Vincent de Leijster

101 Ecosystem services trajectories in coffee agroforestry in Colombia over 40 years 5 with approximately 20 y since agroforestry implementation (Figure 5-3 and Appendix Table A5-4). All other ecosystem services showed no significant relationships; the graphs that visualize their trajectories and the model parameters, including the half-time coefficient and saturation coefficient, of all the ecosystem services are included in the appendix (Table A5-4 and Figure A5-4). Figure 5-3. Development of ecosystem services over time since agroforestry has been implemented. a) The response trajectory of above-ground carbon (AGC) follows an asymptotic shape (P=0.04, R 2 =0.06, n=60). b) Understory vegetation cover (% soil covered with vegetation) follows a negative GLM gamma shape (P=0.03, R 2 =0.0. n=445). c) Butterfly diversity (Shannon index) follows an asymptotic shape (P=0.006, R 2 =0.33, n=19). d) Epiphyte richness (number of taxonomic groups; bromeliads, ferns, lichens, bryophytes) follows an asymptotic shape (P=0.02, R 2 =0.12, n=39). e) Coffee yield reduces over time (GLM gamma) but not significantly (P=0.08, R 2 =0.03, n=54). f) Coffee quality measured in the field (dry weight:fresh weight) increases over time (asymptote), but not significantly (P=0.06, R 2 =0.07, n=39). g) Timber volume (m 3 /plot) follows a sigmoid shape (P=0.03, R 2 =0.03, n=60). Solid lines indicate models with P values < 0.05 and dashed lines models with P values < 0.10. 5.4 Trade-offs and bundles We found a bundle formed by above-ground carbon stocks, erosion control, along with epiphyte richness (Table 5-3). This bundle had a short-term (1-10 y) trade-off with coffee yield and berry borer control, but these trade-offs disappeared on the long-term (11-20 y).