Why tropical island endemics are acutely susceptible to global change

Abstract

Tropical islands are species foundries, formed either as a by-product of volcanism, when previously submerged seabed is thrust upwards by tectonics, or when a peninsula is isolated by rising sea level. After colonisation, the geographical isolation and niche vacancies provide the competitive impetus for an evolutionary radiation of distinct species-island endemics. Yet the very attributes which promote speciation in evolutionary time also leave island endemics highly vulnerable to recent and rapid impacts by modern people. Indeed, the majority of documented human-driven extinctions have been exacted upon island endemics. The causes include over-exploitation, invasive species brought by people and destruction of island’s naturally constrained habitats. Imminent threats include inundation by rising sea levels and other adaptive pressures related to anthropogenic global warming. We review recent work which underscores the susceptibility of island endemics to the drivers of global change, and suggest a methodological framework under which, we argue, the science and mitigation of island extinctions can be most productively advanced.

Appendices

Appendix 1: Modelling the coupled effects of pig exploitation and landscape change in Sulawesi

Babirusa (Babyrousa babyrussa) and wild pig (Sus celebensis) are endemic to the large island of Sulawesi. Concern has been raised about the threat of continued illegal trade of B. babyrussa and the high volume in trade of the unprotected, albeit more resilient species, S. celebensis (Milner-Gulland and Clayton 2002). Both pig species are hunted together for the same end market; consequently, B. babyrussa (the rarer species) are hunted to levels below that at which it is profitable to hunt it alone, threatening extirpation over much of its range (Clayton et al. 1997). B. babyrussa is also sensitive to other human threats (being a forest specialist with a ‘slow life history’), while S. celebenis is hardier (it readily utilises the modified habitat matrix and has a ‘faster life history’). Lattice type spatial models have been used to explore the influences of distances from the end market and road condition on hunting pressure (and local persistence) of B. babyrussa and S. celebenis (Clayton et al. 1997). Adopting a coupled population viability approach would allow the influence of landscape change—primary forests in Sulawesi have been reduced to small isolates (Lee et al. 2007)—and its interaction with exploitation and harvest dynamics to be examined among two sympatric species with different demographic strategies and ecological requirements, by assessing changes in vital life-history traits in response to global change.

Model framework

Following our conceptual method, models are parameterised using life history traits from published field and captive studies and, where necessary, congeneric species (e.g. S. barbatus) and expert opinion (Akçakaya and Brook 2008). Habitat characteristics are linked to a function of habitat suitability, enabling spatial use of the habitat matrix to be simulated (Akçakaya et al. 2004). Market survey data and records from hunters are used to assess exploitation rates (Milner-Gulland and Clayton 2002). The relationship between infrastructure development (e.g. roads, villages, logging camps), habitat fragmentation and pig exploitation, is forecast by developing a predictive model which couples rates and patterns of habitat loss with ecological data (e.g. dispersal, population fluctuations and trends in vital rates). The human-mediated negative relationship between S. celebenis abundance and B. babyrussa survival (due to harvest dynamics of one species mediating hunting effort for the other) is estimated using harvest and survey data (Clayton et al. 1997) and modelled in a dynamic spatial landscape. Future trends in anthropogenic impacts (e.g. habitat loss, exploitation, climate change) are used to forecast population persistence under different management scenarios. Challenges include developing a bio-economic model to predict changes in forest cover/land conversion and exploitation rates in space and time.

Appendix 2: Modelling the impact of landscape change on orang-utans in Borneo

During the Pleistocene, orang-utans (Pongo spp.) were widely dispersed through Southeast Asia (Koeningswald Von 1982). However, today their distribution is restricted to the large islands of Borneo (P. pygmaeus) and Sumatra (P. abelii), where they persist in patches of remaining lowland dipterocarp forests and low-lying freshwater and peat swamps. Previous efforts to model population viability (Singleton et al. 2004) may be overly optimistic because the projections failed to account for the impact of future climate change (e.g. ENSO warming events, sea level rise) and shifts in land use on habitat suitability. For instance: (1) there is a mutually reinforcing association between deforestation, climate (ENSO) and fire (Laurance and Williamson 2001; Siegert et al. 2001), which negatively impacts dipterocarp forests (Curran et al. 2004); (2) the synergistic interaction between deforestation and sea level rise will directly impact orang-utan habitat availability by inundating low-lying freshwater swamps, and indirectly by orchestrating greater agricultural intensification on mid elevation habitats (Millennium Ecosystem Assessment, MA, www.millenniumassessment.org, 2005; Mimura et al. 2007). The Kinabatagan Orang-utan Conservation Project in Sabah Malaysia (Ancrenaz et al. 2007) provides an opportunity to parameterise demographically structured, spatially dynamic, population models for orang-utans (Ancrenaz et al. 2005; Goossens et al. 2006); potentially giving rise to a more ‘ecologically realistic’ framework for assessing the population persistence of orang-utans at local and regional scales under competing management scenarios. With a total population size of about 11,000 individuals in 16 major populations, Sabah is a stronghold for P. pygmaeus (Ancrenaz et al. 2005). More than 60% of P. pygmaeus in the state live outside protected areas, in production forests that have been, or continue to, experience selective logging (Ancrenaz et al. 2005).

Model framework

The movement from traditional population viability approaches (e.g. Singleton et al. 2004) to demographically structured, spatially dynamic, coupled model architectures, which capture many of the ecological complexities and uncertainties impacting species abundance and range, will provide more ‘realistic’ predictive mechanisms to underpin informed orang-utan conservation decision-making, i.e. with the foresight that the biosphere in Southeast Asia is rapidly changing. Conceptually this approach involves (1) extracting vital rates from the published literature and expert opinion (e.g. Wich et al. 2004); (2) drawing on aerial nest surveys (Ancrenaz et al. 2005) to model the relationship between abundance and landscape properties—habitat suitability; (3) using patterns of genetic diversity (Goossens et al. 2006) to integrate dispersal rates between habitat patches; and (4) using paired correlates of human impacts and ecological data to model demographic responses. Future landscape change is simulated by modelling habitat loss or expansion in response to global change.