Community trends from RLQ analysis
RLQ analysis illustrated how traits, species and habitat features varied together. Two axes explained 90.3% of the total inertia in the three tables, also accounting for most variability (> 72-79%) along the first two axes of the environmental variables (R-table) and species’ functional traits (Q-table) separately (Table S1). Traits and environmental variables were particularly strongly related to the first RLQ axis (Fig. 1(a) and (b)).
Both major RLQ axes were related strongly to elevation. The first axis reflected a significant altitudinal trend towards narrower river stretches with faster flows, well vegetated banks and channels with boulders and pebbles while sandy banks, altered riparian cover, human settlements and human activities declined (Fig. 1(a)). Feeding traits correlated significantly with this axis as species using more terrestrial prey from the river margins increased towards higher elevations (e.g. Plumbeous Water Redstart; White-capped Water Redstart (Phoenicurus leucocephalus ) whereas species using a blend of terrestrial and aquatic prey declined (e.g White-throated Kingfisher (Halcyon smyrnensis ) (Fig. 1(b)). Simultaneously, species using aquatic prey solely such as the Brown Dipper (Cinclus pallasii) and Little Forktail(Enicurus scouleri) increased along this axis. Overall, the contribution of ‘aquatic’ feeding was minimal and neutral as aquatic-feeding species persisted at both ends of the axis.
The second axis of the RLQ mostly represented a significant decline in riparian vegetation cover and boulder-strewn banks but an increase in pebble banks and islands at higher elevation – typical of upland braided reaches. Bird traits varying significantly on this axis included an increase in clutch-size, but a decline in body size, tarsus size, bill size and aquatic/terrestrial foraging as species such as River Lapwing, Common Sandpiper (Actitis hypoleucos) , Blue Whistling-thrush (Myophonus caeruleus) and Spotted Forktail(Enicurus maculatus) dropped out of the community (Fig. 1).
Strong, significant relationships among the trait, habitat and species abundance data were corroborated by the global RLQ permutation test (p <0.001 for model 2 of Dray et al. 2014). This held across all regions suggesting a uniform pattern in the species–trait–environment relationship at the community level. Model 2 was rejected (p=0.0005) and Model 4 accepted (p=0.608) together sug­gesting that i) species distributions were influenced by environ­mental conditions, dominantly through changes related to elevation and ii) species composition reflected significant variations in trait character that also tracked elevation on both major axes.