Data collection
Vegetation sampling: I obtained the data on tree communities from an online repository (Osuri and Sankaran, 2016a). The dataset comprised of 50 square plots of 30 x 30 m – 31 plots at eight locations within contiguous forest and 19 plots in eight forest fragments. All plots were censused during Jan-Dec 2011. Sampled fragments ranged from 5-10 ha in size, chosen to avoid ongoing or recent disturbance or human-use. Vegetation plots were placed at least 50-m from fragment edges and 50-m apart from each other. The contiguous forest plots were located within a complex of structurally connected nature reserves adjacent to the fragments. Shade coffee plantations dominated the land-use in the human-modified matrix. For more details of the plot locations and placement, see Osuri and Sankaran (2016b). Species were identified using field keys, regional floras (Pascal and Ramesh 1997; Ramesh et al. 2007) and with input from experienced botanists. We excluded about 3% of individuals as could not be identified to genus or species because they had no leaves or the canopy was obscured by climbers.
Traits: I used four key functional traits of plants—seed size, wood density, specific leaf area (SLA) and maximum height—that were available in the supplementary information included with Osuri & Sankaran (2016b). Trait data was collected between November 2011 and January 2013 following standard protocols (Cornelissen et al. 2003) from trees within or adjacent to the vegetation plots. The data collection methods have been detailed in Osuri and Sankaran (2016b). Here, I briefly repeat the main aspects.
For SLA, five mature, healthy, sun-exposed canopy leaves were collected per tree at the end of the wet season (October–December) for 358 trees comprising 79 species. Leaf areas were estimated using the Black Spot Leaf Area Calculator (Varma and Osuri, 2013), after which leaves were oven-dried at 60°C for 72 h to obtain dry weights. Wood density was estimated by dividing dry weight by fresh volume of trunk wood cores collected with an increment borer, for 352 trees representing 74 species. Thirty-six species for which adequate primary data could not be collected, wood density was also collated from secondary sources (Chave et al., 2009). Seed size was quantified as the length of the primary seed axis, for 34 species (1879 seeds in total). Additionally, seed size data were obtained from two other mid-elevation evergreen forests in the Western Ghats (D. Mudappa, unpublished data). For species without primary measurements, seed lengths were collated from published secondary sources (Matthew 1983; Saldanha & Ramesh 1984; Ramesh et al. 2007). Maximum heights for each species were obtained in the field and from secondary sources (Matthew 1983; Saldanha & Ramesh 1984; Ramesh et al. 2007).
Environmental data: I used mean Climatological Water Deficit (Chave et al., 2014) to quantify site-level variation in seasonal water stress (Condit et al., 2013; Vicente-Serrano et al., 2013). CWD, measured in mm/year, is always a negative number since it is the difference between rainfall and evapotranspiration during dry months only, and more negative values indicate greater water deficit. Data were downloaded from the source cited in Chave et al. (2014; http://chave.ups-tlse.fr/pantropical_allometry.htm#CWD). I also characterized site-level climate aridity with Annual Evapotranspiration (AET) data, available at a spatial resolution of 1 km2, and can be downloaded from the CGIAR-CSI GeoPortal (https://cgiarcsi.community). I also used data from WorldClim, collated per 1 km2, to characterize gradients in precipitation and temperature across the study region (Hijmans et al., 2005). I used mean annual precipitation, mean precipitation of the driest quarter, mean precipitation of the warmest quarter, mean annual temperature, mean temperature of the warmest quarter, and the mean temperature of the driest quarter.