Do Landscape Elements Interfere in Hovenia Dulcis Thunb. Invasions in Subtropical Forest Fragments?

Changes in characteristics of landscape adjacent to forest fragments due to anthropic actions result in increased invasion of non-native species and biodiversity loss. The aim of this study was to evaluate the invasion process of Hovenia dulcis (Thunb.) in forest fragments with different size and shape in order to relate the behavior of the invasion with landscape metrics. As a result, shape and size of the forest fragments do not inuence the abundance of H. dulcis. However, a greater number of H. dulcis individuals are concentrated in the edge areas. The canopy, declivity and land use were not associated with the abundance of H. dulcis. The high abundance at the edge of forest fragments highlights the dynamics of the landscape in the study area that are marked by the agricultural matrix. The invasion process occurs regardless of the size and shape of the forest fragments.

Forest fragments in the southern region of Brazil show an increase in the abundance of Hovenia dulcis Thunb. (Rhamnaceae), a non-native tree species. The species was introduced on rural properties for use as wood, shading, energy production and wind breaks (Carvalho 1994). But actually, any propagation form of H. dulcis (among other species) is prohibited by law. The invasion by H. dulcis is associated with early-stage forest fragments, which are more susceptible due greater canopy opening, less diversity and Studies carried out about pine invasion in South Africa highlight that landscape elements such as the size and shape of forest fragments are less explored, and therefore make it di cult to determine the landscape in uence on the invasion of non-native species (Vilá and Ibánez 2011). In turn, Padilha et al.
(2015) observed that the fragment size had little effect on the abundance of H. dulcis, but the occurrence of the species was associated with fragments with high light. In addition, a higher abundance of H. dulcis was associated with the proximity to the source of propagules (adult individuals) (Padilha et al. 2015).
Habitat invasion can also be determined by the spatial context (Fons 2014)

Materials And Methods
The study area comprises 16 forest fragments located in the South of Brazil ( Figure 1). The economy of the study area is mainly centered on agriculture with the cultivation of soy, corn and wheat, and on The forest fragments were selected from satellite images based on the following criterion: i) Location of the fragment within the topographic limits of study area; ii) Presence of Hovenia dulcis individuals; iii) Fragment size with two area classes, 10 ---100 ha and > 100 ha, with eight fragments selected for each area class; iv) fragment shape index with two classes (≤ 2 and> 2) according to the analysis obtained via Fragstats (forest fragmentation metrics) with four fragments selected, two of which were selected for the area class for each class shape (Supplementary Material Table SI).
The form of a fragment is obtained by Shape, from which a lower value indicates fragments of simple formats, meaning those which are bene cial for conservation. Values close to 1 indicate fragments with shapes that resemble a square, and as the value increases, it indicates fragments with elongated shapes (Oliveira Rodrigues 2011). A radius distance of 50 meters was arbitrarily established to calculate metrics related to the shape, edge and size using the methodology proposed by Murcia (1995); Young and Mitchell (1994); Metzger (2009). The shape and size (Hectares) analysis of the forest fragments used for the study was performed using the FRAGSTATS v4.2.1 program.
After selection with the aid of the MapInfo 8.5 and Idrisi Selva 17.0 software designed in Geographic Coordinate System (GMS) and Datum WGS 84, the selected fragments were veri ed in the eld to verify if they met the established criterion. Thus, 10 plots with a minimum distance of 30 meters were established for each fragment, with 5 plots in the edge area and 5 plots in the interior area of the fragments with a size of 5 x 5m (25m²). A distance of 50 meters from the beginning of the fragment's edge was stipulated for the interior area, as adopted by Murcia (1995).
Data collection H. dulcis individuals classi ed were according to the shoot height in three different classes. Class I corresponds to adult individuals taller than 2.5 meters; Class II corresponds to young individuals between 0.5 to 2.5 meters in height; and Class III to individuals classi ed as seedlings with height < 0.5 meters. All H. dulcis individuals in the different classes counted were in all sampled plots according to the criteria above.
Landscape elements were collected in the plots such as open canopy and slope, and then the forest fragment use and coverage of the surrounding land (500 meters) was later classi ed through Geographic Information Systems.
Canopy opening was assessed using tree hemispheric photographs were obtained in the center of each plot in a perpendicular shot (90º) using a Sony DSC30 high resolution digital camera, coupled to a 180º Rainox sheye lens in order to assess the canopy discontinuity. The images were subsequently analyzed using a Gap Light Analyzer (GLA) program (Frazer et al. 1999), determining the canopy discontinuity percentage (%). Slope data processed were using MapInfo Professional 8.5 and Idrisi Selva 17.0 software, generating a Numerical Terrain Model (NTM) for the entire study region and then selecting the average slope value for the edge and interior areas of the evaluated forest fragments.

Data analysis
The characteristics of the selected fragments (size, shape and the edge/interior area), together with the abundance of H. dulcis (class I, II and III) were analyzed using Split-plot paired measures analysis of variance (ANOVA) to assess whether the invasiveness of H. dulcis is associated with small fragments (shape > 2) and with edge areas. Furthermore, linear regressions performed were to assess whether landscape elements (canopy opening and slope) in uence H. dulcis invasion. It is noteworthy that only adult H. dulcis individuals used were for the canopy opening analysis, as young individuals and seedlings were not found in all evaluated plots.
All tests performed were in the R statistical program using the vegan package (R Core Tean 2018; Oksanem et al. 2018).

Landscape metrics
A total of 1,708 H. dulcis individuals were sampled distributed among the 16 fragments. From this total, 299 individuals were adults (class 1), while 1,409 were young individuals and seedlings (Class II and III). Of these, 1,438 H. dulcis individuals are located in the edge areas (84.20%), and 270 individuals in the interior areas of the forest fragments (15.80%).
The relative abundance of adult H. dulcis individuals (Class I) was not in uenced by size or shape of the forest fragments ( Table 1). The edge and interior areas of the fragments differ in the relative abundance of class I individuals, with higher abundance observed in the edge area of the forest fragments (Table 1 and Figure 2A).
The relative abundance for young H. dulcis individuals (Class II) and seedlings (Class III) does not differ in relation to size or shape of the forest fragment (Table 1). However, a difference was observed in the relative abundance of both class (II and III) between the edge and interior area of the fragments, being higher in the edge ( Figure 2B and C). Table 1 Split-plot analysis of variance (ANOVA) results of Relative abundance of adult, young and seedlings Hovenia dulcis individuals in forest fragments in relation to size, shape and edge-interior. Class I corresponds to adult individuals taller than 2.5 meters; Class II corresponds to young individuals between 0.5 to 2.5 meters in height; and Class III to individuals classi ed as seedlings with height < 0.5 meters.

Landscape elements
Regarding the analyzed landscape elements, it is observed that the canopy opening and slope do not in uence the relative abundance of H. dulcis for all the evaluated classes (Table 2). Class I corresponds to adult individuals taller than 2.5 meters; Class II corresponds to young individuals between 0.5 to 2.5 meters in height; and Class III to individuals classi ed as seedlings with height < 0.5 meters.

Discussion
The size and shape of the forest fragments were not associated with the abundance of H. dulcis individuals from all the size classes evaluated (class I, II and III), partially refuting the rst hypothesis of the study. However, a greater proportion of individuals occurs at the fragment's edges, regardless of class, size or shape, partially corroborating the rst hypothesis of this study. Fragmentation interference (size and shape of forest fragments) drives the edge effect, resulting in a reduction of the forest through a vulnerability of richness and abundance of adult species and seedlings (Tabanez and Viana  It is important to note that the shape with value 1 is considered a standard shape and equivalent to a square. Therefore, more irregular forest fragments are more susceptible to show a greater edge effect, especially those with small size due to greater interaction with the surroundings ( It is also noteworthy that the high proportion of individuals from young classes and seedlings may suggest success in recruiting H. dulcis on the edges, meaning that the species is nding favorable environmental conditions for its establishment and development in these environments. Young H. dulcis individuals and seedlings (Classes II and III) still do not form fruit and have not reached the canopy, at which time they will obtain su cient light for ower and fruit formation, representing an increase in the species' invasiveness in the forest fragment (Gerber 2018).
Regardless size and shape, the presence of H. dulcis is also observed inside the forest fragments. These results highlight that H. dulcis individuals (mainly adults) are also established in environmental conditions such as the interior forest fragments (more closed canopy). Large and regular forest fragments may be less susceptible to the edge effect and, consequently to biological invasion. In this study, the fragments with shape < 2 can be considered the areas with integral ecological characteristics and priority for biodiversity conservation (Ribeiro et al. 2009 The landscape elements (canopy opening and slope) were not associated with the abundance of H. dulcis, refuting the second hypothesis. The forests fragments in the study region are concentrated in areas that have a higher slope and are therefore unsuitable for large scale agricultural production. The Federal Forest Code (2012) can be associated as one of the factors responsible for the considerable increase in small fragments, mainly observed in sloping areas, dispersed in an agricultural matrix. These conditions could facilitate the establishment of H. dulcis, mainly due the increased edge effect.
The results of this study indicated that the landscape elements were not associated with the pattern of occurrence of H. dulcis. However, the observed results highlight that the greater H. dulcis abundance occurs mainly at the edge of forest fragments, regardless of the individuals' size. This pattern of occurrence can be used as an aid in environmental decisions and highlight the potential of H. dulcis for establishing and growth, regardless of the size and shape of the subtropical forest fragments.

Declarations
This work was carried out with the support of the Coordenação de Aperfeiçoamento de Pessoal de Nível Superios (CAPES) -Financing Code 001.

Con icts of interest/Competing interests:
Not applicable Availability of data and material:

Supplementary Material
Code availability: Not applicable Authors' contribution: The authors Marciana Brandalise, Silvia Vendruscolo Miseli, Tanise Luísa Sausen and Vanderlei Secretti Decian declare to be responsible for producing the manuscript entitled "Do landscape elements interfere in Hovenia dulcis Thunb. invasions in subtropical forest fragments?", all of whom participated in producing the article.
-Marciana Brandalise: Data collection, data analysis, statistical analysis, mapping, writing and review of the article.
-Silvia Vendruscolo Milesi: Statistical analysis, writing and review of the article.
-Tanise Luísa Sausen: Guided all stages and participated in data analysis, writing and review of the article.
-Vanderlei Secretti Decian: Guided all stages and participated in the preparation of maps, data analysis, writing and review of the article.  Figure 2