Vincent de Leijster

213 Summary term, this bundle was negatively related to coffee berry borer control. Agroforestry farms with a dispersed tree spatial arrangement provided a higher accumulated ecosystem service supply than farms with trees in alley formation or in living fences. Finally, we found that canopy characteristics were the most important driver of ecosystem services in this system; however, agrochemical input management, altitude and slope also influenced their supply. In this chapter, we concluded that adoption of agroforestry can be used to rehabilitate some ecosystem services, such as carbon storage, erosion control and habitat provisioning, but that the effects on provisioning services are variable. Additionally, we concluded that interactions among ecosystem services often vary over time and that only above-ground carbon and erosion control had a stable positive relationship over time. Chapter 6 followed up on the findings of Chapter 5, as we investigated whether the coffee agroforestry systems were profitable and how costs, benefits and net revenues have developed over time since the adoption of agroforestry. Using a questionnaire, we obtained information about costs, benefits, and management regime along with information about farm, location, and supply chain characteristics. We found that coffee yields significantly decreased over time, but that this did not translate into lower coffee gross revenues, lower coffee net revenues or lower actual net revenues over time. This can be explained by a reduction in harvest and post-harvest costs and an increase in revenues from co-products such as timber and fruit. Specifically in farms with a dispersed tree arrangement, the coffee yields were lower than in monoculture farms, but there were no differences in farms with alley or living fence tree arrangements. Agroforestry farms sold 30% of the total value of timber, fruit and carbon, while the other 70% was not brought to market. The non-marketed share of products was either consumed on the farm, given to labor workers or wasted. This potential revenue was highest for farms with a dispersed tree arrangement. Finally, we found that economic performance was best explained by pest control intensity, altitude, farm size and intermediary choice, and that canopy characteristics were significantly less important. We concluded that agroforestry can and should be implemented in Colombian coffee systems, since net revenues are not negatively affected by the small decreases in coffee yield. There is a large untapped potential for further improvement of the actual revenues of agroforestry farms by better marketing of timber, fruit and carbon. In Chapter 7 I reflect on the findings and address the research questions. Related to the first research question, I found that the agroecological systems we studied had a higher ecological performance than comparable conventional systems. Chapters 3 and 5 showed that ecological performance in almond and coffee plantations improved after agroecological interventions. However, I also found that most agroecological practices promote a specific