Moderation and mediation of land use change! Interplay between β‐diversity and landscape structures in agri‐ and silvicultural modified grasslands of South America

Although land use change is the main driver of biodiversity decline in the South American temperate grasslands, its impacts on β‐diversity have not yet been evaluated. We investigated relationships between β‐diversity, landscape features, and geographical variables by surveying vegetation from 163 plots distributed in eight different land use types in Uruguay. We created land use maps using Landsat images and calculated landscape metrics, determining β‐diversity across all plots and exploring variation of β‐diversity among different land use types. We ran distance decay models to explore relationships between β‐diversity and geographical location, climate, and landscape metrics. Plant species communities were characterized by a high turnover and low nestedness of species, indicating high dissimilarity across Uruguay. Native forest showed higher β‐diversity than grassland, timber plantation, and crops. β‐diversity increased with geographical distance and environmental dissimilarity. At landscape scale, turnover decreased negatively and nestedness increased to contagion, number of patches, and area‐weighted mean. Nestedness increased with timber plantation and species turnover with crop area. Higher Landscape Shape Index of grassland and crops decreased species turnover. An increase of grassland and crop patches in the surrounding landscape was directly related to a higher species turnover. The high β‐diversity across Uruguay resulting from land use change, moderated by landscape configurations, suggests that numerous protected areas for different habitats are urgently required. A more inclusive vision on biodiversity conservation is necessary, extending the predominant focus from native forests to grasslands, and convincing a broad range of stakeholders, especially landowners and managers, about biodiversity‐friendly land‐use approaches.


| INTRODUCTION
Compared to other biomes as a result of land use change, the temperate grassland biome is expected to experience greater changes in biodiversity (Sala et al., 2000). Historically, grasslands have been used for extensive livestock production, but today large areas of South American native grasslands (Pampas and Campos) are subject to expanding afforestation with non-native tree species (Eucalyptus and Pinus) and monocultures of cash crops for the globalized market. In Uruguay, strongly supported by governmental policies, incentives, and investors' expectations (Geary, 2001), these activities have, in recent decades, led to a decrease in the cover of native grasslands to around 7% (Alvarez et al., 2015). However, studies on the impacts on plant diversity are scarce, and existing studies have focused mainly on invasions of grassland by exotics and on phytosociological characterizations (Bresciano et al., 2014;Lezama et al., 2019).
Land use change modifies the dynamics of ecosystems at different spatial scales: communities in South Brazilian grasslands have been affected by non-random loss of species resulting from local variation of environmental conditions (Torchelsen et al., 2019); while at regional scale, novel land uses impose environmental filters that can act as dispersal barriers, altering species distribution and creating new evolutionary processes in the long term (Staude et al., 2018).
β-diversity provides an effective approximation to integrate the variation among species communities from local to regional scale (Bergamin et al., 2017;Socolar et al., 2016), describing the degree to which sites differ in their species compositions on an environmental gradient (Anderson et al., 2011;Flohre et al., 2011;Whittaker, 1960).
In Southernmost Brazil, β-diversity decreased with decreasing grassland cover, increasing landscape fragmentation, and homogenization of plant communities, mainly through a non-random loss of nontolerant species (Staude et al., 2018). In contrast, β-diversity in North American tallgrass prairies increased with the number of grassland patches, resulting in a higher heterogeneity in community composition after disturbance (McGranahan et al., 2018). Thus, as landscape fragmentation leads to distinct β-diversity patterns, choosing an adequate framework of analysis is essential to unravel responses of β-diversity to land use change and to provide insights for landscape planning (Socolar et al., 2016).
In recent years, β-diversity and its response to land use change have received increasing attention (Newbold et al., 2015;Socolar et al., 2016;Soinonen et al., 2017). β-diversity is characterized by the turnover and nestedness of species (Baselga, 2010): turnover describes how species are replaced across sites (Baselga, 2010;Leibold & Mikkelson, 2002), and nestedness shows how local communities consist of subsets of a more diverse regional species assemblage (Baselga, 2010;Leibold & Mikkelson, 2002). The balance between turnover and nestedness provides insights into the trade-offs between dissimilarity and similarity of species composition across sites at regional scale, revealing mechanisms that shape biodiversity patterns in a landscape . Species turnover and nestedness at regional scale are strongly related to local land use (Newbold et al., 2015), with local habitat-loss and the replacement of natural habitats by productive land uses modifying environmental conditions and leading to the homogenization (nestedness) and/or heterogenization (turnover) of metacommunities at regional scale (Myers et al., 2015;Newbold et al., 2015;Staude et al., 2018).
To extend knowledge of how land use change impacts on species assemblages in the Campos regions of the Southern cone, we analyze β-diversity patterns of herbaceous species in 163 plots at different land use types across 44 sites in Uruguay. We hypothesize that (i) nearby sites exhibit a higher degree of nestedness than distant sites and species turnover between sites increases with geographic distance due to climatic gradients, and (ii) the similarity between plant communities (measured by β-diversity) is favored or limited by landscape configuration and composition. Therefore, we aim to answer the following questions: how β-diversity (i) changes across Uruguay; (ii) responds to geographical and environmental variables; and (iii) responds to land use change. of herbaceous species. Plots were located at a ratio of less than 5 km between them per site. We used a stratified random design. In the first step, we randomly selected monitoring sites across the country.

| Study area and sampling design
In the second step, we contacted landowners to explore their willingness to establish a long-term monitoring site. If the owner agreed, plot selection was stratified by land use type (i.e., grassland, native forest, timber plantation, and cropland). The number of different land uses per site depended on local context. Plant surveys were conducted in nine subplots: two subplots at opposite corners, one subplot in the center, and six subplots arranged in a transect from the center to the edge of each permanent observation plot (Figure 1b). The permanent observation plots had the following dimensions: (i) 50 Â 50 m 2 for grassland and cropland, and (ii) 100 Â 100 m 2 for native forest and timber plantations. For grassland and cropland, the size of the subplots was 2 Â 2 m 2 at opposite corners and 3 Â 3 m 2 in the center.
The size of each subplot of the transect was 1 Â 1 m 2 . For native forests and timber plantations, the size of the subplots at the opposite corners was 5 Â 5 m 2 and in the center was 10 Â 20 m 2 . Each subplot of the transect was 3 Â 3 m 2 in size. The presence or absence of species was recorded in a matrix for each permanent observation plot.
We classified the plots on the basis of land use type as: (i) grassland, (ii) forests and shrublands (plot within native forest or ecosystems dominated by scrubs and palms), (iii) timber plantations, and (iv) crops (plots in cash crop monocultures and orchards). We further subdivided grassland plots according to the intensity of use: (i) primary grassland (without grazing), (ii) secondary grassland (with sporadic grazing and low animal charge), and (iii) livestock grassland (with high animal charge). Tree and shrub-dominated natural ecosystems were subdivided based on tree cover: (i) native riparian and hill forests with a closed tree canopy (closed native forest), and (ii) scrubs, palm stands, and park forest with scattered trees or shrubs (open native forest and scrubland). Timber plantations were subdivided based on plantation age in: (i) old plantations (<8 years), and (ii) young plantations (>4 years). Crops consisted of soybean, sorghum, rice, oats, and wheats (for number of plots by land use type and subtype, see Table S1).

| β-diversity partitioned
We partitioned β-diversity into turnover and nestedness of species assemblages, where the total β-diversity has an additive property of turnover and nestedness (Baselga, 2010). We calculated the β-diversity on four levels: (i) among all plots, (ii) among plots within

| Environmental variables and landscape metrics
To evaluate factors constraining the species distribution, we used geographical location and climatic variables (i.e., annual mean temperature, annual precipitation, temperature seasonality, and precipitation seasonality from Worldclim v.2; Fick & Hijmans, 2017). The calculation of geographical distance between sites was based on UTM coordinates from central point of each plot.
We characterized the landscape around the plots with different spatial metrics. We created land use maps for each site within a buffer

| Data analysis
To compare the variation among land use types and subtypes, a multivariate homogeneity of group dispersion was conducted using betadisper in R package vegan (Oksanen et al., 2018). We used a resemblance matrix based on Jaccard's index between plots by land use types and by subtypes (Anderson et al., 2006). The differences in β-diversity were evaluated by nonparametric analysis of variance based on distance to centroid (dc) with 9999 permutations. Differences were considered as significant at p value < 0.05.
To assess the relationship between each environmental variable and each component of β-diversity across sites, we conducted a distance decay analysis, computed using a generalized linear model (GLM) of the decay.model from R package betapart (Baselga & Orme, 2012). The relationship between β-diversity (dependent variable) and environmental variables (independent variable) was analyzed by the power-law function. The best fit of distance decay was assessed by pseudo-r 2 and significance was evaluated by 1000 permutation with p value < 0.05. The pair-wise dissimilarity between all plots for all β-diversity components used in distance decay models was calculated using beta.pair in R package betapart (Baselga & Orme, 2012). The pair-wise distance matrices for independent environmental variables (climatic variables, geographic location and landscape metrics) were computed as Euclidean distance between sites using function dist in R package stat (R Core Team, 2018).
Finally, to assess and compare the β-diversity patterns described by the distance decay model, we ran a multivariate regression on distance matrix (MRM). The MRM was evaluated using intercepts, coefficients and full model significance using function MRM in R Package ecodist (Goslee & Urban, 2007).

| Pattern of β-diversity
We recorded 902 species distributed in 97 families and 421 genera. In general, we found a high species turnover (β sim ), a low nestedness of species assemblages (β sne ), and a high total β-diversity (β sor ) among all plots, within each land use type and subtype (Table S1). Since turnover and total β-diversity were near the maximum dissimilarity (β sim = 0.98, βsor = 0.99) for all plots, nestedness was not relevant at our sites (β sne = 0.01).
Within land use types, the mean of turnover and of the total β-diversity among all land use types was over 30 times higher (β sim = 0.93 ± 0.03, β sor = 0.96 ± 0.02) than nestedness (β sne = 0.03 ± 0.01). We recorded the highest species turnover in grassland and native forest plots, followed by plots in timber plantations and in crops (Table S1).
Within land use subtypes, the means of turnover and of total β-diversity among all land use subtypes were over 20 times higher (β sim = 0.89 ± 0.02, β sor = 0.93 ± 0.01) than nestedness (β sne = 0.05 ± 0.01). The highest species turnover, the lowest nestedness, and the highest total β-diversity were found in native forests, followed by secondary grassland, primary grassland, shrubs, park forests and palm stands, old timber plantation, livestock grassland, crops and young timber plantation (Table S1).
The pair-wise comparisons of mean species dispersion among land use types were significant between native forests and crops (p = 0.0007), and between timber plantations and crops ( p = 0.0087).
Between land use subtypes, the highest distance to centroid (or βdiversity) was detected for old timber plantation (dc = 0.648),

| Response of β-diversity to environmental variables and landscape metrics
The distance decay model showed significant differences between β-diversity components (β sim , β sne , and β sor ) and environmental variables (geographic localization, climate variables, and landscape metrics, Table 1). The mean and standard deviation for the intercept (β-diversity at minimal distance between plots) and pseudo-r 2 (proportion of deviance explained by model) for species turnover were 0.68 ± 0.02 and 0.02 ± 0.02, respectively. The intercept (β-diversity at minimal distance between plots) and pseudo-r 2 were 0.10 ± 0.01 and 0.02 ± 0.02 for nestedness of species and were 0.78 ± 0.02 and 0.01 ± 0.02 for total β-diversity. The best fit between all distance decay models was achieved for total β-diversity concerning geographical distance (pseudo-r 2 = 0.086, Table 1).
In particular, species turnover (β sim ) was associated significantly with geographical distance and seasonality of climate (Table 1, Figure 3). The turnover increased with distance between sites (slope: b > 0). Based on landscape metrics, turnover decreased with an increase of Contagion Index, number of patches, and the areaweighted mean patch size between all patches in the landscape (slope: b < 0, Figure 3). With regard to the proportion of area occupied by each land use type, turnover increased significantly with an increase of crops at landscape scale (slope: b > 0, Table 1, Figure 3). Species turnover decreased significantly with an increase of number of patches of grasslands and crops (slope: b < 0, Table 1, Figure 3).
According to the LSI, turnover decreased significantly with an increase of complexity of grasslands and native forests. This contrasts with the increase in timber plantation complexity, which resulted in an increase of species turnover (Table 1, Figure 3).
Nestedness (β sne ) increased significantly with geographical distance and climatic variables (Table 1, Figure 3). According to landscape metrics, nestedness increased significantly with number of patches and the area-weighted mean patch size between all patches in the landscape (slope: b > 0), decreased significantly with an increase of the proportion of timber plantation in the landscape (slope: b < 0), and increased significantly with an increase of patch number of grassland and crops (slope: b > 0).
Nestedness increased significantly with an increase of LSI of grassland and native forests (slope: b > 0, Table 1, Figure 3). In contrast, it was negatively associated with LSI of timber plantation The distance decay of total β-diversity (β sor ) was positively associated with an increase in both geographical distance and climate (slope: b > 0, Table 1). Total β-diversity was positively associated with Contagion Index and area-weighted mean patch size between all patches in the landscape (slope: b < 0). Total β-diversity increased significantly with an increase of area occupied by agricultural crops (slope: b > 0).
Total β-diversity was not related to the number of patches by land use (Table 1). Based on LSI, the total β-diversity increased with an increase of timber plantation (slope: b > 0). The metrics not mentioned above were non-significant (Table 1).  In the multivariate regression model on distance matrices, the species turnover increased with geographical distance, latitude, and the seasonality of precipitation (Table 2). Landscape metrics showed a significant association with species turnover, where area-weighted mean of patch size between all patches in the landscape had a negative association. Species turnover increased with an increase of LSI of timber plantations, and species nestedness increased with geographical distance and longitude, but was not influenced by climate.
Although metrics at landscape scale did not show relationships with species nestedness, the area-weighted mean showed a positive relation. Species nestedness increased with increasing LSI of native forest, and decreased with increasing LSI of timber plantations (Table 2). Percentage of land covered by a land use type and the number of patches by land use was not related to species turnover or nestedness. Finally, total β-diversity increased with geographical distance (i.e., with latitude and longitude) and with precipitation seasonality. The landscape metrics at all levels did not show any association with total β-diversity (Table 2).

| DISCUSSION
The components of β-diversity (turnover and nestedness) in Uruguayan temperate grassland are being mediated and moderated by geographic features and spatial patterns of composition and configuration of the land use/cover types at the landscape scale.  share a significant amount of species. Since the Uruguayan flora was defined as a transition between the flora of Southern Brazil and of North-Eastern Argentine (Grela, 2004), we expected a major degree of nestedness as a result of non-random species losses due to land use change and increasing habitat fragmentation (Matthews et al., 2015). However, turnover accounted predominately for the high variation in species diversity between our sites, and nestedness was almost absent (Table S1) and indicates the crucial role of environmental filtering and dispersal limitation on colonization or extinction of species across local ecosystems. Predominance of species turnover on grasslands has been shown for birds (Dias et al., 2017), moths (Enkhtur et al., 2021), plants, lichens, mosses, and insects (Fontana et al., 2020). Species turnover increased with landscape heterogeneity, which explains the dissimilarity across local communities (Tscharntke et al., 2012). Expanding monocultures in Uruguay is generating new environmental conditions at landscape scale. A high species turnover is determined by processes that act at different spatial and temporal scales (Baselga, 2010;Gianuca et al., 2017). Environmental filters limit the dispersal and recruitment of species (Staude et al., 2018). Local communities can limit the establishment of new species by competitive exclusion (Bresciano et al., 2014). The high dissimilarity of species assemblages results from the combination of various factors, including management, climate variation, life history traits, and stochastic factors such as land use change (Holyoak et al., 2020;Tiscornia et al., 2019;Tscharntke et al., 2012).

| Geographical distance and seasonality of climate drive species turnover
Species turnover is positively related to geographical distances between sites and also to the variation of the seasonality of temperature and precipitation across the country (Table 1, Figure 3, Table S1).
Though the area covered by a study, and so possible distances between sites, influences turnover and nestedness of species, the responses are diverse (Soinonen et al., 2017). Turnover increase with an increase of sample area and of the distance between sites (Newbold et al., 2016), but nestedness can also increase when widespreading species are present (Staude et al., 2018). Increasing community dissimilarity between our sites shows that landscape heterogeneity is a key factor for biodiversity pattern. The local soil, slope, and humidity limit certain land uses and conserve habitats, so can override the negative effects of fragmentation (Tscharntke et al., 2012).
To our knowledge, this is the first study on how climatic seasonality influences species turnover across Uruguay (Table 1, Figure 3). The local climate is described as highly variable, from scarcity to abundant rainfall at any time throughout the year (Berretta et al., 2000). The climatic seasonality is hypothesized to shape the productivity of grasslands, shrubs, and woodlands (Gallego et al., 2020). Water balance can also affect the composition of grasslands and growth rate of vegetation (Berretta et al., 2000), and there is evidence that water deficits affect plant invasions in the grasslands of Rio Grande do Sul (Guido et al., 2016). The species turnover observed in our study (Table 1, Figure 3) can be linked to the transition from the subtropical to the temperate region, which is related to different rainfall patterns between the North and the South of Uruguay during the austral summer (Barreiro, 2017). The flooding pampas of Argentina are dominated northward by C4 and southward by C3 species (Chaneton et al., 2005).

| β-diversity shaped by land use
The dissimilarity of community composition observed in our study is related to different land uses. Plants in native forest showed higher β-diversity compared to grassland, timber plantation, and crops ( Figure 2a). The higher dissimilarity across groundcover vegetation within native forests compared to other land use types (Table S1) suggests a high local endemism. The canopy of native forest acts as an environmental filter, constraining the establishment of propagules from neighborhood land uses to shade-tolerant species and the germination of specialists from the seed bank (Holyoak et al., 2020). When land use types were categorized into subtypes, the pattern of dissimilarity ( Figure 2b). The community composition of grazed Uruguayan grassland at landscape scale is shaped by edaphic and topographic factors (Lezama et al., 2019). Similar pattern highlighting soil impacts on diversity has been found in the Pampas of Southern Brazil (Andrade et al., 2019) and Argentina (Chaneton et al., 2005).
Grassland composition did not show differences in β-diversity compared to composition of timber plantations and crops (Figure 2a).
This interesting result can be explained by unconsidered facets of land use change. The dominant land cover which, for centuries, has been the temperate grassland has been transformed, in recent decades, not only to crop and timber plantation (Alvarez et al., 2015), but also to artificial grasslands, an often literally "overlooked" land use change.
The establishment of artificial and so-called "improved" grasslands (Berretta et al., 2000;Jaurena et al., 2021;Modernel et al., 2016) is similar to preparation for crop and timber monocultures, involving the complete elimination of existing species with pesticides, followed by sowing of international standard seed mixtures and agrochemical fertilization. In the last decade, previously unknown silage packages reminiscent of intensely used landscapes from the Global North have appeared in the extensively used grasslands of Uruguay. Consequently, despite awareness of conservation efforts, natural near grasslands are disappearing to a greater extent than so far known, and there is evidence that pastures maintain alpha diversity, but are poor in retaining species from primary vegetation (Newbold et al., 2016).
The lack of differentiation between natural grassland and crops may, however, also suggest the "less dramatic" interpretations: that crops still harbor a fraction of species from original grassland which have a greater capacity to adapt to processes of land use changes (Pañella et al. 2020;Rodriguez & Jacobo, 2013). This may also indicate an extinction debt (Tilman et al., 1994) and the recorded species from original grassland that were the most disturbance-tolerant in the short term may disappear in the future (Fischer et al., 2019). Changes of environmental conditions can break the dormancy, and promote germination of some species from seed-bank due to resource release (Holyoak et al., 2020), increasing the local diversity and dissimilarity at regional scale, at least after the disturbance.

| Landscape configuration moderates use impacts on diversity
We demonstrate the influence of composition and configuration of the landscape as a whole on ß diversity (Table 1, Table S1), and help to develop an optimized configuration of the landscape that allows the conservation of species. At landscape scale, not considering each land use, species turnover decreased, and nestedness increased in response to an increase of aggregation and number of patches and an increasing area weighted-mean of patch size between all patches in the landscape. Higher aggregation was found for landscapes with few and large patches (Li & Archer, 1997;McGarigal et al., 2012), suggesting a major homogenization of environmental condition within the landscape. Consequently, large patches can act as pool of recruitment for small patches (MacArthur & Wilson, 1967), thus increasing species nestedness across land uses (Table 1).
An increasing number of grassland patches decreases species turnover and increases nestedness (Figure 3) (Newbold et al., 2016). Species turnover increases significantly if areas covered by crops increase, and nestedness decreases if timber plantation area increases ( Figure 3). Thus, productive land use creates new pattern of species turnover and nestedness (Newbold et al., 2015). Timber plantation can share greater species number from original land use than crops.
Species turnover decreases with an increase in disaggregation of natural land cover (i.e., grassland and native forest) but increases with timber plantation (Table 1, Figure 3). Thus, plant composition of near natural land uses can limit new species recruitment through competitive exclusion (Bresciano et al., 2014). In contrast, a higher disaggregation of timber plantation across the landscape generates new environmental conditions and facilitates germination at local scale, and the dispersal of native and exotic plants due to traffic on new roads in timber plantations.
Nestedness increases with higher disaggregation of grasslands, native forests, and crops, while timber plantations decrease species nestedness at landscape scale (Table 1, Figure 3). The disaggregation of grassland and native forest accounts for the fragmentation across landscape generating patches that are isolated and poorly connected.
Thus, local dynamics of interaction between species results in different pattern of diversity at landscape (Myers et al., 2015;Tscharntke et al., 2012) and can be extended to regional scales (Newbold et al., 2015).

| Mediate biodiversity conservation
Studies based on β-diversity provide an important starting point for creating and designing strategies for biodiversity conservation at different spatial scales. β-diversity links local and regional dynamics (Socolar et al., 2016). The local loss of habitats due to land use change can lead to non-random extinction of some species, but land use changes can also release ecological niches (Bresciano et al., 2014;Guido et al., 2016). Thus, while local ecological communities might show a tendency to homogenization (Staude et al., 2018), the metacommunity at regional scale can reveal heterogeneity between local communities (Bresciano et al., 2014;Myers et al., 2015;Tscharntke et al., 2012). Consequently, when the β-diversity is considered as input for decision-making, different aspects need to be taken into account (Anderson et al., 2011;Socolar et al., 2016). Since a high species turnover shows a high dissimilarity across Uruguay, the creation of a great number of protected areas for different habitats is needed (Bergamin et al., 2017;Socolar et al., 2016). Over the last 30 years, extensively grazed grassland has decreased by around 40%, at the expense of both afforestation and agriculture (Alvarez et al., 2015), and homogenizes community composition across the landscape (Staude et al., 2018). Local policies face the challenge that biodiversity conservation mainly focuses on native forests overlooking the need for conservation of native grasslands and that all (potentially) protected areas are in the hands of private owners (MGAP, 2018), who are mostly oriented towards the exploitation of the land (Jaurena et al., 2021). However, among them, there is also willingness and recognition of the ecological and cultural value that biodiversity provides (Cortés-Capano et al., 2020). This has the potential to empower future planning in the region mediated through a multi-stakeholder approach beyond black and white perspectives to the benefit of all.