Absence of distance decay in the similarity of plots at small extent in an urban brownfield

Absence of distance decay in the similarity of plots at small extent in an urban brownfield
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  Introduction Spatial scale has long been recognised to strongly affectspecies diversity (Arrhenius 1921, Gleason 1922, Connor and McCoy 1979, Palmer and White 1994, Scheiner 2003,Podani and Csontos 2006, Bacaro et al. 2009, Chiarucci et al.2009, Chiarucci et al. 2011). Ecological processes shapingdiversity can be expressed at a certain scale being hidden atanother, and they can also act at multiple spatial scales, lead-ing to the pressing need of developing multi-scale sampling procedures like those proposed by Reed et al. (1993),Stohlgren et al. (1995), Chiarucci et al. (2001), Baffetta et al.(2007), Kalkhan et al. (2007) or Chiarucci et al. (2008). Onemeasure of biodiversity strongly affected by spatial scale isspecies richness (Arrhenius 1921, Gleason 1922, Gaston1996). It is well-known that spatial scale is characterised by twomain components: grain and extent. The first one is related tothe dimension of sampling units (Scheiner et al. 2000, Dun-gan et al. 2002), while the second is represented by the ex-tension of the study area (Wiens 1989). When aiming at sam- pling species information, it must be kept in mind that thenumber of species in a sample is influenced by both grain andextent (Wiens 1989, Palmer and White 1994, Dungan et al.2002). Moreover, given the same data, analysing them at dif-ferent grain and extent is likely to lead to different results(Stohlgren et al. 1997). Thus, any kind of statistical inferenceis restricted to particular scales of analysis (Palmer and White1994, Bacaro et al. 2011).Another core measure of diversity is represented by thedegree of species turnover from one location to another,namely beta-diversity (Whittaker 1972). For biodiversity as- C OMMUNITY  E COLOGY  13(1): 36-44, 2012 1585-8553/$20.00 © Akadémiai Kiadó, Budapest  DOI: 10.1556/ComEc.13.2012.1.5 Absence of distance decay in the similarity of plots at small extent inan urban brownfield G. Bacaro 1,4 , D. Rocchini 2 , C. Duprè 3 , M. Diekmann 3 , F. Carnesecchi 1 , V. Gori 1 and A. Chiarucci 1 1  BIOCONNET, Biodiversity and Conservation Network, Department of Environmental Science “G. Sarfatti”,University of Siena, Via P.A. Mattioli 4, 53100 Siena, Italy 2  Fondazione Edmund Mach, Research and Innovation Centre, Department of Biodiversity and Molecular Ecology,GIS and Remote Sensing Unit, Via E. Mach 1, 38010 S. Michele all’Adige (TN), Italy 3 Vegetation Ecology and Conservation Biology, Institute of Ecology, University of Bremen, Leobener Str., 28359 Bremen, Germany 4 Corresponding author. Email:  Keywords: Beta-diversity, Bremen, Distance decay, DCA, Environmental distance, Monte-Carlo randomizations, Nested plots, PCA, Spatial scales, Vascular plant. Abstract: Similarity in species composition among different areas plays an essential task in biodiversity management andconservation since it allows the identification of those environmental gradients that functionally operate in determining vari-ation in species composition across spatial scale. The decay of compositional similarity with increasing spatial or environmentaldistance derives from: 1) the presence of spatial constraints which create a physical separation among habitats, or 2) the decreasein environmental similarity with increasing distance. Even if the distance decay of compositional similarity represents a wellknown pattern characterising all types of biological communities, few attempts were made to examine this pattern at smallspatial scales with respect to both grain and extent. Aim of this work was to test whether the distance decay of similarity 1) can be observed at a local scale in situations where environmental conditions are relatively homogeneous and ecological barriersare absent, and 2) is dependent on the grain size at which plant community data are recorded. We selected two urban brownfieldslocated at Bremen university campus, Germany, of 40 m   20 m each, systematically divided in nested plots with an increasingspatial scale of 0.25 m 2 , 1 m 2 , 4 m 2  and 16 m 2 . Both plant species composition and soil variables were recorded in each cell.Linear and logarithmic least squares regression models were applied in order to examine the decay of similarity due to spatialdistance (calculated as the Euclidean distance among pairs of plots) and environmental distance (calculated as the Euclideandistance among PCA-transformed soil variables). A general lack of distance decay was observed, irrespective of the type of distance (spatial or environmental) or the grain size. We argue that this is probably due to a random variation both of theimportant environmental parameters and of the local distribution patterns of individual species, the latter mainly caused by thehigh dispersal abilities of the majority of species occurring in the brownfields. Abbreviations: F1 – Sampling field 1; F2 – Sampling Field 2 Nomenclature: Wisskirchen and Haeupler (1998).  sessment and conservation, the concept of beta-diversity plays a key role, since it adds to the simpler concept of alpha-diversity (i.e., the number of species in a site) by introducingthe spatial variation in species composition (Margules andPressey 2000). Recently, another way of looking at beta-di-versity is to consider the distribution of plot-to-plot dissimi-larities within a vegetation sample instead of using a singleindex (Tóthmérész 1998, Legendre et al. 2005, Bacaro et al.2007, Ricotta and Bacaro 2010). When the distribution of  plot-to-plot similarity is related to a measure of spatial or en-vironmental distance, we expect to observe a decay of com- positional similarity with increasing distance between plots,known as the first law of geography (“Everything is relatedto everything else, but near things are more related than dis-tant things”, Tobler 1970, Bjorholm et al. 2008). This lawdrives distance decay patterns at both global and local scales(Palmer 1995, Nekola and White 1999, Palmer 2005,Soininen et al. 2007b).The decay of similarity in species composition may prin-cipally arise from: (i) a decrease in environmental similaritywith distance (niche difference model); (ii) a spatial configu-ration creating spatial barriers among habitats (isolation) thatinfluence species and gene movement across landscapes(Nekola and White 1999). Moreover, Hubbell’s neutral the-ory (Hubbell 2001) predicts the decay of similarity with re-spect to distance only because of random dispersal and eco-logical drift (see also Tuomisto et al. 2003a, Soininen et al.2007a).Since different environmental factors operate at differentspatial scales, patterns of distance decay should depend onthe grain size at which they are analysed (Steinitz et al. 2006).Similarly, extent is expected to influence the pattern of ob-served distance decay, since increasing extent should resultin i) a higher habitat heterogeneity and ii) a higher probabilityto include geographical barriers in the considered area. Efforts were made to demonstrate the effects of spatialdistance on the similarity in species composition at relativelylarge scales, such as floras (Nekola and White 1999, Phillipset al. 2003, La Sorte and McKinney 2007). Moreover, exist-ing studies used pronounced gradients in order to capturemaximum compositional variation between plots across theconsidered distance. Conversely, few attempts were made todemonstrate distance decay patterns for small grain sizes andextents, or for relative homogeneous environments. The aimof this paper was to test whether distance decay in plot simi-larity can be detected in relatively small plots over very shortspatial and environmental distances. As a test system, we ex-amined an urban brownfield characterized by a high propor-tion of pioneer and ruderal species with a well-developed dis- persal ability. Materials and methods Study area, sampling design and sampled data During the summer of 2007, two 20 m   40 m areas wereselected in an urban brownfield at the Campus of the Univer-sity of Bremen (North-Western Germany). The two fieldswere about 100 m apart from each other divided by a path,and were placed as to not include areas recently affected byman. Disturbance included light grazing by rabbits and occa-sional human trampling. The soil consisted of a sandy sub-strate deposited during the building of the University areamore than 30 years ago. The site is characterized by sparsevegetation with many annuals and grasses, dominated byspecies such as  Festuca ovina, F. rubra, Holcus lanatus and  Rumex acetosella . The sampling grid in each area consisted of 50, 4 m   4m modules. A nested series of square quadrats was located inthe north-west corner of each module (see Figure 1a). Their linear dimensions were: i) 0.50 m (0.25 m 2 ); ii) 1 m (1 m 2 );iii) 2 m (4 m 2 ) (see Figure 1b). The presence of vascular plant species was recorded for each of the quadrats described above. Moreover, speciescover was visually assessed and quantitatively estimated byusing the classical Braun-Blanquet cover-abundance scale(Westhoff and Van der Maarel 1978). Plant species that weredifficult to identify in the field were taken to the laboratoryfor later determination. Soil parameters were measured for each field in the cen-tre of alternate modules (starting from the first one located inthe north-west corner of the whole area) for a total of 25 Figure 1.  Sampling designadopted for collecting vegeta-tion and soil data in the twostudy areas at Bremen Univer-sity Campus; a : the total sam- pled area was 800 m 2  (20 m   40 m), divided in 50 modules of 4 m   4 m (16 m 2 ); b : in thenorth-west corner of each mod-ule, a series of nested quadratswas placed, with linear dimen-sions of 0.50 m, 1 m and 2 m.Topsoil samples were takenwith a metal cylinder in the cen-tre of the plot at each spatialscale (black dots in b).   ba Absence of distance decay in homogeneous brownfields37  quadrats for each plot dimension, resulting in a total of 200soil samples. Only topsoil was collected with metal cylinders(5 cm deep and with a diameter of ca. 5 cm, giving a soilvolume of 100 cm 3 ), because previous analyses had shownthat lower soil horizons consisted almost completely of puresand being devoid of any humus particles and nutrients. Onlyone sample was taken per plot for each spatial scale, alwaysin the middle of the plot (see Figure 1). The soils were thenair dried for two days and sifted using a 2 mm sieve and ho-mogenised for chemical analysis. The following variableswere measured:1) Soil moisture content was determined as the amountof available water in % in each soil sample and measured di-rectly in the field by using a Moisture Meter (HH2 type,Delta-T-Devices Ltd) with a Theta Probe sensor, type ML2x.All measurements were carried out on one day after severaldays of dry weather. 2) Carbon (C) and Nitrogen (N) content were determinedin % using an elemental analyser (CN elemental analyzer -EuroEA 3000. HEKAtech GmbH, Wegberg). The values for the two elements were then used to determine the C/N ratio.3) Soil pH was measured in a solution of 10 g of soil and25 ml of 0.01M CaCl 2  with a standard electrode.Among the 50 plots (200 samples) for which soil pa-rameters were determined, 5 samples collected in Field 1showed values strongly deviating from those of the other  plots, most likely caused by some kind of disturbance. Thesewere treated as outliers and excluded from further analyses.  Data analyses In order to explore patterns of compositional similarityfor the two areas, different types of statistical analyses were performed. First, we carried out exploratory Detrended Cor-respondence Analyses (DCA) on plant species compositionto describe the degree of beta diversity in the two study areasin terms of standard deviation units.Second, the similarity in species composition was quan-tified by using the Jaccard index (Legendre and Legendre1998, Koleff et al. 2003), which is suitable for presence–ab-sence data. Quantitatively, this index is expressed for two plots  A  and  B  as:  J = a / ( a+b+c )(1)where a  = number of species in common to both plots; b  =number of species exclusive for plot A; c  = number of speciesexclusive for plot  B . This index determines the proportion of species shared by a pair of sites out of the total number of species present in these sites (see Jaccard 1912, Magurran1988, Southwood and Henderson 2000, Koleff et al. 2003). Jaccard similarity matrices (for each field and plot di-mension) were related to two predictive variables, namelyspatial and environmental distance. The first matrix included plot-to-plot spatial distances based on the geographical coor-dinates of the site corners (north-west), measured by usingthe classical Euclidean distance (Podani 1989, Banerjee2005). Since the two areas were found to have somewhat dif-ferent habitat conditions and were spatially isolated, the rela-tionship between biological similarity and spatial distancewas explored separately for Field 1 and Field 2. With respect to environmental distance, we merged thevariation of the different environmental soil variables into aPrincipal Component Analysis (PCA, see Legendre andLegendre 1998). The coordinates of plots derived from thePCAs were included in the analyses as new environmentalvariables, and the PCA axes were interpreted examining their correlation with the srcinal variables. Once the new vari-ables were extracted from PCA, the plot-to-plot environ-mental distance matrices were calculated using the Euclideanmetric (only the first two PCA axes were considered). Thesematrices were used as explanatory variables for the analysesat all spatial scales (0.25 m 2 , 1 m 2 , 4 m 2 , 16 m 2 ). Simple linear models were applied to test the predictivevalue of spatial and environmental distances with respect to plot compositional similarity. Furthermore, negative expo-nential models were fitted and compared with the linear mod-els, looking for the best model fit. Explicitly, the exponentialdecay in species compositional similarity was described as: S=S  0  e -md  (2)where S   = similarity at distance d  ; S  0  = initial similarity or similarity at distance 0; m  = decay rate. We estimated thenon-linear model transformed into its linear form by takingnatural logarithms of S   for S   > 0 and of 0.00001 for S   = 0 (seeMachado and Santos Silva 2005).The statistical significance of the relationships betweencompositional similarity and the two types of distances (spa-tial and environmental) was tested using Mantel correlations(for both linear and negative exponential models).  P   valueswere calculated using Monte Carlo randomizations (999runs) of the response matrix (species similarity matrix). Allthe analyses were carried out by using the R-package“vegan” by Oksanen et al. (2011). All the above mentionedanalyses were repeated using the Bray-Curtis similarity co-efficients (Bray and Curtis, 1957) for cover-based log-trans-formed vegetation data. Results Both fields were characterized by the same set of domi-nant species (Table 1): the highest frequencies were shown by grasses (e.g.,  Agrostis capillaris ,  Festuca rubra ,  Holcuslanatus ) and perennial species of the Asteraceae (  Achilleamillefolium ,  Erigeron acris ,  Hypochaeris radicata ) and other families ( Oenothera biennis ,  Plantago lanceolata ). Annualswere more common in Field 2 (  Erophila verna , Vicia hir- suta ).The two sampled areas showed consistent differences inmean soil parameters (see Table 2): Field 1 (F1) had a moreacid soil with lower C and N contents than Field 2 (F2). F1also showed higher C/N ratios and lower soil moisture con-38Bacaro et al.  tents. As expected, species richness monotonically increasedwith spatial scale; on average F2 hosted more species thanF1, at any given spatial scale (Table 2). Values of beta diver-sity measured by means of the first four DCA axes showeddifferences between the two areas (with a maximum gradientlength in F1 at 1 m 2 grain size and a minimum reached in thesame field at 4 m 2  dimension, see Table 3). The relationship between spatial distance and composi-tional similarity showed a slope approaching zero at all thefour plot dimensions analysed, indicating the absence of anydistance decay (Table 4 and Figure 2). In accordance, theamount of explained variance (measured by the R  2 coeffi-cient) did not increase with increasing grain size (P-valuesalways > 0.05). Table 1 . Synoptic table of the two study fields. All vascular  plants are given with their percentage of occurrence in the 50 plots, separately for the two fields and four spatial scales. Table 2.  Summary statistics of environmental variables andspecies richness (mean, minimum and maximum) for the quad-rats sampled on different spatial scales in the two study. For amore detailed description of variables and their measurements,see Material and Methods. Table 3 . Gradient lengths in standard deviation units of axis 1of Detrended Correspondence Analyses for the quadrats sam- pled on different spatial scales in the two study areas (F1 andF2). Table 4. For each plot size, the coefficient of determination(R  2 ) was calculated between the Jaccard similarity and spatialdistance for all pair-wise combinations of plots, separately for the two study areas. P-values were calculated by using Manteltests (999 runs) on the two plot-plot matrices and expressed asthe proportion of lower than observed randomised regressioncoefficients. Absence of distance decay in homogeneous brownfields39  The first two PCA axes together explained about 70% of the variance in environmental conditions (the highest amountwas reached for the 16 m 2  dimension, see Table 5). For thefour PCAs (one for each spatial scale), the first axis was con-sistently positively related to the contents of N and C, soilmoisture content and pH, whereas the second PCA axisshowed a positive correlation with C and the C/N ratio (seeTable 6). The observed compositional similarity was uncor-related with environmental distance as expressed in the PCA,for both fields and the four spatial scales (Figure 3). Similarlyas for the spatial distance, in both fields the R  2  values did notincrease with increasing grain size (see Table 7). All theanalyses performed using the Bray-Curtis similarity coeffi-cient mirrored those obtained for presence-absence dataonly. For this reason, these results are not shown here. Discussion For the brownfield communities analysed in this study,compositional similarity did not decay with increasing spa-tial or environmental distance, irrespective of the consideredgrain size and of the type of model applied (linear or logarith-mic). Interestingly, although the gradient lengths of the DCA Figure 2.  Effect of spatial distance on the similarity of plots for the sampled spatial scales (0.25 m 2 , 1 m 2 , 4 m 2 , 16 m 2 ). Each point represents a pair of sites. a ) Field 1; b ) Field 2. Table 5 . Cumulative proportion of explained variance by thePCA axes performed on environmental variables for the ana-lysed spatial scales. Table 6.  Pearson correlation coefficients between the first twoPCA axes and environmental variables (measured for all sam- pled spatial scales). *** p<0.001 ** p<0.01 *p<0.05 Table 7. For each plot size, the coefficient of determination(R  2 ) was calculated between the Jaccard similarity and environ-mental distance for all pair-wise combinations of plots, sepa-rately for the two study areas. P-values were calculated byusing Mantel tests (999 runs) on the two plot-plot matrices andexpressed as the proportion of lower than observed randomisedregression coefficients. ab 40Bacaro et al.
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks