Vincent de Leijster

54 Chapter 3 offs can be drawn. De Giorgio and Lamascese (2005) and Martínez-Mena et al. (2013), did suggest a negative relationship between understory cover and crop yield, as they found lower almond yield in non-tilled plots with natural vegetation, compared to conventionally tilled plots in longer term experiments. There were no other negative correlations among individual ecosystem service indicators. Furthermore, in the PCA analysis we observed three pairs of positively correlated indicators that can be interpreted as early signals for emerging ecosystem service bundles. First, we observed that SOC and soil enzymatic activities are correlated (Figure 3-3), which agrees with the findings of Ramos et al. (2011). Second, we found that crop yield related to the fruit set ratio on the first two axes, and that individual almond fruit weight related to soil enzymatic activities on the PCA1 axis and on both PCA1 and 2 to soil nutrient levels and SOC (Table 3-4). The importance of pollination for almond trees has been shown in Californian almond systems (Klein et al., 2012); another Californian almond study has demonstrated how nutrient availability is essential for fruit development (Muhammad et al., 2015). Nevertheless, the bundles containing yield data are uncertain, due to lower than average yield values. One possible limitation to our ability to detect bundles and trade-offs at the scale of multiple ecosystem services during ecological rehabilitation in woody crop systems could be the length of the study. The rate of ecosystem services interaction, trade-off or bundling, is known to vary between systems and over space and time (Costanza et al., 2017). This suggests that even though certain trade-offs and bundles are not observable at a certain moment, they may yet develop over time. For example, Ponisio et al (2016) found that the ecosystem services pollination and habitat provisioning were higher after more than 10 years of planting hedgerows around intensive agricultural land than in the first 10 years. Applying this hypothesis on dryland tree-crop systems would suggest that trade-offs and/ or bundles as a result of agroecological management implantation may arise over time. Therefore, research on larger spatial and temporal scales is required involving combinations of agroecological practices, such as organic amendments combined with understory vegetation. Each agroecological practice may enhance a specific set of ecosystem services, so by combining them ecosystem service bundles can be expected. Limitations, uncertainties and knowledge gaps We were able to detect rehabilitation of specific ecosystem service provisioning within a short time frame, but the limited time frame might also have prevented the detection of changes in ecosystem services that respond less rapidly (e.g. below-ground carbon stock or arthropod mediated ecosystem services) or that do not respond at all. Therefore, long-