Introduction

The increase in agricultural and urban areas is eliminating most of the tropical and subtropical forest remnants of the world [1, 2]. Deforestation influences global changes that affect biodiversity and causes alteration of the carbon cycle, increased erosion, and lack of connectivity between habitats [3]. There is therefore an urgent need to restore biotopes of terrestrial ecosystems [4, 5]. Many efforts are being made on a global scale to restore diverse environments to their natural, unaltered state [6, 7]. Determining the natural processes of revegetation is extremely important, because little is known about the complex interactions that maintain the stability of tropical and subtropical ecosystems [8], and the diversity of species [9, 10].

Restoration efficiency may be measured by the environmental value of species dependent on the quality of the vegetation, especially birds [1011–12]. Birds are preferentially used to evaluate the effectiveness of restored areas because of their mobility, the speed at which they colonize new environments, their ability to connect habitats through seed dispersal, and their maintenance of gene flow among plant populations [13, 14].

Independent of the forest restoration technique, ecosystems restored under different procedures may gradually converge to form an ecosystem characteristic of the regional flora, due to the climax forest tendency [6, 13]. However, this is not such an obvious pattern, principally because of the stochasticity and randomness of environmental vectors, which tend to increase the floristic richness of the habitat [2, 13]. Nevertheless, the initial restoration plantings may be crucial to future successional processes [8, 17, 18].

Restoration using natural processes may be an important alternative, maintaining interdependence with the fauna and preserving the complex relationships in each phase of succession [18–19]. However, passive restoration is only possible in a highly resilient environment where natural vectors that promote seed dispersal are available nearby [16, 20]. Among the diverse active restoration techniques, tree planting is the most common and can provide great diversity and floristic richness over the long term [19, 21]. Nucleation techniques establish vegetation in small habitat patches, gradually restoring the environment and ecological relationships [22]. These techniques can offer extensive structural complexity through their own spatial configuration and insertion of structural elements into the landscape (e.g., artificial perches, planting of seedlings in nuclei, etc.) [18, 20, 23]. According to Boanares and Azevedo [24], these techniques are more common in Brazil than in other countries, and all possible uses are as yet unknown. Based on the premise that complexity and heterogeneity directly affect the creation of niches, we tested different restoration techniques (nucleation, passive restoration and high diversity planting) and their effects on the richness, abundance and diversity of birds. In addition, we determined the bird assemblage structure and evaluated the degree of individual species' preferences for the experimental treatments.

Methods

Study area

The study was designed and conducted by F.C.B. on the experimental farm of the Universidade Tecnológica Federal do Paraná in the municipality of Dois Vizinhos, state of Paraná, Brazil (Fig. 1). The region was originally dominated by subtropical Atlantic Forest in the transition zone between the Araucária moist forest and seasonal semi-deciduous forest. The climate is Cfa (according to Köppen), with a mean temperature of 20 °C, at least one frost every two years, annual precipitation of about 2,000 mm, an altitude of nearly 500 m, and predominance of generally deep oxisols (Bw). The experimental area was historically used for agricultural purposes and pastures. In October 2010, during the last crop harvest of the year, the 7.2 ha experimental area was cleared using a tractor-mounted brush cutter to begin the treatment planting (restoration techniques) at the same time.

Fig. 1:

Map indicating the location of the study area and arrangement of experimental plots: nucleation (NC), passive restoration (PR) and high diversity plantation (HD).

Experimental design

In a randomized block design, we tested four plot replications (40 × 54 m) of three treatments (each treatment totaling 0.86 ha): 1) passive restoration (PR); 2) nucleation (NC); and 3) high diversity planting (HD) (Fig. 1). A distance of 13±5 SD m (SD = standard deviation) was maintained between plots and 20.6±5.7 SD m from the nearby forest fragment.

The passive restoration (PR) treatment was also considered a control, and its plots, along with the plots of the other treatments, were protected against disturbances (fire, grazing, etc.).

In the nucleation (NC) plots, a set of seven techniques based on Reis et al. [18] was used (Fig. 2a) in six 3 × 40 m-strips occupying 1/3 of the total plot area. We used structural and functional techniques. For the structural sets we built: 1) six artificial shelters for fauna (1 m³-woodpile); 2) two artificial perches (10 m high) made of Eucalyptus poles (including dried crowns and cultivating native climber Sweet-passion-fruit, Passiflora alata). The functional sets consisted of: 3) six 1 m²-topsoil seed bank sod blocks (topsoil collected from a nearby secondary forest remnant - 25°36′83¶ime; S; 53°04′10¶ime; W - and deposited in trays to cultivate regenerating seedlings, which were planted in the field as sod blocks); 4) six 1-m² seed rain sod blocks (seed rain was collected in thirty 1 m²-seed traps in the same nearby forest remnant, and like the seed bank, sowed in trays to cultivate regenerating seedlings, which were planted in the field as sod blocks); 5) Cover-crop of Pigeon pea (Cajanus cajan) was sown in twelve 3 × 4 m-nuclei; 6) six bromeliad (Bromelia antiacantha) islets were planted (five seedlings 0.5 m apart in a “+” shape); and lastly, 7) 24 native tree islets composed of five seedlings planted 1 m apart in a “+” shape formed by four rapid-growth pioneer seedlings at the edges and a shaded non-pioneer species in the center (we used 556 seedlings.ha⁻¹, 12 pioneer species and 24 non-pioneer, listed in Appendix 1).

Fig. 2:

Nucleation (a) and high diversity plantation (b) designs made by F.C.B. The numbers correspond to the species described in the supplementary document (Appendix 1).

The high diversity plantation (HD) design was based on the Brazilian filling and diversity lines technique [17, 25–26], where a total of 70 native tree species were planted (10 fast-shading filling species interspersed within the lines of 60 non-pioneer shaded species, listed in the Appendix 1 supplementary document) in a 3 × 2 m-spacing (Fig. 2b). NC and HD were mowed twice a year using a portable brush cutter, followed by the application of glyphosate (2.5 kg of Roundup WG^©. ha⁻¹ applied using a hand sprayer in dry, non-windy conditions, excluding drift between plots) for weed control, throughout the duration of the study. The mowing and subsequent herbicide application was done each time in the entire area of the HD plots, and just inside the NC area occupied by the six 3 × 40 m-strips. PR received no herbicide, and the herbicide's environmental impacts were not evaluated here.

Data collection

There were 24 samplings (six per season), with a sampling effort of eight hours per plot or 96 hours during the entire experiment, between January and December 2012. A bird census was carried out one year after the beginning of the restoration. The species were recorded only when they occurred within the limits of the plots (e.g., on their perches, on the ground or on shrubs). They were recorded in flight when foraging at a height lower than the artificial perches. We obtained estimates of richness and abundance using straight counting of a single sampling point in the center of each experimental plot, with observers' movements allowed to obtain a visual record of the species [27]. The samplings occurred every fifteen days; however, some observations were delayed by meteorological events (e.g., rain and wind), so that the samplings were standardized on sunny days or days with light rain (< 5 mm). Each sampling lasted 20 minutes (10 in the morning and 10 in the afternoon). The beginning of the census occurred when the sun was at an angle of approximately 5° on the horizon, and in the afternoon at 45°. This arrangement was chosen because 20 consecutive minutes would greatly increase the chance of resampling the same individuals. Dividing the 20 minutes into two 10-minute periods may lead to an overestimation of abundance, but it increases the chance of detecting discrete species [28]. However, abundance should be interpreted with caution due to a bias caused by differences in detectability of species. Even so, it is a relevant metric for comparisons between plots. The nomenclature used to identify the birds is according to the Brazilian Committee of Ornithological Records of 2014.

Data analysis

We did the Shapiro-Wilk test (for normality) and the Bartlett test (for homogeneity of the variances) to explore the data [29]. In addition to richness (S) and abundance (N), diversity was obtained by means of the Shannon-Weaver (H′) index and Pielou estimating for uniformity (e^H/S), according to Krebs [30]. Berger-Parker dominance (D) was estimated for the total of contacts between treatments, with later ordination, using the Whittaker diagram [31]. Based on the rank of dominance (D), the use of the scree test was adapted [32]. This test was used as a criterion to determine which species possess greater representativeness in abundance for each treatment. The mean values per sampling of the parameters of richness, abundance and diversity were subjected to an analysis of variance (ANOVA) factorial design (three treatments × four seasons with six samplings per season). A post hoc Tukey test was carried out next.

Rarefaction curves were carried out for observed richness of both the samplings and the abundance of the individuals, using the software EstimateS^® v.8.2. [ http://viceroy.eeb.uconn.edu/stimates]. Additionally, richness estimates, richness estimates were obtained (estimators Chao 2 and Jacknife 1) using the procedure of 10,000 randomizations, which are indicated for situations in which the sampled populations are composed of several unidentified subpopulations [33].

The total abundance data (sum of the 24 samplings) in the 12 restored plots were grouped (Bray-Curtis coefficient) through the UPGMA (Unweighted Pair Group Method with Arithmetic Mean) in order to show the patterns of similarity in the species composition. The UPGMA was chosen to minimize the distortion of the initial matrix of similarity in the construction of the dendrogram without a defined cutting height, prioritizing natural groupings. The Bray-Curtis coefficient was used because it was considered the most efficient to carry out the analysis of similarity (ANOSIM), using 10,000 permutations and comparing the similarity among the treatments to the post hoc test of pairwise comparisons of Mann-Whitney [34].

In order to evaluate the possibility of individual species' preferences for the treatments [35] we used a species indicator analysis (IndVal). This method combines the degree of specificity of a species to the habitat (in this case, different treatments) and its fidelity, assuming two or more groups established a priori [36]. The indicator index (IV) was obtained for each species (the group with the highest association value was identified). A total of 10,000 permutations were carried out again to test the significance of the values through the Monte-Carlo statistic (α<0.05).

The species were classified according to preferential habitat, based on Ries and Sisk [37] and Scherer-Neto and Toledo [38]: 1) open areas (OA) - species that occur in agricultural areas, abandoned fields, pastures; 2) forest (FO) - species that occur in canopy and understory); and 3) edges (ED) - species common to the margin of the forest, with little sensitivity to the edge effect and tolerant of small gaps. Birds were classified in trophic groups, based on Cueto and Casanave [39] and Telino-Júnior et al. [40], according to the predominant feeding (omnivores, nectarivores, insectivores, granivores, frugivores and carnivores). For the status of their occurrence, based on Cueto and Casanave [39] and Bencke [41], we classified: 1) migrants (M); and 2) residents (R). The frequency of occurrence index [42] was based on the nomenclature used by Lack and Venables [43]: 1) very abundant (80 ├ 100%); 2) abundant (60 ├ 80%); 3) frequent (40 ├ 60%); 4) occasional (20 ├ 40%); 5) rare (1 ├ 20%); and 6) very rare (< 1%). Species abundance was based on Berger-Parker dominance, by grouping in classes of dominance, according to Palissa et al. [44]: 1) eudominant (> 10%); 2) dominant (10 ├ 5%); 3) subdominant (2 ├ 5%); 4) recessive (1 ├ 2%); 5) rare (< 1%). The proportions of species in the categories of preferential habitat, trophic groups, status of occurrence, frequency of occurrence, and frequency of dominance were compared among treatments and categories through the chi-square (χ²) test, with the null hypothesis of equality (α=0.05) using the Yates correction [45].

Results

A total of 58 species were recorded, and 48.28% (n = 28) were present in all treatments. The birds were grouped in 22 families: Thraupidae (n = 14; 24.14%) and Tyrannidae (n = 10; 17.25%) were the most representative (Appendix 2). The highest richness was verified for the NC treatment (S_obs. = 49 ± 2.45 species) and the lowest in HD (S_obs. = 37 ± 3.14 species). The comparison between the mean richness presented a high value of statistical divergence (F _[2; 276] = 61.79; P < 0.01). Abundance and diversity were also superior in the group of nucleation techniques (NC > PR > HD). The Tukey test indicated that the PR treatment at all times assumed intermediate values between NC and HD (Tables 1–2). Seasonality had an influence because higher means of the studied parameters always occurred in the summer and in the spring (Table 1 and Fig. 3a-c).

Table 1.

Descriptive statistics and results of the factorial variance analysis for the parameters richness, abundance and diversity between nucleation (NC), passive restoration (PR) and high diversity plantation (HD). Legend: MS = mean square; F = value of the ANOVA test; P = value of probability (95%); SD = standard deviation. Means followed by the same letter do not differ through the Tukey test (95%).

Table 2.

Descriptive statistics and Tukey test to compare the parameters richness (S), abundance (N) and diversity (Hʹ) between seasons based on general means of the treatments. Means followed by the same letter do not differ through the Tukey test (95%).

Fig. 3:

Graphic representations over the seasons and samplings in relation to the parameters: (a) mean richness - S; (b) mean abundance - N; (c) mean diversity - H′; (d) ordination of Berger-Parker dominance represented by the Whittaker diagram; (e) rarefaction of the observed richness as regards abundance; (f) rarefaction curves of the richness as regards the samplings (Mao Tau); (g) annual fluctuation of uniformity - J′; and (h) dendrogram of similarity (Bray-Curtis) between experimental plots, with ordination through the UPGMA.

Based on the ordination of dominance (Fig. 3d), the scree test indicated the selection of four species that, together, represented 60% of the total of records in the NC treatment: Blue-black Grassquit (Volatinia jacarina, D=16%), Red-crested Finch (Lanio cucullatus, D=16%), Double-collared Seedeater (Sporophila caerulescens, D=15%) and Ruddy Ground-dove (Columbina talpacoti, D=12%). In PR, according to the criteria of ranking, the three most abundant species were selected (62% of the total): Blue-black Grassquit (D=24%), Red-crested Finch (21%) and Double-collared Seedeater (17%). In HD, three species were also selected (47% of the total): Red-crested Finch (D=19%), Ruddy Ground-dove (15%) and White-tipped Dove (Leptotila verreauxi, 13%). Exclusively in the NC treatment were: White-tailed Kite (Elanus leucurus), Yellow-headed Caracara (Milvago chimachima), Picazuro pigeon (Patagioenas picazuro), Boat-billed Flycatcher (Megarynchus pitangua), Variegated Flycatcher (Empidonomus varius), Grassland Sparrow (Ammodramus humeralis), Swallow Tanager (Tersina viridis), and Bay-winged Cowbird (Agelaioides badius). Planalto Hermit (Phaethornis pretrei), Ochre-collared Piculet (Picumnus temminckii), Yellow-browed Tyrant (Satrapa icterophrys), Black-goggled Tanager (Lanio melanops) and Yellow-bellied Seedeater (Sporophila nigricollis) occurred only in PR. Burrowing Owl (Athene cunicularia), Slaty-Breasted Wood-rail (Aramides saracura), and Pauraque (Hydropsalis albicollis) were exclusive to HD.

By means of rarefactions, the highest observed richness occurred in NC (Fig. 3e and f). The estimated richness was also superior in NC [50.38 ± 3.50 species (SD), through Jacknife 1; and 48.14 ± 5.45 species (SD), using Chao 2]. The values were always higher than the observed for the estimator Jacknife 1, with lower estimates for the estimator Chao 2. The graphic analysis of the accumulation curves allows the inference that a satisfactory asymptote of richness was not observed in any of the treatments.

Evenness (J′) was more stable over the samplings in NC [J′ = 0.87 ± 0.06 species (SD)], with a coefficient of variation of 6.84%, very similar to the pattern observed in PR [J′ = 0.84 ± 0.08 species (SD)], with a coefficient of variation of 9.64%. However, the fluctuation of the evenness in HD was highly variable, reaching a mean of J′ = 0.63 ± 0.27 species (SD), with a coefficient of variation reaching 43.10% (Fig. 3g).

The similarity analysis among plots pointed to the formation of two distinct groups [(ANOSIM), R = 0.51; P < 0.01]. One of the formed groups contained restored plots with applied nucleation and passive restoration. Plots restored on high diversity plantations formed an external group, distinct from the other treatments (Fig. 2h).

IndVal analysis allowed the selection of 10 indicator species, with only one, the Striped Cuckoo (Tapera naevia, IV=75) in the PR treatment, while another nine species were associated with NC: Ruddy Ground-dove (IV=59.5), Smooth-billed Ani (Crotophaga ani, IV=61.7), Shiny Cowbird (Molothrus bonariensis, IV=80), Great Kiskadee (Pitangus sulphuratus, IV=67), Roadside Hawk (Rupornis magnirostris, IV=80), Fork-tailed Flycatcher (Tyrannus savana, IV=85), Tropical Kingbird (Tyrannus melancholicus, IV=65), Eared Dove (Zenaida auriculata, IV=73.3) and Yellow-bellied Elaenia (Elaenia flavogaster, IV=75). Seven species were considered migratory (12%). This pattern was maintained among treatments (Table 2).

There was a predominance of 57% of species characteristic of open areas (n = 33), while the characteristics of edges and forest environments (n = 14 and 11, respectively) completed the sampling total. Variations in the quantity of species among the different classes of preferential habitats in the same treatment were detected by means of the χ² test; however, no variations were verified in the proportions within each class among treatments (Appendix 3). A predominance of species characterized as rare occurred in all treatments (NC: χ² = 40.76, df = 5, P < 0.01; PR: χ² = 47.63, df = 5, P = 0.00; HD: χ² = 67.94, df = 5, P < 0.01). There was a variation between treatments only in the class of frequent species, with 10 species in NC (χ² = 8.78, df = 2, P < 0.05). Other classes were constant (Appendix 3).

Discussion

Within 58 bird species recorded, the families Thraupidae and Tyrannidae represented 41% of the richness. These families are characteristic of open, altered, or disturbed environments, common in areas in the early stage of ecological succession, and are also the most abundant in forest habitats in the Neotropical region [46–47]. The elevated species richness of the family Tyrannidae is directly related to the variation observed in the seasonality, because this group corresponds to one third of the austral migrants [48], with preference for open areas [49]. The seasonal variation found for the parameters of diversity has practical implications for ecological restoration, as some studies have indicated a direct relationship between the presence of migratory species and a seasonal increase in the deposition of seeds dispersed by birds [50–51].

Tyrannidae species found in this work are considered by many authors to be insectivores or omnivores (generalists), efficiently dispersing seeds by removing them from canopies and edges and depositing them in viable conditions along open landscapes [52–53]. This ecosystemic function is very important for ecological restoration, as specialized frugivores generally occur in low density or do not occur at all in altered landscapes [9, 13, 54].

The pattern observed for richness, diversity and abundance follows a gradient (NC > PR > HD), corroborating the hypothesis that higher levels of complexity and heterogeneity in NC can increase richness, which influences diversity and abundance (dependent parameters). In summary, studies that link bird assemblages to environmental factors find direct relationships between increase in the structural complexity and increase in the diversity of birds [12, 55, 56].

Dominance analysis demonstrated that only five species represent 47% in HD and up to 60% in NC: Blue-black Grassquit, Red-crested Finch, Double-collared Seedeater, Ruddy Ground-dove and White-tipped Dove. Except for White-tipped Dove, these species commonly occur in open areas [57]. They are ruderal and granivorous and adapt well to environments in early stages of succession [58].

A total of 78% of the species occurring in HD were characteristic of open areas or on the edges. The composition of the assemblage composed of generalist species affected the avian fauna analysis of similarity, where the HD plots recorded a pattern that was different from NC and PR. According to Munro et al. [11], the richness of the avian fauna in forest habitats actively restored by different techniques can be similar, but the faunal composition can be different, with a predominance of more generalist species in large scale plantations. On the other hand, species of forest birds are more associated with plantations that have undergone minor interventions [59, 60].

The occurrence of Ruddy Ground-dove and White-tipped Dove among the higher dominance species in HD and NC treatments is due to the biology of the species themselves, since they benefit from the mowing and chemical weeding of the clearing management procedures (bare soil is the predominant habitat where Columbiformes obtain small fallen fruits and seeds [61, 62]). These data demonstrate that the preference for a certain procedure may involve biological and ecological characteristics of the species, as discussed by Báldi and Batári [63], who pointed out that grassland birds may be benefited by the homogenization of the environment. Specialization by a certain stage of the ecological succession process was observed, corroborating Sanderson et al. [56], who demonstrated the relationship between the decrease in the pioneer vegetation and the decline of some bird populations.

Evenness in HD was more variable. Theoretically, lower levels of evenness are caused by assemblages of a smaller number of species (but with high dominance, common in unstable environments) [9]. Instability in early-restoration habitats can be strongly favored by the edge effect, which is a result of management of the vegetation, as well as environments in the initial stages of ecological succession with variations in environmental conditions (e.g., luminosity and humidity) [15, 17], which restrict the availability of resources for birds [64]. This tendency might be reversed with the growth of the vegetation over the ecological succession [2, 6], but these first two years of monitoring are not sufficiently conclusive to determine the importance of HD in the maintenance of bird diversity.

Articles about the persistence of birds in restored landscapes demonstrate alteration in the community over the ecological succession [14, 56]. This tendency is reinforced by the accumulation curves obtained in this paper, showing that the process of succession and arrival of new colonizers is in full swing. It is important to emphasize that in the dynamics of colonization, in some situations the bird assemblage converges rapidly to a structure similar to that of nearby forest fragments [65]. On other occasions the structure of the avian fauna may take different directions, with species adapted to the structure of the vegetation [12, 66], generally influenced by the distance between forest fragments or by the change in the structure of the vegetation [23, 67].

IndVal analysis demonstrated the preference of nine species for the NC treatment, which when added to the exclusive species, totaled 17 birds (29.3%) that have preferences for different nucleation strategies and management intensities. However, generalizations should be avoided, since some species had few records, while others had a low value of association.

We recorded Ochre-collared Piculet and Yellow-browed Tyrant exclusive to PR. Despite occupying distinct niches, they are predominantly insectivores, a guild shared with Striped Cuckoo [68, 69], which was associated with PR in the IndVal test. Although insectivore richness was the same between NC and PR, the specific composition was different.

Exclusive to HD, Burrowing Owl, Slaty-breasted Wood-rail, and Pauraque have a high tolerance to habitat disturbances and forage in open environments. Burrowing Owl is carnivorous and insectivorous, while Pauraque is insectivorous in open areas. Both are nocturnal [70, 71]. Although the richness of forest species is also similar among the treatments, these species have low sensitivity to habitat disturbance [72] and are probably not good indicators in early succession.

The analysis of the frequency of occurrence demonstrated variations in the category of frequent species (between 40 and 60% of the samples), with a higher quantity in NC. This observation suggests that the habitat supplies resources in a more constant way [73, 74], conferring an important role in the local maintenance of the avifauna.

NC responded sufficiently well to richness, abundance, and diversity. The intermediate vegetation management in NC can be an important variable to be measured in later studies, because it causes less intense disturbance than HD. Increased functional heterogeneity in NC is possibly promoted through modification of the spatial and temporal variability of the resources [75–76]. On the other hand, increasing the structural complexity of the habitat, as well as the spatial arrangement of the techniques used in the NC restoration, allows a rapid increase in available niches, which is controlled by the availability of structural resources [77–78]. Thus, the use of perches, tree islets, and shelters for the fauna, and the recurring exposure of part of the soil, can bring about a larger number of environmental niches, and therefore, more species could benefit, discretely increasing richness and diversity in habitats restored under nucleation. However, monitoring of the three treatments should be continued, to evaluate whether with time, the tendency verified in NC will be maintained or the other methods increase in their ecosystemic value [21, 79].

Implications for conservation

Nucleation techniques presented higher richness, abundance and diversity of birds than the passive and plantation techniques during the first two years of ecological succession. We suggest that both restoration methods responded favorably to the habitat complexity hypotheses and to the dynamic equilibrium hypothesis, and are the probable mechanisms for increasing richness and controlling the diversity of bird species in ecosystems under ecological restoration. In order to guarantee the largest number of niches for birds in restoration, we recommend a mix of the three technologies tested, combined in the same space and time, since each one presented exclusive species and because some bird species still lack studies about their ecological behavior.