Endemic species, one of the qualitative parameters for biodiversity, are one of the key determinants for prioritising conservation program in a region (Bonn et al. 2002; Burlakova et al. 2011). Management of any endemic species requires an in-depth knowledge of its distribution, biogeographic history, ecology and environmental attributes. Ecological Niche Modeling (ENM) of rare, endangered and endemic members can be implicated in multiple aspects, ranging from prioritising conservation measures to understand the evolution-environmental interactions (Loiselle et al. 2003; Ortega-Huerta and Peterson 2004; Engler et al. 2011; Mercer et al. 2013; Yannic et al. 2014). However, the paucity of occurrence data, sampling bias and lack of coverage, ; absence of geographic co-ordinates in historical accounts of distribution (Araujo and Guisan 2006; Hernandez et al. 2008) are the persistent problems for ENM of rare and endemic organisms . Inadequate occurrence data and insufficient coverage of ecological niches often impedes the development of robust models, sensitivity analysis and statistical validation (Duputie et al. 2013; Platts et al. 2014). Although combining various algorithms and data specific validation procedures (eg. Jack-knife leave one out) aid in minimizing the uncertainties, but the need for sufficient occurrence data continues to be a vital component in ENM.
Pertinently, ensemble techniques tend to outperform single algorithm due to multiple trials with different algorithms as well as considerable reduction in modeling uncertainty (Marimon et al. 2009; Buisson et al. 2010). Previously, ensemble modeling on endemics has been applied at macroscale (Thuiller et al. 2006; Li et al. 2013) as well as in species based studies (Lomba et al. 2010; Porfirio et al. 2014; Sousa-Silva et al. 2014) but attempts towards its optimisation for rare/endemic members is scanty till date.
Apart from ensemble modeling approach, absence or pseudo-absence information has significant role in correlative model building strategy. Although presence-only methods (e.g. BIOCLIM, Mahalanobis distance and others) are in use, but are outnumbered in recent times by presence-absence techniques. The absence data helps in the development of discriminative rules which can help to rank habitat suitability based on presence or absence information, thus, restricting over-prediction to some extent. Absence data can be contingent (environmentally suitable but biologically unsuitable), environmental (climatically unsuitable), and methodological (sampling error) depending on the study organism (Lobo et al. 2010). In essence, the collection of reliable absence data is a time and resources demanding task in ecological research. Pseudo-absence (pa) strategy is an alternative to true-absence points that can select multiple points from background and consider them as absence points; however, this could influence considerably the output (Brotons et al. 2004; Lobo et al. 2010; Massin et al. 2012).
In this study, ensemble niche modeling twchnique is attempted with pseudo-absence points for Syzygium travancoricum Gamble. a critically endangered woody endemic tree in southern Western Ghats of India with the objectives to evaluate 1) the performance of ensemble modeling 2) importance of pseudo-absence points in modeling and 3) applicationsm in conservation status assessment.