Novel land uses shape meta-community structures in neighbouring native forests: Dataset across Uruguay

The presented datasets relate to the research article entitled “Native forest meta-community structures in Uruguay shaped by novel land use types in their surroundings” [Ramírez and Säumel; Ecology and Evolution, 2022]. The datasets include field survey data on woody species presence and absence from 384 plots at 32 permanent monitoring sites of native forests across the Oriental Republic of Uruguay (South America). We compiled different methods from meta-community studies, remote sensing and landscape ecology to explore how woody species communities are influenced by land use change from local to regional scale. We describe the diverse woody species composition in native forests across Uruguay and structure of metacommunities of woody species. Data on woody species diversity inform landscape planning, land-use management, policy and governance and can be used for further meta-analysis with other local, regional or global data sets.


Specifications
Ecology Specific subject area Ecology of Metacommunities; Remote Sensing; Species composition analysis; Land cover change Type of data Table; Image; Chart; Graph; Figure How the data were acquired Identification and mapping of woody species during two fieldwork campaigns (from December 2015 to April 2016 and from October 2016 to January 2017) across 32 permanent monitoring sites inside native forest patches of Uruguay; Classification of species occurrence in size/age classes by measurement of dbh, presence based on forest type. Calculation of absolute frequency, relative frequency and cumulative relative frequency of species, elements of Meta-community structure (coherence, turnover and Morisita overlap index) using Matlab [1] , distance between sites using ArcGis v.10.3.1 for Desktop [2] , Jaccard Index (J) using Past 3.16 [3] . Calculation of landscape metrics using Fragstat v.4 [4] : at Landscape scale: number of patches; Landscape shape index, Shannon's evenness index, Aggregation Index; at land use type level: Percentage of the landscape occupied by each land use type; Number of native forest patches within the landscape; Interspersion and juxtaposition index of native forest; Euclidean nearest neighbor distance of native forest; at native forest patch level: Total area of the native forest patch; Perimeter area ratio of native forest patch, Shape index of the native forest patch in a buffer of 3 km from central point of permanent monitoring site Data format Raw and analyzed data Description of data collection We surveyed woody species diversity at 32 plots of native forests across Uruguay (South America

Value of the Data
• The dataset provides relevant information about the main effects of land use change from extensively used grassland to intensively used Eucalyptus plantation and agricultural crops on composition of woody species in neighbouring native forests. • Data on meta-community structure and diversity of woody species are the base to describe the state of the art of the different native forest types and to evaluate how land-use change impacts on these forests. • Insights from the interactions and influences between meta-community patterns and landuse change inform actors involved in territorial planning, land-use management, policy and governance. • Data can be used for example for meta-analysis on land-use change impacts on woody species communities with other data sets regarding changes of woody species diversity and land-use change.

Data Description
The data described in this article show woody species presence and absence, absolute frequency, relative frequency and cumulative relative frequency of species, elements of metacommunity structure (coherence, turnover and Morisita overlap index) from 384 plots at 32 permanent monitoring sites of native forests across Uruguay. Native forests cover around 6% of the country's total surface area [5] . Table 1 shows the absolute, relative and cumulative frequency and traits of woody species recorded at 32 permanent monitoring sites across Uruguay. Species are ordered according to absolute frequency (AF) of all species (Total) from higher to lowest values. Table 1 Frequency and traits of woody species recorded at 32 permanent plots across Uruguay (Ramirez and Säumel 2022). AF: absolute frequency, RF: relative frequency (%) and CRF: cumulative relative frequency. Species are ordered according to absolute frequency (AF) of Total from higher to lowest values. * shrub, + mistletoe and °liana.
We recorded adults of thirteen native species without any presence of juvenile individuals, among them Butia odorata , which is categorized as high priority for conservation ( Table 1 ). All occur with low frequency, except the hemiparasitic mistletoe Tripodanthus acutifolius .
Of the species, 26 were recorded only in the regeneration layer but not among adults. All are native to the region, except the South-East Asian Melia azedarach, the Chinese Poncirus trifoliata and the European Pyracantha coccinea (  Table 1 ). Fig. 1 shows a scheme of idealized pattern of species distribution (checkboard, random, clementsian, gleasonian, evenly-spaced, nested clumped, nested random, nested evenly-spaced and QS or quasi-structures) where columns represent sites and rows represent species, gray square mean specie presence and white mean specie absence (based on [7 , 8] ). Species distribution among sites can follow a discrete or Clementsian pattern [9] , a continuous or Glesonian pattern [10] , a random pattern [11] , a checkboard pattern [12] , evenly-spaced patterns [13] , nested subset [14] , and mixed pattern between nested-random or nested evenly-spaced [8] . Steps to determine the pattern of species distribution are shown: (1) observation of species coherence, (2) evaluation of species turnover and (3) analysis of boundary clumping using Morisita overlap index. NS = non-significant, "+ " = significantly positive, "-" = significantly negative. Fig. 2 shows the distribution of native forests and the 32 permanent monitoring sites in different native forests across Uruguay, South America. photographs show a riverine forest surrounded by grassland and timber plantation (b), a hill forest (c) and an example for park forests (d). Moreover, Fig. 2 e shows the non-metric multi-dimensional scaling (NDMS) using Jaccard distance between native forest.
In Figs. 3-6 we show the matrix ordination by reciprocal averaging of the different frequencies of juveniles, adults and of both age classes together (juveniles and adults) for riverine (23), hill (7) forests and for all (32) native fragments under study (rows) and species recorded (columns). Black cells indicate presence and white cells absence. Table 2 shows the linear distance matrix between permanent plots (or sites) in kilometer (below of diagonal) and Morisita index (above of diagonal). ID number represents each forest Scheme of idealized pattern of species distribution (checkboard, random, clementsian, gleasonian, evenly-spaced, nested clumped, nested random, nested evenly-spaced and QS or quasi-structures) where columns represent sites and rows represent species, gray square mean specie presence and white mean specie absence (based on [7 , 8] ). Steps to determine the pattern of species distribution are (1) observation of species coherence, (2) evaluation of species turnover and (3) analysis of boundary clumping using Morisita overlap index. NS = non-significant, "+ " = significantly positive, "-" = significantly negative.
fragments. The matrix was ordered according linear distance between 1 and 32 sites (see column 1) and Fig. 2 . Table 3 shows the scores of first axis of ordination generated by reciprocal averaging. Table 4 shows the department and sites of each permanent plots (ID) across Uruguay and overview on metadata of landscape metric per site. ID = code native forest fragments ( Fig. 1 ). Means are given at landscape scale for number of patches (NP): number of patches (NP); Landscape shape index (LSI), Shannon's evenness index (SHEI), Aggregation Index (AI) and at land use type level: Percentage of the landscape (P) occupied by each land use type (NF: native forest; GL: Grassland; TF: Timber forest; C: crops); Number of native forest patches within the landscape (NNF); Interspersion and juxtaposition index of native forest (IJI); Euclidean nearest neighbor distance of native forest (ENN) in a buffer of 3 km from central point of permanent monitoring site.
Raw Data are uploaded in the Open Access repository of the Humboldt Universität zu Berlin ( https://edoc.hu-berlin.de ) as Säumel, I. and Ramírez, L. 2021: Land use change impacts on metacommunity structures in Uruguayan native forests.

Experimental Design, Materials and Methods
Woody diversity datasets were obtained from two fieldwork campaigns (from December 2015 to April 2016 and from October 2016 to January 2017) across 32 permanent monitoring sites inside native forest patches of Uruguay ( Fig. 2 ). In general, we used a stratified randomized design. In a first step, we randomly selected monitoring sites across the country and then stratified by different land use types (i.e. native forests, grassland, timber plantation, crops). Second, we asked the potential land owners for their willingness to establish long term monitoring sites. In total we established 32 long-term monitoring plots (100 × 100 m) in different native forests fragments across Uruguay (23 sites with riverine and seven hill forests; Fig. 2 ).
In the vegetations periods 2015/2016 and 2016/2017, we recorded all woody species in two size-classes based on diameter at breast height (dbh). We take the size-classes as a non-invasive proxy measure for tree age to differentiate in adults (dbh ≥ 5 cm) recorded in 3 plots of 10 × 20 m and juveniles (dbh < 5 cm) recorded in 9 plots of 3 × 3 m. Thus, juvenile plots were nested within adults. The woody species in the local forests comprise also multi-stem species, that there are 64 species categorized as trees, 21 as shrubs and 32 that form the growth habit as shrubs or trees depending on site conditions ( e.g. Blepharocalyx salicifolius, Eugenia uniflora or   Maytenus ilicifolia ). Classification in shrubs, trees and those species that can have both growth habits are indicated in Table 1 . All names of species identified were updated using the online database from [15] .
The meta-community structure was described by different elements of meta-community structure (EMS; Fig. 1 ; [7 , 8] ): coherence (i.e. number of interruptions in species distribution across the sites), species turnover (i.e. number of species replacements between two sites) and boundary clumping (i.e. boundaries in species composition across two or more sites based on Morisita overlap index). When coherence is negative or not significant, the meta-community follows a checkboard or random pattern respectively. When coherence is statistically significant and positive ( p < 0.05; less embedded absences than expected by chance), the meta-community is classified into six basic structures evaluating turnover and boundary clumping. When turnover is statistically significant and negative ( p < 0.05; less replacements than expected by chance), the meta-community can follow some nested pattern (i.e. evenly-spaced, clumped or random). When turnover is statistically significant and positive ( p < 0.05; with more replacements than expected by chance), the meta-community is classified as a Clementsian, Gleasonian or evenlyspaced pattern ( Fig. 1 ). The Morisita index (MI) needs to be evaluated to determine boundary clumping between different woody communities (if MI > 1, a Clementsian structure and if MI < 1, an evenly spaced structure; [7 , 8] ). The EMS were calculated with Matlab [1] , using a script developed by Presley and Higgins [16] .
We determined the elements of meta-community structure (EMS) for matrix of adult individuals, juvenile individuals of the regenerating layer and total species (sum of adult and juvenile woody species). The models for matrix ordination were set by reciprocal averaging (Fig. 3-11; [17] ), the null model with fixed species richness per site and equiprobable species occurrence (random 0). The models ran with 10 0 0 iteration and extraction of the scores from first axes of ordination based on reciprocal averaging.
For Table 2 we created a matrix distance-species distribution pattern to explore whether geographic distance influenced the species composition between sites. The distance between sites was calculated using ArcGis v.10.3.1 for Desktop [2] and boundary clumbing between different communities was calculated based on Morisita index (MI; see Fig. 1 ) using Past 3.16 [3] . The matrix distance-species distribution pattern was calculated to both age-classes together.
For Fig. 2 the Jaccard similarity coefficient matrix was subjected to non-metric multidimensional scaling ordination (NMDS) to assess species assemblage among different native forest types. We used a matrix distance-similarity to explore whether geographic distance influenced the similarity of species composition between sites. The distance between sites was calculated using ArcGis v.10.3.1 for Desktop [2] and composition (di)similarity was calculated based on Jaccard Index (J) using Past 3.16 [3] . The matrix distance-similarity was calculated to both ageclasses together.
We classified land use from Landsat 8 OLI satellite image for the year 2017 [17] in a buffer zone of 3 km from central point of each permanent plot, processing atmospheric and geometric correction by Landsat image using Matlab [1] . We combined two techniques of classification: we first used supervised classification using ground control points collected in field across different land uses to capture signature spectral of each land use type, then used tree classification technics based on signature spectral of each land use type with Envy v.5.3 [18] . The land use maps were set to six land use types (i.e. native forest, grassland, timber plantation, agriculture, water Table 2 Linear distance matrix between permanent plots (or sites) in kilometer (below of diagonal) and Morisita index (above of diagonal). ID number represents each forest fragments ( Fig. 2 a). The matrix was ordered according linear distance between 1 and 32 sites (see column 1).  Table 3 The scores of first axis of ordination generated by reciprocal averaging for adults, juvenils and all age classes together. Numbers indicate the ID of permanent monitoring sites (see Fig. 2 a) and the approx. latitude and longitude. We cannot publish the exact location as we agreed to protect owner privacy. body and urban areas). Due to the small area covered by water bodies and urban areas, these land uses were not considered in the analysis. For details see Ramírez and Säumel (2022). At landscape level, we calculated total number of patches (total number of patches in the landscape without considering the identity of the land use type), landscape shape index (standardized measure of total edge that adjusts for the size of the landscape, where index increases without limit as landscape shape becomes more irregular and/or as the length of edge within the landscape increases), Shannon's evenness index (distribution of area among patch types where larger values mean higher landscape diversity), and Aggregation Index (frequency with which different pairs of patch types appear side-by-side on the map). At class level, we calculated percentage of the landscape occupied by each land use type and the metrics of native forests without considering other land use types within of each buffer. We thus determined number of native forest patches (number of native forest patches within the landscape), interspersion and juxtaposition index (measure at class level based on patch adjacencies), Euclidean nearest neighbor distance (mean of the shortest straight-line distance between all native forest patches by landscape). Finally, at patch level, we only considered the native forest fragment where the permanent plots were established, thus calculating total area, perimeter area ratio as a measure of shape complexity, and shape index of patch as a measure of compactness. All spatial metrics were calculated for the 3 km buffer using Fragstat v.4 [4] . Table 4 Department and site each permanent plots (ID) across Uruguay ( Fig. 2 a) and overview on metadata of landscape metric per site. ID = code native forest fragments ( Fig. 1 ). Means are given at landscape scale for number of patches (NP): number of patches (NP); Landscape shape index (LSI), Shannon's evenness index (SHEI), Aggregation Index (AI) and at land use type level: Percentage of the landscape (P) occupied by each land use type (NF: native forest; GL: Grassland; TF: Timber forest; C: crops); Number of native forest patches within the landscape (N NF ); Interspersion and juxtaposition index of native forest (IJI); Euclidean nearest neighbor distance of native forest (ENN) in a buffer of 3 km from central point of permanent monitoring site. The landscape shape index (LSI) by land-use type is given by:

ID
Where e * ik is the total length of edges in the landscape between patches types of land-use type i and k , A is the total area of landscape and 0.25 is the factor of adjustment for raster format.
The Shannon's evenness index (SHEI) is given by: Where P i is the proportion of the landscape occupied by patch type belonging to land use tipe i and m number of patches of land use type in the landscape The aggregation index (AI) expressed in percentage is given by: Where g ii is the number of joins between pixel of patches belonging to land-use type i , max → g ii is the maximum number of joins between pixels of the same land-use type i .
The percentage of landscape occupied by each land-use type is given by: P i = n j=1 a ij A ( 100 ) (4) Where P i is the proportion of the landscape occupied by patches belonging to land-use i , a ij is the area of patch ij , A is the total area of the landscape.
The Interspersion and juxtaposition index (IJI) of native forest is given by: Where e ik is the total edge lenght (m) in the landscape between native forest patches i and k, m is the number of native forest parches present in the landscape The Mean Euclidean nearest neighbor distance (ENN_MN) is given by: Where h ij is the distance (m) from ij to nearest neighboring patch of same type and n is the number of nearest neighbouring distance The ENN_MN is based on patch edge-to-edge distance and computed from cell to cell center of patches The Perimeter area ratio (PARA) is given by: Where Pij is the perimeter (m) of patch ij and aij is the area (m 2 ) of patch ij. The Shape index of patch (SHAPE) is given by: Where Pij is the perimeter (m) of the native forest patch ij and aij is the area (m 2 ) of the native forest patch ij.

Ethics Statements
The authors comply with the ethical guidelines of the journal. Humans, animals or data from social media are not involved in this research.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data Availability
Land use change impacts on metacommunity structures in Uruguayan native forests (Original data) (Edoc HU Berlin).