Although ecological restoration has entered the global agenda to reverse different anthropogenic disturbances, we still know little about how this solution interacts with other conservation strategies, to avoid the progressive loss of species and ecosystem services.

Here we evaluate one of the pioneering restoration programs in the southern Brazilian Atlantic Forest, where the combination of conservation and restoration efforts have been carried out for 20 years. Specifically, we tested how landscape characteristics, restoration strategies  and environmental characteristics affect the results of the restoration of pastures.

We established 65 circular plots (total 4.0 ha) along restoration areas (3-10 years) and sampled trees and shrubs composing the canopy (DBH > 5cm) and understory (DBH< 5cm, height > 1.3m). We analyzed the landscape metrics (proportion of old-growth forests in 200 m, 500m and 1000m buffers around each plot; and area and distance of the nearest-neighboring old-growth forests). We explored the multiple effects of landscape, restoration strategy (reforestation, natural regeneration) and environmental variables (soil, pasture grass types) on the species composition and multiple diversity metrics of restoration areas.

The species composition was very similar among restoration ages and restoration strategies. We found positive and strong effects of old-growth forest (200 m buffer) proportion on the species richness and Shannon diversity (canopy and understory), aboveground biomass (canopy) and functional diversity (understory) of restoration areas. The restoration strategies affected forest structure, and, in general, the reforestation strategy increased aboveground biomass, Shannon, functional and phylogenetic diversities (in canopy), and percentage of endemic species and biomass (understory), when compared to natural regeneration.  

The 20 year-experience in the southern Atlantic Forest showed that programs focused on landscape conservation associated with a mixture of restoration strategies (i.e. natural regeneration in larger areas and active restoration in more disturbed sites) , can be an efficient strategy to ensure biodiversity and ecosystem services in tropical landscapes.

Synthesis and applications: To manage degraded tropical lands and achieve global targets for biodiversity and ecosystem services, it is necessary to first ensure the conservation of natural remnants and then use multiple restoration strategies in less resilient areas.

Restoration experiment: To evaluate the efficiency of the restoration strategies used, we carried out an experiment in partnership with SPVS that authorized the study within its reserves. The experiment consisted of sampling the vegetation in the restored areas in Guaricica and das Águas Natural Reserves (Figure 1a). We established 65 circular plots (radius of 14m; area of 615.7m² each) resulting in a total sampled area of 4.0 ha. Plots were distributed in the two reserves and were selected in a way to account for a gradient of all variables of pasture management techniques, soil and landscape characteristics, and restoration strategies (Figure 1b). The plots were established in areas with ages varying from 2 to 10 years since the sites were abandoned of agricultural use, and in areas with four different soil types (Acrisol, n=6 plots, Cambisol, n=15, Gleysol, n=38, and Fluvisol, n=6),  two restoration strategies (reforestation=24 plots; natural regeneration=41 plots), and 3 types of grass cover combinations (Urochloa cf humidicola and Urochloa arrecta, n=23 plots, Paspalum sp., n=16; and mixed Urochloa and Paspalum, n=20). For the remaining number of sites, we did not have information whether exotic grasses were used and, therefore, they were analyzed in a different sub-group.

The experiment was established in 2010 and since then, some other plots have been incorporated into the sample design and evaluated in several studies (for example, Cequinel et al. 2018; Capellesso et al. 2020). Part of the plots analyzed here (39 plots) were also studied by Kauano et al. (2013) regarding to the effects of the landscape on the canopy of forests ongoing restoration. In this study, we expanded the sample area (here 65 plots) and the vegetation survey (here also considered understorey), incorporating, therefore, a greater environmental gradient where the restoration took place. This field study does not require ethical approval, according to Brazilian standards.

Landscape analysis: The images used in the analysis were obtained from overlapping LANDSAT-5-TM images (from 5^th March 2009) with Ortho-rectified aerial photos (from 2001), and official Parana state geomorphological maps. From these images, landscape related variables were obtained by the application of Vector-based Landscape Analysis Tools (V-Late 2.0) extension (Lang & Blaschke, 2007) for ArcGis 10.8 software (see more details in Kauano et al., 2013). The landscape surrounding the plots is characterized by the presence of young, intermediate and old-growth forests, fluvial areas (wetlands) and pasture and agriculture areas. The classification of land-use categories was done by visual inspection of images associated by comparisons with historical aerial photos. Old-growth forest sites were considered the forest remnants where there were no signals of intensive logging for the past 80 years.

In each plot we obtained the following landscape metrics: distance and the area of the nearest-neighbouring fragment, and proportion of old-growth forest. We chose these metrics because they are related to the amount of available habitat, as well as the distance for the arrival of propagules, which affects regeneration (Magnago et al., 2015). To explore the effects of geographical scales on the proportion of old-growth forest, we considered three buffer distances, with 200 m, 500 m and 1000 m (in radius) around the central coordinate of each plot. Considering the total of plots, the distance to the closest old-growth forest remnant (nearest-neighbor distance) varied from 12 to 488 m, and the area of this remnant varied from 0.36 ha to 4.24 ha. The proportions of old-growth forests for 200 m buffer varied between 0% to 72%, for 500 m buffer varied between 0.13% and 70%, and, for the 1000 m buffer, between 8.5% to 60%.

Vegetation survey and plant community metrics: In all 65 plots of 14m in radius we sampled all trees and shrubs with diameter at breast height (DBH; 1.3m) > 5cm, to characterize the canopy. In a smaller concentric sub-plot (4m in radius, 50.3m²), we measured all individuals (tree saplings and shrubs, which will be referred to hereafter as “understory”) with DBH < 5.0cm and total height >1.30m. For all sampled individuals we determined the species name (by comparing with vouchers in herbarium UPCB) and measured the DBH (or stem base diameter, for understory) and tree total height.

The plant communities (canopy and understory) present in each plot were analyzed in relation to different biodiversity metrics: taxonomic richness (S) and diversity (H’, Shannon’s index), functional diversity (FD, Rao's quadratic entropy), and phylogenetic diversity (PD, Mean Pairwise Distance). Rao Q is a measure of divergence by average functional distance between two individuals (Mouchet et al. 2010) and indicates niche complementarity (Ricotta & Moretti, 2011). MPD provides a global community structure value (Webb, 2000; Webb et al., 2002). Additionally, we considered the percentage of endemic species (%_end) of Atlantic Forest and aboveground biomass (Biom, Mg ha-1), in order to represent multiple dimensions of biodiversity. Functional diversity was calculated using the following plant traits: specific leaf area (SLA, cm2 g-1), leaf dry-matter content (LDMC, mg g-1), wood density (WD, g cm3), and maximum height (Hmax, m). All traits were measured based on locally collected samples whenever possible, following standardized protocols (Pérez-Harguindeguy et al., 2013, see details in Capellesso et al. 2020). The PD was based on the built of ultrametric phylogeny for tree species (see Supporting Information for details). Aboveground biomass was calculated following the general wet tropical forest formulae provided by Chave et al. (2014).

All data are in Excell and can be converted to other formats.