Performance of four ordination techniques assuming three different non-linear species response models

Summary The performances of four ordination techniques (reciprocal averaging, parametric mapping, gaussian ordination and non-metric multidimensional scaling) are evaluated using artificial data based on three vegetation models: the gaussian response curve, -function curves, and an ecological response model. No technique performs well with the artificial data not based on a gaussian model. Non-metric multidimensional scaling seemed marginally better than other techniques. Some of the implications of this failure are discussed.     

Modelling the potential effects of climate change on calcareous grasslands in Atlantic Europe

Aim This study aims to assess the sensitivity of calcareous grassland vegetation to climate change and to indicate the most probable direction of change. Location The study area was a region of Britain, Ireland, France and Spain, centred on the Bay of Biscay, which was defined using a land classification based on climatic criteria. Methods Vegetation was sampled in the field, with additional data collected on soils, climate, management and land cover. The vegetation samples were ordinated by detrended correspondence analysis in order to explore the main gradients present and as a basis for modelling changes. Environmental data were summarized by ordination techniques, with the scores generated used to predict the current vegetation score on the first two ordination axes by multiple regression. The model was then manipulated to represent a 2 °C increase in temperature and resulting shifts in the vegetation samples in terms of their species composition assessed. Results There was a good general agreement between the original vegetation ordination axis scores and those predicted by the model, the latter of which were based on environmental data alone. Following a 2 °C increase in temperature, the predicted changes in the ordination space were demonstrated to be subtle, consisting of small shifts towards vegetation associated with warmer conditions, representing distances 100 km or less on the ground. Main conclusions The models are simple but nevertheless provide a useful basis for the investigation of potential vegetation change. The shifts in the ordination space represent more minor changes than those predicted in previous studies. This suggests that the potential for major change is lower when environmental factors such as soil and management are considered in addition to climate. The potential for change is also reduced when vegetation is considered as a whole rather than on an individual species basis, due to both interspecific interactions and interactions with environmental factors acting as constraints.     

A seriation method for summarizing vegetation-environment relationships

A computer program is described for producing site-species tables termed seriation arrays. Sites can be ordered according to sequences in direct habitat gradients, indirect site ordination axes, and trends oblique to the axes of an indirect site ordination. A ² technique is used to filter species for restriction to segments of each site sequence, retaining only accepted species in the construction of arrays.     

Vegetation gradients in Atlantic Europe: the use of existing phytosociological data in preliminary investigations on the potential effects of climate change on British vegetation

• 1 This paper aims to demonstrate the use of available vegetation data from the phytosociological literature in preliminary analyses to generate hypotheses regarding vegetation and climate change.
  • 2 Data for over 3000 samples of calcareous grassland, mesotrophic grassland, heath and woodland vegetation were taken from the literature for a region in the west of Atlantic Europe and subjected to ordination by detrended correspondence analysis in order to identify the main gradients present.
    • 3 Climate data were obtained at a resolution of 0.5° from an existing database. The relationship between vegetation composition and climate was investigated by the correlation of the mean scores for the first two ordination axes for each 0.5° cell with the climate and location variables.
      • 4 The ordinations resulted in clear geographical gradients for calcareous grasslands, heaths and woodlands but not for mesotrophic grasslands. Significant correlations were shown between some of the vegetation gradients and the climate variables, with the strongest relationships occurring between the calcareous grassland gradients and July temperature, latitude and oceanicity. Some of the vegetation gradients were also inferred to reflect edaphic factors, management and vegetation history.
        • 5 Those gradients that were related to temperature were hypothesized to reflect the influence of a progressively warmer climate on species composition, providing a baseline for further studies on the influence of climate change on species composition.
        • 6 The validity of the literature data was assessed by the collection of an original set of field data for calcareous grasslands and the subsequent ordination of a dataset containing samples from both the literature and the field. The considerable overlap between the samples from the literature and the field suggest that literature data can be used, despite certain limitations. Such preliminary analyses, using readily available data, can thus achieve useful results, thereby saving lengthy and costly field visits.

     

排序和广义线性模型与广义可加模型在植物种与环境关系研究中的应用

排序和广义线性模型(Generalized Linear Model,GLM)与广义可加模型(Goneralized Additive Model,GAM)是研究植物种与环境间关系的重要方法。基于线性模型的排序方法应限定于环境梯度较短的植被数据。而基于单峰模型的排序方法更适用于梯度较长的情况。PCA、CA／RA系列和CCA系列是常用的排序方法。同时进行环境数据和植被数据分析的CCA系列,能清楚地得出植物种与环境间的关系。CCA改进后的DCCA和PCCA,是现今较理想的排序方法。GLM和GAM实质上是用环境变量的高阶多项式来拟合植物种与环境变量的关系。GLM和GAM扩展了植物种与环境变量之间的关系模型,能深入地探讨植物种与环境间的关系。GLM主要是模型决定的,而GAM主要取决于原始数据。一般来说,排序能得出研究区域的主要环境梯度,提供了物种聚集和植物群落的概略描述。GLM与GAM对于深入研究单个植物种与环境间的关系具有优势。在实际研究中,两种方法结合使用能互补不足。     

Effects of aquatic macrophytes on water quality and phytoplankton communities in shallow lakes

We investigated aquatic macrophytes, water quality, and phytoplankton biomass and species composition in three shallow lakes with different levels of vegetation cover and nutrient concentration in Kushiro Moor, during August 2000. Trapa japonica can live in a wide range of nutrient levels. This species forms an environment with a steeper extinction of light, higher concentrations of dissolved organic carbon (DOC), lower concentrations of dissolved oxygen (DO) near the bottom, and lower concentrations of nitrate+nitrite and soluble reactive phosphorus (SRP) than other vegetation types. The pH was much higher in a Polygonum amphibium community, and the DO near the bottom did not decrease compared to a T.japonica community in the summer. The relationship between chlorophyll a and the limiting nutrient (total phosphorus (TP) when total nitrogen (TN):TPis 10 and TN/10 when TN:TP is <10) significantly differed between lakes with and without submerged vegetation. The chlorophyll a concentrations at a given nutrient level were significantly lower in water with submerged macrophytes than in water without them. Correspondence analysis showed that the difference in phytoplankton community structure across sites was largely due to the presence or absence of submerged macrophytes, and the ordination of phytoplankton species in the lakes with submerged macrophytes is best explained by environmental gradients of TN, chlorophyll, pH and SRP.     

A maximum likelihood approach to non-linear ordination

Summary A non-linear method of ordinating vegetation samples based on the fitting of bell-shaped response curves is lescribed. For each species two Gaussian curves were itted, one to quantitative values, where the species was present, the other to probabilities of absence. A maximum likelihood approach was then used to obtain a best approximation of the positions of the samples along a one-dimensional gradient. By an iterative process successively better approximations were obtained.The method was successful in recovering gradients based on hypothetical data. With two sets of real data the gradient produced was more ecologically satisfying and far less distorted than that revealed by principal component analysis.     

Ordination methods and chironomid species in stony bottom streams

Stony bottom streams in Italy were sampled in 369 sites in different years, and 174 chironomid species (or species groups) were identified.The species by sites matrix was submitted to different ordination methods, with the aim of outlining the major factors that are responsible of the observed species composition.The methods considered were principal component analysis (PCA), detrended correspondence analysis (DCA) and non-metric multidimensional scaling (MDSCAL). These methods assume different models describing the species-responses to the environmental gradients.PCA gave poor results, the best species ordination was given by MDSCAL. DCA gave results in agreement with MDSCAL. DCA results were useful as a starting point to perform MDSCAL. MDSCAL was performed with different options. At last Procrustean analysis was carried out to have a single fitted configuration, that summarized the results given by the different MDSCAL methods.All multivariate analyses emphasized that the first ordination axis can be interpreted as a crenonrithron-potamon gradient, probably an oxygen availability and/or a water temperature gradient, whereas the second axis can be considered a water speed gradient, separating lotic from lentic sites.The 3rd and 4th ordination axes had apparently no ecological meaning.Separated clusters of species were never evident, so well defined chironomid species assemblages could not be outlined in this analysis.     

An application of ordination to the identification of forest types

Summary The problems of allocating new vegetation stands to an existing classification are outlined. As one solution, the vegetation samples from an area may be ordinated by reciprocal averaging and classified by dividing the ordination into community-types. The ordination scores for a short list of species present in a new sample may then be averaged to locate that sample on the ordination axes and in one of the community-types. The method is tested by comparing estimates from incomplete species lists with that from a complete list of species in an forest stand in southern Ghana.We gratefully acknowledge the assistance of Mr. M.O. Hill in the correct formulation of the coordinate estimation procedure.     

广东森林群落排序分析

本文应用极点排序和位置向量排序的方法对广东13个森林群落进行排序分析,并分别用极点排序图和位置向量排序的二维图和三维图表示排序的结果。同时对排序图的生态学意义及排序方法的优缺点进行讨论。结果表明,三维位置向量排序图能较好地把性质相近的群落类型聚在一起,可作为植被分类的辅助方法;积点排序图从一定程度上反映了植被的连续变化;极点排序与位置向量排序虽然取得一定结果,但由于同属线性排序,损失的信息量较多,寻求非线性排序方法是研究的方向。     

Can we trust gradients extracted by Detrended Correspondence Analysis?

Abstract. In this study, data from dry temperate grassland vegetation in Denmark are used to compare the predictive power of gradients obtained by ordination. One of the problems of ordination methods based on weighted averaging, namely the assumption of symmetric, unimodal response curves, is investigated by smoothing species responses to simple and complex gradients respectively. It was found that species response types to a pH gradient are diverse and often deviate from a unimodal, symmetrical shape. Bimodal responses were also found, but they disappeared when more influential gradients were considered. Many species showed truncated responses with optima near gradient ends. In order to assess the impact of unrealistic response assumptions and reported problems with instability on the performance of DCA, the predictive potentials of measured variables and coenoclines extracted by DCA were compared. Despite field data violating the assumption of unimodal response, DCA was found to extract gradients predicting species abundance better than the environmental variables available. Shortcomings and advantages of indirect and direct methods in plant ecology are discussed.     

Species response curves along environmental gradients. A case study from SE Norwegian swamp forests

Abstract. Vegetation science has relied on untested paradigms relating to the shape of species response curves along environmental gradients. To advance in this field, we used the HOF approach to model response curves for 112 plant species along six environmental gradients and three ecoclines (as represented by DCA ordination axes) in SE Norwegian swamp forests. Response curve properties were summarized in three binary response variables: (1) model unimodal or monotonous (determinate) vs. indeterminate; (2) for determinate models, unimodal vs. monotonous and (3) for unimodal models, skewed vs. symmetric. We used logistic regression to test the influence, singly and jointly, of seven predictor variables on each of three response variables. Predictor variables included gradient type (environmental or ecocline) and length (compositional turnover); species category (vascular plant, moss, Sphagnum or hepatic), species frequency and richness, tolerance (the fraction of the gradient along which the species occurs) and position of species along each gradient. The probability for fitting a determinate model increased as the main occurrence of species approached gradient extremes and with increasing species tolerance and frequency and gradient length. Appearance of unimodal models was favoured by low species tolerance and disfavoured by closeness of species to gradient extremes. Appearance of skewed models was weakly related to predictors but was slightly favoured by species optima near gradient extremes. Contrary to the results of previous studies, species category, gradient type and variation in species richness along gradients did not contribute independently to model prediction. The overall best predictors of response curve shape were position along the gradient (relative to extremes) and tolerance; the latter also expressing gradient length in units of compositional turnover. This helps predicting species responses to gradients from gradient specific species properties. The low proportion of skewed response curves and the large variation of species response curves along all gradients indicate that skewed response curves is a smaller problem for the performance of ordination methods than often claimed. We find no evidence that DCA ordination increases the unimodality, or symmetry, of species response curves more than expected from the higher compositional turnover along ordination axes. Thus ordination axes may be appropriate proxies for ecoclines, applicable for use in species response modelling.     

松嫩平原南部植物群落的分类和排序

 松嫩平原南部的气候具有温带半湿润大陆性的特点。地带性植被为草甸草原。由于小地形起伏,土壤盐碱和水分含量有差异,影响植物的分布和组合,形成不同的群落。本文采用定性和定量相结合的方法,对该地区的植物群落进行了分类和排序,划分了11个群落类型,归为5类生态组合。通过样地和群落的排序说明植被既是连续的,又是间断的,同时明显看出植物群落随着土壤盐碱和水分含量的变化。     

A comparison of detrended correspondence analysis and principal co-ordinates analysis using four sets of Tasmanian vegetation data

Indirect gradient analysis, which entails the elucidation of relationships between trends in community composition and underlying environmental or successional gradients, is a major objective of ordination in plant ecology. Two ordination techniques, detrended correspondence analysis (DCA) and principal co-ordinates analysis (PCOA), were compared using three sets of Tasmanian vegetation data having known gradients and one set where the vegetation was expected to respond to diverse environmental variables. In every case, the results obtained by DCA were considered superior to, or at least as good as, those of PCOA. Hence, DCA appears to be the more suitable of the two methods for indirect gradient analysis.     

Sahelian woody vegetation in Sénégal

Structure and composition of woody vegetation in Sahelian forest types (not including riparian and riverine types) were studied in Sénégal. The application of classification and ordination resulted in the recognition of seven distinctive woody community types, related to soil and precipitation gradients. The described vegetation types only contained four strictly Saharan species, the rest being of Sahelian and Soudanean affinity, so a floristic change from savanna to desert vegetation after two decades of drought could not be confirmed. Tree density was lower in the northern part than earlier, possibly an indication of degradation. Rejuvenation was greatest at the northernmost arenosolic stations, but low in the southern, ferrouginous areas. This presumably demonstrates the resilience of North-Sahelian vegetation, at the moment being in a dynamic rejuvenation phase, after a period of adverse conditions.     

Evaluating and presenting uncertainty in model‐based unconstrained ordination

Variability in ecological community composition is often analyzed by recording the presence or abundance of taxa in sample units, calculating a symmetric matrix of pairwise distances or dissimilarities among sample units and then mapping the resulting matrix to a low‐dimensional representation through methods collectively called ordination. Unconstrained ordination only uses taxon composition data, without any environmental or experimental covariates, to infer latent compositional gradients associated with the sampling units. Commonly, such distance‐based methods have been used for ordination, but recently there has been a shift toward model‐based approaches. Model‐based unconstrained ordinations are commonly formulated using a Bayesian latent factor model that permits uncertainty assessment for parameters, including the latent factors that correspond to gradients in community composition. While model‐based methods have the additional benefit of addressing uncertainty in the estimated gradients, typically the current practice is to report point estimates without summarizing uncertainty. To demonstrate the uncertainty present in model‐based unconstrained ordination, the well‐known spider and dune data sets were analyzed and shown to have large uncertainty in the ordination projections. Hence to understand the factors that contribute to the uncertainty, simulation studies were conducted to assess the impact of additional sampling units or species to help inform future ordination studies that seek to minimize variability in the latent factors. Accurate reporting of uncertainty is an important part of transparency in the scientific process; thus, a model‐based approach that accounts for uncertainty is valuable. An R package, UncertainOrd , contains visualization tools that accurately represent estimates of the gradients in community composition in the presence of uncertainty.     

Tree species composition and rain forest-environment relationships in the middle Caquetá area,Colombia, NW Amazonia

As part of an integrated forest vegetation and soil survey, tree species composition (DBH 10 cm) was recorded in 95 plots of 0.1 ha, distributed over the principal physiographic units in the middle Caquetá area, Colombian Amazonia. A total of 1077 tree species was found, classified into 271 genera and 60 families. Leguminosae and Sapotaceae show high familial importance values in all physiographic units. Lauraceae, Chrysobalanaceae, Moraceae, and Lecythidaceae are more important on well drained (flood plain or upland) soils, while Palmae, Guttiferae, Bombacaceae, and Apocynaceae are more important in swamps and on podzolised (white sand'rs) soils. Plots on well drained soils show a lower degree of dominance than plots in swamps or on podzolised (white sand) soils. The composition of the most dominant species in the plots changes continuously. Most species (59%) are only recorded in one plot. Individual plot pairs generally show a low overlap of about 2–5 tree species, resulting in Jaccard coefficients below 20%.complementary to a previous forest classification based on TWINSPAN analyses, detrended and canonical correspondence analyses were carried out, using CANOCO 3.1. Despite of a low amount of tree species variance explained (only 6.2% by the first two canonical axes), meaningful patterns of tree species composition were recognised. These are most strongly related to drainage, flooding, humus forms, and soil nutrient status. Forest types are well separated in the CCA ordination diagram. The most frequently found tree species are listed according to their preference with respect to drainage, flooding, and soil nutrient status.Tree species composition in the well drained upland forests was analysed separately. In view of the model explaining high NW Amazonian tree species diversity on the basis of dense community packing and high beta diversity along soil gradients, the canonical analysis here focused on the effect of soils. By means of partial canonical ordination it was found that patterns of tree species composition depended significantly on soil properties, even though the edaphic component explains only a small fraction of the tree species variance. The results show that the well drained uplands of the middle Caquetá area are covered by a complex of two intergrading tree species assemblages. The first assemblage (community of Goupia glabra-Clathrotropis macrocarpa) is associated to some-what less poor, clayey soils developed in Andean origin deposits or Tertiary sediments from the Pebas formation. The second assemblage (community of Swartzia schomburgkii-Clathrotropis macrocarpa) shows affinities to very poor, loamy soils developed in parent materials derived from the Guiana shield. This simple dichotomous pattern of geology, soils, and forest types is incompatible with concepts of high soil heterogeneity and associated beta diversity controlling tree species diversity in well drained uplands of NW Amazonia. The gradient length of tree species in the detrended correspondence analysis was low (3.7 SD), also suggesting a low beta diversity.     

Multivariate analysis of vegetational variation in different habitats at Omayed,Egypt

The application of indicator species analysis has proved useful in classifying stands into groupings coinciding with topographic variations. 41 indicator pseudo-species are identified in the vegetation along a phytosociological gradient: some are characteristic of specific habitats, and the others indicate transition between groups of habitats.The use of qualitative estimations (transformed density records) in multivariate analysis in the present study indicates that they may be more preferable than quantitative estimations, because of their easy and fast recording in the field, and the less comprehensive computations, while yielding precise results.The X-axis of the reciprocal averaging ordination is related to the salinity and fertility gradients, while the Y-axis reflects soil texture.The phytosociological gradients of the canonical variate axes reflect essentially contrasts between groups of species, each correlated with one or the other of the environmental gradients.Comparing the results obtained by the two ordination methods (reciprocal averaging and canonical variates) we find that the first method detects the overall phytosociological changes along strong environmental gradients which would be helpful in studying large surveys. On the other hand, the second technique, which is mainly predictive, is more sensitive to changes within both the phytosociological and environmental gradients and can detect the impact of these changes on the overall variance of each gradient.Nomenclature follows Täckholm (1964).The authors are grateful to Dr J. Jeffers and Mr and Mrs P. Howard of the Institute of Terrestrial Ecology, England, for their help with applying the multivariate techniques to the vegetation data.     

Predicting badger sett numbers: evaluating methods in East Sussex

Abstract. One way in which a species' numbers may be estimated without direct counting is to predict their dispersion and density from more readily available habitat measures, such as landscape variables measured from maps or vegetation variables measured in the field. We compare the power of ordination and regression techniques for predicting badger ( Meles meles L.) numbers at a local scale, using a land class system, map-read landscape variables and field-derived vegetation variables. Sett density was used as a surrogate of badger density. Multiple linear regression using vegetation and landscape variables together gave the most accurate prediction of sett density, while ordination techniques were of lesser value. The addition of vegetation variables to landscape variables did not substantially improve the power of ordination. Outlier Sett Density was predicted more accurately, and by different variables, to Main Sett Density. The relationship between badger ecology and habitat variables that were useful in predicting sett density is discussed.     

The evaluation of multi-species pattern

Abstract. Using artificial data, techniques that have been proposed for analyzing multiple species pattern in vegetation, are compared. No single method was capable of detecting, clearly and unambiguously, all scales of pattern in all cases, and the effectiveness of the different methods was found to depend on the scales of pattern of the component species and on how the patterns of individual species are combined. Some improvements for the application of the popular multi-scale ordination method are suggested. Several sets of field data are analyzed and the results used to illustrate a discussion of the existence and nature of multi-species pattern in vegetation and how it is to be evaluated.