Vincent de Leijster

105 Ecosystem services trajectories in coffee agroforestry in Colombia over 40 years 5 by soil K content, time since pruning coffee plants, pest control management and altitude (P=0.003, R 2 =0.31). Farmer reported-yield was best explained by altitude (P=0.04, R 2 =0.16). Furthermore, we performed correlation analysis between ecosystem service indicators and biotic and abiotic factors, the results of which are presented in a correlation matrix in the appendix (Table A5-8). Tree spatial arrangement We found that canopy closure was higher in ‘dispersed’ farms than in ‘living fence’ farms and did not significantly differ from ‘alley’ farms, but no differences were found for other vegetation characteristics (P<0.001; Appendix Table A5-11). We found higher levels of above-ground carbon in ‘dispersed’ farms (30±18 Mg C ha -1 , n=26) than for farms with ‘living fences’ (15±14 Mg C ha -1 , n=12) and ‘monoculture’ farms (5.2±4.2 Mg C ha -1 , n=11), and ‘alley’ farms (22±11 Mg C ha -1 , n=12) also stored significantly more above-ground carbon than monocultures (F=9.4, df=58, P=0.02; Figure 5-4a and Appendix Table A5-9). We also found that total carbon stocks increased over time since agroforestry was implemented in ‘alley’ farms (R 2 =0.57, P=0.003), and showed an increasing trend for farms with ‘living fences’ (R 2 =0.22, P=0.07), but no changes over time were found for ‘dispersed’ farms (Appendix Table A5-9). Moreover, ‘dispersed’ farms stored 42% more total carbon than ‘monoculture’ farms, 120 Mg C ha -1 and 84 Mg C ha -1 , respectively. Also, ‘dispersed’ farms had 54% lower soil loss than ‘monoculture’ farms (F=4.3, df=41, P=0.01; Figure 5-4b). The diversity of butterflies was 39% higher in ‘dispersed’ farms than in ‘monoculture’ farms and did not differ from the other tree arrangement groups (F=4.3, df=41, P=0.01; Figure 5-4c). We found that ‘monoculture’ farms had an average timber volume of 0.02 m 3 ha -1 , while ‘dispersed’ farms had 66 m 3 ha -1 , ‘alley’ farms 47 m 3 ha -1 and ‘living fence’ farms had 58 m 3 ha -1 . Finally, we found that the coffee yield of ‘living fence’ farms reduced over time since agroforestry was implemented (R 2 =0.39, P=0.02), and that the coffee quality in ‘alley’ farms increased over time (R 2 =0.68, P=0.004)