|
nif:isString
|
-
We searched the scientific literature to identify quantitative measures of variables (ecological indicators) related to the provision of one or more ecosystem services in tropical forests. As biodiversity protection has been considered an ES [50], we also used biodiversity metrics (see below) as ecological indicators. The selection was restricted to studies that comparatively presented quantitative measures of levels of ecological indicators of ES in one of the following conditions: (i) degraded and restored; (ii) restored and reference ecosystem; or (iii) degraded, restored and reference ecosystems. We defined the degraded area as the starting point of restoration, the restored area as being directly or indirectly subjected to restoration actions and the reference ecosystem as the undisturbed area. The search criteria included studies conceived as ecosystem restoration projects (for example, reforestation with native or exotic species, nucleation, natural regeneration), studies designed to maximize forest production (agroforestry) or studies that comparatively surveyed abandoned areas where natural regeneration occurred. In all situations, ecological indicators were locally measured and used for comparisons. We conducted the search in the scientific databases ISI Web of Knowledge and Science Direct using the following terms and combinations, without restriction to year (until May 2017): (tropical* forest) AND (restoration* OR regeneration* OR recuperation* OR rehabilitation* OR restore* OR recovery* OR reforestation native* OR sucession* OR disturbance* OR perturbation). The preliminary search was limited to the following subject areas “climate change”, species richness”, “costa rica,national park”, “tree species”, “tropical forest”, “ecosystem service”, “soil,microbial biomass”, “organic matter”, “Brazil”, “forest ecology”, “atlantic forest”, “microbial community”, “soil organic carbon”, “forest management”, “secondary forest”, “puerto rico" in Science Direct search. With these terms, we obtained 8,764 articles including studies in tropical and subtropical forests. In a prior analysis of the title and summary of each study, we selected 3,190 articles that contained all quantitative variables necessary for the meta-analysis. Then, we searched each article for quantitative variables (mean, standard deviation, sample size and age) of ecological indicators in degraded, restored and/or reference conditions (see below). From this search, 69 articles from 25 countries in five continents were found (Fig 1; Supporting information S1 Table; S1 Fig).
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0208523.g001 Flow diagram of database searching and article screening.The checklist of the total Prisma 2009 requirements is in S2 Table. For each study, we compiled ecological indicators based on Benayas et al. [51] and respective ecosystem services, according to MEA [2], as follows: carbon pool (aboveground biomass, below-ground biomass, dead organic matter and soil organic carbon) and soil attributes (C, Ca, Mg, N, Nitrate, P, pH, cation exchange capacity [CEC], water holding capacity, and soil organic matter). Moreover, we compiled species richness, diversity, density and abundance data as a proxy for evaluating the effect of restoration practices on recovering biodiversity protection (Table 1).
Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0208523.t001 Definition and number of ecological indicators (n) of ecosystem services of parameters considered in the meta-analysis. *CEC: cation exchange capacity.
From these 69 studies, we extracted 866 quantitative measures of ES: 383 independent effect sizes were calculated comparing restoration and degraded areas, and 685 independent effect sizes were calculated comparing restoration areas and reference ecosystems. To explain heterogeneity in effect sizes, we compared the effect sizes among five restoration strategies (natural regeneration, nucleation, reforestation with native species, reforestation with exotic species and agroforestry) and three types of land use (i.e., degradation) prior to restoration (pasture, logging and agriculture) (Table 1). When the sample size was sufficient (pasture and agriculture), we compared the effects of different restoration strategies on ES recovery in the degraded ecosystem. For the land use “logging” the sample size was insufficient for analysis.
We extracted the mean, standard deviation and sample size for each ecological indicator of ES in the primary studies. Using this information, we calculated Hedges’ g effect size, the variance and the bootstrap confidence interval (CI). Hedges’ g (average differences divided by standard deviation) is a variation of Cohen’s d that includes a correction of deviations, which are derived from a small sample [52]. According to our criteria for estimating effect sizes, a positive value means that the amount of ES in restored areas is higher than in degraded areas or a reference ecosystem; a negative value means the opposite. Before summarizing the effect sizes to obtain an overall effect of restoration, we tested the hypothesis of real heterogeneity among studies using Q-statistics [53]. Our data were considered heterogeneous (P (Q) ≤ 0.05), and therefore, random effects models were used to calculate the average effect size (g+). We also tested if heterogeneity among studies could be explained by the type of restoration and the type of degradation using subgroup analyses [52]. Similarly, we performed meta-regression analyses [52] between effect sizes and restoration time to analyze whether the efficacy of ES recovery depended on the time since restoration began. Additionally, we performed a complementary analysis in order to detect any bias in the effect sizes in the metanalysis. For this, we did the fail-safe N of Rosenthal [54] and of Orwin [55] analysis and, also, the trim and fill technique [56].
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}