1HVWHG6SDWLDO3DWWHUQVLQ6HHG%DQNDQG9HJHWDWLRQRI0HGLWHUUDQHDQ2OG)LHOGV $XWKRU V 6/DYRUHO-'/HEUHWRQ0'HEXVVFKH-/HSDUW 6RXUFH -RXUQDORI9HJHWDWLRQ6FLHQFH9RO1R -XQ SS 3XEOLVKHGE\ %ODFNZHOO3XEOLVKLQJ 6WDEOH85/ http://www.jstor.org/stable/3235929 $FFHVVHG Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=black. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact
[email protected].
Blackwell Publishing is collaborating with JSTOR to digitize, preserve and extend access to Journal of Vegetation Science.
http://www.jstor.org
Journal of VegetationScience 2: 367-376, 1991 ? IAVS,OpulusPress Uppsala Printed in the UnitedStates of America
367
Nested spatial patterns in seed bank and vegetation of Mediterranean old-fields Lavorel, S., Lebreton, J. D., Debussche, M. & Lepart, J. Centred'Ecologie Fonctionnelle et Evolutive,C.N.R.S.,B.P. 5051, F-34033 MontpellierCedex, France; Tel: +33 67 61 32 64; Fax: +33 67 41 21 38
Abstract. A nested samplingdesign was used to describethe spatialpatternsfor the species richnessandcompositionin the seed bankandvegetationof threeMediterraneanold-fields (1, 7, and 15 yr after the last ploughing). Three scales were examined hierarchically:sampling units within plots of 0.25 m2for the vegetationand of 0.05 m2for the seed bank, 100 m2 plots withinfields, andfields of 1000 m2.In spite of the strong spatial variationamong sampling units, species richness and compositionof both seed bankand vegetationshowed hierarchically structuredpatternsof heterogeneity,while each oldfield was a homogeneousentity.These spatialpatternstended to be partiallymasked when the data were aggregatedat the scale of the plot. Such results stress the use of a nested samplingdesign for studyingvariationin species richnessand taxonomiccompositionin bothvegetationandseed bank.This design, in combinationwith CCA, also showed that the vegetationshowed a coarsergrainthanthe seed bank,probablyin relationto seed clumping. Keywords: CanonicalCorrespondenceAnalysis; Hierarchy; Samplingdesign; Species richness;Taxonomiccomposition. Nomenclature: Tutin et al. (1964 - 1980). Abbreviations: CA = CorrespondenceAnalysis; CCA = Canonical CorrespondenceAnalysis; MCR = Multivariate CorrelationRatio.
Introduction The soil seed bankcan contributesignificantlyto the dynamics of plant communities (Egler 1954; Harper 1977; Connell& Slatyer 1977). The role of buriedseeds has been intensively investigatedin the context of succession (Major & Pyott 1966; Thompson & Grime 1979; Rabinowitz & Rapp 1980; Rabinowitz 1981; Marks & Mohler 1985), and more recently of patch dynamics(Christensen1985). The soil seed bankcan in particularact as a pool forregenerationafterdisturbances
(Marks 1974; Baskin & Baskin 1978; Uhl et al. 1981; Levassor,Ortega& Peco 1990), and as a buffer against environmental stochastic spatio-temporal variability (Templeton& Levin 1979; Ellner 1984). The pool of germinableseeds and the actual vegetation are thus dynamically linked. Therefore, their spatio-temporalpatternscan be expected to be interdependent.A numberof studies have questionedwhether the spatialpatternsin the seed bank and the vegetation are similar and can be predictedfrom each other. The results vary between systems, habitats, and species (Schenkeveld & Verkaar1984; Thompson 1986; Henderson, Petersen & Redak 1988). From the practical point of view, the distributionof buriedseeds is critical for the experimentaldesign (Bigwood & Inouye 1988), andfor managementandconservationpolicies (e.g. Hill & Stevens 1981; Jalloq 1975; Jefferson& Usher 1987). However, the observed patternsmay depend on the scale of observation (Wiens 1977); thus, the relevant scales of variationneed to be identifiedbeforeprocesses can be studied(Urban,O'Neill & Shugart1987). Therefore, the samplingdesign is crucialto understandingthe dynamicsof vegetation, seed bank and theirrelations. We investigatedthe spatialpatternsof species richness and composition for the soil seed content and the vegetation in a set of three old-fields (1, 7, and 15 yr afterthe last ploughing) in southernFrance.We used a hierarchicalsamplingdesign to describethe degrees of heterogeneityfor species richness and composition on three scales: the samplingunits:plots of 0.25 m2of vegetation, and 0.05 m2in the seed bank;plots of 100 m2 withinfields, andentirefields of 1000 m2. Patternswere compared between vegetation and seed bank and the following hypotheses were tested: 1. Due to the hierarchicalnatureof vegetation patterns (Krummelet al. 1987; O'Neill et al. in press), observations of the spatial structuremay be strongly scaledependent.Therefore,a nested sampling design is appropriateto detect the exact patternsof spatial variability.
Lavorel, S. et al.
368
landscape
,,
In. ,oo-
I. I
Fig. 1. Samplingdesign.
2. Fieldsof differentsuccessionalages arehomogeneous entities for both species richnessandcompositionof the vegetation and seed bank (i.e. their internalvariationis smallerthanthe differencesbetween fields). 3. The patternsof spatial heterogeneity in vegetation and seed bank are related, and similarity values are scale-dependent.Results will be comparedwith previous studies of scales of spatialheterogeneityin the seed bank, and with observationson the relations between patternsin the seed bank and the vegetation. We will also discuss some practical implications concerning samplingdesigns and data aggregation. Methods Studysite The study site is located near Montpellier, in the French Mediterraneanregion (43039' N 3?51' E). The climaie is classified as subhumidMediterraneanwith cool winters(Daget 1977). Rainfall is irregularwith an annual mean of 857 mm. The mean of the minimum temperaturesof the coldest month, January,is 0.6 ?C and the mean of the maximum temperaturesof the warmestmonth,July, is 28.7 ?C (Debussche & Escarre 1983). Frost occurs regularlyfrom mid-November to mid-March.The soil is a 1.5 m deep Mediterraneanred soil on a limestone substrate. The landscapeconsists of a mosaic of successional stages fromploughedfields to coppices of Quercusilex.
The studysite comprisesa set of threeabandonedfields, whichdifferin time since last ploughing(1,7 and 15 yr). They are considered as stages in a successional sere, since theirenvironmentandtheirearlierhistoriesarethe same. A fairly similar successional vegetation in the same region was describedby Escarreet al. (1983). Seed bankand vegetationsampling (Fig. 1) In each of the three fields, three0.01 ha plots were chosen as visually homogeneous in species composition. Within each plot, seven to nine 50 cm x 50 cm squares(referredto as 'subplots') were distributedrandomly, with a minimumof 2 m betweenany two of them (it is estimatedthatover 90% of the seeds of herbaceous species fall within 1 m of the motherplant).The species present in these squares were identified every month from April to October 1988. Additionalspecies outside the subplotsat the levels of the plot andof the field were also recordedduring releves on these scales. Monthly sampling on the scale of the plots from April to June 1989 did not addnew species to those observedin 1988. The tables of species composition of the vegetationon the scale of the subplotsand of the plots will be further referredto as VGSand VGprespectively. The soil was sampled at three dates covering the periods of naturalgermination,in relation to average favourablemoistureconditions:earlyOctoberandearly December 1988, and March 15 1989. At each of these dates, within each plot, ten 500 cm2(22.4 x 22.4 cm), 3 cm deep, subplotswere randomlydistributedbetween
- Spatial patterns in Mediterranean old-fields -
369
Table 1. Hierarchicalde-compositionby CCA of thevarianceof species compositiondatarelativeto thedesign variables,andanalogy with nested ANOVA. I Vegetation;2 Seed bank. Level
Variable
No. of variables
Conditional variable
FIELD
Xf
3
FIELD
PLOT
Xp
9
SUBPLOT
Xs
691 902
PLOT within FIELD SUBPLOT within PLOT
Dimension
Inertia
MCR
2
I(CCAI)
Xp/Xf
6
I(CCA2)
Xs/Xp
60'
I(CCA3)
I(CCA1) /I(CA) I(CCA2) I(CA) I(CCA3) I(CA)
n.
Xf
Total
812
68'
Inertia/ dimension I(CCA )/nI I(CCA2)/n2 I(CCA3)n3
I(CA)
1
I(CA)/n-I
SSE,
SSEI/SSEtot
MSE,
892
Analogy with ANOVA
the vegetation subplots. The great majority of the viable seeds of short-lived species are located within the first few cm for short-lived species (Graham & Hutchings 1988; Schenkeveld & Verkaar 1984). The absence of germination at the locations of the soil samples from the sampling date until the dispersal of new seeds confirmed that all viable seeds had been sampled down to 3 cm deep. Because the aim of the study was the estimation of the potential recolonisation from buried seeds, we determined the 'readily-germinable' seed content (Thompson & Grime 1979), i.e. the seeds capable of germination without any preliminary treatment. Thus the 'seed bank' in fact includes both newly dispersed and older components of the buried seed content (see Thompson & Grime 1979). For this purpose, the soil cores were laid in 20 cm x 30 cm germination trays, which were arranged at random under a transparent plastic shelter and watered daily by misting. The temperatures were close to outdoor conditions, although the diurnal variations were slightly reduced. Most seedlings appeared during the first month. The soil was turned upside down after two months and observed for signs of germination for two more months, but no new species were recorded in any of the soil cores. The seedlings were identified as soon as they were large enough, and pulled out in order to keep the densities low. Seeds were identified at the species level using reference seedlings grown from seeds harvested the previous year, and available weed seedling flora (Hanf 1976; Bailly, Mamarot & Psarski 1983; Mamarot & Psarsky 1983; Moreira et al. 1986). Some species could not be distinguished and were grouped (see App. 1). Since we are interested in the spatial structure of the germinable seeds present over the whole vegetation cycle, we analysed the sum of all presence-absence
records over the three sampling dates (see Lavorel et al. 1991 for an analysis of seasonal variability), and composed a table of abundances of these taxa in relation to the location of the soil cores. The taxonomic composition of the seed bank at the plot level was obtained by simple presence-absence aggregation of the subplots within each plot. We also analysed the taxonomic composition at the level of the plot when only the species present in more than 10% of the subplots (i.e. 3 subplots) were considered. The tables of species composition of buried germinable seeds on the scales of the subplots, of the plots, and of the plots with the species present in at least 10% of the subplots will be further referred to as BKS, BKp, and BKpR respectively. Statistical methods Correspondence Analysis (CA, see e.g. Hill 1973) is a well-known technique which gives simultaneous ordinations of rows and columns (releves x species). In the present context CA was applied to seed counts and species abundance notes, and considered suitable to 0/1 data as well (Prodon & Lebreton 1981). We also applied Canonical Correspondence Analysis (CCA) to constrain CA axes to be linear functions of environmental variables and to combine direct and indirect methods of ordination (ter Braak 1986; see also Lebreton et al. 1988a). In both CA and CCA of ecological data the overall contrasts in species composition between releves are measured by an inertia factor, obtained as the sum of eigenvalues. A comparison of the inertia associated with a constrained analysis, I(CCA), with the inertia of CA, I(CA), indicates how much the environmental variables under consideration are effective in explaining variation in species composition in the data. More precisely I(CCA) / I(CA) is a multivariate multiple
Lavorel, S. et al.
370 15
-- VGs BKs.
C)
0)
._ a) 10 'o .0
z
-4
5
60
--
50
--
a.
40
o'
30
-Z Z
on - -
BKp -A- BKPR
i??~~"-*
100
0 1
7
Fieldage (years)
15
1
7
Field age (years)
a
correlationratio (denotedMCR) (Sabatier,Lebreton& Chessel 1989), which can be interpretedas the squareof a correlationcoefficient (Lebretonet al. 1991). CCA appearsas particularlyefficient when the environmental variablesare categorical(Lebretonet al. 1988b). When two variables(or sets of variables)A and B are considered, one may analyse the effect of B if the effect of A has been accountedfor. This analysis is called a partial CCA (ter Braak 1988; Sabatier, Lebreton & Chessel 1989) and is denotedby CCA with respectto B/A. To account for the hierarchicaldesign of our study we used categorical variables Xf, Xp, Xs, with respectively 3, 9, and 69 (vegetation) or 90 (seed bank) categories, to representthe three levels of the design; we also associateda particularconditionalCCA/to eachlevel i in the design, and thus a particularnumberof axes nl and a particularinertiaI(CCA/). CCA1 is a CCA with respect to Xf, with n1 = 2; CCA2 to Xp/Xf,with n2= 6; CCA3 to Xs/ Xp, with n3 = 60 (vegetation) or 81 (seed bank) (Table 1). The axes of each of these analyses are conditional (ter Braak 1988; Sabatier,Lebreton & Chessel 1989) and arenot correlatedto variationsat higherlevels. As a consequence, the inertias I(CCAl) add up to I(CA). Similarly,the MCRvalues, I(CCA,)/I(CA),addup to 1. The parametersn, and I(CCA,) play the same role as degrees of freedom and sum of squaresrespectively in univariatenested analysisof variance.Hence, the variation in species compositionis brokendown into hierarchical componentsinducedby the design. The role of a particularlevel of sampling is even more precisely quantifiedby accountingfor its numberof axes, i.e. by using I(CCA,)/n,, which is analogous to the Mean Square
Errorin ANOVA. The procedureand the analogy with ANOVA is describedin Table 1. The In, ordinationsobtained(for samples and species respectively) are not correlated, within a given analysis as well as between analyses. Ratherthanlooking for a posterioriuncorrelatedordinationsof decreas-
VGp
15
Fig. 2. Species richnessin the vegetationand the seed bank: a. Speciesrichnesson the scale of thesubplotin the vegetation (VGS) and in the seed bank (BKS). b. Speciesrichnesson the scale of the vegetation plot (VGP), the seed bank (BKP), and the seed bankrestrictedto thespecies presentin at least 10%of the subplots(BKPR).
b
ing importance,like CA does, our approachsplits the informationaccordingto the a prioridesign. In a complementaryway, the variationin the number of species persubplot,respectivelyin the vegetationand in the seed bank, was analysedby a nested analysis of variance,using a mixed model to accountfor the random characterof PLOTswithinFIELDs.Richnessvalues for vegetation and seed bank at the same level of sampling were compared by nonparametricone-way ANOVA (Kruskal-Wallistest).
Results Species richness 118 species were identifiedin the vegetation,with a meanof 7.9 species (range4 - 13) in each 50 cm x 50 cm subplot, and 45.4 species in each plot (range:30 - 65). Species richness on the scale of the subplot increased significantly - F(1,68) = 5.64; p < 0.05 - with the suc-
cessional age of the field (Fig. 2a); this was not so for
richness on the scale of the plot - F(1,8) = 0.12; n.s. -
(Fig. 2c). 84 taxa germinatedfrom buried seeds throughout the vegetationcycle, with a meanof 9.3 species (range3 - 18) per soil core, and a mean of 49.6 species (range44 - 58) perplot. The meandroppedto 40.0 species perplot (range: 34 - 47) when only the 56 species recordedin more than 10% of the sampling units were included. Speciesrichnessincreasedsignificantlywithsuccessional age of the field on the scale of the sampling unit F(1,89) = 7.06; p < 0.01 - (Fig. 2b), as well as on the scale of the plot - F(1,8) = 8.13; p < 0.025 - (Fig. 2d).
The numberof species in vegetation or seed bank dependedsignificantlyon the field (p < 0.05) but not on the plot within a given field, whetherthe data are consideredon the scale of the subplotor on the scale of the plot (Table 2). Thus, the fields can be considered as
- Spatial patterns in Mediterranean old-fields -
371
Table 2. Analysis of varianceof species richnessin vegetationand seed bank.VG5, BKS: counts on the scale of the subplotfor the vegetationandthe seed bankrespectively.VGp, BKp, BKpR:countson the scale of the plot, for vegetation,seed bank,andseed bank restrictedto species presentin at least 10%of the subplotsrespectively.ns = not significant;* =p < 0.05, ** =p < 0.01. Seed bank
Vegetation
Analysis Xf Xp/Xf Correlation/age
VGs
VGp
BKs
BKp
BKpR
F-test df
F-test df
F-test df
F-test df
F-test df
3.15 0.086 5.64
(2,60) * (6,60) ns (1,68) *
6.94 (2,4) * 1.13 (2,4) ns 0.12 (1,8) ns
internally homogeneous for species richness. The seed bank was richer than the vegetation on the small scale (Kruskal-Wallis test KW(1) = 13.34; p < 0.001). On the scale of the plot, seed bank richness was higher than in the vegetation, but this trend was not significant (Kruskal- Wallis test KW(1) = 0.87; p > 0.05). The order was reversed for richnesses on the scale of entire fields, and, as a whole, the vegetation was richer than the seed bank. Hence, the grain for species richness was coarser in the vegetation than in the seed bank. Taxonomic composition
3.00 2.10 7.06
8.26 0.35 5.32
(2,4) * (2,4) ns (1,8) *
(2,4) * (2,4) ns (1,8) *
8.65 0.29 8.13
Table 3. Componentsof the variationof the taxonomiccomposition of vegetationand seed bank, as measuredby the inertia of CanonicalCorrespondenceAnalysis. See text for notations and Table la for dimensions. a. Analyses on the scale of the subplots.b. Analyses on the scale of the plot. a.
Inertia
MCR
Inertia/dim.
Variable
VGs
BKs
VGs
BKs
VGs
BKsI
Xf XJXf Xp
0.931 0.785 1.717
0.594 0.276 0.870
Xs/Xp(residual)
6.035
6.892
0.12 0.10 0.22 0.78
0.078 0.036 0.11 0.89
0.466 0.131 0.215 0.097
0.297 0.046 0.109 0.026
CA
7 752
7.662
1.0
1.0
b.
The inertias of unconstrained CA of the tables of taxonomic composition of the vegetation and the seed bank were much higher than those of the CCA relative to the spatial factors (Table 3a). This showed a strong residual variation on the scale of the subplot. For the data on the scale of the unit subplot, the factor FIELD contributed more to the total variation than the factor PLOT, as shown by the values of the MCR and I(CCA!)/ n, in the respective CCAs (Table 3a). Thus, in biological terms, differences between fields were more important than differences between plots within fields, so most of the spatial variation was due to differences in species composition between fields. The contributions of the FIELD and the PLOT factors were reversed for the data aggregated on the scale of the plot (Table 3b). The elimination of infrequent species from the seed bank data (using BKpF) did not modify this effect of aggregation, although it increased the importance of the variability between fields relative to that within fields (Table 3b). For vegetation seed bank composition on the scale of the subplot, the first axis of the CCA relative to the variable FIELD alone contrasted the 1-yr old field to the two older fields (7-yr and 15-yr old), while the second axis separated the 7-yr old from the 15-yr old field (Fig. 3). The species responsible are those having high loadings on these axes. The vegetation of the 1-yr old field was characterized by a number of annuals, especially from
(2,81) * (6,81) ns (1,68)**
Xf
Var.
Inertia MCR 0.701 0.321 0.200 0.45 0.38 0.51 VGp BKp BKpR VGp BKp BKpR
Xf 0.701 0.321 Xp/Xf 0.845 0.523 1.546 0.844 Xp
CA
0.200 0.193 0.393
1.546 0.844 0.393
0.45 0.38 0.55 0.62 1.00 1.00
0.51 0.49 1.00
Inertia/dim. 0.36 0.16 0.10 VGp BKp BKpR2 0.36 0.14
0.16 0.09
0.10 0.03
0.19
0.11
0.05
1.00 1.00 1.00
'VGs: vegetationcomposition;BKS: seed bank composition; 2VGp:vegetationcomposition;BKp:seed bank composition; BKPR:seed bank composition restrictedto species presentin at least 10%of the samplingunits.
the Asteraceae, commonly met as weeds in the area (Table 4a). The two older fields were opposed to that field by the frequent presence in the vegetation of two grass species, and of four annual forbs (Table 4bl). The vegetation of the 7-yr old field owed its specificity to two grasses, and four perennial species (Table 4b2). The vegetation of the 15-yr old field was characterized by four grasses, one annual legume and one annual forb, two monocarpic perennials, four perennial forbs, and three woody species (Table 4b2). The seed bank of the 1yr old field was distinguished by a variety of short-lived forbs, with again many species from the Asteraceae (Table 4c). The seed banks of the two older fields were distinguished from that of the younger field by the
Lavorel, S. et al.
372
Table 4. Species differentiatingthe vegetationandthe seed bankof the fields. aSpeciesdifferentiatingthe vegetationof the 1-yrold field from the 7-yr and 15-yr old fields. bliSpeciescharacteristicof the vegetationof the7-yr and 15-yrold fields. b2Speciescharacteristicof the vegetationof the 7-yr old field. b3Speciescharacteristicof the vegetationof the 15-yrold field. cSpeciesdifferentiating the seed bank of the 1-yr old field from the 7-yr and 15-yrold fields. dlSpecies characteristicof the seed bank of the 7-yr and 15-yr old fields. d2Speciescharacteristicof the seed bankof the 7-yr old field. d3 Species characteristicof the seed bankof the 15-yrold field. 1-yr olda Cirsiumvulgare Conyza sp pl. Crepispulchra Crepis vesicania C Crepis sancta Geraniumdissectum Papaver dubium Senecio vulgaris Sonchus oleraceus
Z 0 -
r p >
1-yr oldc X
m
X
r~
Amaranthusretroflexus Crepisfoetida Crepispulcher Crepis vesicaria Cirsiumvulgare Diplotaxis erucoides Geraniumdissectum Lactuca serriola Picris hieracioides Solanumnigrum
7+15-yr old bi Atriplexpatula Avena sterilis Bromusmadritensis Carduuspycnocephalus Euphorbiahehoscopia Vicia lutea
Aegilops ovata Avena barbata Dittrichia viscosa
Foeniculumvulgare Lepidiumdraba Plantago lanceolata
7 b2
Bromusmacrostachys Celtis australls Cynodondactylon Daucus carota Echiumvulgare
Galium aparmine Hordeummurinum Jasmlnumfiuticans Lathyruscicera Melica cillata
Rublaperegrina Rubusulmifolius Rumexpulcher 15 b3 Scabiosa maritima Ulmus minor
Centranthusruber Foenlculumvulgare Lepidiumdraba
Muscarl comosum Plantago lanceolata Tr-bulusterrestris
7 d2
Bromusmadrltensis Daucus carota Echiumvulgare Lathyruscicera
Linaria spuila Melica cillata Rubus ulmifollus Torllis nodosa
7+15-yr old dl
Calaminthanepeta Malva alcea Verbascumsinuatum Vicia lutea
presence of one legume and three longer-lived species (Table 4dl). One summer annual and five perennials were specific to the seed bank of the 7-yr old field (Table 4d2). Finally, the seed bank of the 15-yr old field was characterized by two grasses, five short-lived forbs, and one shrub (Table 4d2). The ratios I(CCA,)/n, were higher in the vegetation than in the seed bank for all scales (Table 3), indicating that the vegetation was more heterogeneous than the seed bank. In addition, the relative contribution of the heterogeneity on the scale of the field to the overall spatial variability, as estimated by the inertia/dimension ratio in the CCA relative to FIELD alone, was larger in the vegetation than in the seed bank. This suggests that the grain in the vegetation was coarser than in the seed bank. These differences were less on the scale of the plot as compared with the subplot. Thus, broad scale patterns in the vegetation and the seed bank were similar.
15 d3
position was smaller than the between-field heterogeneity, for the vegetation as well as for the seed bank. From the list of species concerned, we interpret these differences between fields as reflecting their differences in successional age. We conclude that fields had a fairly high internal homogeneity. A nested sampling design such as the one used here minimizes the small-scale variability (i.e. on the scale of the subplot), while maximizing the large scale variability (i.e. on the field scale) (Scherrer 1983), and hence assesses the large-scale variability with a good reproducibility. Our results validate the use of a nested sampling design for providing a reliable representation of the spatial variability in species assemblages having a hierarchical pattern. Such a design will then allow the use of aggregated data on the scale of the entire field for further analyses of species richness and composition. Small-scale variability
Discussion
Nested spatial patterns and samplingdesign In spite of the high between-subplot variation, the within-field heterogeneity in species richness and com-
A considerable small-scale variability was observed in both vegetation and seed bank. This results in part from local clustering effects. Large sampling variances in the seed bank have been assigned to clumped distributions. Seed clumping seems a logical result of seed dispersal, which tends to keep most of the seeds within
- Spatialpatternsin Mediterraneanold-fields a. N
b. VF7
8g
BF7
0
BF1 VF1
373
that the relative weight of rare species is increased. Attentionshouldbe drawnto this artifactlinkedwith the use of presence-absence data, and its impact on the study of spatialpatterns.On the other hand, with presence-absence data emphasis on fine-grain variation, linked in particularto seed clumping, is avoided. Thus, a presence-absenceapproachrepresentsa fair compromise between logistic costs and obtaining a realistic descriptionof spatialvariability. Comparisonof vegetationand seed bankpatterns
VF15
BF15 CCA1
CCA1
Fig. 3. Projectionsof the fields in the plane of the first two components(CCA1, CCA2) of the CCA relativeto FIELD.a. Composition of the vegetation. VF1, VF7, and VF15: 1-yr, 7-yr and 15-yr old fields respectively. b. Compositionof the seed bank.BF1, BF7, andBF15: 1-yr,7-yr and 15-yrold fields respectively.
a shortdistance aroundthe motherplant (e.g. Levin & Kerster1974;Harper1977).However,Thompson(1986) and Bigwood & Inouye (1988) pointed out that the apparentclusteringof seeds may resultfrominadequate sampling.In spite of attemptsfoundin the literature(cf. Thompson 1986; Bigwood & Inouye 1988; and references therein),no optimumsolutionhas yet been found to the problemof logistic limitations,andthe necessary number of subplots. The present results suggest increasingthe numberof subplotson the expense of their size (Roberts 1981; Thompson 1986, in combination with an adequate sampling design such as a nested sampling).Such a design would enablereductionof the totalnumberof subplotsby organizingtheirdistribution on various scales. Improvementsof the samplingstrategy arenecessaryfor studiesrequiringnumerousfields, in particularreplicates among successional ages, if a reliable representationof successional patternsis to be attained. In addition,small-scale variationmay be amplified to some extent by the occurrenceof infrequentspecies in the analyses, although CCA is fairly robust with regardto rarespecies (terBraak1986). The fact thatthe contributionsto the total variationon the scales of the field andof the plot respectivelywerereversedwhen the datawere aggregatedfromthe scale of the subplotto the scale of the plot, may also resultfroman effect similarto that when using presence-absence data. Indeed, the subplot frequencyof species is a sort of coefficient of abundance(Chessel 1978; see Prodon 1988 for a discussion). After reduction to presence-absenceon the scale of the plot this quantitativeinformationis lost, so
The use of similar designs and analyses for the samplingof the vegetationand seed makes it possible to comparetheir spatial structures,that is the partitionof the variationover the scales used. Both species richness and taxonomic composition were found to display a coarser grain in the vegetation than in the seed bank. However, the higher spatialheterogeneityin taxonomic composition in the vegetation could be a statistical consequenceof the differences in species richness. The numberof speciespersubplotin thevegetationis smaller, but drawnfrom a largerpool. Since the possibilities of overlaps among random samples will decrease with theirsizes, independentsubplotsfromthe vegetationare likely to display a variance largerthan that of the subplots from the seed bank.This effect is enhancedby the fact thatthe pool for the seed bankis smallerthanfor the vegetation: in effect, the probability of overlaps between subplotswill be reduced, and the variancemade even smaller. Finally, the subplots for the vegetation were largerthan those for the seed bank, which could explain the differencesobservedin the degrees of small scale variability.However, space may not be perceived in the same way by seeds and adultplants;moreoverthe choice of smaller subplots for the seed bank seems logical to avoid seed clumping effects. The species compositions of the seed bank and the vegetationhave been comparedandoften been found to differ(e.g. Thompson& Grime 1979; Rabinowitz1981; Mayor & Pyott 1986; Thompson 1986; Manders1990). We observed a large degree of overlap in the compositions of the vegetation and seed bank in these Mediterraneanold-fields (see Lavorel& Lebreton1991). On the other hand, studies trying to relate patternsin the seed bankto those in the vegetation are few. Schenkeveld& Verkaar(1984) found for calcareousgrasslandsthatthe relationbetweenvegetationand seed bankstructurewas mostly slight. Henderson,Peterson& Redak (1988) on the otherhand, observed strong correlationsin a desert grassland. However, in the latter case the dominants were mainly annuals, against perennials in the former case. Various biotic and abiotic factors interfere with the initial seed distributionafter seed dispersal,such as
374
Lavorel, S. et al.
seed predation, transport by ants, topographic accumulations, and their effects are more or less species- and density-dependent. As a result, very different patterns may be observed between species within one community (Schenkeveld & Verkaar 1984; Thompson 1986; Henderson, Peterson & Redak 1988), or between communities differing in their vegetation structures (Kellman 1978; Schenkeveld & Verkaar 1984). Here, we also have to take between-year variation into account. In addition, the scale of sampling has to be kept in mind. Patterns observed on smaller scales may disappear as the scale becomes larger and the distribution of seeds should become increasingly similar to that of the parent plants (Bigwood & Inouye 1988). We discussed only a few of the many phenomena modifying the spatial distribution patterns of mother plants in the vegetation as well as those observed for the seeds in the soil. For the time being, the dynamic mechanisms and historical factors which link these two components of the plant community need to be studied case for case, while waiting for a more general predictive scheme of spatial patterns to be developed. The application of Canonical Correspondence Analysis to composition data obtained using an adequate spatial design enables us to de-compose the spatial variability into additive levels, which provides us with a useful tool for the testing of hypotheses.
Christensen,N. L. 1985. Shrublandfire regimes and their evolutionaryconsequences.In: Pickett, S. T. A. & White, P. (eds.) The ecology of natural disturbances and patch dynamics,pp. 86-100. Academic Press, New York. Connell, J. H. & Slatyer,R. 0. 1979. Mechanismsof succession in naturalcommunities and their role in community stabilityand organization.Am. Nat. 111: 1119-1144. Daget, P. 1977. Le bioclimat mediterran6en: caracteres generaux,modes de caract6risation.Vegetatio 34: 1-20. Debussche, M. & Escarr6,J. 1983. Carte des isohyetes interannuelles dans le Montpellierais:documentetablipour la serie 1950-1979. Centre L. Emberger, C.N.R.S., Montpellier. Egler, F. E. 1954. Vegetation science concepts. 1. Initial floristic composition: a factor in old field vegetation development. Vegetatio4: 412-417. Ellner, S. 1984. Asymptotic behavior of some stochastic differenceequationpopulationmodels.J. Math.Biol. 19:169200. Escarre,J., Houssard,C., Debussche, M. & Lepart,J. 1983. Evolutionde la vegetationet du sol apresabandoncultural 6tudede la succession dansles en regionm6diterran6enne: Garriguesdu Montp6llierais(France).Acta Oecol. Oecol. Plant. 4: 221-239. Forman,R. T. T. & Godron, M. 1986. Landscape Ecology. Wiley & Sons, New York. Graham,D. J. & Hutchings,M. J. 1988. Estimationof the seed bankof a chalk grasslandley establishedon formerarable land.J. Appl. Ecol. 25: 241-252. Hanf, M. 1976. Les adventices et leurs plantules. La Maison Rustique, Paris.
Acknowledgements. We thankthe staff fromthe Experimental Stationof the Centred'Ecologie Fonctionnelleet Evolutive, Montpellier.Research was supportedby C.N.R.S., and a fellowship from the French Ministerede la Recherche et de la Technologie.
References Ayala, F. J. 1971. Competitionbetween species andfrequency dependence.Science 171: 820-824. Bailly, R., Mamarot,J. & Psarski,P. 1983. Mauvaisesherbes des grandescultures,2nded. Associationde Coordination TechniqueAgricole. Le Carrousel,Paris. Baskin, J. M. & Baskin, C. C. 1978. The seed bank in a populationof an endemic plant species and its ecological significance.Biol. Conserv. 14: 125-130. Bigwood, D. W. & Inouye,D. W. 1988.Spatialpatternanalysis of seed banks:An improvedmethodandoptimizedsampling. Ecology 69: 497-507. de la disperChessel, D. 1978. Descriptionnon-parametrique sion spatialedes individusd'une espece. In: Legay, J. M. & Thomassone,R. (eds.) Biometrieet Ecologie 1, pp. 45135. Societe Frangaisede Biom6trie,Paris. Chesson,P. L. 1986. Environmentalvariationandthe coexistence of species. In: Diamond, J. & Case, T. (eds.) CommunityEcology,pp. 240-256. HarperandRow, New York.
Harper,J. L. 1977. Population Biology of Plants. Academic Press, London. Henderson,C. B., Petersen,K. E. & RedakR. A. 1988. Spatial andtemporalpatternsin the seed bank andvegetationof a desert grasslandcommunity.J. Ecol 76: 717-728. Hill, M. 0. 1973.Reciprocalaveraging:an eigenvectormethod of ordination.J. Ecol. 61: 237-249. Hill, M. 0. & Stevens, P. A. 1981. The density of viable seed in the soils of forest plantationsof uplandBritain.J. Ecol. 69: 693-709. Jalloq, M. C. 1975. The invasion of molehills by weed as a possible factorof degenerationof reseededpasture.I. The buriedviable seed populationof molehills from reseeded pasturesin West Wales. J Appl. Ecol. 12: 643- 657. Jefferson, R. G. & Usher, M. B. 1987. The seed bank of disused chalk quarriesin the YorkshireWolds, England: Implicationfor conservationmanagement.Biol. Conserv. 42: 287-302. Kellman,M. C. 1978. Microdistributionof viable weed seed in two tropicalsoils. J. Biogeogr. 5: 291-300. Krummel,J. R., Gardner,R. H., Sugihara,G., O'Neill, R. V. & Coleman, P. R. 1987. Landscapepatternsin a disturbed environment.Oikos 48: 321-324. Lavorel, S., Debussche, M., Lebreton,J. D. & Lepart,J. 1991. Seasonal patternsin the seed bank of Mediterraneanold fields. In: Lavorel, S.Spatial structure,disturbances,and dynamicsof coexistenceof plant species 'from experimentation to modelling. The example of Mediterraneanold fields, pp. 65-87. These de Doctorat, Universite des
- Spatial patterns in Mediterranean old-fields Sciences et Techniquesdu Languedoc,Montpellier. Lavorel, S. & Lebreton,J. D. 1991. A field approachfor the theoryof lotteryfor establishment.In: Lavorel,S. Spatial structure, disturbances,and dynamics of coexistence of plant species: from experimentationto modelling. The exampleof Mediterraneanoldfields, pp. 53-63. These de Doctorat, Universite des Sciences et Techniques du Languedoc,Montpellier,France. Lebreton,J. D., Chessel, D., Prodon,R. & Yoccoz, N. 1988a. L'analyse des relations especes-milieu par l'Analyse Canoniquedes Correspondances.I. Variables de milieu quantitatives.Acta Oecologica, Oecol. Gen. 9: 53-67. Lebreton,J. D., Chessel, D., Richardot-Coulet,M. & Yoccoz, N. 1988b. L'analyse des relations especes-milieu par l'Analyse Canoniquedes Correspondances.II. Variables de milieuqualitatives.ActaOecol. Oecol. Gen. 9:137-151. Lebreton,J. D., Sabatier,R., Banco, G. & Bacou, A. M. 1991. Principal component and correspondenceanalyses with respectto instrumentalvariables:an overview of theirrole in studiesof structure- Activity andspecies- environment relationships.In:Devillers,J. & Karcher,W. (eds.)Applied multivariateanalysis in SAR and environmentalstudies. Kluwer,Dordrecht. Levassor,C., Ortega,M. & Peco, B. 1990. Seed bankdynamics of Mediterraneanpastures subjected to mechanical disturbance.J. Veg. Sci. 1: 339-344. Levin, D. A. & Kerster,H. W. 1974. Gene flow in seed plants. Evol. Biol. 7: 139-220. Major, J. & Pyott, W. T. 1966. Buried viable seeds in two Californiabunchgrasssites andtheirbearingon the definition of the flora. Vegetatio 13: 253-282. Mamarot,J. & Psrasky,P. 1983.Mauvaisesherbesdes grandes cultures, Complement. Association de Coordination TechniqueAgricole. Le Carrousel,Paris. Manders, P. T. 1990. Soil seed banks and post-fire seed depostition across a forest-fynbos ecotone in the Cape Province.J. Veg. Sci. 1: 491-498. Marks, P. L. 1974. The role of pin cherry (Prunus pennsylvanica)in the maintenanceof stabilityin northern hardwoodecosystems. Ecol. Monogr.44: 73-88. Marks,P. L. & Mohler,C. L. 1985. Succession afterelimination of buried seeds from a recentlyploughed field. Bull. Torr.Bot. Club 112: 376-382. Moreira,I., Guillerm,J. L., Caixinhas,L., EspiritoSanto,D. & Vasconcelos, T. 1986. Mauvaises herbes des vergers et vignes de l'ouest du bassin mediterraneen.Grafica2000, Lisboa. O'Neill, R. V., Gardner,R. H., Milne, B. T., Turner,M. G. & Jackson,B. in press.Heterogeneityandspatialhierarchies. In: Kolosa, J. & Pickett, S.T.A. (eds.) Spatial heterogeneity. SpringerVerlag, New York. Prodon,R. 1988.Dynamiquedes systemesavifaune-vegetation apres deprise rurale et incendies dans les Pyrenees mediterraneennessiliceuses. These de Doctorat d'Etat, Paris. Prodon, R. & Lebreton,J. D. 1981. Breeding avifauna of a Mediterraneansuccession: the holm oak and cork oak series in the easternPyrenees, 1. Analysis and modelling of the structuregradient.Oikos 37: 21-38. Rabinowitz,D. 1981. Buriedviable seeds in a NorthAmerican
375
tall-grassprairie:the resemblanceof theirabundanceand composition to dispersingseeds. Oikos 36: 191-195. Rabinowitz, D. & Rapp, J. K. 1980. Seed rain in a North Americantall-grassprairie.J. Appl. Ecol. 17: 793-802. Roberts,H. A. 1981. Seed bankin soils. Adv.Appl.Biol. 6: 155. Sabatier,R., Lebreton,J. D. & Chessel, D. 1989. Principal ComponentAnalysis with instrumentalvariablesas a tool for modelling compositiondata.In: Coppi, R. & Bolasco, S. (eds.) Multiway data analysis, pp. 341-352. North HollandPubl., Amsterdam. Schenkeveld, A. J. & Verkaar,H. J. 1984. The ecology of short-livedforbs in chalk grasslands:distributionof germinativeseeds andits significancefor seedlingemergence. J. Biogeogr. 11: 251-260. Scherrer,B. 1983. Techniques de sondage en ecologie. In: Frontier,S. (ed.) Strategiesde chantillonnageen ecologie, pp. 63-162. Masson,Paris,andLes Pressesde l'Universite du Quebec, Montreal. Templeton,D. R. & Levin, D. A. 1979. Evolutionaryconsequences of seed pools. Am.Nat. 114: 232-249. ter Braak,C. J. F. 1986. Canonicalcorrespondenceanalysis:a new eigenvectortechniquefor multivariatedirectgradient analysis. Ecology 67: 1167-1179. ter Braak C. J. F. 1988. Partial canonical correspondence analysis. In: Bock, H. H. (ed.) Classificationand related methods of data analysis, pp. 551-558. North Holland Publ., Amsterdam. Thompson, K. 1986. Small-scale heterogeneity in the seed bank of an acidic grassland.J. Ecol. 74: 733-738. Thompson,K. & Grime J. P. 1979. Seasonal variationin the seed banks of herbaceousspecies in ten contrastinghabitats.J. Ecol. 67: 893-921. Tutin, T. G. et al. 1964-1980. Flora Europaea. Cambridge UniversityPress, Cambridge. Uhl, C., Clark,K., Clark,H. & Murphy,P. 1981. Early plant succession after cutting and burning in the Upper Rio Negro region of the Amazon Basin. J. Ecol. 69: 631-649. Urban, D. L., O'Neill, R. V. & Shugart Jr., H. H., 1987. Landscapeecology: A hierarchicalperspective can help scientistsunderstandspatialpatterns.Bioscience 37: 119127. Warner,R. R. & Chesson, P. L. 1985. Coexistence mediated by recruitmentfluctuations:a field guide to the storage effect. Am. Nat. 125: 769-787. Wiens, J. A. 1977. On competitionandvariableenvironments. Am. Sci. 65: 592-597. Received 4 November 1990; Revision received 29 April 1991; Accepted 6 May 1991.
App. 1 on p. 376.
376
Lavorel, S. et al.
Appendix 1. List of the species observedin the vegetationandthe seed bankof Mediterraneanold-fields. Groupsof seed banktaxa are indicatedwith brackets. Family Species
Family Species
Family Species
Family Species
Poaceae Avena sterilis } Avenafatua + Aegilops ovata Dactylis hispanica Bromusmadritensis } } Bromussterlilis Bromusmacrostachys ) Bromusmollis Hordeummurinum Loliummultiflorum Cynodondactylon Melica ciliata Gastridiumventricosum Brachypodiumphoenicoides
Ranunculaceae Nigella damascena
Portulacaceae Portulaca oleracea
Plantaginaceae Plantago lanceolata
Papaveraceae Papaver dublum
Aplaceae Foeniculumvulgare Daucus carota Torilis nodosa Scandixaustrallhs Pastinaca sativa
Amaranthaceae Amaranthusretroflexus
Leguminosae Medicago polycarpa Medicago minima Medicago orbicularis Medicago turbinata Trifoliumhirtum Astragalushamosus Vicia lutea Viciaperigrina Vicia sativa Vicia hybrida Lathyruscicera
Fumarlaceae Fumaria officinalls Brassicaceae Diplotaxis erucoides Lepidiumdraba Lepidiumgraminifolium Capsella bursa-pastoris Sinapis arvensis Cardaminehirsuta Resedaceae Resedaphyteuma
}
Caryophyllaceae Silene noctiflora Silene italica Arenariaserpyllifolia Stellaria media Cerastiumsemidecandrum
}
}
Malvaceae Malva alcea
Asteraceae Cercis silicastrum Conyzacanadensis } Conyzaalbida Senecto vulgaris Dittrichia viscosa Filago arvensis Calendulaarvensis Anthemisarvensis Carlina corymbosa Cirsiumvulgare Carduuspycnocephalus Centaureascabiosa Centaureacollina Cichorlumintybus Picris hieracioides Urospermumdalechampil Podospermumlaciniatum Tragopogonporrifolius Lactuca serriola Sonchusoleraceus Sonchusasper Picridiumvulgare Crepisfoetida Crepis vesicaria Crepis setosa } Crepis sancta
}
}
Geraniaceae Geraniumrotundifollum Geraniumdissectum Erodiummalacoides Hypericaceae Hypericumperforatum Vitaceae Vitis vinifera Oxalidac eac
Oxalis corniculata Zygophyllaceae Tribulusterrestris Rosaceae Rubusulmifolius Prunusspinosa Prunusdulcis Poterlummagnoli Onagraceae Epilobiumtetragonum
}
Chenopodiaceae Atriplexpatula Polygonaceae Rumexcrispus Rumexpulcher Polygonumaviculare Polygonumconvolvulus
Caprifoliaceae Viburnumtinus Rubiaceae Rublaperegrina Galiumaparine Galhumlucidum Sherardiaarvensis Valerianaceae Centranthusruber
Euphorbiaceae Euphorbiahelioscopia Euphorbiachamaesyce Euphorbiasegetalis Euphorbiaserrata Mercurialisannua
Dipsacaceae Scabiosa maritima Cephalarialeucantha
Ulmaceae Celtis australis Ulmusminor
}
}
Campanulaceae Campanularapunculoides
Fagaceae Quercus coccilera
Primulaceae Anagallis arvensis
Liliaceae I Muscarl commosum Muscarl neglectum Allium oleraceum Alliumpolyanthum Ornithogalumumbellatum Asparagus acutifollus
Oleaceae Jasminumfructicans Olea europaea Convolvulaceae Convolvulusarvensis Boraginaceae Echiumvulgare Heliotropiumeuropaeum Solanaceae Solanumnigrum Scrophularlaceae Antirrhinumorunt:um Linarla spuila Verbascumsinuatum Veronicahederaefolia 1Veronica arvensis Lamlaceae Thymusvulgaris Calaminthanepeta Lamiumamplexicaule
}
