Vincent de Leijster

110 Chapter 5 these negative effects on coffee yield (Farfán-Valencia and Baute-Balcázar, 2010). Moreover, we found in the second period after transition (11-20 y) a synergy between erosion control and coffee yields, suggesting that soil loss on the long-term reduces coffee yields (Table 5-3). This could be caused by the loss of nutrients from the top layer of the soil due to erosion, which results in lower production potential of the soil (Pimentel et al., 1995). 5.5.3 Factors explaining ecosystem service supply In the current study we found that tree layer characteristics (canopy cover), management intensity (weed and pest control management) and farm location characteristics (slope and altitude) explained ecosystem service supply. We also expected soil conditions to be related to ecosystem service supply, based on earlier findings (Méndez et al., 2009), but we did not find this. Canopy cover characteristics are widely reported to be related to ecosystem service supply in coffee systems, for example to pollination (Klein et al., 2003b), habitat provisioning for butterflies and epiphytes (Jezeer et al., 2019; Mas and Dietsch, 2003), carbon stock (Jezeer et al., 2019) and erosion control (Cannavo et al., 2011). Farm management intensity also explained ecosystem supply, which is in line with previous findings in coffee systems in Costa Rica (Cerda et al., 2017). In our study more intensive pest control management related to higher coffee productivity and more weed control management related to more herb cover and less litter cover. The relation between pest control management and coffee production can be explained by the coffee berry borer, which was a sincere problem in this region (Escobar-Ramírez et al., 2019). Therefore, we expect that on farms where coffee plants where protected more actively, fewer coffee beans were lost due to infestation by berry borers. We expect that in farms where there was a faster growth of herbaceous plants, farmers applied more management to control this. Further, we found interaction among pruning of coffee plants, canopy cover and coffee production, since more recently pruned coffee plants had lower coffee production than more mature coffee plants, but agroforests with denser canopies tended to have older coffee plants (Appendix Table A5-8). 5.5.4 Limitations Space-for-time substitutions are less time consuming than repeated measurements over time, however, previous research has demonstrated that they are not sensitive enough to measure high level of details. A study comparing the two methods concludes that repeated measurements provide more precise information, but also that the strongest trends are found by space-for-time substitutions (Rácz et al., 2013). This can be explained by variation