Vincent de Leijster

109 Ecosystem services trajectories in coffee agroforestry in Colombia over 40 years 5 2002). Therefore, we expect that the opposite developments of understory cover against canopy cover and litter cover canceled out any changes in potential soil loss in the temporal analysis. We also did not find differences in pest control with time since agroforestry in this study. This could be because our chosen method, space-for-time substitution, was not sensitive enough to detect changes in this service over time since agroforestry was implemented. We found that pest incidence was related to altitude, which in turn is related to temperature (Appendix Table A5-8), and colder temperatures are known to suppress coffee berry borer incidence (Jaramillo et al., 2011). Therefore, we expect the environment in which the research site (farm) was embedded to have had a stronger influence on pest control than the in-situ treatment. 5.5.2 Ecosystem service trade-off and bundle development We found that most ecosystem service interactions are variable over time, since the positive relationship between erosion control and carbon stock was the only consistent relationship between time periods (Table 5-3). This is in line with earlier findings on historical dynamics of ecosystem services in mixed-use landscapes in Canada, where only 3 out of the 27 ecosystem service interactions were consistent cross time (Renard et al., 2015). The interaction between canopy cover, and their corresponding ecosystem services, and coffee yield remains a topic of debate. Shade trees have been found to negatively affect coffee yields (Campanha et al., 2005; Farfán-Valencia and Urrego, 2004), but also to not have an effect on coffee yields (Cerda et al., 2017; Clough et al., 2016; Jezeer et al., 2019). This variation in results is most probably caused by variation in local factors (i.e. soil conditions, rainfall patterns, solar radiations, etc.), which affects the trade-off relationship (Farfán and Jaramillo, 2009; Farfán V., 2014). In this study we did not find significant differences in coffee yields between monoculture coffee farms and agroforestry farms (Appendix Table A5-9), neither did we find a negative correlation between canopy cover and coffee yield (Appendix Table A5-8), however, we did find a trade-off between coffee yield and above-ground carbon stocks in the first period (0-10 y) and we found that canopy height negatively related to coffee productivity (Appendix Table A5-8). Further we also found that coffee yield significantly declined over time only for agroforestry farms that had trees planted in borders (Appendix Table A5-10). These ‘border’ agroforestry farms had the highest timber tree density and the lowest tree species richness (Appendix Table A5-9). In our study C. alliodora was the most abundant timber tree, known to exert very strong effects on coffee yields (Farfán-Valencia and Urrego, 2004). Reducing the densities of C. alliodora and using more legume trees could remove