The percentage of complex habitat within 1 km across study sites was negatively correlated with crop cover

To evaluate whether the availability of land use types influenced bee abundance and diversity, we ran Pearson’s chi-square tests of independence. We tested whether the frequency of occurrence of each land use type across study sites was associated with the number of 1) captured bees, 2) abundance of the three most common bee tribes, 3) bee genera, and 4) bee tribes. All analyses were conducted in R. We present evidence for the effects of different local and landscape factors on bee abundance and diversity in the Colombian Andes, a region for which this type of studies are scarce. Bee abundance and diversity were influenced by several habitat factors including flower availability, elevation, and unshaded crop cover within 1 km. Contrary to our hypotheses, bee abundance decreased, although diversity increased in farms with higher habitat complexity, and both local and landscape factors greatly influenced bee community composition. Our first research question examined which local and landscape factors influenced bee abundance and diversity. In general, bee abundance was predicted by flower abundance and elevation. Our results coincide with other studies documenting a positive response of bee density to floral resources , and the unresponsiveness of some groups such as Meliponini to flower availability . We found overall bee abundance was greatly influenced by the abundance of Apis and Trigona bees in areas with high flower abundance. Other studies have documented the positive responses of Apis to the spatial aggregation of floral resources and mass-flowering crops , and have suggested Apis can exclude other bees from accessing flowers through interference or exploitative competition . Like Apis, Meliponini are social and generalist bees. While there is niche overlap among Meliponini species,grow raspberries in a pot and between Meliponini and Apis bees, there is resource partitioning among Meliponini but not between Meliponini and Apis . Thus, a possible explanation for the differential influence of flower abundance on different bee groups may be mediated by their competitive interactions with Apis.

Other factors influencing bee abundance may be related to traits specific to bee groups. For example, Augochlorini abundance decreased with pasture cover within 1 km and increased with % vegetation between 1-3 m. In our study, percent of vegetation cover between 1-3 m was negatively correlated with the percent of vegetation between 0-1 m, which is the strata at which we found more flowering herbs in unshaded land uses. Further, pasture cover within 1 km was negatively correlated with forest or agroforest cover within 1 km. In general, Augochlorini are associated to forest areas , are soil or wood nesters, and use more flowering herbs and vines as feeding resources in comparison with eusocial bees . Thus, use of nesting and food resources by this tribe may explain their abundance patterns. Elevation had a strong influence on the abundance of Apis and Trigona bees. Elevation may act as a filter structuring biological communities because of its inverse relationship to temperature, which may influence species distributions based on their tolerance to cold environments and to climatic fluctuations, among other biophysical variables . Apis and native Trigona cf. amalthea and T. cf. fulviventris have broad altitudinal and geographic ranges as well as high reproductive capacities with colonies of ~10000 and >2500 individuals respectively . The combination of these two factors may explain the high abundance of these genera in our study region, yet it still does not totally explain their positive response to high elevations. It is interesting to consider that, when we excluded these hyperabundant genera, bee abundance was not responsive to elevation. This suggests other mechanisms that favor some groups and undermine others may be at play in the highlands of this region. Bee richness and evenness were also predicted by elevation. We found a strong gradient of bee richness within a relatively narrow altitudinal range , which is considered a hump in general altitudinal-diversity gradients . Potts et al. found richness of bee taxa decreased with elevation in the Alps due to thermal tolerance of bees found along the elevation gradient.

This is consistent with the richness gradient we found, which may also be influenced by the response of different groups to other conditions changing with elevation such as agricultural disturbance. In our study region elevation was positively correlated with the percent of eroded soils at the landscape scale , and the number of land use units on pasture and conventionally managed crops increased with elevation. This suggests rates of disturbance increase with elevation, possibly acting as a strong environmental filter excluding species associated with forests and complex habitats and favoring species with high tolerance to disturbance. This may explain the reduction of rare species and the high presence of Apis and Trigona bees in high elevations, which is reflected in an inverse relationship between evenness and elevation. Bee richness and evenness decreased and dominance increased with unshaded crop cover at the landscape scale. Thus, our results suggest that changes in dominance within communities are associated with the availability of complex habitat at the landscape scale, also found in other studies . Changes in dominance may be associated with the negative responses of rare solitary species to landscape simplification , and with the ability of some groups to equally use complex or simplified habitats. Apis and T. spinipes, closely related to T. cf. amalthea, have been considered hyper-generalist species that do not show negative responses to environmental disturbance and occupy simplified lands , thereby increasing their abundance within the community in areas unfavorable for other bees . Thus, although we cannot differentiate effects of elevation, agricultural disturbance and landscape simplification, these factors and their interaction may sort bee groups in and out of local bee assemblages in this region.

Vertical structure of the vegetation and flower abundance influenced bee richness as well. The stratum at which we found bees was 0-3 m, thus bees at higher strata were not sampled and these results may reflect our sampling bias. However, a study conducted along an ecological succession gradient using different sampling methods reported bee richness and abundance increased in low vegetational strata, and decreased in areas with dense canopy cover . In general, dense canopy reduces sunlight reaching the understory, which in turn influences flowering of herbs . Bee activity may follow the distribution of flowers in the vertical strata , which may explain our results. Our second research question addressed changes in bee community composition. Changes in the composition of bee communities were influenced by elevation and flower abundance. This is consistent with the elevation-richness gradient we found, and may be explained by the degree of specialization and distribution of different bees in our study region. Specialization of plant-pollinator interactions declines with increasing elevation , and biological groups in mountainous regions have narrower ecological niches and altitudinal distributions as elevation decreases . The degree of specialization along elevation gradients may also interact with negative effects of disturbance, thus the changes in community composition in our results may indicate either genera turnover or differential loss of species along the altitudinal range. This has additional implications in light of climate change. Studies have shown increasing temperatures and their influence on thermal tolerance have driven range shifts of different organisms towards higher latitudes and elevation . However, shifts do not only depend on physiological responses but are conditioned on species interactions such as competition and mutualism, and on variation in dispersal abilities . Further studies could target range shifts of bee communities in Anolaima, where environmental change and dominance of highly competitive species are concentrated in areas with higher elevation, possibly influencing range shifts. Our third research question evaluated whether the availability of different land uses was associated with differences in the diversity and abundance of bee genera and tribes. The influence of local and landscape habitat factors on bee abundance and diversity can be linked to current land uses in the region. In general,best grow pots abundance and richness was higher in low-impact land uses, and in areas associated with human constructions. Unshaded traditional crops and fallow lands can have high floral abundance and diversity . Thus, plant richness may beget bee richness in these land uses. Also, we found nests of different groups in human constructions, and foraging on flower and medicinal gardens and on forbs surrounding houses.

Constructions offer areas with favorable features for nest thermoregulation and unmanaged flowering plants may represent continuous floral resources. Thus, human resources may have inadvertent yet important positive impacts on bee populations. We did not find high bee abundance or richness in habitats with high structural complexity i.e. forest or agroforests. Most areas in this land use correspond to shaded coffee crops. In this region, farmers manually exclude forbs that grow despite the low incidence of light into the understory, and rain-fed coffee shrubs typically bloom synchronously only during two or three days a year. This means the understory of this land use does not offer a continuous availability of floral resources for bees. When flowering, trees may offer feeding resources at the canopy level. Tree availability positively influences bee diversity and the long-term availability of flowering trees positively influences abundance of solitary bees in shaded coffee agroforests in Mexico . Some solitary bees are associated with high forest strata, even at the end of the flowering season, perhaps due to supplemental floral resources provided by honeydew and sap from the canopy . Besides food sources, trees also offer nesting resources. Nates et al. found living trees were the most frequent nesting substrate for stingless bees in eastern Colombia. We frequently found Meliponini nests in Inga trees or in abandoned bird nests in citrus trees found in shaded coffee areas. Other studies have found Augochlorini nest in wood and even in bromeliads growing on trees . Therefore, despite the under story of agroforests is not greatly used by bees, the canopy may offer important resources for bees in Anolaima. However, a dense canopy or the presence of high flowering resources distributed across the landscape may influence negatively the local provision of pollination services to coffee shrubs . We also found evidence for the negative effect of conventional crops for bee richness, despite their smaller scales when compared to other agricultural landscapes. Conventional crops in this region are both monocrops and polycrops managed with high agrochemical use . For example, chlorpyrifos and neonicotinoids are systematically used twice a week on tomato for protection from white flies. In extreme cases, application mixtures also include antibiotics to treat cattle from Dermatobia flies. Apis and Trigona were using floral resources in these crops. This suggests bees have different degrees of tolerance to chemical disturbance. For example, while Apis bees and the stingless bee Partamona helleri have different resistance to certain pesticides, both are negatively impacted by mixes of biocides . Thus, species less sensitive to pesticides can thrive in areas with high-impact management, subsidizing the pollination of plants where other bees cannot tolerate agrochemical disturbance. However, areas with high agrochemical disturbance represent a sink for rare bee species yet also a potential threat for bees with relative high resistance to agrochemicals in Anolaima. Different local and landscape factors influence bee abundance and diversity in Anolaima. Some factors were associated with the increase in abundance of two hypergeneralist groups, Apis and Trigona, and with reductions in the representation of rare species, reflected in changes in evenness and dominance within local bee assemblages. This suggests a process of biotic homogenization with the loss of some species and the spread of others, especially in high-elevation areas. In addition, we found that factors that may directly affect bee physiology, such as elevation, interact with resource availability and space to influence the composition of bee communities. We found that land use types such as unshaded conventional crops have negative impacts on bee abundance and diversity, despite the high heterogeneity of agroecosystems in the study region, and that other land uses such as pastures may not benefit some bee groups but are not adverse for others. This suggests bee communities are highly responsive to agricultural management in small-scale farming systems. Our study calls for attention to assess the effect of environmental change on bee communities in mountainous regions where climate change may influence elevational range shifts, such as in the Colombian Andes.