a b s t r a c tCoffea arabica shrubs are indigenous to the understorey of the moist evergreen montane rainforest of Ethiopia. Semi-forest coffee is harvested from semi-wild plants in forest fragments where farmers thin the upper canopy and annually slash the undergrowth. This traditional method of coffee cultivation is a driver for preservation of indigenous forest cover, differing from other forms of agriculture and land use which tend to reduce forest cover. Because coffee farmers are primarily interested in optimizing coffee productivity, understanding how coffee yield is maximized is necessary to evaluate how, and to what extent, coffee production can be compatible with forest conservation.Abiotic variables and biotic variables of the canopy were recorded in 26 plots within 20 forest fragments managed as semi-forest coffee systems near Jimma, SW Ethiopia. In each plot, coffee shrub characteristics and coffee yield were recorded for four coffee shrubs. Cluster and indicator species analyses were used to differentiate plant communities of shade trees. A multilevel linear mixed model approach was then used to evaluate the effect of abiotic soil variables, shade tree plant community, canopy and stand variables, coffee density and coffee shrub size variables on coffee yield.Climax species of the rainforest were underrepresented in the canopy. There were three impoverished shade tree communities, which differed in tree species composition but did not exhibit significant differences in abiotic soil variables, and did not directly influence coffee yield. Coffee yield was primarily determined by coffee shrub branchiness and basal diameter. At the stand level a reduced crown closure increased coffee yield. Yield was highest for coffee shrubs in stands with crown closure less than median (49 ± 1%). All stands showed a reduced number of stems and a lower canopy compared to values reported for undisturbed moist evergreen montane rainforests.Traditional coffee cultivation is associated to low tree species diversity and simplified forest structure: few stems, low canopy height and low crown closure. Despite intensive human interference some of the climax species are still present and may escape local extinction if they are tolerated and allowed to regenerate. The restoration of healthy populations of climax species is critical to preserve the biodiversity, regeneration capacity, vitality and ecosystem functions of the Ethiopian coffee forests.

The effect of arabica coffee management intensity on composition, structure, and regeneration of moist evergreen Afromontane forests was studied in three traditional coffee-management systems of southwest Ethiopia: semiplantation coffee, semiforest coffee, and forest coffee. Vegetation and environmental data were collected in 84 plots from forests varying in intensity of coffee management. After controlling for environmental variation (altitude, aspect, slope, soil nutrient availability, and soil depth), differences in woody species composition, forest structure, and regeneration potential among management systems were compared using one way analysis of variance. The study showed that intensification of forest coffee cultivation to maximize coffee production negatively affects diversity and structure of Ethiopian moist evergreen Afromontane forests. Intensification of coffee productivity starts with the conversion of forest coffee to semiforest coffee, which has significant negative effects on tree seedling abundance. Further intensification leads to the conversion of semiforest to semiplantation coffee, causing significant diversity losses and the collapse of forest structure (decrease of stem density, basal area, crown closure, crown cover, and dominant tree height). Our study underlines the need for shade certification schemes to include variables other than canopy cover and that the loss of species diversity in intensively managed coffee systems may jeopardize the sustainability of coffee production itself through the decrease of ecosystem resilience and disruption of ecosystem services related to coffee yield, such as pollination and pest control.

The montane rainforests of SW Ethiopia are the primary centre of diversity of Coffea arabica and the origin of all Arabica coffee cultivated worldwide. This wild gene pool is potentially threatened by forest fragmentation and degradation, and by introgressive hybridization with locally improved coffee varieties. We genotyped 703 coffee shrubs from unmanaged and managed coffee populations, using 24 microsatellite loci. Additionally, we genotyped 90 individuals representing 23 Ethiopian cultivars resistant to coffee berry disease (CBD). We determined population genetic diversity, genetic structure, and admixture of cultivar alleles in the in situ gene pool. We found strong genetic differentiation between managed and unmanaged coffee populations, but without significant differences in within-population genetic diversity. The widespread planting of coffee seedlings including CBD-resistant cultivars most likely offsets losses of genetic variation attributable to genetic drift and inbreeding. Mixing cultivars with original coffee genotypes, however, leaves ample opportunity for hybridization and replacement of the original coffee gene pool, which already shows signs of admixture. In situ conservation of the wild gene pool of C. arabica must therefore focus on limiting coffee production in the remaining wild populations, as intensification threatens the genetic integrity of the gene pool by exposing wild genotypes to cultivars.

Summary1. Understanding landscape scale variation in reproductive and pollination success is a major aim of plant population biology. A potential determinant of reproductive success that has received surprisingly little attention so far is variation in nectar chemical composition across plant individuals and populations. Using the Lepidoptera pollinated fragrant orchid (Gymnadenia conopsea) as a model species, we aimed at providing a first account of landscape scale variation in nectar amino acid and sugar composition, its environmental drivers, and how it may affect plant fruit set. 2. Nectar was sampled from 986 flowers from 296 individuals across eleven discrete populations of the G. conopsea. The proportions of sugars and amino acids were determined using high-performance anion-exchange chromatography. Nectar composition was related to soil characteristics of the populations, and to fruit set, using linear mixed models.3. Approximately, 20% of the variance in nectar traits was situated between populations, whereas the highest proportion of variance (c. 45%) was found among flowers within individuals. Soil carbon and nitrogen content affected both nectar concentration and composition. Furthermore, fruit set of G. conopsea individuals was found to be significantly related to nectar amino acid and sugar composition. 4. Synthesis. These results show that landscape scale variation in nectar amino acid and sugar composition should be taken into account in future studies of plant reproductive success. Furthermore, there is also high within-plant variation in both nectar amino acid and sugar composition. This variation may reduce geitonogamous pollination, but it may also limit rapid pollinator-mediated selection on nectar composition.

Floral traits have evolved to maximize reproductive success by attracting pollinators and facilitating pollination. Highly attractive floral traits may, however, also increase the degree of self-pollination, which could become detrimental for plant fitness through inbreeding depression. Floral nectar is a trait that is known to strongly mediate pollinator attraction and plant reproductive success, but the particular role of the nectar amino acid (AA) composition is poorly understood. Therefore, we experimentally manipulated the nectar AA composition and abundance of the Lepidoptera-pollinated orchid Gymnadenia conopsea through soil fertilization, and we quantified AA content and AA composition through high performance anion exchange chromatography with pulsed amperometric detection. Mixed models were then used to evaluate differences in pollinia removal, fruit set, seed set and degree of selfing between fertilized and control individuals. Selfing rates were estimated using microsatellite markers. We found that fertilized individuals had a significantly higher nectar AA content and an altered AA composition, whereas plant height, number of flowers, nectar volume and sugar concentration remained unchanged. Fertilized individuals also had significantly more pollinia removed and a higher fruit set, whereas control plants that did not receive the fertilization treatment had significantly fewer selfed seeds, and more viable seeds. Although we cannot exclude a role of changes in floral scent following the fertilization treatment, our results strongly suggest a relation among nectar AA composition, fruiting success and selfing rates. Our results also indicate potential consequences of nutrient pollution for plant reproductive success, through the induced changes in nectar AA composition.