of 2024, ∼700 ha of native vegetation (∼8.5% of all native remnants) is undergoing restoration in that way on the island (Republic of Mauritius, 2024).

Two of the forest areas undergoing ecological restoration for biodiversity conservation and known to host native epiphyte and liana communities (Fig. 1) were surveyed between August 2023 and August 2024. The first one, Mount Camizard, (20°19′51″to 20°20′00″S and 57°42′52″to 57°43′02″E, 250–320 m asl) is located in the island’s South-East within an area of Mountain Reserve inside the Bamboo Mountains forest block. The native forest at the site has been undergoing ecological restoration mainly through invasive alien plant control since 2005 and is close to the lower elevational range of H. madagascariensis on Mauritius. The second site, Brise-Fer, occurs close to the higher elevational range of H. madagascariensis in the South-West of Mauritius (20°22′10″to 20°22′30″S and 57°25′55″to 57°26′20″E, 560–600 m asl) within the Black River Gorges National Park. There, weeding of invasive alien plants has been done in different phases for conservation management since 1986. We selected the largest area, weeded in 1996, for our sampling. Mount Camizard and Brise Fer receive comparable annual rainfall of respectively 2.5 and 2.4 m, and no permanent water sources, but some storm streams, cross the study areas. Surveys in the Black River Gorges National Park and Mount Camizard were conducted with permission from The National Parks and Conservation Services of the Ministry of Agro-Industry, Food Security, Blue Economy and Fisheries, and Mr. Owen L. Griffiths and Mrs. Mary-Ann Griffiths respectively.

Data collection

Naturally growing (non-planted) trees of H. madagascariensis (52 at Mount Camizard and 21 at Brise Fer) were randomly sampled along with the closest individual of another native tree species to each of them that was of (1) similar trunk diameter (measured at 1.3 m above ground, along the stem) and (2) similar age (Figs. 2 and 3). The decision to focus on individual trees rather than specific host species was in part guided by previous studies demonstrating that epiphyte assemblages are often more strongly influenced by tree-level structural attributes than by host identity (Zotz & Vollrath, 2003; Flores-Palacios & García-Franco, 2006; Komada et al., 2022), and that host specificity among vascular epiphytes and lianas is generally weak or absent (Garrido-Perez & Burnham, 2010; Wagner, Mendieta-Leiva & Zotz, 2015). The choice to include tree age in our analysis was guided by previous studies demonstrating the importance of including the dynamic of habitat-building trees in epiphyte studies (Merwin, Rentmeester & Nadkarni, 2003; Wagner & Zotz, 2020) and to ensure the validity of our comparisons between trees of different ecology and longevity. In Mount Camizard, each potential H. madagascariensis tree was selected using a coin-flip method, whereas at Brise Fer, trees were chosen from individuals previously recorded in a long-term monitoring plot of one hectare using random number attribution.

Trees of similar age were chosen from within the ten most important species of each site (Florens et al., 2012) to ensure their nearby occurrence, their most comparable ecological importance to H. madagascariensis which itself is relatively common, and to also reflect the most commonly planted species by conservation practitioners after they cut H. madagascariensis. In all, 219 woody plants were sampled, including 73 trees of each category, which served as the sampling units. For each of the 219 potential phorophytes, GPS coordinates were recorded (with GPSMap^® 65, Garmin), the species identified, and its diameter at breast height (DBH) recorded. Plant reproductive status and size were used as proxies of epiphytes and lianas’ fitness following Tremblay et al. (1998), Zhang, Nurvianto & Harrison (2010) and Wagner & Zotz (2020). The reproductive status of each epiphyte and liana was recorded as either vegetative or reproductive (bearing flower buds, open flower, unripe fruit, ripe fruit or showing traces of fallen fruits for angiosperms and fertile fronds for pteridophytes). The number of leaves or fronds of each epiphyte, and the stem diameter of each liana, were recorded to assess plant size.

Vascular epiphytes and lianas were counted on all sampled trees from the ground, with the aid of an 8 × 42 pair of binoculars when necessary, and identified using identification keys from the regional Flora (Bosser et al., 1976–2023) and other relevant work (Hermans et al., 2025) (illustrations of selected species are provided in supplementary material (Fig. S1)). Identifications were validated by the second author, CB, and third author, FBVF, who are experts in Mauritius plant taxonomy. Sampled trees ranged from 1.3 m to 15.0 m in height, with H. madagascariensis reaching on average similar heights than other native trees of similar diameter (8.98 m and 8.36 m respectively). Trees of similar age were on average 3 m high. For larger trees (>8 m height), we restricted our census up to the first section of the canopy (e.g., 1/3 of the branches length), equivalent to “Zone 3” (Johansson, 1974), to avoid observation bias due to the high probability of missing individuals higher up. Hemiepiphytes and hemiparasites were not considered due to their different ecology from epiphytes and lianas and also because of their rarity or absence in the study areas. We defined an individual epiphyte as an assemblage of rhizomes and leaves forming a clearly bounded stand (Sanford, 1968) due to the difficulty of delineating individual epiphytes when multiple shoots occur in close proximity. For species exhibiting a creeping growth form, individuals were considered separate if physically separated rhizome segments were growing on distinct branches or if no visible connection was discernible between them. Finally, each clearly separated clump of epiphytic filmy ferns (Hymenophyllaceae) observed was defined as a single individual due to the impossibility to delimitate individuals otherwise in the field.

Estimation of tree age

We estimated tree age using the best long-term monitoring census data available for the island supplied by the Mauritius Herbarium and comprising ∼19,000 individual native woody plants belonging to ∼100 species from Brise Fer. It includes tree measurements from 2005, 2010 and 2022 (spanning a maximum of 17 years) collected in 3 ha permanent plots, which provided the most reliable results compared to the smaller datasets available (including from Mount Camizard) that were collected during a shorter time period and that would therefore include a larger margin of error. Furthermore, the similarity in climate and forest structure and composition between the two sites and the extremely low growth rate of the studied species (0.15 cm/year on average, see Table S1) limits the potential bias that the lack of long-term data from Mount Camizard could have on our results. We calculated the annual DBH increments of H. madagascariensis and of the ten most important species along which it grows in both study sites using two segments of stem diameter monitoring (from 2005–2010 and from 2010–2023). The ten most important species were determined following Importance Values from Florens et al. (2012). Changes in stem diameters were used to estimate the average annual growth rate of each species of interest. A growth rate ratio between H. madagascariensis and each of the other species was then computed to estimate the diameter of an individual of the ten other most important species that would be of similar age to the individual of H. madagascariensis being sampled (Table S1). Finally, we sampled the nearest similar-aged plant to each H. madagascariensis studied. This indirect method was chosen as the census data used were the best available data for tree age estimation, and more reliable methods such as tree coring could not be performed due to the risk of damage (Florens, 2014; Tsen, Sitzia & Webber, 2016) which would not have been justifiable in our context.

Data analysis

Only native species were used in the data analysis, because introduced species encountered represented only individuals that were missed during invasive alien plant control campaigns and are therefore not characteristic of the study sites, and transient in nature until removed at a future weeding campaign. In all, 81% (1,805 of 2,229 individuals) of all epiphytes and lianas observed on the 219 sampled potential phorophytes were identified to species level directly in situ or at the National Herbarium of Mauritius based on photographs taken in-situ. Beside these, Hymenophyllaceae were treated as a single group due to their small size and the difficulty to identify them to species level. Observations were grouped under morphospecies groups for species that were indistinguishable from each other either because of the lack of distinctive characters on immature individuals or due to the absence of visible distinctive characters. This was the case for Angraecum calceolus, Angraecum caulescens and Angraecum multiflorum which have been grouped under the “Angraecum spp. Group” (n = 383); Bulbophyllum spp. grouped under the “Bulbophyllum spp. Group” (n = 4); Haplopteris spp. in “Haplopteris spp. Group” (n = 3), Polystachia mauritiana and Polystachia virescens grouped in “Polystachia mauritiana s.l. Group” (n = 119), and Selaginella spp. in “Selaginella spp. Group” (n = 19). Four observations of ferns could not be associated to any genus and those were excluded from the data analysis, on the basis that they could have been immature alien species. The final dataset used for analysis comprised the respective taxa classified as morphospecies, with 2,229 records of epiphytes and lianas collected from 219 sampled phorophytes.

We used R version 4.4.1 (R Core Team, 2024) to do all statistical analysis and graphs. Hill numbers were calculated per tree across phorophyte category and site (orders q = 0, 1, 2) using the R package ‘HillR’ (Li, 2018) to capture species richness, effective number of common species (Shannon), and dominant species (Simpson) (Hill, 1973; Chao, Chiu & Jost, 2014). Structural variables such as epiphyte abundance (N, epiphytes tree⁻¹), and species richness were also computed. We assessed differences in epiphyte species richness and abundance among tree categories (defined by diameter at breast height) using generalized linear models (GLMs), with tree category as the main predictor, and site as a covariate. Counts of epiphytes were explored for overdispersion and excess zeros using residual diagnostics using ‘DHARMa’ package (Hartig, 2024). Species richness was analysed using Quasi-Poisson GLMs, while abundance was modelled using zero-inflated negative binomial (ZINB) GLMs to account for overdispersion and excess zeros. Model fit was evaluated using pseudo R² (1 − residual deviance/null deviance) (Zuur et al., 2009). Post hoc pairwise comparisons between tree categories were performed using estimated marginal means (EMMs) with Tukey adjustment via the ‘emmeans’ package (Lenth, 2025). To evaluate differences in phorophyte suitability for epiphytes, total abundance and total species richness of all epiphyte and liana individuals per tree were used as response variables. Graphs were produced using the R package ‘ggplot2’ (Wickham, 2016).

Description of the epiphyte and liana communities in each tree category was made through its species composition and the relative importance value of all species. To this end, we carried out an indicator species analysis using the multipatt function in the indicspecies package (version 1.7.15) (De Cáceres, Legendre & Moretti, 2010) to identify species that are good indicators for one or several tree categories (Dufrêne & Legendre, 1997). In addition, we chose the Orchidaceae, the most abundant epiphyte family recorded (1,181 individuals), to compare the abundance of reproductive individuals (N, reproductive orchids tree⁻¹) growing on H. madagascariensis and other phorophytes of similar age and size. The most abundant orchid species, namely Angraecum pectinatum (523 individuals), and the most abundant species of liana recorded, namely Piper borbonense (226 individuals) were used to compare the sizes of epiphytes and lianas growing on the different phorophytes. Extreme values of reproducing orchid abundance and plant size were removed prior to analysis to reduce the influence of extreme outliers. The distribution of reproductive status and plant size data was assessed using the Shapiro–Wilk test. As the data deviated from normality, non-parametric tests were applied in subsequent analyses. Reproductive orchid abundance and the size of epiphyte/liana were compared between the different tree categories using Kruskal–Wallis rank sum tests with the post-hoc Dunn’s tests of multiple comparisons with Bonferroni adjustment using the R packages ‘rcompanion’ and ‘dunn.test’ (Mangiafico, 2024; Dinno, 2024). Graphs were produced using the R package ‘ggplot2’ (Wickham, 2016).

Ecological implications

The native pioneer tree H. madagascariensis which grows naturally best in disturbed areas, precisely where other potential phorophytes are rare, constitutes a better phorophyte compared to other potential native phorophytes that grow alongside it in the Mauritian native forests. Furthermore, the epiphyte and liana communities that H. madagascariensis can support in just two to three decades of its growth, are comparable to those assembling on often much-slower growing and much older, often multi-centennial trees of comparable size to H. madagascariensis. This further stresses the importance of H. madagascariensis for epiphytes and lianas establishment following a disturbance, reinforcing previous findings showing that tree age and size strongly influence phorophytic function in various ways that depend on tree species’ ecology (pioneer versus later successional species) (Catling & Lefkovitch, 1989; Wolf, 1994; Annaselvam & Parthasarathy, 2001; Bernal, Valverde & Hernández-Rosas, 2005; José Válka Alves, Kolbek & Becker, 2008).

Moreover, we showed that, within two to three decades of a disturbance, the fitness of native epiphytes (as assessed using commonly used proxies of fitness) can be substantially superior for those establishing on H. madagascariensis compared to other potential phorophytes close by. This double superiority as phorophyte (hosting greater diversity of epiphytes and fostering their higher fitness) is apparent even when compared to much older other species of the same trunk diameter as the H. madagascariensis. This situation appears linked to the fact that the bark of H. madagascariensis is relatively thick and spongy compared to most other native phorophytes. Such a bark retains moisture for longer periods and probably provide more nutrients, thereby promoting epiphyte establishment and their faster growth and maturation compared to most other Mauritian native trees. Hence, H. madagascariensis can not only quickly provide large surface areas suitable for epiphyte establishment and maintenance, but also offer a suitable habitat for their relatively rapid growth and earlier maturation. Those results corroborate previous studies showing that pioneer trees can be suitable phorophytes for epiphytes (Callaway et al., 2002; Cascante Marín et al., 2008; Einzmann et al., 2015; Besi et al., 2023; Pie et al., 2023; Wysocki et al., 2024), and lianas (Putz, 1984; Letcher, 2015; Schnitzer, 2015). However, it is important to note that site conditions modulate the benefits that pioneer trees like H. madagascariensis can bring, with greater positive impacts on boosting epiphytes and lianas in sites where greater species richness and abundance of epiphytes and lianas are found (e.g., Brise Fer compared to Mount Camizard).

Finally, the indicator analysis identified 21.4% of the epiphyte and liana species recorded in this study (N = 6) as significantly associated with H. madagascariensis, compared with only 10.7% (N = 3) with trees of the same size. These proportions are consistent with previous studies reporting that only a minority (15–30%) of epiphyte species show non-random host associations (Laube & Zotz, 2006; Wagner, Mendieta-Leiva & Zotz, 2015; Wagner, Wanek & Zotz, 2021), but further stress the important role that H. madagascariensis plays in supporting specific native epiphyte and liana species compared to other native trees that grow in conditions closest to those where H. madagascariensis grows. Therefore, H. madagascariensis trees provide a suitable habitat for epiphytes and lianas relatively early following a disturbance. Importantly, the patterns observed in Mauritius are likely applicable across the extensive native range of H. madagascariensis (over 12 million km²; Baguette, Baider & Florens, 2025), wherever the species occurs within the natural distribution of epiphytic orchids, ferns, and lianas, given the broad similarity in the ecological niches and requirements of these plant guilds.

Implications for ecological restoration and biodiversity conservation

The extreme invasion of Mauritius native forests by alien plants (Florens et al., 2016) has driven a high rate of native tree mortality (Florens et al., 2017) including some of the largest canopy species (Baider & Florens, 2006). As a result, when alien plants are removed, scanty native trees often remain to foster ecological restoration within the substantial gaps created in the forest canopy. These gaps form ideal habitat for H. madagascariensis which grows naturally from the seedbank to recreate a canopy reaching ∼12 m high within four to six years (Swaine & Hall, 1983; Ndam & Healey, 2001; Manjaribe et al., 2013) before starting to decline after ∼10 years (Hervé et al., 2015). Our results show that, where it grows, this pioneer tree does not suppress native epiphytes and lianas but on the contrary, is highly beneficial to native epiphytes species richness, abundance and fitness, more so than other species that grow alongside it. Yet, all major conservation practitioners of Mauritius have been, at one point or another, cutting back large numbers of H. madagascariensis in areas undergoing ecological restoration for biodiversity conservation (Fig. 8), contra best available evidence of its benefits for ecological restoration (Florens & Baider, 2013; Baguette, Baider & Florens, 2025), and without evidence of its presumed negative impact on native biodiversity. Here, we show that in addition to such practices hindering restoration progress of woody plant cover and fostering re-invasion by invasive alien plants (Florens & Baider, 2013), it also slows broader ecosystem recovery by: (1) removing native structural epiphytes already established on H. madagascariensis; and (2) reducing future recruitment of structural epiphytes, as alternative phorophytes generally provide lower-quality habitat than H. madagascariensis.

Among the species significantly associated with H. madagascariensis, four are orchids. Orchids constitute a major group of the island flora as it is the island’s most species-rich family of flowering plants, and is dominated by species endemic to the biodiversity hotspot region (80%), followed by species endemic to the Mascarene archipelago (41%), including those endemic to the island (10%) (Baider & Florens, 2022). The Orchidaceae is also the native angiosperm family that has sustained the highest extinction rate on Mauritius, with 22% of Mauritian native orchids driven extinct over the last 2.5 centuries or so (Baider & Florens, 2022) and many species are now extremely rare (e.g., Baider et al., 2012; Pailler et al., 2020a). Furthermore, known species have been found for the first time on the island relatively recently (Roberts et al., 2004) and new species are still being discovered even more recently (Fournel, Micheneau & Baider, 2015; Pailler & Baider, 2020; Pailler et al., 2020b). For all these reasons, conservation of native orchids in Mauritius should be a priority and our results show that H. madagascariensis can greatly help to enhance their conservation in wet forests by providing advantageous habitats for their colonisation and maturation. It is thus particularly unfortunate that most conservation managers cut back H. madagascariensis from areas undergoing restoration. Importantly, H. madagascariensis germinates and grows naturally in wet forests undergoing restoration such that no additional investment after invasive plants weeding is required from conservation managers for its establishment.

Finally, it is important to stress that epiphyte support diverse ecological interactions with animals (Nadkarni & Matelson, 1989; Stuntz et al., 2002; Boechat, da Silva & Nunes-Freitas, 2019; Spicer & Woods, 2022), as lianas also do (Yanoviak, 2015; Odell, Stork & Kitching, 2019). In particular, Piper borbonense, the most abundant native liana growing on H. madagascariensis, produces many fleshy fruits eaten by and benefiting native vertebrates (Heinen et al., 2023) including the threatened endemic Mauritius Bulbul (Hypsipetes olivaceus), also recently found to be a key pollinator of a fast declining endemic plant (Bissessur et al., 2019), thereby stressing on the judiciousness of considering broader ecological interconnectivity when devising ecological restoration policy. Mauritius is the only place within H. madagascariensis’ ∼12 million km² native range where conservationists cut the tree (Baguette, Baider & Florens, 2025), based on justifications contradicting best available evidence, including the unsubstantiated claim that it harms native biodiversity. Here, we show the opposite to be true regarding the neglected and threatened guilds of epiphytes and lianas. We hope that our findings may help practitioners to reallocate scarce conservation resources away from management that harm native biodiversity, and provide them additional evidence of the benefits that pioneer species can bring for ecological restoration.

Policy implications

Mauritius is globally well-known for some resounding conservation successes notably of birds (Sodhi et al., 2011) that were enabled by painstaking evidence-based approaches (Jones & Swinnerton, 1997; Swinnerton et al., 2004; Jones, 2008). More recently, however, some management of threatened species by conservation practitioners were applied that contradicted best available evidence (Florens, 2015; Florens, 2016), failed to reach the intended objectives and instead worsened the situation for threatened native biodiversity (Florens & Baider, 2019). A similar situation also emerged concerning ecological restoration for biodiversity conservation of the island’s most diverse native ecosystems whereby a policy of cutting back native plants was implemented. Indeed, following control of invasive alien plants, many native plants like woody lianas, strangler figs (FBV Florens, pers. obs., 2005 to present) and native pioneer trees like H. madagascariensis (Florens & Baider, 2013) that regenerated, were removed.

Mauritius was recently found to be the only place in the 12 or so million km² of native range of H. madagascariensis where such a policy is applied (Baguette, Baider & Florens, 2025). By cutting back H. madagascariensis, the policy, at a cost, sets back restoration progress of the native woody plant community and increases the re-invasion by alien plants, hence the costs of maintenance weeding (Florens & Baider, 2013). While this policy has been supported and/or implemented by government conservation services and most of the main local conservation NGOs for over three decades, its proponents have in that time not yet been able to present evidence that it somehow benefits native or threatened biodiversity. In this context, we now present novel evidence that the policy of conservation managers cutting back native pioneer trees from areas undergoing ecological restoration, is in fact harmful to the native biodiversity of epiphytes and lianas in addition to the native trees themselves in addition to diverting scarce conservation resources elsewhere from where it is proven to be impactful, such as the control of invasive alien plants (Monty Florens & Baider, 2013; Krivek et al., 2020). We therefore recommend that the enduring practice of cutting or lopping branches of H. madagascariensis be stopped and be replaced instead by planting of the species where it can grow but where seed bank or natural germination is lacking. Some encouraging signs have started appearing with the first conservation organisation shifting its policy in 2025 from cutting to planting the species (Ferney Ltd., 2025).