Orchids are the most heavily traded plant group globally, putting pressure on wild populations in many source countries like the Philippines. Despite its rich orchid diversity, there remains a notable gap in understanding the factors driving orchid trade within the country. To address this knowledge gap and support orchid conservation efforts, we utilized a five-year orchid diversity dataset extensively collected through floristic field and village garden surveys in one of the largest key biodiversity areas in the southern Philippines. We employed a trait-based approach to investigate ecological drivers of local orchid collection within this source area. Our results show that around 36% of local orchid diversity have predicted collection risks of ≥50%. Notably, locally collected orchid species exhibited multiple, large, and conspicuously colored flowers that are found in low-elevation forests and higher up in forest stratum. Elevational distribution and flower size emerged as the strongest predictors, potentially influencing collection preferences. Our analysis of predicted collection risks underscores the vulnerability of both threatened and non-threatened orchid species to local collection pressures. Moreover, we highlight the practical utility of our trait-based approach in predicting risks and informing management strategies for local orchid conservation. This research marks a significant step towards identifying ecological drivers influencing orchid trade at its source, providing insights that can inform targeted conservation strategies across many key biodiversity areas for this highly diverse, charismatic, and threatened plant family.

Study area and sampling method

The Mt Busa Key Biodiversity Area (KBA 196) stands as one of the largest and partially protected KBAs in southern Mindanao, Philippines, spanning approximately 1,162 km². It encompasses the major mountains of Busa, Three Kings, Melibengoy, Malibato, and Tasaday, with Busa peak as its highest point, reaching 2064 masl (Figure 1). The KBA 196 has mixed advanced secondary and primary lowland evergreen, lower montane, and upper montane forests (Saavedra & Pitogo, 2021), which are source areas for locally collected orchids in the region.

From July 2019–July 2023, we conducted ~2,495 person-hours of floristic field surveys within KBA 196 forests, spanning elevations from 200 to 2064 masl and encompassing both wet and dry seasons (Table S1). These surveys aimed to inventory orchid species while accounting for temporal and spatial variations in species distribution. Over this five-year period, we employed a combination of 2-km transect surveys upslope along the primary trails of major mountains and opportunistic surveys in randomly chosen sampling points outside of transects (e.g., forest interior, forest patches, riparian areas, ravines; Saavedra & Pitogo, 2021) to ensure a good representative of local orchid diversity in KBA 196.

Five 2-km transects were extensively surveyed, strategically established on both the southern (Mt Busa) and northern (Mt Tebotu and Three Kings Mountain) slopes of the primary mountain range, with additional transects on different ridges of nearby, unconnected Mt Melibengoy to cover the diversity of forest types and habitats present (Figure 1; Table S1). A total of 33 transect surveys were conducted throughout the data gathering period, with varying frequencies of survey per site (Table S1). During each transect survey, 4–5 observers recorded all orchid species up to about 5 m from trail, stopping every 250 m for more thorough observations for about 30 minutes. Our surveys also included extensive opportunistic examinations from the ground up to about 4 m above the forest floor for terrestrial and epiphytic orchids, with the caveat that species distributed above the understorey layer may have been missed. Although previous studies demonstrate that this ground-based survey approach may underestimate epiphytic species richness (Flores-Palacios & García-Franco, 2001; Burns & Dawson, 2005), logistical constraints prevented us from employing climbing-based survey techniques. However, we controlled potential observer bias by meticulously examining epiphytes on all encountered fallen trees and branches. Our methodology thus illustrates the thoroughness of our field efforts, although it does not necessarily guarantee completeness of orchid diversity within the KBA.   

Home garden surveys were conducted in 14 small villages, situated 300–500 meters from nearby forests within or directly adjacent to KBA 196, to supplement our records from the wild (see Figure 1 and Table S1 for the number of villages per mountain surveyed). Given that numerous small villages within KBA 196 are accessible only by foot or specific modes of transportation, village selection prioritized logistical efficiency and security, while still ensuring that villages selected are not biased to certain locations. These villages typically consist of 20 to 40 clustered houses, with <10 outliers built over 1 km away from the main cluster, often near their farmland. Primarily inhabited by Tboli and Ubo indigenous communities, whose way of life is deeply intertwined with their natural surroundings, many maintain small home gardens of wild-sourced ornamental, medicinal, and food plants. Although our survey of home gardens per village focused on the main cluster of houses for logistical efficiency, any houses located outside of cluster were still surveyed if encountered. Each house within the cluster of all 14 villages was inspected once per village survey, with a total of 66 village surveys conducted between 2019 and 2023 (Table S1). Due to the opportunistic nature of our village surveys, we were unable to track the number of houses surveyed per visit for each village. Our survey entailed thorough inspections of home gardens, if ever present, at every house to search for flowering orchids. To ensure that the orchid species inventoried from home gardens originated from KBA 196, we verified this information with the houseowner and visually inspected each plant. Indications of recent removal from substrate, such as fresh algae, moss, and soil materials attached to roots, were taken as evidence.

All flowering plants were photographed using a digital camera equipped with a macro lens (Canon EOS 850D, 100 mm macro lens), while floral specimens of some species were collected to facilitate accurate identification. Flowering plants were identified at species level by using published literature and photographic guides (Cootes, 2001, 2011; Pelser et al., 2011 onwards), as well as the citizen science platform of Co’s Digital Flora of the Philippines (CDFP; Barcelona et al., 2013). All species underwent secondary identification by consulting with orchid taxonomists (see Acknowledgments). Taxonomic treatment generally follows Plants of the World Online (POWO, 2023), except that we recognized Dendrochilum, Chelonistele, and Pholidota at the generic rank, as it is by CDFP (Pelser et al., 2011 onwards).

2.2 Trait database

We selected species from the five-year dataset of orchid diversity in KBA 196 for which the required trait data were available. The trait database consists of 178 species from 65 genera, ecologically and morphologically representative of about 250 species recorded in KBA 196 over the five-year data gathering period (Table S2). The dataset includes seven morpho-ecological traits related to species’ accessibility (elevation and vertical distributions) and visibility (plant size, number of flowers, flower size, flower color diversity and conspicuousness).

2.2.1 Elevational and vertical distributions

We recorded the lowest and highest elevations at which each species was observed in KBA 196 using a Global Positioning System (Garmin 64s, USA). Our record was augmented with data from CDFP (Pelser et al., 2011 onwards) and published literature (Cootes 2001, 2011; Betanio & Buenavista, 2018) to cover documented elevational range of each species. We defined vertical distribution as the range from forest floor (0 meters) to forest canopy, assessing the height at which orchids were observed through visual estimation relative to the height of second author. This estimation method was also applied to orchids observed on fallen trees. Vertical placements per individual of a species may vary, but this estimate provided quantitative description of ground and epiphytic species. For species observed in houses but not in the wild, elevation data were obtained from available published literature and databases, while vertical placement was estimated by the collector. Elevational and vertical distributions used for each species were calculated as the average of their lowest and highest values.

2.2.2 Plant size, flower size and number

Given that orchid flowers exhibit bilateral symmetry, we utilized reported length of the longest petal/sepal as a proxy for flower size. Data on petal/sepal length, flower number, and plant size (sum of maximum reported lengths of leaves and pseudobulb/stem) were obtained from original descriptions and available literature (e.g., Cootes 2001, 2011). In some cases, we supplemented or augmented these data through direct measurements during our fieldwork: number of flowers in an inflorescence was directly counted, floral measurements for some orchid species were measured using digital caliper (Mitutoyo 500-196-30, Japan), and plant size was measured using a measuring tape. As flower number can vary among individuals of a species, we approximated values and utilized published data (Cootes 2001, 2011) to provide quantitative descriptions for single-flowered, few-flowered, and many-flowered species. In instances where data were lacking for a particular orchid species, we used values from closely related and comparable species.

2.2.3 Flower color

For objective assessment of flower color, we employed quantitative color pattern analyses on each of 178 orchid species to determine color pattern complexity (also referred to here as color diversity) and conspicuousness. Color data were extracted from a digital image of a single flower for each species taken in situ in KBA 196 using a macro lens (Canon EOS 850D, 100 mm macro lens), except for 18 species whose in situ photos were obtained from CDFP (Nickrent et al., 2006 onwards). We ensured that the image selected for each species showed most, if not all, colors of perianth and labellum. These images were imported into Adobe Photoshop to mask manually the background using transparency and saved as PNG files.

Color segmentation was conducted for each processed species image using recolorize package in R statistical software (Weller et al., 2022). This package can handle a wide range of species images captured under different lighting conditions, and it does not require specification of the number of color classes, thereby accounting for diversity of colors present in the species. The recolorize2() function in the package was used for each species image to generate color histograms in CIE Lab color space, which is a perceptually uniform color space for human vision, with three bins per color channel (3³= 27 possible color bins per image). Color bins with Euclidean distance of less than 20 were then clustered to produce the final color map for color pattern analyses.

Using the recolorize package, we ran adjacency and boundary strength analyses for resulting color maps of each species image to quantify color pattern complexity and conspicuousness. A single set of CIE Lab color space and HSL color distances that correspond to perceptual distances of human visual were used for all species images. Color pattern complexity was measured using Simpson color class diversity (S_c), while conspicuousness is estimated by the weighted mean of chromatic boundary strength (Cr m_ΔS). Higher S_c values indicate more complex color patterns (Endler, 2012), and higher Cr m_ΔS values suggest greater visibility and likelihood of detection (Endler et al., 2018).