CONSERVATION ASSESSMENT AND SPATIAL DISTRIBUTION OF ENDEMIC ORCHIDS IN SABAH, BORNEO

1Universiti Malaysia Sabah, Malaysia e-mail: *suzikajuiling@gmail.com, tsensandy@gmail.com, yewleung@gmail.com, bs14161036@alum.ums.edu.my, colinr@ums.edu.my 2Sabah Forestry Department, Malaysia e-mail: gmirus4@gmail.com, biodipterocarp@gmail.com, john.sugau@sabah.gov.my, rnilus70@gmail.com, joan.pereira18@gmail.com, Alviana.Damit@sabah.gov.my 3Sabah, Malaysia e-mail: oberonia@gmail.com 4Sabah Parks, Malaysia e-mail: skybyn@gmail.com, hanrymujih@gmail.com


Introduction
Loss of plant diversity is a major global concern (Jackson & Kennedy, 2009). Land use change is the major cause of plant diversity loss, especially in the wet tropics (Giam et al., 2010;Brummitt et al., 2015;Corlett, 2016). High rates of forest loss coupled with high levels of biodiversity can lead to the loss of many plant species (Giam et al., 2010). As a response to this extinction crisis, the global community has called for actions to halt the current and continuing loss of plant diversity (Jackson & Kennedy, 2009). Assessing which species are most at risk of extinction and identifying priority areas for plant protection are an integral part of this approach (Callmander et al., 2005).
The Orchidaceae is one of the largest families of flowering plants and contains ca. 30 000 species (Hassler & Rheinheimer, 2020). Their distinctive flowers make them popular among collectors and prone to overcollecting (Lamb, 1991a;Vermeulen & Lamb, 2011;Ballantyne & Pickering, 2012). Orchids are also sensitive to habitat loss and disturbance (Wraith & Pickering, 2018). As most orchids are epiphytic and their existence depends on the availability of suitable host plants, this makes them susceptible to forest loss (Lamb, 1991a;Hietz, 1999). Despite being among the most threatened plants, only 5% of the world's orchids are listed on the IUCN Red List of Threatened Species (Fay, 2018;IUCN, 2020).
The island of Borneo is one of the centres of orchid diversity (Hassler & Rheinheimer, 2020). The estimated number of orchid species in Borneo varied widely. Past estimates suggest that there were between 1400 and 3000 orchid species in Borneo (Lamb, 1991b;Wood & Cribb, 1994). However, the most recent figure suggests about 1700 species, of which more than 1000 species are endemic to the island (Hassler & Rheinheimer, https://dx.doi.org/10.24189/ncr.2020.053 2020). Within Borneo, the Malaysian State of Sabah has the richest orchid flora with an estimated 1300 species, of which approximately 250 species are endemic to Sabah (Sabah Forestry Department, unpublished data). This high species richness is partially related to the presence of Mount Kinabalu, the highest mountain in the Malay Archipelago. Mount Kinabalu is one of the global hotspots for plant diversity, and more than 860 orchid species have been recorded from the mountain (Ent, 2013).
Over the last forty years, Sabah has lost 39.5% of its forest cover (Gaveau et al., 2014). Much of the remaining forest has been selectively logged, with only 19.1% of the remaining forest considered intact (Gaveau et al., 2014). How these disturbances affected the plant diversity of Sabah is only just starting to be understood. Currently, only 11% of the estimated 8000 plant species in Sabah have a Red List assessment and only eight of ca. 250 orchid species endemic to Sabah are listed on the IUCN Red List (IUCN, 2020). This study aims to 1) assess the threat status of the endemic orchids of Sabah using the IUCN Red List Categories and Criteria (Version 3.1), and 2) develop species distribution models (SDMs) using Maximum Entropy (MaxEnt) to identify their potential distribution across Sabah. As Sabah is in the process of increasing its Totally Protected Area (TPA) to 30% of Sabah's area (Williams et al., 2020), these SDMs will be used to identify areas with a high species richness of endemic orchids that are currently outside of Sabah's Protected Area network.

Study site
Sabah is located in the northern part of Borneo (05.420278° N, 116.796667° E, Fig. 1). It has a total land area of 73 371 km 2 , accounting for approximately 10% of Borneo's total land area (Marsh & Greer, 1992). In Sabah, the altitude ranges from sea level up to 4095 m a.s.l., with Mount Kinabalu being the highest point. The average annual temperature varies between 32°C in the lowlands and 21°C in the highlands. The mean annual rainfall is between 2500 mm and 3500 mm, with the highest rainfall occurring in the mountainous areas of Mount Kinabalu and the Crocker Range. Fiftyone distinct soil associations have been described based on differences in soil texture, nutrient availability and drainage (Acres et al., 1975). These variations in climate, altitude and soils factors interact and determine the vegetation structure and diversity (Ashton & Hall, 1992;Chan et al., 1994), with nineteen distinct floristics associations recognised within the State of Sabah (Sabah Forestry Department, 2004). The lowland and upland areas with well drained and moderately fertile to fertile soils are dominated by mixed dipterocarp forests. Kerangas (heath) forests are found on areas with podzolic soils in the lowland, upland and montane regions. Freshwater swamp and peat swamp forests occur in the lowland on soils with an impeded drainage. Ultramafic forests are found on serpentine soils. These soils have high concentrations of iron and nickel and low concentrations of plant nutrients (Chan et al., 1994;Proctor, 2003). Lower montane forests occur on the moderate fertile soils at altitudes higher than 1000 m a.s.l.. This transition to upper montane forest occurs at ca. 2500 m a.s.l. It is replaced by a sub-alpine vegetation at altitudes higher than 3500 m a.s.l. on the upper reaches of Mount Kinabalu (Sabah Forestry Department, 2004). These different floristic associations give rise to a diverse orchid flora. Terrestrial orchids such as Calanthe R. Br., Claderia Hook. f., Arundina Blume, Bromheadia Lindl., and Phaius Lour. are mostly found in lowland mixed dipterocarp forest near streams and on exposed ridges. The dense canopy layers of the dipterocarp forest favours epiphytic orchids such as Phalaenopsis Blume, Bulbophyllum Thouars and Coelogyne Lindl. The orchid flora is less diverse on the nutrient-poor soil of the Kerangas and Ultramafic forests. However, most orchids found in these areas are endemic to these habitats (Chan et al., 1994). High endemism also occurs in the montane regions due to different climatic zones and a habitat that restrict orchid distribution (Wood et al., 1993).

Data compilation
To determine the conservation status and develop species distribution models for the endemic orchids of Sabah, we gathered data from the local herbaria (Sandakan Herbarium and Kinabalu Park Herbarium), living collections (Kipandi Butterfly Farm, Tenom Agricultural Park, Rainforest Discovery Centre and Poring Orchid Garden), the Global Biodiversity Information Facility (GBIF), and literature (Chan et al., 1994;Vermeulen, 1994;Wood, 1997Wood, , 2003Wood, , 2010Wood et al., 2011;O'Byrne, 2016). Species information including the locality data were extracted and stored in a database (Chapman & Wieczorek, 2006). Species occurrences without co-ordinates were georeferenced following the standard GBIF georeferencing protocols (Chapman & Wieczorek, 2006) using a combination of topographic maps, soil maps for Sabah (Acres et al., 1975), the Sabah gazetteer (Tangah & Wong, 1995), Mapcarta, Google Earth and Map of Mount Kinabalu (Beaman et al., 2003). The georeferenced datasets were used to produce a conservation assessment for the species, and a species distribution model was generated for species with more than five data points after data cleaning.

Conservation assessment
We used the IUCN Red List Categories and Criteria (Version 3.1) to assess the threat status for each species (IUCN Standards and Petitions Committee, 2019). The datasets were imported into the open source website Geospatial Conservation Assessment Tool (GeoCAT: http://geocat.kew.org) and an initial Area of Occupancy (AOO) and Extent of Occurrence (EOO) calculated for each species (Bachman et al., 2011). The EOOs were measured using a minimum convex polygon for species with at least three occurrences localities, and the AOOs were calculated using a 2 km × 2 km grid (cell area: 4 km 2 ) overlaid on each occurrence locality (IUCN Standards and Petitions Committee, 2019). To estimate habitat loss, we performed a reduction analysis by removing collection localities that no longer have a natural vegetation cover. The species were then assigned to a threat category based on this information after a group discussion by members of the Sabah Plant Red List working group. All members of this group have completed at least one IUCN Red List training course.

Species distribution modelling (SDM)
To reduce potential collection biases, we cleaned the datasets before modelling. Data cleaning involved removing points that were less than 2 km apart, except where the points occurred on different soil types, or were separated by more than 300 m in altitude (Frazier et al., 2004;Williams et al., 2020). For species with more than five collection localities after data cleaning, we modelled their distributions using MaxEnt (Version 3.4.1: https://www.cs.princeton.edu/~schapire/ maxent/). We used a mixture of bioclimatic, altitude and edaphic layers as predictor variables to generate the SDMs. The bioclimatic data was obtained from WorldClim (https://www.worldclim.org/), the altitude data from the STRM DEM (http://srtm.csi.cgiar.org) and the edaphic layers from the Soils of Sabah maps (Acres et al., 1975). All layers were standardised at 6 arc second resolution (Maycock et al., 2012).
We generated SDMs using the presence-only data and selected environmental variables. The layers included the annual mean temperature, annual precipitation, precipitation of the wettest month, and precipitation of driest month, soil associations and altitude. We ran MaxEnt using the cross validate function with ten replicate runs and 10 000 random background points used as pseudo-absence. A detailed description of the modelling approach can be found in Williams et al. (2020). Model accuracy was evaluated using the area under the receiver operating characteristic curve (AUC) generated from the test data. This is a commonly used metric and generally accepted for model evaluation (Merow et al., 2013). All SDMs with an AUC of more than 0.8 value were retained and converted from logistic format to a presence-absence model using the Maximum training sensitivity plus specificity logistic threshold. The SDMs were then stacked in DIVA GIS to create a heatmap to show areas with high species richness of endemic orchids.

Conservation status
A total of 136 species out of about 250 endemic orchid species in Sabah were assessed, of which about 83% are threatened ( Fig. 2; Electronic Supplement 1). The majority of species (67%) were assigned to the Vulnerable D2 category based on having an AOO of less than 20 km 2 or number of locations ≤ 5, and a potential threat that could drive them to a higher threat category. Twenty-two species were assigned to higher threat categories (Critically Endangered and Endangered) primarily due to a restricted distribution and a continuing decline in suitable habitat (Table 1; Electronic Supplement 1). Five species (Aerides sukauensis Shim, Appendicula clemensiorum J.J.Wood, Bulbophyllum cerebellum J.J.Verm., B. elachanthe J.J.Verm., and Zeuxine papillosa Carr) were assessed as Data Deficient due to a lack of recent collections and uncertainties associated with the original collection localities.  EN A2c Note: Criterion A2c refers to a past reduction in population size, estimated as reduction in a decline in area of occupancy (AOO), extent of occurrence (EOO) and/or habitat quality. Criteria B2 refers to the extent of the geographic range measured as the area of occupancy, with (a) a severely fragmented or small number of locations and (b) a continuing decline in either the (i) extent of occurrence; (ii) area of occupancy; (iii) area, extent and/or quality of habitat; or (iv) number of locations or subpopulations (IUCN, 2020).

Species distribution model (SDM)
We have produced species distribution models for 47 species. The AUCs for these models range from 0.714 to 0.998 (Electronic Supplement 2). The overlaid SDMs for the 45 models with an AUC higher than 0.8 shows that all species occur in the montane regions of western Sabah (Fig. 3). Areas with a high species richness of endemic orchids include Mount Kinabalu (45 species), Mount Tembuyuken (45 species), Mount Alab (40 species), and Mount Trusmadi (28 species). Other areas that contained a range of Sabah endemic orchids were the ultramafic areas extending from central Sabah eastward towards Mount Silam, the Tawau Hill Park area in the southeast of Sabah and the Ulu Sipitang area in the southwest of Sabah (Fig. 3). With the exception of the Ulu Sipitang region and the midaltitude areas around Kinabalu and Crocker Range Parks, these areas are gazetted as TPAs.

Discussion
Of the 136 species evaluated in this study, 113, i.e. 83%, are threatened (Fig. 2). Previous studies have reported that between 43.5 and 56.5% of orchids examined were threatened (Kull et al., 2016;Gale et al., 2018;Liu et al., 2020). These studies had either a broader geographical scope (Kull et al., 2016), examined the complete orchid flora of a region (Liu et al., 2020) or were a summary of global assessments (Gale et al., 2018). In contrast, we assessed species that were known to have narrow ranges and had a limited number of collections. As narrow range endemics tend to have higher threat levels and are more prone to extinction (Cribb et al., 2003;Swarts & Dixon, 2009), this probably explains the higher percentage of threatened species in our study. With further assessments of the more widely distributed endemics, we envisage that the percentage of species listed as threatened should decline. Despite having a higher overall percentage of threatened species, we found a lower percentage in the higher threat categories compared to the species currently listed on the IUCN Red List of Threatened Species (IUCN, 2020). Currently, of the ca. 1600 orchid species listed on the Red List, 34% are listed as Critically Endangered or Endangered (IUCN, 2020). In our study, only 16% were assessed as either Critically Endangered or Endangered (Fig. 2). These differences could result from the different approaches. The IUCN data is a summary of the threat status of all orchids currently listed on the Red List, of the 542 species currently listed as CR or EN, 127 species are from Madagascar (IUCN, 2020), a region that has lost 44% of its forest cover in the last 60 years (Vieilledent et al., 2018). In contrast, most of the orchids examined in our study are species that are restricted to the montane areas of Sabah. The majority of this habitat is gazetted as TPAs, and has experienced a minimal habitat loss (Gaveau et al., 2014). So, while illegal collecting for the orchid trade is a potential threat to some of the species examined in this study, the loss of habit is negligible for most of them. However, we have not assessed how distribution might change due to climate change. Previous studies in Sabah suggested that narrow-range endemic restricted to the upper montane and subalpine forests of Sabah are at higher risk from climate change, than more widespread species (Maycock et al., 2011). Further study is required to investigate the potential impacts of climate change on the distribution of endemic orchids (Wraith & Pickering, 2018).
The heat map produced from the stacked SDMs, shows that most of the areas with a high species richness fell within the existing TPAs (Fig. 3). All orchid species modelled occurred in the Kinabalu Park (Mount Kinabalu and Mount Tembuyuken) and the neighbouring Crocker Range Park (Mt. Alab). A few endemic orchid species were predicted to occur in the lowland areas. Previous studies also showed a higher species orchid richness at higher altitude, compared to lowland sites, with species richness then declining with the transition from upper montane to subalpine vegetation (Wood et al., 1993;Grytnes & Beaman, 2006;Acharya et al., 2011;Zhang et al., 2015). This higher species richness is due to the heterogeneous habitat with various physical barriers on mountains that leads to a higher degree of isolation and results in a higher species diversification (Zhang et al., 2015). However, not all studies demonstrated a higher species richness of endemic orchids at higher altitudes (Acharya et al., 2011), as other factors, such as rainfall, temperature, habitat, soil and vegetation, contribute to endemic orchid distribution (Acharya et al., 2011;Hsu et al., 2011).
One limitation of our approach is that most of the highly threatened species had too few collections to model. Of the 22 species assessed as Critically Endangered or Endangered (Table  1), only two species (Bulbophyllum alatum J.J.Verm. and B. apheles J.J.Verm.) are included in the heatmap (Fig. 3). Most of the species assessed as Critically Endangered or Endangered were found outside of the TPA network. Six of the highly threatened species ( (Fig. 3). The areas adjacent to the Kinabalu and Crocker Range Park boundaries are mainly owned by local communities, so there is limited potential to expand the park boundaries (Wood et al., 1993). In contrast, most of the Ulu Sipitang area is gazetted as a production forest, under the jurisdiction of the Sabah Forestry Department. Therefore, we recommend the regazettement of this area as a TPA to increase plant protection in Sabah.
Several lowland endemics orchids were also assessed as Critically Endangered or Endangered ( were recorded from the production forests around Kalabakan in southeastern Sabah. Dendrobium paitanense J.J.Wood is known from Sungai Paitan (River Paitan) in northeastern Sabah. These lowland areas are highly disturbed (Gaveau et al., 2014). And although some orchids are able to survive in disturbed habitats (Majit et al., 2014), it is uncertain how these species have been affected by these disturbances. Further field work is required to investigate whether these species are still extant in these areas.
In situ conservation is important to protect threatened species in their natural habitat. However, ex situ conservation is also appropriate for the Critically Endangered species, especially for those where there are limited or no in situ options available. We would like to suggest collaborations with any orchid garden in Sabah such as Kipandi Butterfly Farm, Poring Orchid Garden and Tenom Agriculture Park as well as local community for ex situ conservation, especially for threatened species. These orchids could potentially be reintroduced back into their natural habitats. Long-term monitoring of the endemic orchids and further collecting is required to gain a better understanding on population dynamics and threats, as well as to get living material for the development of ex situ conservation collections. While there are some on-going efforts in these areas, more extensive work is required to prevent species extinction.

Conclusions
In this study, we produced provisional IUCN Red List assessments for 136 species of orchids endemic to Sabah. Of these, 83% were assessed as being threatened. Our results highlight that the montane regions in western Sabah are a hotspot for diversity and conservation of endemic orchids. For most species, habitat loss was not the major threat as most occurred in the TPA network in Sabah. However, illegal collections for the horticulture trade and climate change are potential threats that need further monitoring.
While this is a desktop assessment, it is preferable to have a detailed assessment that include information on species habitat preferences, ecology, population structures and dy-namics, demography and threats (Kumar et al., 2001). These provisional assessments provide an overview of the threat status of the endemic orchids of Sabah. The information is being utilised by on-going conservation efforts in Sabah including the selection of new TPAs and developing species management plans to protect and conserve the most endangered endemic orchids in Sabah.

Supporting Information
The list of endemic orchids used in the study and assessed according to IUCN guidelines (Electronic Supplement 1: Endemic orchids of Sabah used for analysis in the present study), and MaxEnt AUC values for each of orchid species used in SDM modelling (Electronic Supplement 2: List of species used in SDM modelling with MaxEnt AUC value) may be found in the Supporting Information here.