Search

Attributes of each dataset used in the analyses. Locations of each plot are provided as *.pdf file (Additional file 2). N - Number of plots used for the analyses (we excluded plots that measured at less than eight sampling points). Year - Year of field measurements. Researcher - AB, Andrew Burt; ACS, Aida Cuni-Sanchez; AG, Alemu Gonsamo; AL, Alicia Ledo; ARM, Andrew R Marshall; BW, Beatrice Wedeux; DD, Dereje Denu; DS, Deo Shirima; HS, Hamidu Seki; JGT, Jose Gonzalez de Tanago Menaca; KC, Kim Calders; LC, Luis Cayuela; LAS, Lau Alvaro Sarmiento; MJM, Manuel J Macia; MP, Marion Pfeifer; ND, Nicolas Deere; PO, Pieter Olivier; PKEP, Petri Pellikka; PJP, Philip J Platts; RT, Rebecca Trevithick; RH, Robin Hayward; RM, Robert Marchant; TP, Timothy Paine; WW, Woodgate William. Figure S1. Example maps of human population pressure, calculated from human population density grids using a range of sigma values (σ = 5, 15, 25, 50). Colours are graduated on a log base 2 scale (light colours, low pressure; dark colours, high pressure). The maps provide scope for capturing human-driven pressures at a variety of spatial scales (Platts 2012). For example, if σ = 5 then the weight given to remote populations (relative to the local population) halves over a distance of ~4 km, nearing zero by ~15 km, whereas if σ = 25 then the weight halves over a distance of ~20 km, nearing zero by ~60 km. We imposed a maximum distance of 100 km, beyond which no pressure is exerted. Figure S2. Relationships between Annual Moisture Index (AMI) and Mean Annual Precipitation (MAP) and canopy attributes LAI, fAPAR and FCover. We fitted linear, polynomial and nonlinear (nls model 1: y ~ a + b * I(x^z); nls model 2: y~a/(1 + exp.(−(b + c*x))) models. Upper panel: polynomial models fitted to LAI ~ MAP, FCover - MAP and fAPAR - MAP relationships. The polynomial (RSS 1.464) and sigmoidal growth models (RSS 1.464) produced slightly better fits to the LAI data compared to the fits produced by the linear (RSS 1.47) and exponential (RSS 1.467) models. The polynomial model produced the best fit to the FCover (RSS 24.76) and fAPAR (RSS 0.2371) data. Lower panel: nls model 2 fitted to LAI ~ MAP, FCover - MAP and fAPAR - MAP relationships. The logistic growth model produced the best fit to the LAI data (RSS 1.347), the FCover data (RSS 22.95) and the fAPAR data (RSS 0.2191). (DOC 590 kb)

Attributes of each dataset used in the analyses. Locations of each plot are provided as *.pdf file (Additional file 2). N - Number of plots used for the analyses (we excluded plots that measured at less than eight sampling points). Year - Year of field measurements. Researcher - AB, Andrew Burt; ACS, Aida Cuni-Sanchez; AG, Alemu Gonsamo; AL, Alicia Ledo; ARM, Andrew R Marshall; BW, Beatrice Wedeux; DD, Dereje Denu; DS, Deo Shirima; HS, Hamidu Seki; JGT, Jose Gonzalez de Tanago Menaca; KC, Kim Calders; LC, Luis Cayuela; LAS, Lau Alvaro Sarmiento; MJM, Manuel J Macia; MP, Marion Pfeifer; ND, Nicolas Deere; PO, Pieter Olivier; PKEP, Petri Pellikka; PJP, Philip J Platts; RT, Rebecca Trevithick; RH, Robin Hayward; RM, Robert Marchant; TP, Timothy Paine; WW, Woodgate William. Figure S1. Example maps of human population pressure, calculated from human population density grids using a range of sigma values (σ = 5, 15, 25, 50). Colours are graduated on a log base 2 scale (light colours, low pressure; dark colours, high pressure). The maps provide scope for capturing human-driven pressures at a variety of spatial scales (Platts 2012). For example, if σ = 5 then the weight given to remote populations (relative to the local population) halves over a distance of ~4 km, nearing zero by ~15 km, whereas if σ = 25 then the weight halves over a distance of ~20 km, nearing zero by ~60 km. We imposed a maximum distance of 100 km, beyond which no pressure is exerted. Figure S2. Relationships between Annual Moisture Index (AMI) and Mean Annual Precipitation (MAP) and canopy attributes LAI, fAPAR and FCover. We fitted linear, polynomial and nonlinear (nls model 1: y ~ a + b * I(x^z); nls model 2: y~a/(1 + exp.(−(b + c*x))) models. Upper panel: polynomial models fitted to LAI ~ MAP, FCover - MAP and fAPAR - MAP relationships. The polynomial (RSS 1.464) and sigmoidal growth models (RSS 1.464) produced slightly better fits to the LAI data compared to the fits produced by the linear (RSS 1.47) and exponential (RSS 1.467) models. The polynomial model produced the best fit to the FCover (RSS 24.76) and fAPAR (RSS 0.2371) data. Lower panel: nls model 2 fitted to LAI ~ MAP, FCover - MAP and fAPAR - MAP relationships. The logistic growth model produced the best fit to the LAI data (RSS 1.347), the FCover data (RSS 22.95) and the fAPAR data (RSS 0.2191). (DOC 590 kb)

The high degree of endemism on Sulawesi has previously been suggested to have vicariant origins, dating back 40 Ma. Recent studies, however, suggest that much of Sulawesi's fauna assembled over the last 15 Myr. Here, we test the hypothesis that more recent uplift of previously submerged portions of land on Sulawesi promoted diversification and that much of its faunal assemblage is much younger than the island itself. To do so, we combined palaeogeographical reconstructions with genetic and morphometric datasets derived from Sulawesi's three largest mammals: the Babirusa, Anoa and Sulawesi warty pig. Our results indicate that although these species most likely colonized the area that is now Sulawesi at different times (14 Ma to 2–3 Ma), they experienced an almost synchronous expansion from the central part of the island. Geological reconstructions indicate that this area was above sea level for most of the last 4 Myr, unlike most parts of the island. We conclude that emergence of land on Sulawesi (approx. 1–2 Myr) may have allowed species to expand synchronously. Altogether, our results indicate that the establishment of the highly endemic faunal assemblage on Sulawesi was driven by geological events over the last few million years.

Locations of each plot. (PDF 730Â kb)

Locations of each plot. (PDF 730Â kb)