286 CHAPTER 8 ficients) of data from grassland vegetation, and for three soil variables. Soil depth is clearly related to the first component and soil potassium to the third component, but the relationship of soil phosphate is more complex; it has some affinity with the trend surfaces of all three components. Gittins pointed out that when a surface giving a satisfactory fit has been derived, it may still be profitable to examine the distribution of deviation from the calculated surface. These will represent very local variations and reflect pattern in the vegetation, in the sense discussed in Chapter 3. The need to place further stands into an existing arrangement probably arises much less frequently with ordination than with classification, but it may do so with extensive and continuing surveys (e.g. Hall & Swaine, 1976) or in geographical comparison of floras, as in Lawson’s (1978) study of the marine algal floras of the Atlantic. With some ordinational procedures this can readily (a) Linear (b) Quadratic ost’ ge (c) Cubic R= 250 | Fig. 73. Trend-surfaces fitted to surface relief for data of Goodall (19Sdc). (From Gittins, 1968, by courtesy of Journal of Ecology.)PLANT COMMUNITIES—ORDINATION 287 (a) First principal component (b) Second principal component (e) Extractable soil phosphate (£] Exchangeable soil potassium (ppm P) (m Eq /100 dry soil ) Fig. 74, Trend surfaces fitted to component scores of stands and to soil variables for data from limestone grassland. (a), (b), (d), (e) Third degree surfaces, (c), (f) second degree surfaces. (From Gittins, 1968, by courtesy of Journal of Ecology.)288 CHAPTER 8 be done. Distances from the reference stands can be calculated in polar ordination. In most forms of principal component analysis, species loadings can be used to calculate the appropriate stand position. For reciprocal averaging, position for a further stand can be calculated from a regression of stand position on mean score for species occurring in the stand (Swaine & Hall, 1976). Swaine and Hall found, usefully, that a random sub-set of the species ina new stand was normally adequate to given a satisfactory estimate of stand position. With other procedures no straightforward method of placing further stands is available, at least if more than one axis is involved, and recourse may have to be made to estimating distances from a sample of stands of known position to give an approximate position. When a set of vegetation data have been classified it may be useful to ordinate the resulting groups. Although hierarchical classifications display the degree of dissimilarity between the sub-groups at each node, they do not show the interrelationships of the final groups, nor do procedures normally provide any estimate of similarities between final groups. An ordination will provide information on inter-group relationships and a basis for exploring the correlation between groups and environmental factors additional to that provided by imposing environmental factors on the hierarchy (p. 173). A straightforward way of ordinating groups of stands is to calculate the mean composition of cach group and apply an ordination technique to these means, treating them as if cach mean represented a single stand. If the original stand data are for presence only, the mean composition for a group will necessarily be in quantitative form, as frequency of occurrence of species. Even very widely dissimilar vegetation types can be included in one ordination in this way; Curtis (1959) produced a three-dimensional ordination of 28 native terrestrial plant communities. He used the one-complement of the Czekanowski measure of similarity to construct the ordination directly by triangulation from acentral type, but other methods could equally well be used, e.g. principal coordinate analysis of a suitable distance measure. Ordinations of groups should be interpreted with caution. Since each group is represented by its mean, the ordination conveys no information on its range of composition and what appear as two quite distinctive groups may in reality be much less distinctive. If the techniques of classification and ordination used are not strictly compatible, groups may even overlap in the ordination space. Grigal and Goldstein (1971; Goldstein & Grigal, 1972) used multiple discriminant analysis (canonical variate analysis) to ordinate groups. Blackith and Rayment (1971) give a useful account of this technique (see also Gittins, 1979). The first axis is in the direction which accounts for as much as possible ofPLANT COMMUNITIES—ORDINATION 289 the variability between the means of the groups, the second is orthogonal to the first, and, within this constraint, in the direction which accounts for as much as possible of the remaining variability, and so on, as in principal component analysis. Provided that the data are substantially multivariate normal and the dispersion matrices (variances and covariances) of the different groups are the same, statistical tests of the significance of the differences between groups, and of the assignment of a stand to a group, can be made. These assumptions are unlikely to be even approximately satisfied by vegetational data. However, the technique appears to be moderately robust in this respect and it can be used to compare the variability within groups (Matthews, 1979). If an initial classifi- cation is compatible with the comparison of within- and between-group variances on which multiple discriminant analysis is based, it also provides a useful method of revising the initial classification (del Moral, 1975).CHAPTER 9 Vegetation and Environment The ultimate objective of classification may be and that of ordination commonly is to generate hypotheses about the relationship between vegeta- tion and environment. It is convenient to consider techniques of correlating vegetational and environmental data in relation to classification and ordi- nation together, as some of the problems are common to the two approaches. The subject has been carefully reviewed by Noy-Meir (1970) and the following account largely follows his discussion, Either or both of the vegetational and environmental data may have been generalized by classification or ordination before their relationship is examined. Vegetational and environmental data may be either qualitative (discontinuous) or quantitative (continuous) or include both. Qualitative variables may be binary (c.g. presence/absence of a species) or multistate (c.g. soil types). There are thus a variety of combinations of data to be considered. For convenience, following Noy-Meir, generalization of vegetation and environment is symbolized by V* and E* respectively, and discontinuous and continuous data by D and C respectively. The simplest case is V/E, where the occurrence of single species is examined in relation to separate environmental factors. The use of con- tingency tables for D/D data, of analysis of variance for C/D and D/C data and of regression for C/C data has been outlined in Chapter 5. Regression analysis is not confined to C/C data, but can be modified to use binary or multistate} variables (Noy-Meir et al., 1973). The earlier uses of regression (e.g. Blackman & Rutter, 1946; Rutter, 1955) involved only a small number of independent variables; the computation involved precluded the inclusion of large numbers. With the present speed and capacity of computers there is now no effective limit on the number of variables that can be included. This does not, however, remove the need for careful prior assessment of the relevance of the variables to be included; + A multistate variable can be treated as several binary variables. Alternatively, ifa multistate variable is the only one being considered, the relation between it and quantitative data for a species can be examined by analysis of variance, 290VEGETATION AND ENVIRONMENT 291 unnecessary elaboration of the field and observational programme is as wasteful as is unnecessary computation. If a number of independent variables are initially included, the question arises how many can be climinated without seriously reducing the efficiency of the regression (normally assessed as the proportion of the total sum of squares of the dependent variable accounted for by the regression). In some fields, the objective of regression analysis is primarily predictive, to estimate the value of the dependent variable corresponding to stated values of the independent variables. This may be truc where the functioning of an ecosystem is being examined (cf. Jeffers, 1978), but in the present context we are concerned rather with detecting those environmental variables with which plant performance is correlated, as a basis for hypotheses about the factors which determine the plant performance. When multiple regression is used as a means of exploring the relationship between variables, rather than to derive a predictive equation for the dependent variable, it is mostly usefully applied in a stepwise manner; the best fit to a single independent variable is selected, and then the best combination of a second independent variable with this one is selected, and further independent variables added to the regression equation until the proportion of the variance of the dependent variable accounted for is considered adequate. For binary multiple regression the environmental attribute which has the highest correlation with the species being examined (measured by |r| for C/D data, or z? for D/D) is the first independent variable to be included in the regression equation. At each subsequent step the one of the remaining environmental attributes with the highest partial correlation with the species is included. General aspects of the use of regression in ecological contexts have been usefully discussed by, among others, Austin (1971, 1972), Mead (1971) and Yarranton (1967d, 1969b, 1971). Yarranton suggested the inclusion of other species as well as the environmental attributes in the regression equations for species, i.e.a V/(V + E) approach. This makes allowance for interactive effects between species, resulting from modifications of the microcnvironment by vegetation. Interpretation in other than species-poor situations is, however, likely to be difficult. If the relationship is markedly non-linear, the inclusion of linear terms only in a multiple regression on environmental attributes may fail to indicate the possible importance of particular attributes in determining the representa- tion of a species. Difficulty also arises from the strong correlation between environmental attributes that is commonly found. Multiple regression takes account of such correlation between independent variables, so that a multiple regression equation will be satisfactory as an empirical predictor, but may be292 CHAPTER 9 little help as a means of exploring the importance of environmental attributes. Thus Austin (1971), commenting on Yarranton’s (1970) regression analysis of cryptogamic species growing on limestone pavement, pointed out that the independent variables such as light, humidity and temperature are function- ally related and form part of a factor-complex determining the evapo- transpiration rate which is the primary determinant of plant moisture stress. The use of multiple regression is less straightforward than may appear. As Austin put it, ‘judgement is required based on: (a) knowledge of the numerical assumptions and possibilities of the technique, (b) field experience of the vegetation studied, (c) knowledge of the processes influencing plant growth’. An alternative to multiple regression is predictive attribute analysis (Macnaughton-Smith, 1963, 1965; Noy-Meir et al., 1973). This was developed for binary ‘internal’ (here, environmental) attributes, but Lance and Williams (1968b) showed that it can be extended to include attributes of mixed C and D form. The first step, as in binary multiple regression, is to identify the environmental attribute showing the maximum correlation with the species, but the data are then divided into two sets, with and without this environmental attribute, and the correlations between the species and other environmental attributes examined in the two sets separately. Correlation of species with environment is thus examined hierarchically. This may be more informative than a single multiple regression equation, at least provided that the original environmental variables are of binary form; if multistate variables have been divided into binary ones, difficulties of interpretation may arise. For data of C/C form, Westman (1980) suggested, as a means of identi- fying the environmental factors most influencing the performance of species, fitting Gaussian curves to the response of each species to the values of each environmental factor. The percentage variance accounted for by the fit toa Gaussian curve is determined and the most influential environmental factors are taken to be those for which the mean percentage variance accounted for is greatest. Only actual occurrences of species are taken into account in fitting the Gaussian curves. Westman emphasized that not all, or even most, of the important influencing factors will necessarily be identified because of the multifactorial control of species performance. Nevertheless, this is a promis- ing approach to a preliminary sorting of environmental variables to decide which are worth further investigation, particularly where data are available for a large number of variables. The V*/E case, where the vegetation has been generalized but environ- mental attributes are examined individually, has been the most widely used approach. Commonly, environmental attributes have been plotted onto an ordination of stands or the average values for environmental attributes in theVEGETATION AND ENVIRONMENT 293 two subgroups at a node of a classification have been examined (Figs 43, 44, 31). Nealand Kershaw (1973) suggested calculating trend-surfaces (p. 284) for asingle environmental variable in relation to the coordinates of a vegetational ordination as an improvement on direct plotting. The general failure to examine relationships with the environment any more critically is open to the criticism that the advantages of refined numerical methods of simplifying vegetation data are largely nullified by the subsequent casual approach to correlation with environment. Straightforward plotting onto the ordination does, however, have the advantage that it is unaffected by non-linearity and indeed may expose non-linearity where it has not initially been recognized. The approaches available for the V/E case are applicable, though not all have been used, and are subject to the same limitations. With vegetational ordination, i.e. C/C or C/D data, the situation is exactly comparable; vegetational components may be regressed on environmental attributes, or mean component scores for discontinuous environmental attributes may be examined by analysis of variance (or ¢ test for binary attributes)+. If vegetation has been classified, though the same analyses can be used as with individual species, interpretation is complicated by the fact that the groups are mutually exclusive in nearly all procedures of classification—a stand belongs to one group only. It is thus not as useful to know fora group of stands as it is for a species what environmental attributes are correlated with its occurrence or non-occurrence; interest lies rather in what environmental attributes differentiate between the groups and it may be useful to ordinate the groups and examine this ordination in relation to environment (e.g. Greig-Smith er al., 1967). The V/E* approach has not been widely used, and when it has, species correlation with environmental components has been examined by plotting species presence or abundance onto the ordination (e.g. Goldsmith, 1973). The same procedures of more precise analysis as for V*/E are, in principle, available. In view of the limitations of environmental ordinations discussed above (p. 278), it is not surprising that Austin (1968b), applying both V*/E and V/E* approaches to data from chalk grassland, found the vegetational ordination more informative, although the environmental ordination pro- vided some further insight Initial simplification of both vegetational and environmental data, a V*/E* approach, is an evident possibility, and several investigations have compared vegetational and environmental ordinations of the same data (e.g. + Intheir pioneer ordinational study, Bray and Curtis(1957) calculated correlation coefficients between axis position of stands and environmental attributes.294 CHAPTER 9 Dagnelie, 1960; Austin, 1968b; Austin et a/., 1972), but the results have been rather disappointing. The most thorough test of this approach is that of Austin (1968b). The highest correlation between any of the first three vegetational and the first three environmental components was only 0-536 (Table 26)*. Plots of environmental components onto the vegetational ordination (Fig. 75) show that the relationships are not linear and suggest that acurvilinear multiple regression would need to be complex and would thus be difficult to interpret usefully. Table 26. Correlation coefficients between vegetational and environmental components in an analysis of data from chalk grassland (Austin, 1968b) Environmental components Vegetational components 1 2 3 1 —0-229 —0-109 —0-152 2 0-244 0536 © 0-064 3 —0-191 0-189 0-149 A vegetational classification could be compared with environmental components in the same ways as comparison with single environmental attributes. Classification of both vegetational and environmental data is unlikely to be used. Environmental data do not generally lend themselves to classification within the range likely to be encountered in a single set of data (but cf. Goldsmith, 1973) and the approach is, in any case, unlikely to be an informative one; procedures are available for estimating the similarity of two hierarchical classifications of the same set of individuals (see, for example, Sneath & Sokal, 1973; Rohlf, 1974; Smartt et al., 1974; Williams, 1976) butit is not very helpful to know merely how similar the vegetational and environ- mental classifications are. Instead of separate principal components or factor analysis ordinations of vegetation and environment, both may be included in a single analysis ((V+E)*) (e.g. Ferrari et a/., 1957; Dagnelie, 1960; Gittins, 1969; Walker & Wehrhahn, 1971; Schnell et al., 1977; Dye & Walker, 1980). The environ- mental attributes having high correlation with a particular species may be expected to have similar loadings to those of the species. Examination of the resulting attribute ordination can thus be used to gain an overall impression of * The highest correlation obtained by Cassie and Michael (1968) in a study of marine fauna and environment was still lower (0-26).295 VEGETATION AND ENVIRONMENT (Abojoxq fo prunos Jo ksmanos Kq ‘qgy6t “USN Wor) ’sa1098 JWsUOdutOD [RIUSUTUOLAUD Jo BUR ‘,——9 ‘puR[sseIT 4LeYD Jo UONEUIPIO [eUONRaTaA e UO JUaUOdUIOD |e UAUIUOMAUA PUODAS (q) PUL 1814 (F) J0J SalO9s Jo WIONe_ ‘SL “Bly296 CHAPTER 9 the relative importance of environmental attributes and of those most closely correlated with each species. Gittins (1968, 1969) examined the limestone grassland data he had earlier analysed by other procedures (Gittins, 1965a, b, c) (Fig. 76). His earlier recognition of two major trends of variation related to soil depth and soil nutrients respectively was confirmed by the high loadings of soil depth on axis 1 and of soil phosphate and soil potassium on axis 2. The inclusion of environmental attributes in the analysis imposes the constraint that the data must be centred and standardized by attribute*, and this may not ePbertoloni = 1 G verume glomereta, ‘soil P 0-4 Soil k o2| © mpexo H pubescons en cosion oF 95 Brace) 1 repens TErsengorum @6 Sporiem AR squorrosise p © Soil depth i “oz peru e Tprotense H tonots Caray wit radicats elt rinino |& 000MM, «cucumbers @ .meaio 4. tenuis 04 eH pilosella @H.chamoecistus eLeompetiit —_ng 9c covootyleo 8 P lorceotors ot. comicatonse © Psonguisorba © chocco Fig. 76. Attribute ordination of species and three soil variables for stands of limestone grassland. (From Gittins, 1968, by courtesy of Journal of Ecology.) * Walker and Wehrhahn (1971) (sce also Jeglum et al., 1971) made the interesting modification of dividing the values for environmental variables by an arbitrary large number before entering them in the data for principal component analysis ‘thereby inactivating these attributes in the analysis (but retaining them for calculation of loadings and ensuring that the extracted axes are determined by the vegetation data alone’ (Dye & Walker, 1980). The environmental data were scaled before analysis to be of approximately the same order of magnitude. The principal component analysis was of otherwise unstandardized data; the need to standardize the data by attribute was thus avoidedVEGETATION AND ENVIRONMENT 297 Table 27. Factor analysis of 50 stands of agricultural grassland based on yield of grass and various environmental variables (Ferrari er al., 1957) Loadings on four factors 1 2 3 4 1 Soil pH 020° 024 «= 0-41 0.09 2 Organic matter content 038 075 «= 002-015 of soil 3 Clay content of soil —0:10 0-04 = —020 4 Fine sand content of soil O17 013 013 5 ‘U figure’ (soil surface) 0-04 0-14 013 6 P status of soil 0-42 0-45 —0-20 7 K status of soil 0-62 006 © —0-04 8 Mg status of soil 0-27 0-44 0-02 9 Thickness of humus layer 043 002 022 10 Distance from farm —0:34 —0-09 —0:34 11 Water table 052 009 001 12 Fluctuation in water table —0-29 —0-05 017 13 N fertilizer 059 045 Ol 14 P fertilizer 037 084 = 0-00 15 K fertilizer 043 071 004 16 Pasture quality 057 004 020 17 N content of grass 0-36 0-00 063 18 PO, content of grass 0-47 O14 O58 19 K,O content of grass 068 008 030 20 MgO content of grass —003 022 Os 21 Spring yield of grass (% of 090 -009 -031 annual yield) 22 Spring yield of grass 098 = 003. 002-011 23 Autumn yield of grass 075 007 «003 038 Percentage variance accounted for 24-6 IS1 92 65s be appropriate to the vegetational data, and introduces the general difficulties of choice of environmental variables. Ferrari e¢ al.’s (1957) analysis of agricultural grassland (Table 27) provides an example of a component of environmental variation with little relation to yegetation. The vegetation attributes included (annual yield, spring yield as percentage of annual yield, magnesium, potassium, phosphorus and nitrogen content of grass and pasture quality) all have low loadings on the second axis, which nevertheless accounts for over 60% as much variance as the first axis. It seems unlikely that a (V + E)* approach will be helpful unless there is already considerable information about which environmental attributes are298 CHAPTER 9 most relevant or, as Noy-Meir (1970) suggested, integration of vegetation and environment is specifically required in the definition of land systems. The examination of correlation between components of separate principal component analyses of vegetational and environmental data (V*/E*) leads on the theoretically promising approach of canonical correlation analysis. The latter involves the extraction of components in each set of data such that the correlation between equivalent components of the two sets is maximized ((V/E)*). Thus species with high loadings on the first vegetational component will be those whose occurrence is most closely correlated with environmental variables having high loadings on the first environmental component. Gittins (1979) has provided a detailed account and assessment of the technique, the application of which to vegetational data was suggested independently by Dagnelie (1961) and Hughes (1961). Trials with field data have mostly not been rewarding (Austin, 1968b; Barkham & Norris, 1970; Gauch & Wentworth, 1976) though Gittins’ assessment is more encouraging, at least for data with a broadly linear and continuous structure. Barkham and Norris (1970), examining data from woodland, obtained an interpretable analysis, but concluded that correlation of the components from separate principal component analyses was more informative. Kercher and Goldstein (1977) described a modification, ‘canonical group correlation’, operating on pre- viously established groups of stands rather than individual stands. Canonical correlation analysis is subject to the difficulties of non-linearity in vegetational components and of choice of variables and possible inclusion of environmental variables with little or no effect on the vegetation in environmental components, already discussed. Barkham and Norris (1970) further suggested that correlation between vegetational and environmental components may be non-linear. Williams and Lance (1968) suggested a possible strategy of separate ordination of vegetation and environment, followed by a canonical correlation analysis of the relation between them (V*/E*)*). As Noy-Meir (1970) suggested, preliminary definition of vegeta~ tion and environment in terms of a small number of components could allow easier extension of the canonical analysis to include non-linear relationships.CHAPTER 10 Practical Considerations The previous chapters have attempted to summarize the very wide range of techniques available for the analysis of data on the composition and variation of vegetation. Many techniques have not been used in other than very limited trials, often in simplified field situations or with unrealistic simulated data. There has been some discussion of the considerations involved in the choice of appropriate techniques for a particular investigation (e.g. Goodall, 1970; Greig-Smith, 1971a; Lambert, 1972; Lambert & Dale, 1964; Noy-Meir & Whittaker, 1977) but the greatly increased speed and capacity of computers as well as the further development of methodology have made much of the earlier discussion of limited value. The form of the data to be collected and the type of data analysis to be used in a vegetational study are closely interrelated. The type of data to be collected obviously depends on the question or questions being asked. The method of analysis to be used is equally dependent on the objective of the investigation, but the practicality of collecting appropriate data may place constraints on the choice of method. In contrast to many biological fields there is rarely technical difficulty in obtaining data of the desired form, but the cost in terms of time may be unacceptable, or there may be other objections, for example to destructive sampling such as that involved in biomass determinations. Acceptance of a suboptimal type of data may involve a modification of method of analysis. Because of the heavy investment of time almost always involved in data gathering before any worthwhile analysis can be made, it is important that superfluous gathering of data be avoided. Against this, the mechanics of field work may be such that it may be expensive to return to sites and it may be better to gather data that may be required at a later stage, even at the risk that it may not be used. The nature of vegetation places constraints on the options available. For example, any method of analysing pattern that depends on recognition of discrete individuals is of limited use because many species are not amenable to recording in this way (p. 5), or such density measures may be biologically unrealistic because individuals vary very widely in size. Moreover, in practice, many pattern studies require concomitant assessment of pattern of environ- 299300 CHAPTER 10 mental factors (cf. Greig-Smith, 1979), and the method of analysis used must be capable of handling environmental as well as species data Problems of choice of method are greatest in relation to investigations of variation in overall composition of vegetation. Three interconnected decisions are required: the type of data to be collected; the layout of stands; the type of analysis, whether by classification or ordination, or both, and the precise procedure of classification or ordination. These are in turn influenced by the objectives of the investigation. Three principal possible objectives may be identified: simplification of the observed range of vegetation as a basis of mapping or inventory, either as an objective in itself or as a basis of management; search for the existence of discontinuities in the range of composition; detection of correlation between composition and environ- mental (in the broadest sense) factors as an aid to generation of hypotheses about the causation of differences in composition (Greig-Smith, 1980). Though it is possible for an investigation to serve more than one of these objectives, to do so is likely to require compromise in choice of method, so that a firm decision at the planning stage is important. Possible forms of data to be collected from each stand range froma simple species list, through estimates of the contribution of each species to the total vegetation of the stand (by one or more of a variety of measures), to records of some aspect of the size or performance of each individual plant (as in some forest enumerations). The cost, in terms of time, of recording a stand varies greatly. Preparation of a species list is, in most vegetation, very much quicker than any form of quantitative measurement, but different forms of quantitat- ive measure themselves vary in the time required. It is commonly assumed that the recording of quantitative measures adds greatly to the time taken to record a stand (e.g. Lambert & Dale, 1964) but there is little precise information available on how serious this is. Noy-Meir (1971p) estimated, rather than measured, cover, density and height of each species in semi-arid vegetation in south-eastern Australia and found that this approximately doubled recording time for a stand; precise measures, rather than visual estimates, would have greatly increased the time necessary. The extra labour involved in recording quantitative data will clearly vary with the type of vegetation. If large trees only are being recorded in species-rich rain forest, where it is necessary to examine each individual separately for identification, the additional time needed to record density and diameter is negligible; ina moderately species-rich limestone grassland the time needed to note the species present is negligible compared with that needed to obtain quantitative data. The additional labour involved depends also on the measure used, e.g. frequency is normally quicker than a cover measure. If aPRACTICAL CONSIDERATIONS 301 fixed time is available for gathering data, the time taken to record a stand must be considered in relation to that needed to move between and demarcate stands, which may be considerable in extensive surveys in difficult terrain. If the time saved at each stand by recording presence and absence only is relatively small compared with the time needed to travel to a further stand and demarcate it, the gain in number of stands recorded may not outweigh the loss in total information resulting from the lack of quantitative data. The data from each stand should be the minimum in terms of time required that will be adequate to answer the questions posed. This will depend on the range of variation being examined and the level of detail of variation that is of interest. The relative contribution of qualitative and quantitative data have been discussed above (p. 224), but it is worth emphasizing that qualitative data only (i.., a species list from each stand), even when not as efficient as some form of quantitative measure, may lead to an adequate sorting of the stands. Although this may appear surprising, the reason why it should be so is clear. Consider the simplified situation of a single gradient of environmental factors to which species show unimodal response curves (Fig. 77). Although an 2 4 5 3 6 8 7 0 9 10. ul Fig. 77. Unimodal response curves of species along an environmental gradient with corresponding species presence.302 CHAPTER 10 observed amount of any one species in a stand gives a more precise indication of the position of that stand on the gradient than its presence would do, any particular combination of species presences also corresponds to a short segment of the gradient only (provided a reasonably large number of species are involved). If no species have a high representation in any part of the data, analysis of a quantitative measure is likely to be dominated by the species with the greatest ranges of representation (unless an appropriate standardization is used) and a quantitative measure may give a /ess satisfactory analysis than presence/absence (Austin & Greig-Smith, 1968). If a quantitative measure is to be used, the choice of measure depends firstly on the objective. For example, if interest centres ori productivity of different species, either a direct measure of biomass, or one which correlates well with it, must be used. Only if a proposed measure is satisfactory on this account, should considerations of economy of time determine the choice. A further consideration is that partially bounded data (p. 18) may give results that are so similar to qualitative data that the greater cost is not balanced by the increase in efficiency (Smartt et al., 1974). This is a disadvantage in the use of frequency, otherwise attractive on account of its relative ease of recording. It should also be borne in mind that completely bounded data are, in effect, standardized by stand and this may or may not be desirable. In complex vegetation, consideration should be given to whether it is necessary to record species of all growth forms. It has long been customary to examine forest composition in terms of the tree layer only, and comparison of analyses of various sets of growth forms with analysis of the complete angiosperm complement confirmed the validity of this practice for one set of data (Webb ef al., 1967). On the other hand, inclusion of a greater range of life forms increases the information available from a given sample area and thus allows the use of a smaller stand (Hall & Swaine, 1976). The problem of whether to record all species has mainly been considered in relation to forest, especially tropical forest, but similar considerations apply, for instance, to the inclusion of bryophytes in data from much temperate vegetation, though this is rarely discussed. Before leaving the question of type of data to be recorded, it must be emphasized that quite different considerations apply to description of a vegetation type as a starting point for understanding structure and function- ing. This will always need more information than is needed for efficient derivation of classification or ordination. If fuller descriptions of the vegetation are needed, it may be profitable to delay these and examine in morePRACTICAL CONSIDERATIONS 303 detail a more limited number of stands selected on the basis of the initial analysis. If qualitative data only are to be collected, no problems of sampling within the stand arise—the whole stand must be examined. Apart from tree density and basal area, normally enumerated over the whole stand, a quantitative measure will normally be estimated by samples taken within the stand. The placement of samples to ensure an unbiased estimate has been discussed in Chapter 2. Although the estimate must be unbiased, only if tests of significance between estimates for different stands are to be made, so that a standard error of the estimate is required, need the sampling be random. Other sampling schemes, ¢.g. systematic or stratified unaligned systematic (p. 23), may give a more precise estimate, though at the cost that the precision cannot be estimated. In most survey investigations subsequent significance tests are unlikely to be made. This contrasts with field experiments in vegetation, where subsequent analysis does normally involve significance tests and appropriate random sampling within stands is then necessary. Stand size is rarely critical, though broad limits can be drawn. Stands must be large enough to cover all phases of any mosaic pattern that may be present, otherwise analysis will reflect this pattern and may be dominated by it. Conversely, if ordination is being used to elucidate pattern within otherwise homogeneous vegetation (p. 103), stands must be small enough to include one phase only of the mosaic. Stands should be large enough to include a reasonable proportion of the total species complement, particularly if subsequent standardization of the data will emphasize rarer species; pre- liminary examination of a species/area curve may be useful. They should not be so large that a single stand is obviously including markedly different habitats". It is not even essential that all stands should be of the same size or shape, though this is desirable if the nature of the habitat and vegetation permits. Sometimes, however, this is not possible, e.g. in study of cliff ledge vegetation (cf. Bunce, 1968) or of relict vegetation in a predominantly managed landscape. If stands of varying size are recorded qualitatively the subsequent analysis will be influenced by stand size through the effect of size on species number. With quantitative measures, unless standardization emphasizes rare species, the analysis will be relatively little affected, but any * Goff and Mitchell (1975) compared ordinations of individual plots or point- centred samples within large stands (up to several hectares) with ordinations based on the lumped data from stands and concluded that they showed little difference. These results should not be taken as a recommendation of such large stands, but do suggest that forest inventory data from large stands may validly be used in vegetation studies.304 CHAPTER 10 subsequent interpretation in terms of species- caution. Williams e7 al. (1969) used the nth nearest neighbours to an individual asa ‘stand’ in a study of pattern, classifying the overlapping ‘stands’ centred on all individuals in the single plot examined. This approach might be useful in broader studies of forest composition, eliminating the often time-consuming delimitation of plots, and equalizing the amount of information obtained from stands when tree density is variable from stand to stand. Another alternative to the use of plots is to use single transects as stands. Robertson (1977) showed that the results from qualitative data, of species touching a line transect, in sand dune vegetation were comparable to those from plot stands. If full description of the vegetation is required, rather than the minimal data needed for efficient analysis, a larger stand will generally be necessary both to include the majority of species and to allow reasonably precise quantitative estimates for more of the species. Density estimates present a particular problem in description; since the accuracy of a density estimate, even for random distribution of individuals—the most favourable case— depends on the number of individuals counted (p. 26), even complete enumeration may give low precision for all but the commonest species. The problem is acute in vegetation, such as much tropical rain forest, where no species have high densities, and the size of homogeneous stand theoretically desirable may be unattainable (Greig-Smith, 1965). The considerations involved in the siting of stands have apparently often been poorly understood. This may, in part, be because the situation is deceptively similar to the design of sampling surveys and censuses aimed at estimates of some parameter for a heterogencous area (Greig-Smith, 1971c). Whether the situations are really comparable depends on the objectives of the vegetation survey. If the objective is primarily to simplify the observed range of vegetation, constraints on the siting of stands depend on what is required of the results. If the aim is inventory, e.g. to determine what proportion of the total area is occupied by different vegetation types, the stands must be sited in such a way that an unbiased estimate of amounts of different species or vegetation types in the area as a whole is obtained; the stands are themselves regarded as samples. A form of systematic sampling is likely to give a more precise estimate unless the scale of the sampling grid happens to coincide with that of periodic variation in the vegetation. Any risk of this happening is likely to be evident in the field. Smartt and Grainger (1974) have demonstrated the greater efficiency of stratified systematic unaligned sampling, but in some vegetation types, e.g. forest, the much greater time needed to site stands on this system ichness must be viewed withPRACTICAL CONSIDERATIONS 305 may make it uneconomic; greater efficiency may result from more stands sited strictly systematically. Only if confidence limits for estimates are required, cither to test for differences between parts of the area being investigated, or to compare with other areas, is random siting necessary. If the primary aim is to produce a vegetation map, stands should be placed as evenly as possible so that the uncertainty of the position of boundaries between one vegetation type and another is more or less constant over the area, i.e. strict systematic siting should be used. If the ‘grain’ of the vegetation, the average size of patches of different vegetation types, is evidently different in different parts of the whole area, it is useful to stratify the area and site stands systematically at an appropriate spacing in each stratum. If the primary objective is to search for discontinuities in the range of composition of vegetation, systematic siting is better avoided, because even relatively slight correlation between grid interval and periodic variation will tend tosharpen the boundaries between noda. However, critical delineation of discontinuities in composition is unlikely to be the principal or only objective and appropriate analysis presents problems (see below); useful indications of the discreteness or otherwise of the vegetation types may be obtained from systematically sited stands. If elucidation of correlations between vegetation and environment is the primary objective, stands should, as far as possible, include all variants of vegetation and an equal representation of all variants. No mechanical system of siting can achieve this dual objective. Both random and, with the usual proviso about periodic variation in the vegetation, systematic siting will, if sufficient stands are examined, ensure that all variants are examined, but unless variants are equally represented in the area, more information will be obtained about some variants than others. There is no complete solution to this dilemma. If there are obvious broad differences in the vegetation, stratification in relation to these will be helpful*, but the same difficulties apply to variation within the broad types. If there is reason to suspect that obvious environmental difference may play an important part in determining the composition of the vegetation, even if corresponding vegetational differences are not obvious, it is sensible to stratify in relation to these and, for example, in undulating topography ensure that equal numbers of stands are sited in valleys, on slopes and on ridges. Note that stratification here aims at equal numbers of stands from different strata, not equal density of stands. There is no objection to including stands because they appear to be in some * In contrast to the undesirability of stratifying samples rather than stands in relation to major vegetational differences (p. 22).306 CHAPTER 10 way unlike those already recorded, provided that no attempt is made to draw conclusions from the subsequent analysis about the sharpness of community boundaries or the relative abundance of different vegetation types. A workable approach, if the terrain allows, is to record stands initially by a set scheme, stratifying as above if appropriate, and then add any stands that appear to be different from those already recorded*. The major decision over analysis of the data is between classification and ordination. This depends partly on the objective of the investigation and partly on the nature of the data. Preparation of a vegetation map or of an inventory of vegetation clearly requires a classification. Classification can be used to identify correlations of vegetation with environment by comparing the average environmental characteristics of the two subgroupsat each division in the hierarchy. It does, however, lose information on any correlation between members of a single final group and their environments. This information is, in principle, retained by ordination, but in practice the more detailed correlation may not be detectable if the data represent more than one independent or partially independent gradients of composition. This suggests, for very heterogeneous data, an initial classification down to a level that provides reasonably homogeneous groups, followed both by an ordination of the groups to clarify the relationships between them and by separate ordina- tions of the members of each group (cf. Greig-Smith e# a/., 1967), or Noy-Meit’s (1971b) nodal ordination (p. 264). In practice, classification and ordination, considered separately, often both contribute to understanding of a data set. Choice of technique of classification or ordination is not easy; there is no best’ method. Every simplification of a complex set of data involves ignoring some of the information in the data, and different techniques preserve different information, as is evident from the analysis of the same data set by different techniques (e.g. Austin & Greig-Smith, 1968; Crawford et a/., 1970; Ivimey-Cook e¢ al., 1975; Robertson, 1978; Clymo, 1980; del Moral, 1980). Which technique is most appropriate to a particular investigation will depend on which retains the most relevant information, but it is often precisely this which is not known in advance. There is therefore much to be said for analysing data by several techniques. There has been considerable discussion of the efficiency of different techniques’. Though an understanding of the assumptions and procedures of * Useful discussions of sampling schemes actually used in extensive surveys have been given by, among others, Bunce and Shaw (1973), Curtis (1959, pp. 63-79), Hall and Swaine (1976) and Noy-Meir (1970, 19716). * Noy-Meir and Whittaker (1977) gave an excellent account of the merits and limitations of techniques up to 1976.PRACTICAL CONSIDERATIONS 307 a technique may indicate its suitability or otherwise in a particular case, there can be no objective assessment of efficiency (Greig-Smith, 1980). How far the results confirm preconceptions is no test of efficiency, though many discus- sions implicitly adopt this criterion. Blackith and Reyment (1971) put it well *....., there are no objective criteria against which the classifications can be judged. There is therefore a tendency for multivariate techniques to be condemned when they disagree with conventional methods, and regarded as superfluous when they agree’. The investigator is, perhaps, less likely to have preconceptions about the ordination of a set of data. Efficiency of an ordination has sometimes been judged by the extent to which the variation in the initial data is ‘accounted for’ in the ordination, but the information retained is not necessarily the most useful; even a low percentage retention of the initial variation may be very informative if there is a large amount of irrelevant information in the original data. An alternative approach to assessment of efficiency is to apply a technique to a set of simulated data and see how far the original structure of the data is retrieved. It is commonly assumed that species response curves to an environmental gradient have a Gaussian form, but there is good reason to doubt whether this is generally so (p. 269). It is thus difficult to produce realistic simulated data related to a single environmental gradient, and more so for response to two or more at least partially independent gradients, a situation more like that commonly found in the field. A common misconcep- tion about ordination techniques is that interpretation must be in terms of the axes, ie. that an axis represents the response of the vegetation to an environmental gradient. This is unnecessarily restrictive; if a pattern of an environmental factor is apparent on a vegetational ordination either by a straight line at an angle to the axes, or by a curved line, there is evidence of correlation between composition of the vegetation and that factor. With only slight exaggeration, it may be said that the axes of an ordination are mere constructional artefacts—it is the display of relative stand positions that is valuable (Greig-Smith, 1971c). Although direct tests of the efficiency of techniques cannot be made*, they can be judged, albeit subjectively, by how satisfactory they prove to be as working tools and how productive they prove to be in generating hypotheses which can then be tested by further field observation or by experiment, e.g. if analysis of survey data suggests a relationship between composition of vegetation and available level of an inorganic ion in the soil, an obvious * Wilson (1981) suggested tests for the consistency of ordinations of random subsets of a data set. These test whether an ordination represents a real property of the data, but not whether it is retaining the most relevant information.308 CHAPTER 10 experiment is to examine the performance of selected species grown in different concentrations of that ion. In the earlier applications of numerical methods, the amount of compu- tation required was a serious obstacle to the use of some techniques. With the rapid increase in speed and capacity of computers this is much less often a constraint. Thus association analysis was put forward not as the most satisfactory technique that could be proposed at the time, but as one that was computationally feasible. Now, with its inherent disadvantage of built-in standardization, liability to reversals in the hierarchy and tendency to chain, its use would not be seriously considered. Although the computational load is now a less serious consideration, the simplest technique that will meet the requirements of the investigation should be used; just as the gathering of data in the field has a cost in time and expense, so has the subsequent data analysis. The question also arises whether field data, at least in extensive surveys, are accurate enough to justify very refined methods of analysis (Greig-Smith, 1980); Hall and Okali (1978) demonstrated the considerable extent to which data in a complex vegetation type are affected by season and by the experience of the observers. Little has been said above specifically about analysis of data when the objective is the detection of ‘real’ entities. In the strict sense of detecting complete discontinuities in the range of composition, if they exist, recognition of such entities is not possible by numerical analysis. Where the expressed aim is detection of real entities, the objective is a less demanding one, to identify species combinations of much commoner occurrence, intermediates between which are very rare. This implies the erection of a classification which will have general validity for other stands of comparable vegetation and is more demanding than the extraction of noda to provide convenient reference points in mapping or inventory (and still more so than the use of classification on a particular set of data as a means of examining correlation of composition with environment). Superficially it might appear that this could be achieved by examining the degree of clustering. In a classification those stands which fuse ata low levelin a hierarchy and are only linked to others at a much higher level would be recognized as forming a real entity. Alternatively, such an entity would be recognized on an ordination as a cluster of points more or less isolated from others. Extreme caution must, however, be exercised. It is abundantly clear that the degree of clustering depends on the exact technique of analysis used and is a property of the technique rather than of the data; compare, for instance, the effect of nearest-neighbour and furthest-neighbour sorting in agglomerative classification (p. 205-6). If a classification of general validity is needed, a considerable number ofPRACTICAL CONSIDERATIONS 309 stands should be included to minimize the effects of chance in producing groups, and classes represented by a small number of stands only should be regarded as tentative. If at all practicable, the analysis should be repeated ona separate sct of similar number of stands, ideally from a different area, and only if the two results are comparable should the classification be accepted as of general validity. Two particular situations, not covered above, call for brief comment. The most direct approach to elucidation of change in vegetation with time is to record the composition of permanent plots at successive intervals. The information that results is limited to the length of time over which observations are made. If plots initially at different stages in a succession are available and are recorded over one or more time intervals ordination can be used to summarize the information on change. Each recording of a plot is regarded as a separate ‘stand’ and a suitable ordination is produced. If the course of change is comparable for different plots, their trajectories in the ordination display will be closely parallel for plots at the same stage and form a sequence for plots at different stages. It is thus possible to reconstruct the whole sequence of changes from shorter term observations and to detect any tendency for plots to converge or diverge in composition (Van der Maarel, 1969; Austin, 1977) (Fig. 78). In the special case of forest vegetation, where tree diameter for a particular species can be regarded as a measure of age, even one set of observations on stands at different successional stages can be used to elucidate the course of change (Goff, 1968; Goff & Zedler, 1968, 1972; Carleton & Maycock, 1978). Swaine and Greig-Smith (1980) suggested an alternative ordination approach where a series of observations on the same plots are available. From a matrix of species x time for cach stand an appropriate cross-product (covariance or correlation coefficient) is calculated for all pairs of species. The values for each species-pair are summed and used as input to principal component analysis. Both species loadings and stand scores can be calculated in the usual way and differential behaviour of stands can be detected. The method has the advantage that initial differences between stands are ignored and analysis concentrates on change, but is only applicable to relatively homogeneous sets of sites, on which the correlations between species may be expected to be the same. Another approach based on observed short term changes is to estimate transition probabilities for change either from an individual of one species to an individual of another species (as for forest trees, e.g. Enright & Ogden, 1979) or from one vegetation type, defined by a preliminary classification, to another (e.g. Austin, 1980b; Usher, 1981). From the resulting transition310 CHAPTER 10 320. Trampled 280 Untrampled Fig. 78. Ordination of eight individual quadrats on a lawn, recorded on six successive occasions. (From Austin, 1977, by courtesy of Vegetatio.) matrix alternative pathways of change can be elucidated ( Fig. 79). Successive multiplication of the initial composition of the vegetation of an area by the transition matrix can then, on the somewhat limiting assumption that the transition probabilities remain constant, be used to make predictions of the future composition. Usher (1981) has a useful critical assessment of this Markovian approach. It is potentially useful especially for elucidating the later stages of succession, where change at any one point is not necessarily unidirectional, and the cyclic changes in vegetation having a mosaic structure of different phases (Watt, 1947) (cf. Legg, 1980). If the objective of analysis is to identify optimal boundaries between vegetation types in the field (as opposed to discontinuities in the range of composition of the vegetation) a form of agglomerative classification in which fusion is only permitted between stands which are adjacent on the ground might be used. Such methods have been used in the analysis of pollen- analytical and other palacoecological data from profiles (Gordon & Birks, 1972; Birks & Birks, 1980). These methods could be applied directly toPRACTICAL CONSIDERATIONS 311 Ungrazed Grazed Fig. 79. Diagrams showing main transition probabilities between cight vegetation types, based on a preliminary classification, for grazed and ungrazed grassland. (—) Largest 15% of probabilities; (--) next 15% of probabilities. (From Usher, 1981, by courtesy of Vegetatio.) transect data and comparable methods allow fusion of adjacent stands on a grid (see Gordon, 1981). Little or no mention has been made in this chapter of tests of significance or probability statements. Tests of significance are appropriate when a hypothesis is being tested, but this is rarely the case with survey data. The objective is much more commonly either the erection of an empirical framework of description, to be judged on its practical utility, as with some uses of classification, or data exploration asa means of generating hypotheses. In neither case is there a null hypothesis to be tested and questions of significance testing and probability do not arise.Appendix: Tables Table 1. Variance of binomial distribution after angular transformation. (a) For various values of n (In part from Robinson, 1955) Approximate value 04 03 02 Of (2) n os 06 07 08 09 a 2 10125 972.0 850-5 648.0 364-5 4103 3 5757 566-4 530-8 449-3 2888 273-6 4 3656 = 369-4 3726) 348R 2501 205-2 5 2536 = 261-6 = 279-2 2828 = 2221 164-2 9 1067 1106 = 1233 148-1 1489 912 10 94-0 966 1064 1259 1360 82:1 16 552 55-7 578 658 816 513 20 434 BS 445° 488 63:5 410 25 343 346 35:3 36:7 473 328 30 28:2 28-4 28:7 299 367 274 Table 2, Mean range of samples of different sizes (after Pearson & Hartley, 1954). Range expressed in units of population standard deviation Sample Sample Sample Sample size Range size Range size Range size Range 2 11289 2970 16 353223 3-858 3 1693 10 307817 3588 24 3895 4 2059 317318 3640 25 3931 5 23260 12 325819 3689 50 4-498 6 253413 3336-20 3735 100 S015 7 270414 340721 3-778 8 284715 3472-22 3819313 APPENDIX: TABLES 9ST DEST 6 EST EST 6PST LSst 99ST O-LST 8857 7-097 v0 9-197 E97 897 9-997 $897 T0LT OTLT 6 ELT LSlT SLLT £0 T6LT7 L087 787 EE87 E787 O87 Py S8T S587 1-$87 TH8T 70 8-787 8-087 BLT 9+bLT £017 TSI 06ST LAST TET f€7 10 1-7 £607 8-61 BLT £091 66E1 tLI 1-76 £9 Ltt 0 O10 600 800 100 90-0 0-0 10-0 £00 7-0 10-0 d (paystjqndun ‘uooy “f “y) $ = ¥ 405 919% popudixg (4) C(Q) 1 atqny xipuaddy )APPENDIX: TABLES 314 $0000. 90000-90000. 40000-_ L0000-_- 80000-60000: 01000--— 11000: Z1.000- 6 +1000: ‘$1000- L1000- 81000- 07000 77000 $Z000° £7000 0£000- ‘£000 8 L£000: TP000 SP000- 0S000° $S000° 19000- 89000 $4000: €8000° 16000- L 0100: T100- T100- 100° S100- L100- 8100- 0200- 7Z00- $Z00- 9 L200 0€00- €£00- ££00- Tr00- ‘Sb00- 0s00- $S00: 1900: £900: s 400: 800° 1600- TOI0- II10- €710- 9ET0° OsT0- 9910- £810° Y 7020: ‘7Z0- Lbz0- €120- TOEO- veto: 69€0- 8060: Ost0- 86P0- £ Osco: 2090: Z190- £6L0- 1280- 4060: £001- 801T- SzzI- ecer- t 9661 S91 Lzst- 6102- Tezz- 9982. Stuz- z10¢ 6zeE GL9E- I 990P: orp 996% 88PS- $909 £019: 80PL: L818- 8P06- 00001 =o 60 80 L0 90 so 0 £0 70 10 oo mu E8pL 8SSL- PEL TILL 88LL: 9981: SPOL- S708 9018- L818 ra) 0Lz8: £seg- LEpS: T7s8- L098- 698: T8L8- 6988: 8968 8h06- 10 6ET6. 1&7 v7e6- 8Ib6- ZIS6- 8096- POL6- 2086- 0066- 00001 =o. 600 3800 400 90-0 0-0 10 £00 00 100 om wi? “ENGELAPPENDIX: TABLES 315 Table 4. N 10 “4714 100 1421 20 “3244 150 LLS9 30 “2626 200 1003 40 2265 250 “08962 50. -2020 300 08179 60 “1841 400 07080 70 “1703 500 06331 80 “1591 1000 04474 90 “1499 Significant points for @ = 2nonz/n,* (no, m1, z = number of quadrats containing 0, 1, 2, individuals respectively) (from Moore, 1953, by courtesy of Annals of Botany) N = total number of quadrats _ Rg try +n N R x 100 The figure given is the mean of ¢ plus twice its standard error, ie. approximately the 5% point Mean number- per quadrat 05 10 15 20 25 R 99 92 81 68 54 270 240 246 2-66 298 216 195) 1993234 1:80 166 168 1-77 1-92 165 153 SS 162-7 155 146 147 154 1-63 149 141 142 148 1:56APPENDIX: TABLES 316 8L069 SIT = «THORS SITS EROTS «= CODTTS «BING BTR SOIL 6-66 8:66 £66 9-66 $-66 7-66 £66 7-66 166% SOV OTIS = DOSE RITE LSHHZ ETB «= LHS ~—«sLSTSZ—CSLOMT:«SCMROETCCG ELOCZ —LOTTT_—TOPOT_—=199GT = TORT. «= OZERT. = OZLLT_— HILT = -LO99-1 609-108. 909ST TPIS T = LO9HT LTT = CORBET = TAPED «= CODE = OLLZT LETT. «OPO. OL CILIT = POETT —LROT-T = BROT = 86P0T LITO «= £BH6 9496 OTH TD 9168: S198: Ors Ize: S86L S9LL- OSL: OPEL «ETL! «TED s £EL9- 659: 6rE9- Z919 8465: 86LS- 129s Lipps. OLTS- OTS! th £r6t- ORL 079: £9bb- 80€b: Sip S00 SRE Te L9SE o¢ Ste S87e Lele: Toe: LL8Z: Pelt: P19 SRT! LSE: lew Ot Lolz: S861 £98 bell Szor 80ST. €6ET BLTT- SOIT SOL OT £760: r80(9ZLO- 6190 = 150-800" SOO Z0Z0: 1010- 0 0 6 8 L 9 s € z 1 0 %A Svonngisisip wopues wi sorouenboxy eBeyussied yuasayjip 0} Burpuodsass0o sanisuaq *9 a|qe317 APPENDIX: TABLES LOT 80-1 801 60-1 anury aod 9) £60 760 160 160 yay s9MOT EST 0871 ZEIT Pz01 wopaasy Jo saasiiag OOF OT OT OT TET eh eb et St LET yruny sadd) 160 160 060 680 680 88:0 880 180 980 S80 $80 wun 19M07 096 968 89L_ OD STIS ORHSBHHCBE COE. BRT wopeasy jo soas9q OL IT kT FT OTT OTT LTT ORT ELT SET SET OFT wuny s0dd) 780 180 60 80 LL0 S20 SLO $0 140 010 890 190 Vu J9MOT vet 61M: RTO]: TT Hi LC aD wopaasy Jo soaazaq OFl FhT BRT IST SST LST OST OT LLT SLT O8T ERT yuauy sod p) 99) 990 190 650 LEO 950 SSO 70 8hO SrO chO ZrO wun] 12M07 9S 8h hE ET $I wopoaay Jo saaatoq L81 P61 SOT INT GIT OTT IT LST BLT TE 69E ZS wun sod, Or0 SEO ZED OED L7H FD «ZO LTO 710400 £00000 du 1207 rl owls 8 L 9 s roo z I wopaday Jo saaataq (4601093 fo jounor Jo ksarin09 £q ‘q1961 “YUG-2191, Wosy) WONed Jo siskyeue ur sourier paisedxe 01 peaiasgo Jo ones ay) jo UoNNgUISIP WopUES 40} SIU] BUaPYUOD "66 “L a1dUL,318 APPENDIX: TABLES Table 8. Probability that § (for 2) attains or exceeds a specified value. (Shown only for positive values. Negative values obtainable by symmetry.) (From Kendall, 1948, by courtesy of Charles Griffin & Co. Ltd) Values of » Values of n Ss o4 5 8 9 S 6 7 10 0 0625 0592 0548 «= 054010500 500-0500 2 0375 0-408 0-452. 0460 303600386 0-431 4 0167 0-242 0-360 0381502385281 0364 6 0042 0117 0274 «= 0306S 7) 0136 0191-0300 8 0042 199023890068) 119-0242 10 00783 0138 «©0179 11.0028) 0068 0190 12 0089 013013. 0.0783 0035 (0146 14 0054 0090 «15 00714 0015 0-108 16 0-031 0060 «17 00754 0-078 18 0-016 0038 19 00714 0-054 20 0071 002221 00°20 0.036 22 00728 «0012-23, 0.023 24 00987 0076325 0014 26 00°19 0072927 0.0783 28 0025 0071229 0.0746 30 00°43 31 0.0723 32 00712 33 00711 34 0025 35 00747 36 00°28 37 00°18 39 0.058 41 00415 43 00°28 45 00°28 Repeated zeros are indicated by powers, e.g. 00°47 stands for 0-00047.References Aberdeen, J. E. C. (1954) Estimation of basal or cover areas in plant ecology. Aust. J. Sci. 17, 35-6. Aberdeen, J. E. C. (1955) Quantitative methods for estimating the distribution of soil fungi. Pap. Dep. Bot. Univ. Qd, 3, 83-96. Aberdeen, J. E. C. (1958) The effect of quadrat size, plant size and plant distribution on frequency estimates in plant ecology. Aust. J. Bot. 6, 47-58. Agnew, A. D. Q. (1961) The ecology of Juncus effusus L. in North Wales. J. Ecol. 49, 83-102. Anderson, A. J. B. (1971) Ordination models in ecology. J. Ecol.. 59, 713-26. Anderson, D. J. (1963) The structure of some upland plant communities in Caernarvonshire. III. The continuum analysis. J. Ecol. §1, 403-14 Anderson, D. J. (1965) Studies on structure in plant communities. I. An analysis of limestone grassland in Monk's Dale, Derbyshire. J. Ecol. 53, 97-107. Archibald, E. E. A. (1948) Plant populations. I. A new application of Neyman’s contagious distribution. Ann. Bot. 12, 221-35. Archibald, E.E.A. (1949) The specific character of plant communities. I. Herbaceous communities. J. Ecol. 37, 260-73. Archibald, E. E. A. (1949b) The specific character of plant communities. I. A quantitative approach. J. Ecol. 37, 274-88 Archibald, E. E. A. (1950) Plant populations. II. The estimation of the number of individuals per unit area of species in heterogeneous plant populations. Ann. Bot. 14, 7-21. Archibald, E. E, A. (1952) A possible method for estimating the area covered by the basal parts of plants. S. Afr. J. Sci. 48, 286-92. Ashby, E. (1935) The quantitative analysis of vegetation. Ann. Bot. 49, 779-802. Ashby, E. (1936) Statistical ecology. Bor. Rev. 2, 221-35. ‘Ashby, E. (1948) Statistical ecology. Il. A reassessment. Bot. Rev. 14, 222-34. Ashby, W. C. (1972) Distance measurements in vegetation study. Ecology, 53, 980-1. Ashton, P. S. (1964) Ecological studies in the mixed dipterocarp forests of Brunei State. Oxf. For. Mem. 28. Austin, M. P. (1968a) Pattern in a Zerna erecta dominated community. J. Ecol. 56, 197-218. ‘Austin, M. P. (1968b) An ordination study of a chalk grassland community. J. Ecol. 56, 139-57. Austin, M.P. (1971) Role of regression analysis in plant ecology. Proc. ecol. Soc. Austr. 6, 63-75. Austin, M.P. (1972) Models and analysis of descriptive vegetation data. In:320 REFERENCES Mathematical Models in Ecology (Ed. by J.N. R. Jeffers), pp. 61-86. Blackwell Scientific Publications, Oxford. Austin, M. P. (1976a) On non-linear species response models in ordination. Vegetatio, 33, 33-41. Austin, M. P. (1976b) Performance of four ordination techniques assuming three different non-linear species response models. Vegetatio, 33, 43-9. Austin, M. P. (1977) Use of ordination and other multivariate descriptive methods to study succession. Vegetatio, 35, 165-75. Austin, M. P. (1980a) Searching for a model for use in vegetation analysis. Vegetatio, 42, 11-21. Austin, M. P. (1980b) An exploratory analysis of grassland dynamics: an example ofa lawn succession. Vegetatio, 43, 87-94. Austin, M. P. (1981) The role of certain diversity properties in vegetation classifi- cation. In: Vegetation Classification in the Australian Region (Ed. by A.N. Gillison & D.J. Anderson), pp. 125-40. C.S.LR.O. and Australian National University Press, Canberra. Austin, M. P. & Noy-Meir, I. (1971) The problem of non-linearity in ordination: experiments with two-gradient models. J. Ecol. 59, 763-73. Austin, M. P., Ashton, P. S, & Greig-Smith, P. (1972) The application of quantitative methods to vegetation survey. III. A re-examination of rain forest data from Brunei. J. Ecol. 60, 305-24. Austin, M. P. & Greig-Smith, P. (1968) The application of quantitative methods in vegetation survey. II. Some methodological problems of data from rain forest. J. Ecol. 56, 827-44. Austin, M. P. & Orloci, L. (1966) Geometric models in ecology. II. An evaluation of some ordination techniques. J. Ecol. 54, 217-27. Barkham, J. P. & Norris, J. M. (1970) Multivariate procedures in an investigation of vegetation and soil relations of two beech woodlands, Cotswold Hills, England. Ecology, 51, 630-9. Barnes, H. & Stanbury, F. A. (1951) A statistical study of plant distribution during the colonization and early development of vegetation on china clay residues. J. Ecol. 39, 171-81. Bartlett, M. S. (1936) Some examples of statistical methods of research in agriculture and applied biology. Suppl. JI R. statist. Soc. 4, 137-83. Bartlett, M. S. (1963) The spectral analysis of point processes. JI R. statist. Soc. B, 25, 264-96. Bartlett, M.S. (1964) The spectral analysis of two-dimensional point processes. Biometrika, 51, 299-311. Batcheler, C. L. (1971) Estimation of density from a sample of joint point-and nearest- neighbour distances. Ecology, 52, 703-9. Batcheler, C.L. & Bell, D. J. (1970) Experiments in estimating density from point- and nearest-neighbour distance samples. Proc. N.Z. ecol. Soc. 17, 111-7. Beals, E. W. (1960) Forest bird communities in the Apostle Islands of Wisconsin. Wilson Bull. 72, 156-181. Beals, E. W. (1965) Species patterns in a Lebanese Poterictum. Vegetatio, 13, 68-87. Beals, E. W. (1973) Ordination: mathematical elegance and ecological naiveté. J. Ecol. 61, 23-35.REFERENCES 321 Beard, J. S. (1973) The physiognomic approach. In: Ordination and Classification of Communities. Handbook of Vegetation Science, 5, (Ed. by R. H. Whittaker), pp. 355-86. Junk, The Hague. Benninghoff, W.S. & Southworth, W.C. (1964) Ordering of tabular arrays of phytosociological data by digital computer. Abstracts Xth Int. bot. Congr. pp. 331-2. Benzecri, J. P. (1973) L’Analyse des Données. Vol. 2. L'Analyse des Correspondances. Dunod, Paris. Beshir, M. E. (1968) A study of the accuracy of estimates of the specific composition of vegetation. M.Sc. thesis, University of Wales. Bieleski, R. L. (1959) Factors affecting growth and distribution of Kauri (Agathis australis Salisb.). 1. Effect of light on the establishment of Kauri and of Phyllocladus trichomanoides D. Don. Aust. J. Bot. 7, 252-67. Birks, H. J. B. & Birks, H. H. (1980) Quarternary Palaeoecology. Arnold, London. Bitterlich, W. (1948) Die Winkelzihlprobe. Allg. Forst- u. Holzw. Zig, 59, 4-5. Blackith, R.E. & Reyment, R.A. (1971) Multivariate Morphometrics. Academic Press, London. Blackman, G. E. (1935) A study by statistical methods of the distribution of species in grassland associations. Ann. Bot. 49, 749-77. Blackman, G. E. (1942) Statistical and ecological studies in the distribution of species. in plant communities. I, Dispersion as a factor in the study of changes in plant populations. Ann. Bot. 6, 351-70. Blackman, G.E. & Rutter, A.J. (1946) Physiological and ecological studies in the analysis of plant environment. I. The light factor and the distribution of the bluebell (Scilla non-scripta) in woodland communities. Ann. Bot. 10, 361-90. Bormann, F. H. (1953) The statistical efficiency of sample plot size and shape in forest. ecology. Ecology, 34, 474-87. Bottomley, J. (1971) Some statistical problems arising from the use of the information statistic in numerical classification. J. Ecol. 59, 339-42. Bourdeau, P. F. (1953) A test of random versus systematic ecological sampling. Ecology, 34, 499-512. Bray, J. R. (1956) A study of mutual occurrence of plant species. Ecology, 37, 21-8. Bray, J. R. (1962) Use of non-area analytic data to determine species dispersion. Ecology, 43, 328-33 Bray, J. R. & Curtis, J. T. (1957) An ordination of the upland forest communities of southern Wisconsin. Ecol. Monogr. 27, 325-49. Brillouin, L. (1962) Science and Information Theory, 2nd edn. Academic Press, New York Brown, D. (1954) Methods of surveying and measuring vegetation. Bull. Bur. Past., Hurley, 42. Brown, R. T. & Curtis, J. T. (1952) The upland conifer-hardwood forests of northern Wisconsin. Ecol. Monogr. 22, 217-34. Bullock, J. A. (1971) The investigation of samples containing many species. I. Sample description. Biol. J. Linn, Soc. 3, 1-21. Bunce, R. G. H. (1968) An ecological study of Ysgolion Duon, a mountain cliff in Snowdonia. J. Ecol. 56, 59-75.322 REFERENCES Bunce, R.G.H. & Shaw, M. W. (1973) A standardized procedure for ecological survey. J. environ. Mgmt, 1, 239-58. Burr, E. J. (1970) Cluster sorting with mixed character types. II. Fusion strategies. ‘Aust. Comput. J. 2, 98-103. Cain, S. A, (1934) Studies on virgin hardwood forest. Il. A comparison of quadrat sizes in a quantitative phytosociological study of Nash’s Woods, Posey County, Indiana. Amer. Midl. Nat. 15, 529-66. Cain, S. A. (1938) The species-area curve. Amer. Midl. Nat. 19, 573-81. Cain, S. A. (1943) Sample-plot technique applied to alpine vegetation in Wyoming. Amer, J. Bot. 30, 240-47. Carleton, T. J. (1979) Floristic variation and zonation in the boreal forest south of James Bay: a cluster seeking approach. Vegetatio, 39, 147-60. Carleton, T.J. (1980) Non-centered component analysis of vegetation data: a comparison of orthogonal and oblique rotation. Vegetatio, 42, 59-66. Carleton, T. J. & Maycock, P. F. (1978) Dynamics of the boreal forest south of James Bay. Can. J. Bot. 56, 1157-73. Carleton, T. J. & Maycock, P. F. (1980) Vegetation of the boreal forests south of James. Bay: non-centered component analysis of the vascular flora. Ecology, 61, 1199— 212. Cassie, R. M.& Michael, A. D. (1968) Fauna and sediments of an intertidal mud flat: a multivariate analysis. J. exp. mar. Biol. Ecol. 2, 1-23. CeSka, A. & Roemer, H. (1971) A computer program for identifying species-relevé groups in vegetation studies. Vegetatio, 23, 255-17. Chadwick, M. J. (1960) Nardus stricta, a weed of hill grazings. In: The Biology of Weeds (Ed. by J.-L. Harper), pp. 246-36. Blackwell Scientific Publications, Oxford. Christian, C. 8. & Perry, R. A. (1953) The systematic description of plant communities by the use of symbols. J. Ecol. 41, 100-5. Clapham, A. R. (1932) The form of the observational unit in quantitative ecology. J. Ecol. 20, 192-97. Clapham, A. R. (1936) Over-dispersion in grassland communities and the use of statistical methods in plant ecology. J. Ecol. 24, 232-51 Clark, P. J. (1956) Grouping in spatial distributions. Science, 123, 373-4. Clark, P. J. & Evans, F. C. (1954a) Distance to nearest neighbour as a measure of spatial relationships in populations. Ecology, 35, 445-53. Clark, P. J. & Evans, F. C. (1954b) On some aspects of spatial pattern in biological populations. Science, 121, 397-8 Clausen, J. J. (1957) A comparison of some methods of establishing plant community patterns. Bot. Tidsskr. 53, 253-18. Clements, F. E. (1916) Plant succession: an analysis of the development of vegetation. Publ. Carneg. Instn. 242. Cliff, A. D. & Ord, J. K. (1973) Spatial Autocorrelation. Pion, London. Clymo, R. S. (1980) Preliminary survey of the peat-bog Hummell Knowe Moss using various numerical methods. Vegetatio, 42, 129-48. Cochran, W.G. (1954) Some methods for strengthening the common 7? tests. Biometrics, 4, 417—S1.REFERENCES 323 Cochran, W. G. (1963) Sampling Techniques, 2nd edn. Wiley, New York and London. Cochran, W. G. & Cox, G. M. (1944) Experimental Design. Mimeographed. (Quoted from Snedecor, 1946). Conway, V. M. (1938) Studies in the autecology of Cladium mariscus R. Br. IV. Growth rates of the leaves. New Phytol. 37, 254-78. Cooper, C. F. (1957) The variable plot method for estimating shrub density. J. Range Mgmt, 10, 111-15. ‘Cooper, C. F. (1963) An evaluation of variable plot sampling in shrub and herbaceous vegetation. Ecology, 44, 565-9. ‘Cormack, R. M. (1971) A review of classification. J/ R. statist. Soc. A. 134, 321-67. ‘Cormack, R. M. & Ord, J. K. (Eds) (1979) Spatial and Temporal Analysis in Ecology. Statistical Ecology, 8. International Co-operative Publishing House, Fairland, Md, U.S.A. Cottam, G. (1947) A point method for making rapid surveys of woodlands. Bull. ecol. Soc. Amer. 28, 60. Cottam, G. & Curtis, J. T. (1949) A method for making rapid surveys of woodlands by means of pairs of randomly selected trees. Ecology, 30, 101-4. Cottam, G. & Curtis, J.T. (1955) Correction for various exclusion angles in the random pairs method. Ecology, 36, 767. Cottam, G. & Curtis, J. T. (1956) The use of distance measures in phytosociological sampling. Ecology, 37, 451-60. Cottam, G., Curtis, J. T. & Catana, A. J. (1957) Some sampling characteristics of a series of aggregated populations. Ecology, 38, 610-22. Cottam, G., Curtis, J.T. & Hale, B, W. (1953) Some sampling characteristics of a population of randomly dispersed individuals. Ecology, 34, 741-57. Cottam, G., Goff, F.G. & Whittaker, R. H. (1973) Wisconsin comparative ordi- nation. In: Ordination and Classification of Communities. Handbook of Vegetation Science, 5, (Ed. by R. H. Whittaker), pp. 195-221. Junk, The Hague. Crawford, R.M.M. & Wishart, D. (1967) A rapid multivariate method for the detection and classification of groups of ecologically related species. J. Ecol. 55, 505-24, Crawford, R.M.M. & Wishart, D. (1968) A rapid classification and ordination method and its application to vegetation mapping. J. Ecol. 56, 385-404. Crawford, R. M. M., Wishart, D. & Campbell, R. M. (1970) A numerical analysis of high altitude scrub vegetation in relation to soil erosion in the Eastern Cordillera of Peru. J. Ecol. 58, 173-91. Crovello, T. J. (1970) Analysis of character variation in ecology and systematics. 4. Rev. Ecol. Syst. 1, 55-98. Curtis, J. T. (1947) The palo verde forest type near Gonaives, Haiti, and its relation to the surrounding vegetation. Caribb. Forester, 8, 1-26. Curtis, J. T. (1955) A note on recent work dealing with the spatial distribution of plants. J. Ecol. 43, 309. Curtis, J. T. (1959) The Vegetation of Wisconsin: An Ordination of Plant Communities. University of Wisconsin Press, Madison, Wisconsin. Curtis, J.T. & McIntosh, R. P. (1950) The interrelation of certain analytic and synthetic phytosociological characters. Ecology, 31, 434-55.324 REFERENCES Curtis, J. T. & McIntosh, R. P. (1951) An upland forest continuum in the prairie-forest border region of Wisconsin. Ecology, 32, 476-96. Czekanowski, J. (1913) Zarys Metod Statystycznyck, Warsaw. Dagnelie, P. (1960) Contribution a l’étude des communautés végetales par l'analyse factorielle. Bull. Serv. Carte phytogéogr. Sér. B, 5, 7-71 and 93-195 Dagnelie, P. (1961) L’application de l’'analyse multi-variable a l'étude des com- munautés végétales. Bull. Inst. int. Statist. 39. Dagnelie, P. (1962) Etude statistique d'une pelouse a Brachypodium ramosum. IV. La distribution des fréquences des espéces. Bull. Serv. Carte phytogéogr. Sér. B, 7, 99-109. Dagnelic, P. (1965a) L’étude des communautés végetales par l'analyse statistique des liaisons entre les espéces et les variables écologique: principes fondamentaux. Biometrics, 21, 345-61. Dagnelic, P. (1965b) L’étude des communautés végétales par I'analyse des liaisons entreles espéces et les variables écologiques: un example. Biometrics, 21, 890-907. Dagnelie, P. (1973) L'analyse factorielle. In: Ordination and Classification of Communities. Handbook of Vegetation Science, 5 (Ed. by R. H. Whittaker), pp. 223-48, Junk, The Hague. Dale, M. B. (1971) Information analysis of quantitative data. In: Statistical Ecology (Ed. by G. P. Patil et a/.), Vol. 3, pp. 133-48. Pennsylvania State University Press. Dale, M. B. (1975) On objectives of methods of ordination. Vegetatio, 30, 15-32. Dale, M.B. & Anderson, D. J. (1972) Qualitative and quantitative information analysis. J. Ecol. 60, 639-53. Dale, M. B, & Anderson, D. J. (1973) Inosculate analysis of vegetation data. Aust. J. Bot. 21, 253-76. Dale, M. B. & Clifford, H. T. (1976) On the effectiveness of higher taxonomic ranks for vegetation analysis. Aust. J. Ecol. 1, 37-62. Dale, M. B., Lance, G. N. & Albrecht, L. (1971) Extensions of information analysis. Aust. Comput. J. 3, 29-34. Dale, M. B. & Quadraccia, L. (1970) Computer-assisted tabular sorting of phyto- sociological data. Vegetatio, 28, 57-13. Dale, M.B. & Webb, L. J. (1975) Numerical methods for the establishment of associations. Vegetatio, 30, 77-87. Dansereau, P. (Ed.) (1968) The continuum concept of vegetation: responses. Bol. Rev. 34, 253-332. David, F.N. & Moore, P. G. (1954) Notes on contagious distributions in plant populations. Ann. Bot. 18, 47-53. David, F.N. & Moore, P. G. (1957) A bivariate test for the clumping of supposedly random individuals. Ann. Bot. 21, 315-20. Davis, T. A. W. & Richards, P. W. (1933-4) The vegetation of Moraballi Creek, British Guiana: an ecological study of a limited area of tropical rain forest. J. Ecol. 21, 350-84, 22, 106-55. Dawson, G. W. P. (1951) A method for investigating the relationship between the distribution of individuals of different species in a plant community. Ecology, 32, 332-4. De Jong, P., Aarssen, L. W. & Turkington, R. (1980) The analysis of contact sampling data. Oecologia, 45, 322-4.REFERENCES 325 De Jong, P., Aarssen, L. W. & Turkington, R. (1983) The use of contact sampling in studies of association in vegetation. J. Ecol. 71 (in press). De Vries, D. M. (1953) Objective combinations of species. Acta bot. neerl. 1, 497-9. De Vries, D.M. (1954) Constellation of frequent herbage plants, based on their correlation in occurrence. Vegetatio, 5-6, 105-11. De Vries, D. M. & De Boer, T. A. (1959) Methods used in botanical grassland research in the Netherlands and their application. Herb, Abstr. 29, 1-7. Del Moral, R. (1975) Vegetation clustering by means of ISODATA: revision by multiple discriminant analysis. Vegetatio, 29, 179-90. Del Moral, R. (1980) On selecting indirect ordination methods. Vegetatio, 42, 75—84. Del Moral, R. & Denton, M. F. (1977) Analysis and classification of vegetation based on family composition. Vegetatio, 34, 155-65. Denning, W. E. & Stephan, F. F. (1940) On a least squares adjustment of a samples frequency table when the expected marginal totals are known. Aun. math. Statist. 11, 427-44. Dice, L. R. (1945) Measures of the amount of ecologic association between species. Ecology, 26, 297-302. Dice, L. R. (1952) Measure of the spacing between individuals within a population. Contr. Lab. Vertebr. Biol. Univ. Mich. 55, 1-23. Dirven, J.G.P., Hoogers, B. J. & De Vries, D. M. (1969) Interrelation between frequency, dominance and dry-weight percentages of species in grassland vegetations. Netherlands J. agric. Sci. 17, 161-6. Dix, R. L. (1961) An application of the point-centred quarter method to the sampling of grassland vegetation. J. Range Mgmt, 14, 63-9. Dix, R. L. & Butler, J. E. (1960) A phytosociological study of a small prairie in Wisconsin. Ecology, 41, 316-27. Dixon, W.J. & Massey, F. J. (1957) Introduction to Statistical Analysis, 2nd edn. McGraw-Hill, New York. Dony, J. G. (1977) Species-area relationships in an area of intermediate size. J. Ecol. 65, 475-84, Du Rietz, G. E. (1921) Zur methodologischen Grundlage der modernen Pflanzen- soziologie. Akad. Afh., Uppsala. Du Rietz, G. E. (1931) Life-forms of terrestrial flowering plants. Acia phytogeogr. suec. 3, 1-95. Dye, P. J. & Walker, B. H. (1980) Vegetation-environment relations on sodic soils of Zimbabwe. J. Ecol. 68, 589-606. Eberhardt, L. L. (1967) Some developments in ‘distance sampling’. Biometrics, 23, 207-16. Edwards, A.W. F. & Cavalli-Sforza, L. L. (1965) A method for cluster analysis, Biometrics, 21, 362-75. Ellison, L. (1942) A comparison of methods of quadratting short-grass vegetation. J. agric. Res. 64, 595-614. Emmett, H. E. G. & Ashby, E. (1934) Some observations on the relation between the hy- drogen-ion concentration of the soil and plant distribution. Ann. Bot. 48, 869-76. Enright, N. & Ogden, J. (1979) Applications of transition matrix models in forest dynamics; Araucaria in Papua New Guinea and Nothofagus in New Zealand. Aust. J. Ecol. 4, 3-23.326 REFERENCES Erickson, R. O. & Stehn, J. R. (1945) A technique for analysis of population density data. Amer. midl. Nat. 33, 781-7. Errington, J. C. (1973) The effect of regular and random distributions on the analysis of pattern. J. Ecol. 61, 99-105. Evans, D. A. (1953) Experimental evidence concerning contagious distributions in ecology. Biometrika, 40, 186-211. Evans, F.C. (1952) The influence of size of quadrat on the distributional patterns of plant populations. Conir. Lab. Vertebr. Biol. Univ. Mich, 54, \—15. Everitt, B.S. (1977) The Analysis of Contingency Tables. Chapman and Hall, London Ewer, S. J. (1932) Life forms of Illinois plants. Trans. Ill. Acad. Sci. 24, 108-21 Fager, E. W. (1972) Diversity: a sampling study. Am. Nat. 106, 293-310. Falinski, J. (1960) Zastosowanie taksonomii wrosclawskiej do fitosocjologii. (Anwendung der sog. “Breslauer Taxonomie’ in der Pflanzensoziologie). Acta Soc. Bot. Polon. 29, 333-61. Fasham, M.J.R. (1977) A comparison of non-metric multidimensional scaling, principal components and reciprocal averaging for the ordination of simulated coenoclines and coenoplanes. Ecology, 58, 551-61. Feller, W. (1943) On a general class of ‘contagious’ distributions, Ann. math. Statist. 14, 389-400. Feoli, E. (1977) On the. resolving power of principal component analysis in plant community ordination, Vegetatio, 33, 119-25. Feoli, E. & Orloci, L. (1979) Analysis of concentration and detection of underlying factors in structured tables. Vegetatio, 40, 49-54, Ferrari, H. P., Pijl, H. & Venekamp, J. T. N. (1957) Factor analysis in agricultural research. Netherlands J. agric. Sci. §, 211-21. Field, J. G. (1969) The use of the information statisticin the numerical classification of heterogeneous systems. J. Ecol. 57, 565-9. Finney, D. J. (1948) Random and systematic sampling in timber surveys. Forestry, 22, 64-99. Finney, D. J. (1950) An example of periodic variation in forest sampling. Forestry, 23, 96-111. Finney, D. J., Latscha, R., Bennet, B,M. & Hsu, P. (1963) Tables for Testing Significance in a 2x 2 Contingency Table. Cambridge University Press, London. Fisher, R. A. (1941) Statistical Methods for Research Workers, 8th edn. Oliver and Boyd, Edinburgh Fisher, R. A., Corbet, A. S. & Williams, C. B. (1943) The relation between the number of species and the number of individuals in a random sample of an animal population. J. anim. Ecol. 12, 42-38. Fisher, R.A. & Yates, F. (1943) Statistical Tables for Biological, Agricultural and Medical Research, 2nd edn. Oliver and Boyd, Edinburgh. Fisser, H. G. & Van Dyne, G. M. (1966) Influence of number and spacing of points on accuracy and precision of basal cover estimates. J. Range Mgmt, 19, 205-11. Fracker, $. B. & Brischle, H. A. (1944) Measuring the local distribution of Ribes. Ecology, 28, 283-303 Freeman, M. F. & Tukey, J. W. (1950) Transformations related to the angular and the square root. Ann. math. Siatist. 21, 607-11 Garrison, G. A. (1949) Uses and modifications for the ‘moosehorn’ crown closure estimator. J. For. 47, 733-5.REFERENCES 327 Gauch, H. G. (1973) The relationship between sample similarity and ecological distance. Ecology, $4, 618-22. Gauch, H.G. (1980) Rapid initial cluste 103-11. Gauch, H. G. & Chase, G. B. (1974) Fitting the Gaussian curve to ecological data. Ecology, 85, 1377-81. Gauch, H. G., Chase, G. B. & Whittaker, R. H. (1974) Ordination of vegetation samples by Gaussian species distribution. Ecology, 55, 1382-90. Gauch, H. G. & Scruggs, W. M. (1979) Variants of polar ordination. Vegetatio, 40, 147—53. Gauch, H.G. & Wentworth, T.R. (1976) Canonical correlation analysis as an ordination technique. Vegetatio, 33, 17-22. Gauch, H.G. & Whittaker, R. H. (1972a) Coenocline simulation, Ecology, 53, 446-51. Gauch, H.G. & Whittaker, R. H. (1972b) Comparison of ordination techniques. Ecology. 53, 868-75. Gauch. H.G. & Whittaker, R.H. (1976) Simulation of community patterns. Vegetatio, 33, 13-16. Gauch, H. G., Whittaker, R. H. & Wentworth, T. R. (1977) A comparative study of reciprocal averaging and other ordination techniques. J. Ecol. 65, 157-74. Ghent, A. W. (1963) Kendall's ‘tau’ coefficient as an index of similarity in comparisons of plant or animal communities. Canad. Ent. 95, 568-75. Ghent, A. W. (1972) A simplified computation procedure for Kendall's Tau suited to extensive species-density comparisons. Am. Midl. Nai. 87, 459-71. Gilbert, N. & Wells, T. C. E. (1966) Analysis of quadrat data. J. Ecol. 54, 675-85. Gimingham, C. H. (1961) Northern European heath communities: a ‘network of variation’. J. Ecol. 49, 655-94. Gimingham, C.H., Gemmell, A. R. & Greig-Smith, P. (1948) The vegetation of a sand-dune system in the Outer Hebrides. Trans. bot. Soc. Edinb., 25, 82-96. Gittins, R. (1965a) Multivariate approaches to a limestone grassland community. I. A stand ordination. J. Ecol. 53, 385-401. Gittins, R. (1965b) Multivariate approaches to a limestone grassland community. II. A direct species ordination. J. Ecol, 53, 403-9. Gittins, R. (1965c) Multivariate approaches to a limestone grassland community. III. A comparative study of ordination and association-analysis. J. Ecol. 53, 411-25. Gittins, R. (1968) Trend-surface analysis of ecological data. J. Ecol. $6, 845-69. Gittins, R. (1969) The application of ordination techniques. In: Ecological Aspects of Mineral Nutrition in Plants (Ed. by 1. H. Rorison), pp. 37-66. Blackwell Scientific Publications, Oxford. . Gittins, R. (1979) Ecological applications of canonical analysis. In: Multivariate Methods in Ecological Work (Ed. by L. Orloci, C. R. Rao and W. M. Stiteler). Statistical Ecology, 7, 309-535. International Co-operative Publishing House, Fairland, Md. U.S.A. Gleason, H. A. (1926) The individualistic concept of the plant association. Bull. Torrey bot. Cl. $3, 7-26. Goff. F. G. (1968) Use of size stratification and differential weighting to measure forest trends, Am. Midl. Nat. 79, 402-18 Goff, F. G. & Cottam, G. (1967) Gradient analysis: the use of species and synthetic indices. Ecology, 48, 793-806. ng of large data sets. Vegetatio, 42,328 REFERENCES Golf, F. G. & Mitchell, R. (1975) A comparison of species ordination results from plot and stand data. Vegetatio, 31, 15-22. Goff, F. G. & Zedler, P. H. (1972) Derivation of species succession vectors. Am. Midl. Nat. 81, 397-412, Goff, F. G. & Zedler, P. H. (1968) Structural gradient analysis of upland forests in the western Great Lakes area. Ecol. Monogr. 38, 65-86. Goldsmith, F. B. (1973) The vegetation of exposed sea cliffs at South Stack, Anglesey. I. The multivariate approach. J. Ecol. 61, 787-818. Goldstein, R. A. & Grigal, D. F. (1972) Definition of vegetation structure by canonical analysis. J. Ecol. 60, 277-84. Good, I. J. (1953) The population frequencies of species and the estimation of population parameters. Biometrika, 40, 237-64, Goodall, D. W. (1952a) Quantitative aspects of plant distribution. Biol. Rev. 27, 194— 245. Goodall, D. W. (1952b) Some considerations in the use of point quadrats for the analysis of vegetation. Aust. J. sci. Res. Ser. B, 5, 1-41. Goodall, D. W. (1953a) Objective methods for the classification of vegetation. I. The use of positive interspecific correlation. Aust. J. Bot. 1, 39-63. Goodall, D. W. (1953b) Objective methods for the classification of vegetation. II. Fidelity and indicator value. Aust. J. Bot. 1, 434-56. Goodall, D. W. (1953c) Point quadrat methods for the analysis of vegetation. The treatment of data for tussock grasses. Aust. J. Bot. 1, 457-67. Goodall, D. W. (1954a) Vegetational classification and vegetational continua. Angew. Pflanzensoz. (Wien), Festschr. Aichinger, 1, 168-82. Goodall, D. W. (1954b) Minimal area: a new approach. Int. bot. Congr. 8, Rap. Sect. 7 and 8, 19-21. Goodall, D. W. (1954c) Objective methods for the classification of vegetation. IIT. An essay in the use of factor analysis. Aust. J. Bot. 2, 304-24. Goodall, D. W. (1961) Objective methods for the classification of vegetation. IV. Pattern and minimal area. Aust. J. Bot. 9, 162-96. Goodall, D. W. (1962) Bibliography of statistical plant sociology. Excerpta bot. Sect. B, 4, 253-322. Goodall, D. W. (1963) Pattern analysis and minimal area—some further comments. J. Ecol. 51, 705-10. Goodall, D. W. (1965) Plot-less tests of interspecific association. J. Ecol. 53, 197-210. Goodall, D. W. (1970) Statistical plant ecology. A. Rev. Ecol. Syst. 1, 99-124. Goodall, D. W. (1973) Sample similarity and species correlation. In: Ordination and Classification of Communities, Handbook of Vegetation Science, § (Ed. by R. H. Whittaker), pp. 105-56. Junk, The Hague. Goodall, D. W. (1974) A new method for the analysis of spatial pattern by random pairing of quadrats. Vegetatio, 29, 135-46. Goodall, D. W. & West, N. E. (1979) A comparison of techniques for assessing dispersion pattern. Vegetatio, 40, 15-27. Gordon, A.D. (1981) Classification: Methods for the Exploratory Analysis of Multivariate Data, Chapman and Hall, London. Gordon, A.D. & Birks, H. J. B. (1972) Numerical methods in quaternary palaeoe- cology. I. Zonation of pollen diagrams. New Phytol. 71, 961-79.REFERENCES 329 Gordon, A. D. & Birks, H. J.B. (1974) Numerical methods in quaternary palacoe- cology. II. Comparison of pollen diagrams. New Phytol. 73, 221-49, Gounot, M. (1962) Etude statistique d'une pelouse 4 Brachypodium ramosum. Ml. Etude de la distribution des espéces au moyen d'un test non-parametrique. Bull. Serv. Carte phytogéogr. Sér. B, 1, 65-84. Gounot, M. (1969) Méthodes d'Etude Quantitative de la Végétation. Masson, Paris. Gower, J. C. (1966) Some distance properties of latent root and vector methods used in multivariate analysis. Biometrika, 53, 325—38. Gower, J.C. (1967a) Multivariate analysis and multidimensional geometry. Statistician, 17, 13-28. Gower, J. C. (1967b) A comparison of some methods of cluster analysis. Biometrics, 23, 623-37. Gower, J.C. (1971) A general coefficient of similarity and some of its properties. Biometrics, 27, 857-74. Gower, J.C. & Ross, G. J. S. (1969) Minimum spanning trees and single linkage cluster analysis. Appl. Statist. 18, 54-64. Greig-Smith, P. (1952a) The use of random and contiguous quadrats in the study of the structure of plant communities. Ann. Bot. 16, 293-316. Greig-Smith, (1952b) Ecological observations on degraded and secondary forest in Trinidad, British West Indies. II. Structure of the communities. J. Ecol. 40, 316— 30. Greig-Smith (1961a) The use of pattern analysis is ecological investigations. Recent ‘Advances in Botany, 2, pp. 1354-8. University of Toronto Press, Toronto. Greig-Smith (1961b) Data on pattern within plant communities. I. The analysis of pattern. J. Ecol. 49, 695—702 Greig-Smith (1965) Notes on the quantitative description of humid tropical forest. ‘Symposium on Ecological Research in Humid Tropics vegetation, Kuching, Sarawak, pp. 227-34. UNESCO Science Co-operation Office for Southeast Asia. Greig-Smith (19712) Analysis of vegetation data: the user viewpoint. In: Statistical Ecology (Ed. by G.P. Patil er ai.), Vol. 3, pp. 149-66. Pennsylvania State University Press. Greig-Smith, P. (1971b). Application of numerical methods to tropical forests. In: Statistical Ecology (Ed. by G. P. Patil et al.), Vol. 3, pp. 195-206. Pennsylvania State University Press. Greig-Smith, P. (1971c). Some problems of analytical data. Statistician, 21, 215-19. Greig-Smith, P. (1979) Pattern in vegetation: presidential address to the British Ecological Society. J. Ecol. 67, 755-79. Greig-Smith (1980) The development of numerical classification and ordination. Vegetatio, 42, 1-9. Greig-Smith, P., Austin, M.P. & Whitmore, T.C. (1967) The application of quantitative methods to vegetation survey. I. Association-analysis and principal component ordination of rain forest. J. Ecol. 55, 483-503. Greig-Smith, P. & Chadwick, M. J. (1965) Data on pattern within plant communities. IIL. Acacia-Capparis semi-desert scrub in the Sudan. J. Ecol. 53, 465-74. Greig-Smith, P., Kershaw, K. A. & Anderson, D. J. (1963) The analysis of pattern in ‘vegetation: a comment on a paper by D. W. Goodall. J. Ecol. 51, 223-9.330 REFERENCES Grigal, D. F. & Goldstein, R.A. (1971) An integrated ordination-classification analysis of an intensively sampled oak-hicKory forest. J. Ecol. 59, 481-92. Grime, J. P. & Lloyd, P. S. (1973) An Ecological Atlas of Grassland Plants. Arnold, London. Groenewoud, H. van (1965) Ordination and classification of Swiss and Canadian coniferous forests by various biometric and other methods. Ber. geobot. Inst. ETH, Stiftg. Ritbel, Ziirich, 36, 28-102. Groenewoud, H. van (1976) Theoretical considerations on the covariation of plant species along ecological gradients with regard to multivariate analysis. J. Ecol. 64, 837-47. Grunow, J. O. (1967) Objective classification of plant communities: a synecological study in the sourish mixed bushveld of Transvaal. J. Ecol. 55, 691-710. Haberman, S. J. (1973) The analysis of residuals in cross-classified tables. Biometrics, 29, 205-20. Hairston, N. G. (1964) Species abundance and community organization. J. Ecol. (Suppl), 52, 227-39. Hall, A. V. (1970) A computer-based method for showing continua and communities in ecology. J. Ecol. 58, 591-602. Hall, J.B. & Swaine, M. D. (1976) Classification and ecology of closed-canopy forest in Ghana. J. Ecol. 64, 913-51. Hall, John B. & Okali, D. U. U. (1978) Observer-bias in a survey of complex tropical vegetation. J. Ecol. 66, 241-9. Hall, John B. & Okali, D. U. U. (1979) A structural and floristic analysis of woody fallow vegetation near Ibadan, Nigeria. J. Ecol. 67, 321-46. Hamilton, K.C. & Buchholtz, K.P. (1955) Effect of rhizomes of quackgrass (Agropyron repens) and shading on the seedling development of weedy species. Ecology, 36, 304-8. Hansen, K. & Jensen, J. (1974) Edaphic conditions and plant-soil relationships on roadsides in Denmark. Dansk bot. Ark. 28 (3), 1-143. Harberd, D. J. (1962) Application of a multivariate technique to ecological survey. J. Ecol. 50, 1-17. Harman, H. H. (1976) Modern Factor Analysis, 3rd edn. Chicago University Press. Hasel, A. A. (1938) Sampling error in timber surveys. J. agric. Res. 57, 713-36. Hazen, W. E. (1966) Analysis of spatial pattern in epiphytes. Ecology, 47, 634-5 Hecke, P. van, Impens, I., Goossens, R. & Hebrant, F. (1980) Multivariate analysis of multispectral remote sensing data on grasslands from different soil types. Vegetatio, 42, 165-70. Heslop-Harrison, J. (1964) Forty years of genecology. Adv. ecol. Res. 2, 159-247. Hill, M. O. (1973a) The intensity of spatial pattern in plant communities. J. Ecol. 61, 225-35, Hill, M. O. (1973b) Reciprocal averaging: an eigenvector method of ordination. J. Ecol. 61, 237-249. Hill, M. O. (1973c) Diversity and evenness: a unifying notation and its consequences. Ecology, 54, 427-32. Hill, M. O. (1974) Correspondence analysis: a neglected multivariate method. Appl. Statist. 23, 340-34.REFERENCES 331 Hill, M. O. (1979a) TWINSPAN-a FORTRAN program for arranging multivariate data in an ordered two-way table by classification of the individuals and attributes. Ecology and Systematics, Cornell University, Ithaca, New York. Hill, M. O. (1979b) DECORANA~—a FORTRAN program for detrended correspon- dence analysis and reciprocal averaging. Ecology and Systematics, Cornell University, Ithaca, New York. Hill, M. O., Bunce, R.G. H. & Shaw, M. W. (1975) Indicator species analysis, a divisive polythetic method of classification, and its application to a survey of native pinewoods in Scotland. J. Ecol. 63, 597-613. Hill, M. O. & Gauch, H. G. (1980) Detrended correspondence analysis: an improved ordination technique. Vegetatio, 42, 47-58. Hill, M. O. & Smith, A. J. E. (1976) Principal component analysis of taxonomic data with multi-state discrete characters. Taxon, 25, 249-55. Hole, F. D. & Hironaka, M. (1960) An experiment in ordination of some soil profiles. Proc. Soil Sci. Soc. Am. 24, 309-12. Holgate, P. (1965a) Some new tests of randomness. J. Ecol. $3, 261-6. Holgate, P. (1965b) Tests of randomness based on distance methods. Biometrika, 52, 345-53. Holland, P. G. (1969) The plant patterns of different seasons in two stands of Mallee vegetation. J. Ecol. 57, 323-33. Hope-Simpson, J. F. (1940) On the errors in the ordinary use of subjective frequency estimations in grassland. J. Ecol. 28, 193-209. Hopkins, B. (1954) A new method for determining the type of distribution of plant individuals. Ann. Bot. 18, 213—27. Hopkins, B. (1955) The species-area relations of plant communities. J. Ecol. 43, 409— 26. Hopkins, B. (1957) Pattern in the plant community. J. Ecol. 48, 451-63. Hora, F. B. (1947) The pH range of some cliff plants on rocks of different geological origin in the Cader Idris area of North Wales. J. Ecol. 38, 158-65. Hughes, R. E. (1961) The application of certain aspects of multivariate analysis to plant ecology. Recent Advances in Botany, 2, pp. 1350-4. University of Toronto Press, Toronto Hughes, R. E, & Lindley, D. V. (1955) Application of biometric methods to problems of classification in ecology. Nature, Lond. 175, 806-7. Hurlbert, S. H. (1971) The nonconcept of species diversity: a critique and alternative parameters. Ecology, 52, 577-86. Hutchings, S. S. & Holmgren, R. C. (1959) Interpretation of loop-frequency data as a measure of plant cover. Ecology, 40, 668-77. Thm, P. & Groenewoud, H. van (1975) A multivariate ordering of vegetation data based on Gaussian type gradient response curves. J. Ecol. 63, 767-77. Ivimey-Cook, R. B. & Proctor, M. C. F. (1967) Factor analysis of data from an east Devon heath: a comparison of principal component and rotated solutions. J. Ecol. 55, 405-13. Ivimey-Cook, R. B., Proctor, M. C. F. & Wigston, D. L. (1969) On the problem of the *R/Q’ terminology in multivariate analyses of biological data. J. Ecol. 57, 573-5332 REFERENCES Ivimey-Cook, R. B.,Proctor, M. C. F. & Rowland, D. M. (1975) Analyses of the plant communities of a heathland site: Aylesbeare Common, Devon, England Vegetatio, 31, 33-45 Jaceard, P. (1912) The distribution of the flora in the alpine zone. New Phytol. 11, 37— 50. James, F. C. (1971) Ordination of habitat relationships among breeding birds. Wilson Bull. 83, 215-36. Jancey, R. C. (1979) Species ordering on a variance criterion. Vegetatio, 39, 59-63, Janssen, J. G. M. (1975) A simple clustering procedure for preliminary classification of very large sets of phytosociological relevés. Vegetatio, 30, 67-71. Jeffers, J.N.R. (1978) An Introduction 10 Systems Analysis: with Ecological Applications. Arnold, London. Jeglum, J. K., Wehrhahn, C.F. & Swan, J. M.A. (1971) Comparisons of environ- mental ordinations with principal component vegetational ordinations for sets of data having different degrees of complexity. Can. J. For. Res. 1, 99-112. Jensen, D.R., Beus, G.B. & Storm, G. (1968) Simultaneous statistical tests on categorical data. J. exp. Educ. 36(4), 46-56. Jensen, $. (1978) Influences of transformation of cover values on classification and ordination of lake vegetation. Vegeratio, 37, 19-31 Johnson, R. W. & Goodall, D. W. (1979) A maximum likelihood approach to non- linear ordination. Vegetatio, 41, 133-42. Johnston, A. (1957) A comparison of the line interception, vertical point quadrat, and loop methods as used in measuring basal area of grassland vegetation. Can. J. Pl. Sci. 37, 34-42. Jones, E. W. (1955-6) Ecological studies on the rain forest of southern Nigeria IV. The plateau forest of the Okomu Forest Reserve. J. Ecol. 43, 564-94, 44, 83 117, Joweit, G. H. & Scurfield, G. (19492) Statistical test for optimal conditions: note on a paper of Emmett and Ashby. J. Ecol. 37, 65-7. Joweit, G. H. & Scurfield, G. (19496) A statistical investigation into the distribution of Holcus mollis L. and Deschampsia flexuosa (L.) Trin. J. Ecol. 37, 68-81 Kaiser, H. F. (1958) The Varimax criterion for analytic rotation in factor analysis. Psychometrika, 23, 187-200. Kemp, C. D. & Kemp, A. W. (1956) The analysis of point quadrat data. Aust. J. Bol. 4, 167-74. Kendall, M. G. (1948) Rank Correlation Methods. Griffin, London. Kendall, M. G. (1957) 4 Course in Multivariate Analysis. Griffin, London. Kendall, M. G. (1966) Discrimination and classification. Proc. Symp. Multiv. Analysis, Drayton, Ohio, (Ed. by P. R. Krishnaiah), pp. 165-85. Academic Press, New York. Kercher, J. R. & Goldstein, R. A. (1977) Analysis of a East Tennessee oak hickory forest by canonical correlation of species and environmental parameters. Vegetatio, 38, 153-63 Kershaw, K. A. (1957a) The use of cover and frequency in the detection of pattern in plant communities. Ecology, 38, 291-9. Kershaw, K. A. (1957b) A study of pattern in certain plant communities. Ph.D. thi University of Wales.REFERENCES 333 Kershaw, K. A. (1958) An investigation of the structure of a grassland community. 1. The pattern of Agrostis tenuis, J. Ecol. 46, 571-92. Kershaw, K. A. (1959) An investigation of the structure of a grassland community. Il. The pattern of Dactylis glomerata, Lolium perenne and Trifolium repens. ILL. Discussion and conclusions. J. Ecol. 47, 31—S3. Kershaw, K. A. (1960) The detection of pattern and association. J. Ecol. 48, 233— 42. Kershaw, K. A. (1961) Association and co-variance analysis of plant communities. J. Ecol. 49, 643-54, Kershaw, K. A. (1968) Classification and ordination of Nigerian savanna vegetation. J. Ecol. 56, 467-82. Kershaw, K. A. (1970) An empirical approach to the estimation of pattern intensity from density and cover data. Ecology, 51, 729-34. Kilburn, P. D. (1966) Analysis of the species-area relation. Ecology, 47, 831-43. King, J. (1962) The Festuca-Agrostis grassland complex in south-east Scotland. J. Ecol. 50, 321-56. Knight, D. H. (1965) A gradient analysis of Wisconsin prairie vegetation on the basis of plant structure and function. Ecology, 46, 744-7. Knight, D. H. & Loucks, O. L. (1969) A quantitative analysis of Wisconsin forest vegetation on the basis of plant function and gross morphology. Ecology, 50, 219-34, Kooijman, S. A. L. M. (1976) Some remarks on the statistical analysis of grids especially with respect to ecology. Anns Syst. Res. 5, 113-32. Krishna Iyer, P. V. (1948) The theory of probability distribution of points on a line. J. Indian Soc. agric. Statist. 1, 173-95. Krishna Iyer, P. V. (1950) The theory of probability distributions of points on a lattice. Ann. math. Statist. 21, 198-217 Kruskal, J. B. (1964a) Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis. Psychometrika, 29, 1-27. Kruskal, J. B. (1964b) Nonmetric multidimensional scaling: a numerical method. Psychometrika, 29, 115-29. Kruskal, J. B. & Carmone, F. (1971) How to use M-D-SCAL (Version 5M ) and other useful information. Bell Telephone Laboratory, Murray Hill, New Jersey, U.S.A. and University of Waterloo, Waterloo, Ontario, Canada. (Mimeographed). (Quoted from Orloci, 1978). Kruskal, J. B. & Carroll, J. D. (1969) Geometric models and badness of fit functions. In: Multivariate Analysis If, (Ed. by P. R. Krishnaiah), pp. 639-71. Academic Press, London. Kullback, S. (1959) Information Theory and Statistics. Wiley, Chapman and Hall, New York. Kylin, H. (1926) Uber Begriffsbildung und Statistik in der Pflanzensoziologie. Bot. Notiser (1926), 81—180. Lambert, J. M. (1972) Theoretical models for large-scale vegetation survey. In: Mathematical Models in Ecology, (Ed. by J. N. R. Jeffers), pp. 87-109. Blackwell Scientific Publications, Oxford. Lambert, J. M. & Dale, M. B. (1964) The use of statistics in phytosociology. Adv. ecol. Res. 2, 59-99.334 REFERENCES Lambert, J. M., Meacock, S. E., Barrs, J. & Smarit, P. F. M. (1973) AXOR and MONIT: two new polythetic-divisive strategies for hierarchical classification Taxon, 22, 173-6. Lambert, J. M. & Williams, W. T. (1962) Multivariate methods in plant ecology. IV. Nodal analysis. J. Ecol. 50, 775-802. Lambert, J. M. & Williams, W. T. (1966) Multivariate methods in plant ecology. VI. Comparison of information-analysis and association-analysis. J. Ecol. 54, 635-64. Lance, G. N., Milne, P. W. & Williams, W. T. (1968) Mixed-data classificatory programs, III. Diagnostic systems. Aust. Comput. J. 1, 178-81. Lance, G. N. & Williams, W. T. (1965) Computer programs for monothetic classifi- cation (“Association analysis”). Comput. J. 8, 246-9. Lance, G.N, & Williams, W.T. (1966a) Computer programs for hierarchical polythetic classification (“Similarity analyses”Y’ Comput. J. 9, 60-4. Lance, G. N. & Williams, W. T. (1966b) A generalized sorting strategy for computer classifications. Nature, Lond. 212, 218. Lance, G. N. & Williams, W. T. (1967a) A general theory of classificatory sorting strategies. I. Hierarchical systems. Comput. J. 9, 373-80. Lance, G.N. & Williams, W.T. (19676) Mixed-data classificatory programs. I. Agglomerative systems. Aust. Comput. J. 1, 15-20. Lance, G.N. & Williams, W.T. (1968a) Note on a new information-statistic classificatory program. Comput. J. 11, 195. Lance, G.N. & Williams, W.T. (1968b) Mixed-data classificatory programs. IL. Divisive systems. Aust. Comput. J. 1, 82-99. Lange, R. T. (1968) Influence analysis in vegetation. Aust. J. Bot. 16, 555—64. Lange, R.T. (1971) Influence analysis and prescriptive management of rangeland vegetations. Proc. ecol. Soc. Austr. 6, 153-8. Lange, R. T.,Stenhouse, N. 8. & Offler, C. E, (1965) Experimental appraisal of certain procedures for the classification of data. Aust. J. biol. Sci. 18, 1189-205. Langlet, O. (1971) Two hundred years genecology. Taxon, 20, 653-721. Lawley, D,N. & Maxwell, A. E. (1963) Factor Analysis as a Statistical Method. Butterworth, London. Lawson, G. W. (1978) The distribution of seaweed floras in the tropical and subtropical Atlantic Ocean: a quantitative approach. Bot, J. Linn. Soc. 76, 177— 93. Lefkovitch, L. P. (1966) An index of spatial distribution. Res. Popul. Ecol. 8, 89-92. Legg, C. J. (1980) A Markovian approach to the study of heath vegetation dynamics. Bull. Ecol. 11, 393-404. Levy, E. B. (1933) Technique employed in grassland research in New Zealand. I. Strain testing and strain building. Bull. Bur. Pl. Genet. Aberystw. 11, 6-16. Levy. E. B. & Madden, E. A. (1933) The point method of pasture analysis. V. Z. JI Agric. 46, 267-79. Lieth, H. & Moore, G.W. (1971) Computerized clustering of species in phyto- sociological tables and its utilization for field work. In: Statistical Ecology (Ed. by G. P. Patil ef al.), Vol. 1, pp. 403-22. Pennsylvania State University Press. Lindsey, A. A., Barton, J. D. & Miles, S. R. (1958) Field efficiencies of forest sampling methods. Ecology, 39, 428-44.REFERENCES 335 Lloyd, M. (1967) ‘Mean crowding’. J. anim. Ecol. 36, 1-30. Lloyd, M. & Ghelardi, R. J. (1964) A table for calculating the equitability component of species diversity. J. anim. Ecol. 33, 217-25. Lloyd, M., Zar, J.H. & Karr, J. R. (1968) On the calculation of information— theoretical measures of diversity. Am. Midl. Nat. 79, 257-72. Lloyd, P. S. (1972) The grassland vegetation of the Sheffield region. II. Classification of grassland types. J. Ecol. 60, 739-76. Loetsch, F., Zéhrer, F. & Haller, K. E. (1973) Forest Inventory Vol. 2. BLV Verlagsgesellschaft, Munich. Loucks, O. L. (1962) Ordinating forest communities by means of environmental scalars and phytosociological indices. Ecol. Monogr. 32, 137-66. Louppen, J. M. W. & van der Maarel, E. (1979) CLUSLA: a computer programme for the clustering of large phytosociological data sets. Vegetatio, 40, 107-14. Lowe, R. G. (1967) Competition and Thinning Studies in Nigeria, 16 pp. Federal Dept. of Forest Research, Ibadan, Nigeria. Ludwig, J. A. & Goodall, D. W. (1978) A comparison of paired- with blocked-quadrat variance methods for the analysis of spatial pattern. Vegetatio, 38, 49-59. Lynch, D.W. & Schumacher, F. X. (1941) Concerning the dispersion of natural regeneration. J. For. 39, 49—S1. Maarel, E. van der (1969) On the use of ordination models in phytosociology. Vegetatio, 19, 21-46. Maarel, E. van der (1972) Ordination of plant communities on the basis of their plant genus, family and order relationships. In: Grundfragen und Methoden in der Pflanzensoziologie, (Ed. By E. van der Maarel and R. Tiixen), pp. 183-92. Junk, The Hague. Maarel, E. van der (1979) Transformation of cover-abundance values in phyto- sociology and its effects on community similarity. Vegetatio, 39, 97-114. Maarel, E. van der, Janssen, J. G.M. & Louppen, J. M. W. (1978) TABORD, @ program for structuring phytosociological tables. Vegeratio, 38, 143-56. MacArthur, R. H. (1957) On the relative abundance of bird species. Proc. natn. Acad. Sci, U.S.A. 43, 293-5. MacArthur, R. H. (1960) On the relative abundance of species. Am. Nat. 94, 25-36. MacArthur, R. H. (1965) Patterns of species diversity. Biol. Rev. 40, 510-33. MacArthur, R. H. & MacArthur, J. W. (1961) On bird species diversity. Ecology, 42, 594-8, McGinnies, W. G. (1934) The relation between frequency index and abundance as applied to plant populations in a semi-arid region. Ecology, 15, 263-82. Melntosh, R. P. (1962) Pattern in a forest community. Ecology, 43, 25-33. McIntosh, R. P. (1967a) An index of diversity and the relation of certain concepts to diversity. Ecology, 48, 392-404. McIntosh, R. P. (1967b) The continuum concept of vegetation. Bot. Rev. 33, 130- 87. McIntosh, R. P. (1973) Matrix and plexus techniques. In: Ordination and Classification of Communities. Handbook of Vegetation Science, 5 (Ed. by R. H. Whittaker), pp. 159-91. Junk, The Hague. McIntyre, G. A. (1953) Estimation of plant density using line transects. J. Ecol. 41, 319-30.