Vincent de Leijster

108 Chapter 5 2013; Yang et al., 2019) and that may be a reason why we did not (yet) find an effect here. However, soil carbon content also depends on other factors, such as soil conditions and former or current land-use (Paul et al., 2002), which may have varied between our study locations. In developing cocoa agroforests in Cameroon the below-ground carbon content (at 0-15 cm depth) increased over a period of 80 years after implementation (Nijmeijer and Harmand, 2019), which suggests that also below-ground changes in carbon stock can be expected. Moreover, we found that canopy height was positively related to soil organic carbon (Appendix Table A5-8), which also suggests that below-ground carbon is related to above- ground biomass production and that this response has yet to occur. We found that both butterflies and epiphytes responded rapidly. Butterflies are sensitive to changes in micro-climatic and habitat conditions and since they are mobile, they can rapidly respond to changes in these conditions (Meyer and Sisk, 2001). A study in Uganda showed that butterfly diversity rapidly increased (as of the first year) after forest restoration activities (Nyafwono et al., 2014). Epiphyte colonization speed depends on presence of woody vegetation to host on, and therefore the rehabilitation speed is slower than the growth of woody vegetation (Cruz-Angón and Greenberg, 2005). This is also reflected in our results as the epiphyte richness directly related to canopy cover (Appendix Table A5-8). In a temporal analysis study in shaded coffee systems in Nicaragua, the authors found that shade tree densities declined over time, while epiphyte richness increased (Goodall et al., 2015). They suggested that for epiphytes the time to colonize and/or canopy height is more important than tree density (Goodall et al., 2015). The latter was also supported by our results as epiphytes were related to canopy closure and canopy height (woody vegetation development), but not to tree density or tree species richness (Appendix Table A5-8). The presence of mature trees therefore seems to play a crucial role for epiphyte habitat provisioning. Erosion control was found to be stable over time since implementation of agroforestry. This could mean that agroforestry does not affect this ecosystem service or that the speed of the response was either too fast or too slow to be captured in the study timeframe. We suspect that changes in erosion control were too fast to be captured in our study, as we found that the agroforestry systems had on average 48% lower potential soil losses than the monoculture coffee farms (P=0.008; Appendix Table A5-8), which suggests that implementation of agroforestry should play a role. Similar results were found in Nicaragua where coffee agroforestry had more soil covered with litter, reducing soil erosion (Blanco Sepúlveda and Aguilar Carrillo, 2015). We found that canopy cover significantly increased over time, while understory vegetation cover reduced over time, which can be explained by a limitation for light which reduces the growth of herbaceous vegetation (Soto-Pinto et al.,