Mineland rehabilitation aims to restitute biodiversity and ecosystem functioning as closely as possible to premining levels (Gastauer et al., 2018a), representing an important pillar of the mitigation hierarchy (World Economic Forum, 2016) for reconciling economic development with the conservation of biodiversity and ecosystem services (Lovejoy and Nobre, 2019). When implemented successfully, such rehabilitation promotes self-perpetuating communities and ecosystems (Gann et al., 2019) and becomes an integral part of conservation planning and implementation (Volis, 2019). Continuous systematic monitoring of ecosystem rehabilitation is necessary to detect changes in plant communities along rehabilitation trajectories, measure the success or failure of activities and provide important information about the return of ecosystem functions and biodiversity after mining (Gwenzi, 2021).

Traditional environmental monitoring programs survey field variables related to species composition, abundance, and taxonomic diversity as well as vegetation structure and key ecosystem functions such as carbon sequestration to track the trajectory of rehabilitating minelands (Gann et al., 2019). Methods for integrating different environmental properties into a measure of rehabilitation success are available (Gastauer et al., 2021, see methods for details), and phylogenetic ecology, which links evolutionary history and short-term ecological processes, may provide additional tools for the assessment of rehabilitating minelands (Gastauer et al., 2018b; Hipp et al., 2015). This is because phylogenetic diversity (PD) can be a useful proxy for high-dimensional functional trait diversity (Tucker et al., 2018), maximizing communities' abilities to resist and/or recover from disturbances (Augeraud-Véron et al., 2019). Thus, the lack of phylogenetic groups in rehabilitating minelands may lead to ecological function decay, hinder the return of ecosystem services, and reduce community resistance and resilience (Maza-Villalobos et al., 2011).

Specific traits and adaptations may be required for plant recruitment in degraded minelands, commonly characterized by low nutrient availability, a lack of mutualism, high temperatures and elevated radiation (Mukhopadhyay et al., 2016). If such adaptations occur in only a few evolutionary lineages, such environmental filtering may cause phylogenetic clustering, i.e., the coexistence of closely related species within the community (Ginocchio et al., 2017). In contrast, interspecific interactions such as competition or facilitation limit niche overlap between coexisting species (Carrión et al., 2017). When ecological niches are conserved within evolutionary lineages, such that close relatives tend to resemble each other, the phylogenetic distance between coexisting species tends to increase when interaction-mediated coexistence is relevant for community assembly (Godoy et al., 2014).

Because environmental constraints of degraded minelands are expected to be alleviated with the establishment of vegetation cover and the return of soil function (Gastauer et al., 2018a,b,c), the recovery of plant–plant interactions should shift the phylogenetic structure of plant communities from clustering to overdispersion (Swenson and Enquist, 2009). The return of interactions is furthermore expected to contribute to the success, stability, resistance, and resilience of the rehabilitating areas (Lake, 2013). Accordingly, the evaluation of phylogenetic relatedness in plant communities along rehabilitation trajectories may be important, especially when we aim to infer community assembly along rehabilitation chronosequences. In such efforts, the integration of phylogenetic indices into monitoring frameworks is paramount.

Here, we present an analysis of phylogenetic structure and diversity along a rehabilitating mineland chronosequence in the Carajás National Forest, eastern Amazon, Brazil. This region harbors the world’s most important reserves of iron ore, and comprehensive open-cast mining requires large-scale mineland rehabilitation. The objective was to evaluate these features and provide feedback to enhance the rehabilitation processes, understand the underlying processes of community assembly and outline how phylogenetic measures complement existing monitoring frameworks. We expected that (i) tree communities in initial rehabilitation stages would be dominated by only a few phylogenetic clades, while clades would be more evenly distributed in older stages and at reference sites (random or overdispersed); (ii) phylogenetic clustering would decrease with the advancement of rehabilitation, such that phylogenetic structure would become more similar to reference values from natural ecosystems over time; and (iii) metrics of diversity and phylogenetic structure would be related to the success of rehabilitation activities.

In total, 1536 trees of 232 species and 21 orders were recorded. With 37 species, Fabales was the most species-rich order (Fig. 2). With 179 species from 18 orders, the species richness at the reference sites was higher than that at the rehabilitation sites (total of 67 species, Supplementary Material 1). Among the rehabilitation sites, the advanced stage included 49 species and 11 orders, followed by the intermediate stage with 43 species in 12 orders and the initial stage with 7 species and 4 orders.

Fig. 2.

Phylogeny of angiosperm tree species sampled in this study. Positive and negative signals indicate more or fewer daughter taxa than expected by chance, respectively, from nodesig analysis within distinct rehabilitation stages (p value ≤ 0.05; for details on nodesig analysis, see Supplementary Material 3).

(0.27MB).

All orders except Solanales and Asterales were detected at the reference sites, where more descendants than expected by chance were detected in the orders Sapindales, Myrtales and Caryophyllales (Fig. 2), indicating a lack of these orders in rehabilitation stages. Trees of only four orders were found in the initial rehabilitation stages, including Fabales and Asterales, with more descendants than expected by chance. Oxalidales, Gentianales, Ericales, Caryophyllales, Proteales, Arecales (palms) and Laurales were lacking in the intermediate and advanced rehabilitation stages, where Fabales, Rosales, Malpighiales and Myrtales were represented by more species than expected by chance, indicating their importance for phylogenetic clustering. The abundance of angiosperm orders varied considerably among and within the rehabilitation stages (Fig. 2).

Reference areas showed higher PD than rehabilitating sites (Fig. 3), and increases in PD from initial to intermediate and advanced stages were detected. MPDt and MPDi values were higher at reference sites than at rehabilitating sites, but no differences between rehabilitation stages were found. Regarding phylogenetic structure, marginal overdispersion (0.1 < p value < 0.05) was found in reference forests (Fig. 3). In the initial stage, ses.MPDi indicated phylogenetic clustering, and ses.PD and ses.MPDt indicated marginal phylogenetic clustering. The low significance level is most likely due to the small number of plots with two or more shrub species necessary to compute these phylogenetic metrics (Supplementary Material 1). ses.PD and ses.MPDt indicated phylogenetic clustering in the intermediate and advanced stages, whereas phylogenetic structure measured by ses.MPDi in the advanced stage did not differ from the null expectation. ses.PD and ses.MPDt values were higher in the reference forests than in the rehabilitation stages. When measured by ses.MPDi, advanced stages did not differ from reference sites or from initial or intermediate rehabilitation stages (Fig. 3). All the metrics of phylogenetic structure and PD were positively associated with rehabilitation status, and correlation coefficients were higher for PD and MPDi than for MPDt or the standard effect transformations of the metrics (Fig. 4).

Fig. 3.

Phylogenetic diversity (A. PD - phylogenetic diversity, B. MPDt - mean pairwise phylogenetic distance among taxa and C. MPDi - mean pairwise phylogenetic distance) and structure (D. ses.PD - standard effect size of phylogenetic diversity, E. ses.MPDt - standard effect size of mean pairwise distance among taxa, F. ses.MPDi - standard effect size of mean pairwise distance) comparisons among initial, intermediate and advanced rehabilitation stages and reference sites. Different letters indicate significant differences between sites (p ≤ 0.05). Darkly shaded bars indicate phylogenetic structures differing from the null expectation (p ≤ 0.05), and lightly shaded bars differ marginally from the null expectation (0.1 ≤ p < 0.05).

(0.21MB).

Fig. 4.

Rehabilitation status in comparison to phylogenetic diversity and structure (A. PD - phylogenetic diversity; B. MPDt - mean pairwise phylogenetic distance among taxa; C. MPDi - MPD among individuals; D. ses.PD - standard effect size of PD; E. ses.MPDt - standard effect size of MPD among taxa; F. ses.MPDi - standard effect size of MPD among individuals).

(0.29MB).

As expected, we observed shifts in tree community composition and phylogenetic metrics along the analyzed rehabilitation chronosequence. First, Fabales and Malvales dominated tree communities beginning in the initial rehabilitation stage, and further clades established with rehabilitation advancement. Second, (marginal) phylogenetic clustering (0.05 < p values ≤ 0.1) in early rehabilitation stages indicated the presence of mechanisms that limit the establishment of species belonging to a few evolutionary clades. With time, all phylogenetic metrics used here increased and became more similar to values found at reference sites, indicating that actual rehabilitation efforts met desired trajectories (Gastauer et al., 2021). If this trend continues, rehabilitating minelands may indeed become habitats for species from distinct phylogenetic lineages. Third, phylogenetic metrics were positively associated with the environmental quality of the rehabilitation sites and should be further explored to outline the importance of evolutionary history for vegetation structure, community diversity and ecological processes.

More daughter lineages from the Fabales order than expected by chance caused phylogenetic clustering in early rehabilitation stages. Due to their adaptations to high-radiation and nitrogen- and water-limited soils (Gei et al., 2018), higher seed viability and higher growth rates than those of seeds from other families (Ramos et al., 2019), nitrogen-fixing legume species perform better in these environments than other families and are recommended for further mineland rehabilitation purposes (Jia et al., 2020). Following these recommendations, a large portion of seeds of nitrogen-fixing species from the Fabales clade were applied during the hydroseeding procedure (see Supplementary Material 2 for details), contributing to the observed dominance, on the one hand.

On the other hand, similar patterns, i.e., phylogenetic clustering (Satdichanh et al., 2019; Yu et al., 2019) and high abundances of species from the Fabaceae family (Maza-Villalobos et al., 2011), have been reported for early rehabilitation stages in the literature, including in naturally regenerating ecosystems, and have been associated with environmental filtering. Thus, the observed phylogenetic clustering is also consistent with the presence of environmental filters that select for species with specific traits (Gastauer et al., 2018b). Increases in MPD values and shifts toward phylogenetic randomness, as detected by abundance-weighted measures of phylogenetic structure (ses.MPDi), within the rehabilitation chronosequence indicate the colonization of the sites by distantly related clades. In addition to the colonization by evolutionary lineages new to the sites, larger increases in abundance-weighted MPDi than in MPDt values (based on presence-abundance data only) with rehabilitation advancement highlight the above-average recruitment of lineages that were not highly represented during the initial stage.

Although the colonization of distantly related species with rehabilitation advancement may result from longer growth periods of hydroseeded species from further clades to achieve inclusion criteria, this indicates that eventual environmental filters during early stages are rapidly alleviated with rehabilitation advancement. Thus, mineland colonization may indeed occur randomly or uniformly (and not in a clustered manner) throughout the phylogenetic tree, and diverse plant communities containing representatives of distinct evolutionary clades can rapidly establish. The observed switch from phylogenetically clustered to phylogenetically random or overdispersed communities over time further indicates the emergence of interaction-mediated coexistence in the advanced rehabilitation stage (Martínez-Ramos et al., 2021). This is because plant–plant interactions such as competition or facilitation tend to limit the phylogenetic similarity among coexisting species when ecological niches are conserved within evolutionary lineages (Webb et al., 2002), contributing to niche diversification with rehabilitation advancement (Cadotte et al., 2016).

Correlations between phylogenetic metrics and overall rehabilitation status demonstrated partial recovery of phylogenetic community structure and PD simultaneously with improvements in environmental quality. Considering shifts from phylogenetic clustering to overdispersion with rehabilitation advancement, consistent with the alleviation of environmental effects and the emergence of coexistence-stabilizing plant–plant interactions, our data highlight the importance of niche diversification for the ecosystem performance of rehabilitating minelands (Enquist et al., 2015), although the underlying mechanisms require further investigation.

The findings of this study have two major implications. First, the correlations between phylogenetic metrics and overall environmental quality show that phylogenetic structure and PD recover during the rehabilitation process together with further ecosystem properties. Thus, mineland rehabilitation may indeed contribute to mining impact mitigation, as these habitats may become home to a wide range of species from distinct evolutionary clades and a considerable portion of PD. Second, mechanisms restricting the colonization and establishment of phylogenetically distinct taxa in early mineland rehabilitation stages may hinder the enhancement of tree diversity found to be paramount in maximizing rehabilitation success (Gastauer et al., 2021). Therefore, using a narrow set of species to initiate the rehabilitation process may alleviate the environmental filters and over time allow a more diverse set of species to colonize. When needed, enrichment planting of a phylogenetically diverse set of species at later stages can help enhance diversity and increase rehabilitation success. Candidate species for enrichment plantings may be found in the orders Oxalidales, Ericales, Caryophyllales, Proteales and Arecales, which were found in the reference forests but not at the rehabilitation sites.

Our study confirms previous findings regarding rehabilitating chronosequences, such as the convergence of important ecosystem characteristics toward those of reference sites. Eventual environmental filters or further mechanisms hindering the establishment of some lineages in the initial stages of mineland rehabilitation tend to be overcome with time, although important ecosystem properties such as PD are not completely recovered within the observed periods. Strong associations between the overall environmental quality of rehabilitating minelands and phylogenetic indicators highlight the importance of niche diversification in the ecosystem performance of rehabilitating minelands.

None declared.