Introduction
Species distribution models (SDMs), as a capital method in biogeography research, are widely used in the response of target species under global climate change, the potential predicted distribution of invasive species, and the identification of conservation priorities (Araújo & Guisan, 2006; Austin & Van Niel, 2011; Franklin, 2013; Guisan et al., 2013; Guo et al., 2020). Known as ecological niche models, it’s a mathematical model established by species occurrence data and environmental information, estimating the ecological niche requirements of species based on statistical information provided by sampling sites and projecting to specific spatial and temporal regions to reflect the degree of habitat preference of species in a probabilistic form and the model results reflect the suitable habitats in the geospatial distribution (Franklin, 2013; Guillera-Arroita et al., 2015; Guisan & Thuiller, 2005; Guo et al., 2020; Naimi et al., 2014; Phillips et al., 2006).
The correspondence of species locality and environmental information is the crucial factor affecting the model performance (Abrahms et al., 2019; El-Gabbas A et al., 2021; McCune & Baraloto, 2016; Ranc et al., 2017). Since orchids could be considered flagship taxa for biodiversity conservation due to their advanced evolution status, their biogeography studies via SDMs are instrumental in identifying conservation priorities and analyzing biogeographic patterns in biodiversity hotspots (Crain & Fernandez, 2020; Luo et al., 2003). However, in the orchid SDMs, researchers always use all occurrence data as sample inputs to the model (Crain & Fernandez, 2020; Djordjevic et al., 2020; Djordjevic et al., 2016; Faruk et al., 2021; Tsiftsis & Tsiripidis, 2020; Wan et al., 2014). Orchids have broad ecological fitness and are dependent more on the microenvironment (Kelly et al., 2013; Souza Rocha & Luiz Waechter, 2010). While biogeographic studies based on SDMs are usually conducted at a large spatial scale, such as a global biodiversity hotspot, signifying the target space is likely to contain enough heterogeneous environments to provide enormous ecological space for a variety of orchids. Meanwhile, physiological studies of orchids have shown distinct ecological requirements between various orchids (McCormick & Jacquemyn, 2014; Zhang et al., 2018). The most recognized is that physiological differences exist in different orchids’ lifeforms. Hence, from a statistical point of view, the rough sample input may blur the mathematical relationships corresponding to the occurrence data and environmental information established by the model. Causing a bias would increase the model uncertainty and then affect the model accuracy and suitability maps.
Human activity is an issue of considerable concern in the biogeographic research of orchids, often regarded as one of the threats limiting their geographic distribution (Anibaba et al., 2022; Crain & Fernandez, 2020; Djordjevic et al., 2020; Guisan & Thuiller, 2005; McCune & Baraloto, 2016; Pilar A. Hernandez, 2006). However, there is no exact method to quantify or assess the impact of human activities on orchid distribution in current research. This shortcoming may result in biases between the predicted potential habitat generated by the lack of anthropogenic dispersal constraints in predicting species distributions and the potential geographic distributions (Franklin, 2023). In orchid distribution pattern studies in Central America, model results show that most orchid hotspots occur in the most densely populated provinces (Crain & Fernandez, 2020). Although they can indicate the threat level to orchids outside protected areas, the absence of verification of ground truthing still does not rule out the possibility of prediction uncertainty in models only under natural predictors (Eyre et al., 2022).
To further validate and confirm these issues, we used the case of the Hengduan Mountains. This region is one of the global biodiversity hotspots with a prominent representation of orchids amongst its flora. Different modeling approaches and validation methods were employed to explore the role of physiological characteristics and human activities in orchid species distribution models (SDMs). The following questions were addressed: 1. How to physiological characteristics and human activities impact orchid SDMs? 2. How do these factors affect orchid suitability prediction maps? 3. What are the orchid geographic distribution patterns and critical locations in the Hengduan Mountains based on different modeling strategies? These studies provide valuable insight into the geographic distribution patterns of orchids in the Hengduan Mountains and aid in assessing protected areas. Furthermore, the results can inform the modeling process for other species or regions.