Conversion of rainforests into agriculture resulted in massive changes in species diversity and community structure. Although the conservation of the remaining rainforests is of utmost importance, identifying and creating biodiversity-friendly agriculture landscape is vital for preserving biodiversity and their functions.

Biodiversity studies in agriculture have often been conducted at low elevations. In this study, we compared the functional diversity (FD), phylogenetic diversity (PD), and community structure of birds along an interacting gradient of land-use (protected rainforest, reserve buffer, and agriculture) and elevation (low, middle and high) in Sri Lanka. Then, we measured compositional change by identifying how ecological traits (dietary guild, vertical strata, body mass, and dispersal ability) and conservation characteristics (forest-dependence and threatened status) responded to land-use types.

Elevation and land-use interacted with each other to shape bird FD. Depending on the elevation, FD in agriculture was either higher or similar to forest. However, PD was similar across all elevation and land-use types. Bird community structure in forest was functionally and phylogenetically clustered in comparison to agriculture. Insectivorous birds declined from forest to agriculture, and so did understorey and middle-storey birds. But frugivorous and canopy birds did not change across land-use types, while nectarivores, granivores and carnivores proliferated in agriculture. Forests were dominated by birds with low dispersal abilities, but birds in agriculture had more evenly distributed dispersal abilities. About half of all the individuals in agriculture was composed of forest species, several of which were threatened.

Synthesis and applications. Most farmers in Sri Lanka practice agriculture on small farms (c. 2 ha) and rely on services (e.g., pest-control and pollination) provided by biodiversity for their livelihoods. Our results underline the important role of these heterogenous agriculture landscapes in maintaining high functional diversity (FD) and harbouring several threatened species. While FD in agriculture was comparatively high, conservation decisions based on land-use alone cannot be reliable, because land-use effects were elevation dependent. Thus, priority setting exercises aimed at designing optimal agriculture landscapes should consider landscape features, in combination with elevation, to benefit both people and wildlife outside protected areas.

In this study, we sampled the full elevation (90 m to 2180 m) and land-use (protected rainforest to agriculture) gradients in the wet zone of Sri Lanka. We sampled birds in the low elevation (90–799 m) habitats of Sinharaja western sector, middle elevation (800–1499 m) habitats of Sinharaja eastern sector, and high elevation (>1499 m) habitats of Nuwara Eliya region between December 2006 and December 2007 (24, two-kilometre transects; see Figure S1 in Supporting Information). Later, between January 2008 and January 2009, we sampled the full elevation gradient from the Gillimalle Forest reserve to Horton Plains reserve (17, one-kilometre transects). See Appendix S1 of supporting information in our paper for further details on study design.

The README.xlsx file explains all the variables in the dataset.