In most other studies (except that in ref. 13), variables related
to shade (i.e., shade tree diversity and density, and shade cover)
were positively associated with pollinator abundance and diversity
(Table S1). This was not the case in Kodagu, where shade
did not affect bee abundance at coffee, although native tree
density (highly correlated with tree species richness) negatively
affected bee abundance. It is possible that floral resources
available within the many forest patches across the landscape,
as well as the flowering shade trees within coffee agroforests,
are sufficiently abundant in Kodagu that resources offered by
coffee flowers fail to attract pollinators. Indeed, the distribution
of forest patches in Kodagu, in terms of their distance to coffee
plantations, has little effect on pollinator visitation to coffee
(10), a result contrary to other studies that suggest declining pollinator services with increasing distance to forest (14, 15).
This is probably because forest cover is still extensive across
Kodagu, and even plantations far from forest patches are not
sufficiently isolated to suffer from deficiencies in pollinator visitation.
Bee diversity and abundance at coffee flowers might,
therefore, be a function of the abundance and distribution of
alternative floral resources within agroforests or across the landscape,
and where these resources are abundant the attractiveness
of coffee as a resource might be diminished (10, 16). Should this
be the case, then the validity of encouraging forest conservation
through an associated crop pollination service will depend on the
context of land cover distributions.