Assessing forest restoration effectiveness in the Seasonal Semideciduous Forest in the Upper Paraná Atlantic Forest ecoregion using epigaeic ant assemblages

Abstract Itaipu Hydroelectric Power Plant initiated a large reforestation programme after the expropriation of the areas destined for the formation of the reservoir. This study aimed to evaluate the effectiveness of forest restoration of the Seasonal Semideciduous Forest in the Upper Paraná Atlantic Forest ecoregion, Brazil, using epigaeic ant assemblages as bioindicators, by comparing ant species richness and composition in the Reservoir Protection Strip with adjacent areas, such as the primary forest of the Iguaçu National Park and the Permanent Preservation Area located on a rural property and agricultural areas. In total, 171 ant species were identified. Ant species richness was higher in forest than in agricultural areas and did not differ among forest areas. However, ant species composition in forest areas, regardless of the restoration technique used, was not similar to the primary forest, possibly due to variation in forest recovery time. This study highlights the great value of the Iguaçu National Park as a conservation unit. Also, it reveals that the efforts for the creation and maintenance of the Reservoir Protection Strip, which remains without anthropic interventions for years, might indeed lead to a complete recovery of the ant species composition over time, reinforcing their great importance for biodiversity conservation.


Introduction
The last few centuries have been marked by a major conversion of tropical forests into mosaics of habitats altered by human action, driven mainly by the human population growth and socioeconomic pressures (Gascon et al. 2002).The Atlantic Forest is the second largest tropical rain forest on the American continent, which originally stretched continuously along the coast of 17 Brazilian states, extending into the east of Paraguay and northeastern Argentina in its southern portion.Considered a global centre of endemism, the Atlantic Forest of South America is among the most diverse tropical forests.However, after centuries of exploitation, this forest has lost more than 93% of its area, placing it among the world's highest priorities for conservation (Galindo-Leal & Câmara 2003, Myers et al. 2000).Forest loss and degradation have led to widespread biodiversity loss (Solar et al. 2016), which generate concern and awareness regarding the need of natural resource conservation (Malhi et al. 2014).
The Upper Paraná Atlantic Forest ecoregion is the largest among the 15 ecoregions identified in the Atlantic Forest biome, with the Seasonal Semideciduous Forest as the dominant vegetation type (Di Bitetti et al. 2003).The Iguaçu National Park stands out as the largest integral protection conservation unit of the Upper Paraná Atlantic Forest ecoregion, located in the western region of the State of Paraná, in southern Brazil (Di Bitetti et al. 2003), adjacent to the Iguazú National Park, in Argentina.This region includes the Itaipu Binacional Hydroelectric Power Plant, on the Paraná River, at the border between Brazil and Paraguay.The dam was built by the two countries between 1975 and 1982.Since 1979, after the expropriation of the areas destined for the formation of the reservoir, Itaipu developed the largest reforestation programme ever conducted by a hydroelectric power plant in the world.Today, more than 99% of the Permanent Preservation Areas, which currently consist of the so-called Reservoir Protection Strip, are restored (Itaipu 2015).
Forest restoration is a strategy to reverse forest loss and degradation, recovering the same or very close conditions of the original forest.It is mainly carried out through active revegetation, natural regeneration or mixed techniques, which can be carried out by planting seedlings of native and/or exotic species, natural regeneration, assisted natural regeneration, or even the establishment of agroforestry systems (Stanturf et al. 2014a, b).And, environmental indicators can provide useful information for monitoring management practices, aiming to rehabilitate degraded ecosystems.The use of bioindicators can provide evidence of the development level of an environment in different stages of reconstitution, which can be evaluated by the structure of certain species and/or the composition of species present in each environment (Majer 1983, Ribas et al. 2012).
Ants are ecologically important components of natural and disturbed ecosystems, providing a variety of ecological functions in almost all trophic levels, given their diversity and behavioural plasticity in nesting habits, feeding spectrum, and association with numerous species of plants and animals (Elizalde et al. 2020, Folgarait 1998).Several studies documented how ant community composition is influenced by plant diversity and vegetation physiognomy and therefore can serve as an indicator taxon for the invertebrate fauna, as well as soil conditions (e.g., Costa-Milanez et al. 2014, Segat et al. 2017and Solar et al. 2016).Ant assemblages can change over the course of plant succession.That is, ant diversity tends to increase with increasing diversity of plant communities, the availability of micro-habitats, and consequently a greater availability of food and shelter resources (Majer 1983, Mauda et al. 2018, Solar et al. 2016).In particular, epigaeic ants are the most sensitive ant assemblages to forest recovery (Schmidt et al. 2013), and they were more efficiently sampled with pitfall traps (a rather simple and cheap survey method) than other sampling methods, such as Winkler (Donoso, 2017, Parr & Chown 2001).
Epigaeic ant assemblages have not been studied to date in the Iguaçu National Park, and in the Reservoir Protection Strips of the Itaipu, on the Brazilian side, even though they are considered important species source sites for biodiversity conservation.This study aimed to evaluate the effectiveness of forest restoration of the Seasonal Semideciduous Forest in the Upper Paraná Atlantic Forest ecoregion, Brazil, using epigaeic ant assemblages, comparing ant species richness and composition in the Reservoir Protection Strips of the Itaipu, with adjacent areas, such as the primary forest of the Iguaçu National Park, and disturbed areas, such as Permanent Preservation Areas located on a rural property and agricultural areas.Our hypothesis is that forest restoration techniques used in the Reservoir Protection Strips, which remain without anthropic interventions for approximately 35 years, were effective and ant richness and composition may be more similar to primary forest than highly disturbed areas.

Study area
The research was carried out in the western portion of the State of Paraná, Brazil, in the municipalities of Foz do Iguaçu, PR, Santa Terezinha de Itaipu, PR, and São Miguel do Iguaçu, PR (Figure 1).According to the global classification of Köppen, the climate is of the Cfa type, characterised as humid subtropical temperate, with well-defined winter and summer seasons, where rainfall is equally distributed throughout the year.The local temperature varies between maximum 40º C and minimum 3º C, with an annual maximum average of 26º C and minimum of 15º C. The average annual rainfall is 1,700 mm, with a predominantly high relative humidity, rarely below 80%, even in the driest periods of the year.The average altitude is 192 m, and the predominant vegetation is the Seasonal Semideciduous Forest (Ibama 1999).
At the time of the reservoir formation, a study carried out by Itaipu in the Brazilian territory revealed that 23% of secondary forests, 24.7% of highly exploited forests undergoing natural rehabilitation and 50.3% of agricultural areas were restored.Itaipu also made possible the establishment of the Santa Maria Ecological Corridor, which connects the riparian forests of the reservoir and other protected areas, with the Iguaçu National Park, extending through the Private Reserve of Natural Heritage of the Santa Maria Farm.This corridor allowed the connection between two of the most extensive conservation areas in Southern Brazil, the Iguaçu and Ilha Grande National Parks, maintaining genetic exchange among different populations (Itaipu 2015).
The Reservoir Protection Strips have been without human intervention for 35 years.Secondary forest located on a rural property until now suffers from anthropogenic disturbances.However, all forest areas selected for the study are surrounded by agricultural areas.

Sampling ants
We established 12 transects of 125 metres in each area, separated by at least 500 m.In total, we sampled 72 transects (sampling units).We sampled ants in three transects per season of 2017, using epigaeic pitfall traps.Pitfall traps consisted of plastic recipients (8 cm diameter and 11 cm height) half filled with 300 ml of liquid solution of water, glycerol (5%), and salt (5%), which collected and killed the ants.Traps were buried so that they were flush with the soil and remained in the field for 48 h (Martins et al. 2021, Schmidt et al. 2013, Solar et al. 2016).Along each transect, we installed five epigaeic pitfall traps spaced by 25 m.The first trap was placed 25 m away from the edge in each area.A pitfall trap was also placed in each transect in the transition between forest areas and agriculture, totalling 420 epigaeic pitfall traps.
We sorted, mounted and identified the ants to genus level using available taxonomic key in Baccaro et al. (2015).For species identification, comparisons were made with material deposited at the Padre Jesus Santiago Moure Entomological Collection (DZUP)
in the Universidade Federal do Paraná, Curitiba, PR, Brazil, where vouchers were deposited.We also applied species names using updated taxonomic revisions for some of the ant genera recorded here.Taxonomic sources for species identification included Acromyrmex Morphospecies were assigned number codes that apply only to this study.Ant collections were authorised by the licence 55313-1 (Brazilian Biodiversity Information and Authorization System -SISBIO).Access to the genetic heritage was also registered by the record number ACDFB38 (National Management System of the Genetic Heritage -SisGen).

Environmental variables
Vegetation cover was used as an estimate of resources available for epigaeic ants.The percentage of vegetation cover was measured using digital images taken with a camera at the height of 1.3 m, with the lens facing up and next to where the epigaeic trap was placed.We also recorded richness and composition of plant species in all forest strata.To determine the number of plant species, 10 x 20 m plots were delimited in five transects from each forest area for the identification and measurement of tree and shrub individuals taller than 1.0 m and/or with more than 3 cm in circumference at breast height, measured at 1.30 m above ground.The plant species identification was carried out with the help of The Plant List platform (version 1.1, 2013).

Statistical analyses
Ant species accumulation curves and extrapolated sample-based rarefaction curves were obtained considering presence/absence data and incidence data in each sampling unit, reducing a potential bias caused by rarely sampled species (Chao et al. 2014).To assess whether ant species richness differs among treatments, a one-way analysis of variance (ANOVA) was performed, considering the data normality that was analysed by the Shapiro-Wilk test.In addition, a non-metric multidimensional scaling (NMDS) analysis was performed to test for changes in ant species composition among areas.The ordering was based on a species presenceabsence data matrix and Jaccard's dissimilarity index.After a visual inspection of the data, a PERMANOVA was performed to test for significant clustering of the areas.In order to test the significance of possible differences in ant species composition among areas, the distribution of similarities between each pairwise area was also compared through an ANOSIM similarity analysis.By ANOSIM similarity analysis, high R-values mean high dissimilarities among the areas.
The value of each ant species as an indicator of each area was calculated using the Indicator Value (IndVal) method of Dufrêne and Legendre (1997).This index combines a measure of the habitat specificity of a species to a level of disturbance, or a disturbance state, with its fidelity within that state.Species with high specificity and fidelity within an area will have a high indicator value.Significance was tested using the Monte Carlo test with 10,000 permutations.
To assess plant species richness and the percentage of forest cover among areas, a one-way ANOVA and Shapiro-Wilk test were performed.And the NMDS analysis was also performed to assess differences in tree composition.To analyse plant species composition, we selected only species with more than 10 cm in diameter at breast height.
Ants collected in the transition between forest areas and agriculture were not included in the analyses, as there was only one pitfall in the transition per transect, but they were included in the taxonomic list.Also, agricultural areas were not included in the analysis of the environmental variables, since they were either soybean or corn monocultures.All analyses were performed using the R software environment (R Core Team 2019), except interpolation and extrapolation curves performed in the iNEXT online (Chao et al. 2016).
The secondary forest located on a rural property accumulated the largest number of species (94), followed by the primary forest (87), but ant species richness did not vary within forest areas (Figure 2, F 5,66 = 13.74,p < 0.001).However, ant species richness was significantly higher in forest areas, irrespective of the rehabilitation technique used, than in agricultural areas.The interpolation and extrapolation ant species accumulation curves also showed that the agricultural area accumulated the lowest ant species (Figure 3).In the transition between forest areas and agriculture, 86 ant species were sampled in total, with 9 being sampled only in the transition.(Appendix 1).
Ant species composition also varied among areas.NMDS revealed that ant species composition differed in forest and agricultural areas (Figure 4, PERMANOVA F 5,66 = 6.63, p < 0.001).ANOSIM revealed greater similarity within forest  areas than agricultural areas.However, there are differences in ant species composition evenly among forest areas (Table 1).For all combinations, the R values were significant, but there was a greater similarity in ant species composition between the secondary forest and the natural regeneration area.In addition, there was a greater similarity between the natural regeneration and the reforestation area (Table 1).
The IndVal results showed a total of 11 significant indicator species for the primary forest, 1 species for the secondary forest, 2 species for the reforestation area, 2 species for the secondary forest located on a rural property, and 4 species for the agricultural area (Table 2).

Environmental variables
We sampled 110 plant species belonging to 82 genera and 34 families.Of these, 45 species were arboreal with more than 10 cm diameter at breast height (Appendix 2).

Discussion
Through monitoring ant communities in regenerating forested areas, it is possible to evaluate the methodologies and the effectiveness of revegetation techniques in maintaining local diversity and, consequently, the self-sustainability of these environments (Pereira et al. 2007).Ant species richness did not differ among forested areas in our study.The structural complexity of the habitat is one of the main factors driving species richness and composition on a regional scale.The main factors that contribute to the high diversity of ants in forested areas are forest cover, the diversity of nesting sites, the amount of food available, the foraging area and interspecific competition (Braga et al. 2010, Coelho et al. 2009, Mauda et al. 2018, Ribas et al. 2011, Solar et al. 2016).A minimum of established forest conditions is sufficient for ants to colonise a forest regardless of its age, causing species richness to be similar among the forest remnants with different recovery times, although species composition could differ (Schmidt et al. 2013).
Ant species richness in forest areas is significantly greater than in agricultural areas.The lower species richness in agricultural areas is expected since agricultural areas have less structural diversity, a fact observed in several studies (e.g., Falcão et al. 2015, Martello et al. 2018, Solar et al. 2016).Deforestation, mainly for agricultural development, pastures or forest monoculture plantations, is widely recognised as the most serious anthropogenic threat to terrestrial biodiversity (Sala et al. 2000).Degraded environments or environments with low plant diversity (e.g., monocultures)   present limitations to the organisms due to the lack of resources provided by these environments.Thus, in these areas, ant communities may present low species diversity (Pereira et al. 2007).In these conditions, most ant species are usually generalist species that can nest in several nest sites and use varied food sources, such as some Pheidole and Solenopsis species found in the present study (Braga et al. 2010, Diehl et al. 2004).In addition, the lower ant diversity in agricultural areas can be attributed to the application of pesticides for the control of pests, diseases and weeds, the level of soil compactness caused by human intervention in these areas, the absence or low amount of litter, which possibly decreases the quantity and quality of available resources, and the exposure of species to high thermal amplitude (Dias et al. 2008, Lapola & Fowler 2008).Ant species composition is considered the best indicator for assessing changes in habitat quality, as it changes according to variation in land use classes, from undisturbed primary forest to highly disturbed areas, such as intensive agriculture (Majer & 2016).In the present study, ant species composition in the agricultural area is different from ant species composition in the forest areas.However, among forest areas, ant species composition also differed.
After approximately 35 years of forest recovery, ant species composition in the permanent preservation areas that suffered disturbances differs significantly from the composition of the Iguaçu Nacional Park Primary Forest.Studies show after 13 (Falcão et al. 2015) or 25 years (Silva et al. 2007) of forest regeneration, the number of species and the composition profile between the primary forest and the disturbed areas still show substantial differences.In tropical forests, the complete recovery of ant species richness is estimated to occur 39 years after land abandonment.However, the recovery of species composition appears to take substantially longer than the recovery of species richness (Dunn 2004), as observed in the present study.Estimates indicate a time frame for recovery of 50 to several hundred years for a complete recovery in ant species composition in secondary forests (Bihn et al. 2008), since in the present study, we also observed arboreal plant composition in areas under recovery also still differ from the primary forest.
There was a greater similarity in ant species composition between the secondary forest and natural regeneration area.As the level of disturbance increased, the dissimilarities in the composition of ant species also increased compared to the secondary forest area.For example, the secondary forest located on a rural property, which still suffers from anthropogenic disturbances, such as the presence of cattle, presents greater dissimilarity in the composition of ant species than the reforestation area, which was predominantly reforested with a mix of native species of the Mata Atlantica biome and which has been without human intervention for at least 35 years.Ant communities in secondary forests might recover more quickly in areas where the forest was less disturbed at the beginning of the succession than when it was established on former pasture (e.g., Bihn et al. 2008).The reforestation was an agricultural area at the beginning of the reservoir's formation, so it had to be revegetated.But currently, this area is structurally similar to other forest areas in the recovery process, so there were no differences in plant species richness and percentage of plant cover due to the arboreal size of the planted species.These factors contributed to ant species richness did not differ among forest areas, despite the percentage of vegetation cover being lower in the forest area that was naturally regenerated.Similar results were  observed in a 28-year unmanaged eucalyptus plantation, where ant species richness was similar to native forests, yet ant species composition was more similar to managed plantations (Martello et al. 2018).
This is the first study on epigaeic ant assemblages in forest and agricultural areas in the region between the Iguaçu National Park and the Reservoir Protection Strip of the Itaipu, on the Brazilian side.Pheidole gigaflavens Wilson (2003) and Pheidole mosenopsis Wilson (2003) (Guénard et al. 2017, Janicki et al. 2016).Furthermore, three species are new to science (Pheidole sp.n. 1, 2 and 3, ACF unpublished data).
Undisturbed primary forests have a unique ant species composition, are a key driver of species richness at the landscape scale, and may be an important species source for biodiversity conservation at local and regional scales (Pacheco et al. 2013, Solar et al. 2016).In the Iguaçu Nacional Park, 87 ant species were sampled through pitfall traps in the present study.On the Argentine side, another study revealed 172 ant species in the Iguazú National Park, sampled with different methods from 1998 to 2011 (Hanisch et al. 2015).Our study adds 18 new records, for a total of 191 ant species to this important protection conservation unit of the Upper Paraná Atlantic Forest ecoregion.
The most significant number of indicator ant species were detected in the primary forest: five species of Pheidole, two species of Camponotus and Linepithema each, and one species of Acromyrmex and Holcoponera.Of these, Holcoponera striatula (Mayr 1884) has already been recorded as an indicator of primary forests in studies carried out in the Amazon forest (Solar et al. 2016, Vasconcelos & Vilhena 2006).For the other species, there are no records as indicator species, but it is worth mentioning that this is the first study using ants as bioindicators in this region.The species Pheidole radoszkowskii (Mayr 1884) is considered an indicator of the secondary forest.This species has already been recorded as an indicator of habitats with discreet human impact on the southeast coast of Bahia, Brazil (Delabie et al. 2006).In the natural regeneration area, no indicator species was determined, which can be explained by the greater similarity of this area with the secondary forest and the area that was reforested.In the reforestation, A. subterraneus and P. leonina were the indicator species.Acromyrmex subterraneus has already been considered an indicator in fragments of Atlantic Forest in São Paulo, Brazil (Lapola & Fowler 2008) and also in areas of eucalyptus reforestation (Marinho et al. 2002).Pheidole leonina was also recorded in areas of the Amazon Forest in Acre, Brazil (Oliveira et al. 2009).In the secondary forest located on a rural property, W. auropunctata and P. sigillata are considered indicators of this area.Wasmannia auropunctata is considered an indicator of production areas, such as pastures (Solar et al. 2016), but this species has also been reported as an indicator of areas with an intermediate stage of succession in the Vale do Rio Doce region, Minas Gerais, Brazil (Coelho et al. 2009).Pheidole sigillata has also been recorded in areas of urban parks and fragments of the Atlantic Forest in the state of São Paulo (Fernandes et al. 2018).In the AGR, Solenopsis sp. 5, D. brunneus, S. invicta and P. cornicula are considered indicators of this environment.Dorymyrmex brunneus and Pheidole sp. have already been considered indicators of agricultural areas (Pacheco et al. 2013).Dorymyrmex brunneus is also commonly found in urban environments (Farneda et al. 2007).Solenopsis invicta is common in disturbed or undisturbed areas, presenting enormous adaptive plasticity and possibly having the genic potential to exploit a great variety of environments (Diehl et al. 2004).In general, Pheidole also is a generalist genus, having species characteristic of forest environments or open environments (Braga et al. 2010), as observed in the present study.
In the present study, the effectiveness of the rehabilitation techniques used in the Reservoir Protection Strip of the Itaipu was evaluated, showing there were no differences in ant species richness among forest areas in the process of recovery and the primary forest of the Iguaçu National Park.It demonstrates the great importance of the Reservoir Protection Strip and the Permanent Preservation Areas in rural properties to conserve biodiversity.The discovery of three ant species new to science sampled in these areas reinforces these environments' value.However, regardless of the rehabilitation technique used, ant species composition in disturbed forest areas is not similar to the primary forest due to the short forest recovery time.This highlights the great value of the Iguaçu Nacional Park as a conservation unit within the Atlantic Forest biome, given that they represent unique biodiversity in this region.Also, the present study revealed that the efforts for the formation and maintenance of the protection strips around the Itaipu reservoir, which remain without anthropic interventions for several years, might indeed lead to a complete recovery of the composition of ant species over time, reinforcing their great importance for biodiversity conservation.

Figure 2 .
Figure 2. Average ant species richness per transect in each area.Different letters indicate significant differences among treatments (p < 0.001).Bars are standard errors.

Figure 3 .
Figure 3. Interpolation and extrapolation ant species accumulation curves per area, with 95% confidence intervals.

Figure 5 .
Figure 5. Average plant species richness per transect in each area.Treatments underneath the same letter are not statistically different (p > 0.05).Bars are standard errors.

Figure 7 .
Figure 7. Vegetation cover percentage per transect in each area.Different letters indicate significant differences among treatments (p < 0.001).Bars are standard errors.

Table 1 .
Dissimilarity (R-values)among the areas obtained by ANOSIM analysis for ant species composition.High R-values mean high dissimilarities among the areas

Table 2 .
Ant species indicators of each area according IndVal analyses.Only significant indicator species are shown IndVal: indicator value.pprobability, resulting of the permutation test.