ES recovery
Response variables were related to a wide variety of ES, so multiple RR-ES combinations were included as separate rows in the database (Table S1). The parallel assessment of these multiple associations allowed us to capture the simultaneous supply of several ES [14], [15]. To avoid counting the same data more than once in a meta-analysis, we performed a separate meta-analysis for each ES using a random-effects model. We considered this approach suitable because we wanted to evaluate each ES separately, rather than the heterogeneity among different ES.
Correlation between biodiversity and ES recovery
We assessed the correlation between biodiversity recovery and ES recovery using the Spearman rank coefficient to quantify the correlation between the corresponding RRs. We used only RRs from studies that evaluated both biodiversity and ES, and we treated each of these studies as an independent sample. When the same study reported multiple measures of biodiversity or ES, the related RRs were averaged to generate an overall RR for biodiversity and an overall RR for ES for each study, thereby minimizing the risk of pseudo-replication. This approach led us to combine the four major ES types in order to ensure adequate sample size