2.1. Study area

The “Parque Estadual da Serra Furada” (PAESF) is located between Orleans and Grão Pará, two municipalities in the southern part of the Brazilian state of Santa Catarina (Figure 1). Historically, the area now occupied by PAESF experienced unsupervised timber extraction and human disturbance until the creation of the park on June 20, 1980 (Decree No 11.233; ^{Guislon et al., 2017}). Hunting and collection of economically valuable species were widespread and led to the loss of much of the original biodiversity. Subsistence livestock farming and logging were widely practiced before the creation of the park and substantially altered the forest structure (^{Guislon et al., 2017}). The vegetation of the study area presents high diversity of tree species, distributed across a mosaic of heterogeneous habitats, and includes species such as Alchornea triplinervia (Spreng.) M. Arg., Ocotea catharinensis Mez., Actinostemon concolor (Spreng.) Müll.Arg. and Guapira opposita (Vell.) Reitz (^{Guislon et al., 2017}).

Figure 1 “Parque Estadual da Serra Furada” (PAESF) located between Orleans and Grão Pará, in the southern part of Santa Catarina, Brazil. 

The Köppen climate classification for the PAESF region is humid subtropical with no defined dry season and hot summers (Cfa), with average annual temperature varying from 17.0 °C to 19.3 °C at lower altitudes, and balmy summers (Cfb) with average annual temperature ranging from 13.8 °C to 15.8 °C at higher altitudes (^{Alvares et al., 2013}). Average annual rainfall ranges from 1,300 mm to 2,500 mm (^{Alvares et al., 2013}). The PAESF terrain is rugged, characterized by craggy areas and steep valleys with altitudes from 400 to 1480 m. The vegetation in the park is classified as Dense Ombrophilous Montane and High-Montane Forests (^{IBGE, 2012}). In terms of geology, PAESF is characterized by extensive rock areas, such as siltstones, shales and sandstones included in the following lithostratigraphic classifications: São Bento Group, represented by diabase intrusion and by Serra Geral and Botucatu formations; Passa Dois Group, composed of Rio do Rastro, Terezina, Serra Alta and Irati formations; and Guatá Group, with non-glacial sediments and coal layers, composed of the Palermo Formation (^{IBGE, 2012}).

2.2. Data collection

A palm community survey was carried out in Dense Ombrophilous Forest within five 20 m × 100 m sampling units (SU), or plots, totaling 1ha (Figure 2). The first sampling unit (A) was installed five meters from the forest edge in an area with history of intense exploitation by past residents and located close to PAESF headquarters. The second sampling unit (B) was installed within the forest interior on sloping ground with little history of human intervention. The third sampling unit (C) was installed nearby streams and waterfalls and surrounded by rugged terrain. The fourth and fifth sampling units (D and E) were located in selectively logged forest at higher altitudes and far from any water bodies. Sampling units ranged in altitude from 470 to 650 m (^{Guislon et al., 2017}). All palms from seedlings to adults were measured and identified in each plot. Height at the top of leaf crown was measured using a laser measure tape (Bosch GLM 40), and Diameter at Soil Height (DSH) of each palm was measured using a digital caliper (MTX – 150mm).

Figure 2 Sampling Unit (SU) locations within the “Parque Estadual da Serra Furada” (PAESF) boundary, located between municipalities of Orleans and Grão Pará, Santa Catarina, Brazil. 

2.3. Data analysis

To analyze the horizontal and vertical structure of PAESF palm populations, the DSH distribution (Diameter at Soil Height) and height classes separated into 15 size classes based on ^{Sturges (1926)} were examined.

Data were standardized, and non-metric multidimensional ordination methods (NMDS) (^{Clarke, 1993}) and the Bray-Curtis index were used to determine associations and distinct groups among SUs using species abundance as variable. Analysis of similarity (ANOSIM) (^{Clarke, 1993}) was then applied with 999 permutations to determine significance of assigned groups. All palm species were included in these analyses.

To characterize the horizontal structure of each group, as determined by NMDS and ANOSIM, Importance Value (IV) was calculated, which is the sum of the relative frequency, density and dominance values for each species. Finally, the species that most contributed to differences between groups were identified using analysis of similarity (SIMPER) (^{Clarke, 1993}). Statistical analyses were run in the R computing environment (^{R Development Core Team, 2018}).

3.1. Distribution on sampling units

Palms occurred throughout all SUs; however, ordination analysis identified two distinct palm communities, separating SU (A) from all others (B, C, D, and E) (Figure 3). Differences were related to past land use history. Before PAESF was officially demarcated, SU (A) experienced heavy exploitation – mainly the harvest of E. edulis palm heart. Even after demarcation, forest reserves were still plagued by illegal harvesting, a common practice in Santa Catarina (^{Gasper et al., 2011}). For example, ^{Silva & Fisch (2012)} reported illegal palm heart extraction in protected areas of the “Parque Estadual da Serra do Mar” in Ubatuba, along the northern coast of the state of São Paulo.

Figure 3 Multidimensional non-metric scaling for sampled palm species, according to the different Sampling Units (A, B, C, D, and E). Stress: 0.044 (2 dimensions). 

SUs (B, C, D and E) were grouped together despite different intervention histories and distinct environmental contexts. One of the main reasons for this procedure was the considerable distance of each SU from the park boundary and difficult access to these areas. ANOSIM also confirmed the separation of SUs into the same two distinct groups (R=0.80; P=0.001): environments with high intensity exploitation (A) were significantly different from those with low impact exploitation (B, C, D and E).

Relative frequency, density, dominance, and importance values showed no clear differences among the four SUs (B, C, D and E), while absolute palm density in SU (A) was, on average, five times lower than in the other SUs. This contrast reflects land use history, especially the extraction of E. edulis stems for palm heart and the exploitation of other economically valuable species, which altered luminosity and the presence of propagules in the area, reducing the abundance of species typical to these environments, as observed in other forest fragments of Santa Catarina (^{Gasper et al., 2011}).

SIMPER analysis pointed E. edulis as key for distinction among environments, corresponding to 84% of dissimilarity, followed by G. gamiova (14%). In addition, E. edulis presented the highest IVI across all SUs (Table 3). This species dominates all layers of the forest strata and at times, it appears monodominant. ^{Reitz (1974)} also emphasized the importance of E. edulis in the Dense Ombrophilous Forests of Santa Catarina, and our results corroborate ^{Oliveira et al. (2014)}, who identified E. edulis as the most abundant palm species of Ubatuba, state of São Paulo.

Geonoma gamiova was also important to the palm community surveyed here. In total, 1219 individuals of this thin-stemmed species were recorded in nearly all SUs. Geonoma gamiova was spared during PAESF’s logging period, due to its use as a non-timber forest product, such as rope and roofing that crafted from leaves (^{Reitz, 1974}). The species has few environmental restrictions (^{Oliveira et al., 2014}) and is regularly distributed throughout the Dense Ombrophilous Forest understory (^{Henderson, 2011}; ^{Elias et al., 2018}).

Bactris setosa and G. schottiana occurred at low densities with only 74 and 70 individuals, respectively. Both species are selective in terms of habitat conditions. The rugged terrain found within PAESF does not provide ideal establishment conditions for B. setosa, which prefers wet soils, nor for G. schottiana, which develops better at higher altitudes (^{Reitz, 1974}; ^{Elias et al., 2018}).

Syagrus romanzoffiana was the least abundant palm species in PAESF, with only 12 individuals all shorter than 1.2 m. This large arborescent palm can reach 15 m and shows natural regeneration inside the forest, especially in forest gaps. The dispersion of S. romanzoffiana seeds into PAESF originates from only a few seed sources, although mammal feces containing many seeds are frequently found. We believe that with time, S. romanzoffiana will become increasingly common in PAESF, as well as throughout the state of Santa Catarina (^{Elias et al., 2016}).