Résumé
The composition of agro-ecological landscapes is thought to have important implications for the production of major crops through its effects on pollinator abundance and behaviour. We explored the roles of land cover and land cover heterogeneity on bee nest distribution for the giant honeybee Apis dorsata, a key species for coffee pollination, in a complex agroforest landscape. We emphasized scaling and non-uniform effects by combining two different approaches of spatial analysis, the point-pattern analysis and surface-pattern analysis. We found non-exclusive, positive effects of agroforests, forest fragments and land-cover heterogeneity on the presence and number of nests. The distribution of nests responded to habitat heterogeneity at small scale (<100 m), forest fragments at medium scale (<300 m) and to agroforest at larger scales (500 m to 2 km). Our multiple approaches highlight that the landscape effects were neither linear nor uniform within the study zone. Nests were consistently located in areas of medium agroforest density or medium to high forest density, but were absent where forest fragments are the most concentrated. The agroforest matrix was particularly important in shaping the size of nest aggregates. Nests tended to be few when there is low tree cover at broad scale, while nests were numerous when agroforest patches are abundant within the bees’ foraging range. Synthesis and applications. Our study revealed that structurally complex landscapes appear to support bee populations. The spatial arrangement of different land covers affected honeybee nest distributions by providing nesting and foraging resources across multiple scales. The results suggest that continued intensification of small forest fragments and expansion of large monospecies plantations will be deleterious to the populations of giant honeybees A. dorsata. Fragmentation of the agroforestry matrix at small scales (100s m) does not, however, appear detrimental for A. dorsata as long as sufficient diversified resources are available at the landscape scale (kms).