ARTICLE IN PRESS

Basic and Applied Ecology 8 (2007) 332—342

1439-1791/$ - sdoi:10.1016/j.

�CorrespondE-mail addr

www.elsevier.de/baae

Multivariate analysis of the response of overgrownsemi-natural calcareous grasslands to restorativeshrub cutting

Simona Maccherinia,�, Michela Marignania,Paolo Castagninia, Paul J. van den Brinkb,c

aDipartimento di Scienze Ambientali, ‘‘G. Sarfatti’’ Universita di Siena, Via P. A. Mattioli 4, 53100 Siena, ItalybAlterra, Wageningen University and Research Centre, P.O. Box 47, 6700 AA Wageningen, The NetherlandscDepartment of Aquatic Ecology and Water Quality Management, Wageningen University and Research Centre,Wageningen University, P.O. Box 8080, 6700 DD Wageningen, The Netherlands

Received 5 December 2005; accepted 26 June 2006

KEYWORDSBARCI design;Principal responsecurves;Prunus spinosa;Restoration;Communitystructure;Monitoring design;Overgrown grassland

ee front matter & 2006baae.2006.06.002

ing author. Tel. +39 577ess: maccherini@unisi.i

SummaryMonitoring and evaluation of restoration management often suffers from poorsampling design and a lack of statistical rigour, seldom considering the spatial andtemporal variability of habitat. We tested the effectiveness of shrub cutting on therestoration of calcareous grasslands, using a 4-year Before–After/Reference-Control-Impact (BARCI) project design with replicated patches of habitat. Departure fromthe control sites and convergence with the reference sites were analysed using thePrincipal Response Curves (PRC) method.

The structure of communities was compared 1 year before and 3 years aftermanagement on control scrubland, restored ex-arable land, restored and referencegrassland. Results show that shrub cutting did not suffice to restore the communitycomposition of the semi-natural calcareous grasslands. The restored ex-arable areasmaintained a community structure extremely similar to the control scrubland.

The restored grassland, on the other hand, deviated from the control and slightlyconverged to the reference grassland, though not significantly. The dominant scrubspecies that was cut, Prunus spinosa, showed higher cover values compared to thereference grassland in any of the treatments, even the first year after cutting.Species typical of xeric and/or calcareous grassland were more abundant in thereference than in the restored grassland, while arable and ruderal species weremore frequent in the restored sites.

This study has demonstrated that the BARCI approach is a powerful tool for theevaluation of restoration management, as it was possible to evaluate not only

Gesellschaft fur Okologie. Published by Elsevier GmbH. All rights reserved.

232874; fax: +39 577 232896.t (S. Maccherini).

ARTICLE IN PRESS

Assessing grassland restoration success 333

departure from the unmanaged control, but also convergence with the referencecommunity. The PRC method provided a comprehensive overview of the response ofthe 153 species involved in the study.

Our results also indicate that the second PRC should be considered, whensignificant, to not exclude important information from the assessment.& 2006 Gesellschaft fur Okologie. Published by Elsevier GmbH. All rights reserved.

ZusammenfassungDie Erfolgskontrolle von okologischen Restaurationsmaßnahmen leidet haufig anungenugendem Design und mangelnder statistischer Genauigkeit, denn seltenwerden die raumliche und zeitliche Habitatvariabilitat berucksichtigt. Wir pruftendie Effektivitat der Entbuschung fur die Restauration von Kalkgrasland in einemvierjahrigen ‘Before-After/Reference-Control-Impact (BARCI) Untersuchungsdesignmit replizierten Probestellen. Abweichung von den Kontrollflachen und Annaherungan die Referenzflachen wurden mit Hauptreaktionskurven (PRC) analysiert.

Die Struktur der Gemeinschaften wurde ein Jahr vor und drei Jahre nach derEntbuschung untersucht: die Probestellen lagen auf Gebuschflachen (Kontrolle),restaurierten ehemaligen Ackerflachen, in restauriertem Grasland und im Referenz-Grasland.

Die Entbuschung genugte nicht, um die Gemeinschaft von halbnaturlichemKalkgrasland wiederherzustellen. Die restaurierten ehemaligen Ackerflachen be-hielten eine Gemeinschaftsstruktur, die der der Gebuschflachen außerst ahnlich war.Die restaurierten Graslander wichen dagegen von den Kontrollen ab und nahertensich leicht, wenn auch nicht signifikant, dem Referenz-Grasland an. Die dominanteBuschart, die entfernt wurde, Prunus spinosa, erreichte in jeder Behandlung hohereBedeckungsgrade als im Referenz-Grasland, sogar im ersten Jahr nach derEntbuschung. Arten der trockenen und/oder Kalkgraslander waren auf denReferenzflachen zahlreicher als auf den restaurierten Graslandern, wahrend hierAcker- und Ruderalarten haufiger vorkamen.

Diese Untersuchung zeigt, daß der BARCI-Ansatz ein effektives Mittel fur dieEvaluation von Restaurationsmaßnahmen ist, da nicht nur die Abweichung von derunbehandelten Kontrolle, sondern auch die Annaherung an die Referenzge-meinschaft ermittelt werden kann. Die PRC-Methode bietet einen umfassendenUberblick uber die Reaktionen der 153 Arten, die mit dieser Untersuchung erfaßtwurden. Unsere Ergebnisse zeigen, daß auch die zweite PRC, soweit signifikant,ausgewertet werden sollte, um nicht fur die Bewertung wichtige Informationenaußer acht zu lassen.& 2006 Gesellschaft fur Okologie. Published by Elsevier GmbH. All rights reserved.

Introduction

Semi-natural grassland communities are of greatinterest for their landscape value, abundance ofrare and threatened species and high speciesrichness (Willems, 2001). However, semi-naturalgrasslands have strongly decreased due to changesin land use throughout Europe (Poschlod, Bakker, &Kahmen, 2005; Stoate et al., 2001). Abandonmentor insufficient management of pastures and mea-dows are the main factors that trigger thesecondary succession of semi-natural grasslandsinto increasingly dense scrublands, characterisedby lower species richness and fewer rare species(Barbaro, Dutoit, & Cozic, 2001), contributing to anoverall homogenisation of the landscape (De-bussche, Lepart, & Dervieux, 1999), and leading

to a loss of habitat and species diversity (GallegoFernandez, Garcıa Mora, & Garcıa Novo, 2004).

From 1992 to 2005, 45 projects on semi-naturaldry grasslands (Festuco-Brometalia) have beenfinanced by the European Union (LIFE ProjectsDatabase, 2005). Moreover, there is plenty ofinformation about grassland restoration projects(Muller, Dutoit, Alard, & Grevilliot, 1998; Ormerod,2003; Ruiz-Jaen, & Aide, 2005); however, we oftencannot correctly measure the ecological outcomesof those projects. Many authors have recentlyemphasised the need for a powerful samplingdesign and statistical analysis to evaluate restora-tion experiments, considering spatial-temporalvariability of habitat (Block, Franklin, Ward, Ganey,& White, 2001; Chapman, 1999; Chapman & Under-wood, 2000; Michener, 1997; Palmer et al., 2005;

ARTICLE IN PRESS

S. Maccherini et al.334

White & Walker, 1997; Young, Petersen, & Clary,2005). The most powerful approach, Before–After /Reference-Control-Impact design (BARCI, Lake,2001), includes references and control sites inorder to unambiguously assess the success of arestoration programme. Using this design, datafrom the restored sites and reference sites (areaswhich have the desired end conditions) should becompared to control locations (degraded areas thatare not being restored) (Chapman & Underwood,2000).

Several studies have incorporated restored,reference and control sites to evaluate the restora-tion of semi-natural grasslands (Pywell et al., 2002;Pykala, 2005) but no study, to our knowledge,applied a complete Before–After sampling design(BARCI).

Principal Response Curves (PRC) method is amultivariate repeated measurement techniquethat is able to show differences between timeseries in a single diagram. This method is widelyapplied in aquatic ecotoxicology and environmentalsciences (see Frampton, Van den Brink, & Gould,2000; Heegaard & Vandvik, 2004) and is consideredas an analysis that has a potential for widerapplication in experimental community ecology(Leps & Smilauer, 2003; Vandvik, 2004).

In this study we evaluated the success of shrubcutting on the restoration of semi-natural calcar-eous grasslands as an example of a highly spatialheterogeneous and species rich communities, usinga 4-year BARCI design. Departure from the controlsites and convergence with the reference sites interms of plant community composition were ana-lysed by the PRC method. The null hypothesis thatwas tested using Monte Carlo permutation tests wasthat the temporal trends in the species composition

Figure 1. Study area. Different treatments location are localanalysis, are represented in grey.

is independent of the treatment (i.e. Reference-Control-Impact).

Materials and methods

Study area

The research was conducted near the MonteLabbro summit (4214901300 N, 1113103300 E) aprevalently calcareous massif on the Uccellina-Monte Amiata ridge, Tuscany (Lazzarotto, 1993)(Fig. 1). The area is included in a proposed Site ofCommunity Interest (pSCI, ‘‘Monte Labbro andUpper Albegna Valley’’). Mean annual rainfall inthe area is 1002–1469mm and the average annualtemperature is 12.1–12.5 1C. Monte Labbro summithas been grazed for centuries, mostly by sheep, andhuman traces in the area date back to the BronzeAge (Pistoi, 1989). After the1960s the grazingpressure decreased and arable land abandoned.

The study area, ranging in altitude from 1089 to1193m, is characterised by a mosaic of communitytypes: semi-natural calcareous grasslands with lowor no presence of shrubs; overgrown grasslands(Prunus spinosa scrublands); dense scrublands onsmall abandoned ex-arable areas located on karsticdoline where stones have been removed in thepast, with Prunus spinosa, Rubus ulmifolius andCytisus scoparius and small woodlands with Acercampestre and A. monspessulanum. The area ismainly grazed by sheep (stocking rate estimated as2/3 sheep ha�1 yr�1; Comunita Montana, unpub-lished data).

The restoration project, financed by a LIFE-NatureProject, started in the area in 1999 and has beenmanaged by a single authority (Comunita Montana

ised by different symbols. Woodlands, not included in the

ARTICLE IN PRESS

Assessing grassland restoration success 335

del Monte Amiata Zona ‘‘I1’’, Area Grossetana) sincethen. The project included the cutting of shrubs onovergrown grasslands and dense scrublands onabandoned ex-arable areas.

Sampling design and data collection

To test the effects of shrub cutting on thecommunity structure by means of species composi-tion of overgrown grassland we conducted a 4-yearBARCI experiment with replicated patches ofhabitat to take into account natural spatialvariability (Chapman, 1999; Dutilleul, 1993; John-son, 2002; Ruiz-Jaen &Aide, 2005).

In late June 2000, we allocated 40 1� 1-m squareplots by randomly generating spatial coordinates inthe study area to gather the field data before theintervention. According to the LIFE project design,in the period comprising late summer of 2000 toearly spring 2001, the labourers manually cut allshrubs (except all individuals of Juniperus commu-nis and some of Prunus spinosa, Rosa canina andCrataegus monogyna, for faunistic conservationpurposes) in three out of five patches of overgrowngrassland and in the two patches of scrubland onex-arable areas, for a total area of about 8 ha.Shrub clearing was not repeated.

Plots were divided into four treatments, namely:control scrubland, restored ex-arable, restoredgrassland and reference grassland (Fig. 1 andTable 1). The number of plots are proportional tothe dimension of patches (Elzinga, Salzer, Willough-by, & Gibbs, 2001). We decided to differentiate twomanagement treatments (restored ex-arable andrestored grassland) because we expected a differ-ent response following the shrubs cutting, due todifferences in topographic features and land usehistory. Two patches of overgrown grasslandsdominated by Prunus spinosa were taken as a‘‘control’’ to assess the changes regarding commu-nity structure of the restored ex-arable andrestored grassland, as they represented ‘‘sitessimilar to the site being restored prior to restora-tion’’ Chapman (1999). As reference, we adopted

Table 1. The total number of 40 plots is divided into four trtwo patches

Treatments

Control scrubland: dense scrubland where shrubs were not cRestored ex-arable: scrubland on ex-arable sites where shrubRestored grassland: shrubby grassland where shrubs were cuReference grassland: target grassland with scattered shrubs

Plot amount always allows to perform statistical analysis with a po

the interspersed patches of semi-natural calcar-eous grasslands recognised in the study area, toavoid the bias of changing spatial context andenvironmental distance from the site to be restored(White & Walker, 1997).

Floristic measurements were made annually inthe 40 plots in late June—early July from 2000,before the treatment, to 2003. All vascular plantsgrowing in each plot were recorded and their coverestimated using a point-quadrant method (Moore &Chapman, 1986) with a density of 100 pins/m2.Species present in a plot but not touched by anypins were recorded with an arbitrary cover of 0.1%.The nomenclature of species follows Pignatti (1982)and Bechi (1998) for Cerastium arvense ssp.arvense var. etruscum.

Statistical analysis

Departure of the managed plots from control siteand convergence to reference site conditions wereanalysed by the PRC technique (van den Brink & terBraak, 1997, 1998, 1999).This technique, derivedfrom the redundancy analysis (RDA) constrainedordination method, plots the principal componentsof the effects of treatment, expressed as deviationsfrom the control treatment, against time (van denBrink & ter Braak, 1998).

PRC consists of the canonical coefficients of apartial redundancy analysis, in which the input dataset are the sample-by-species matrix of log-abundance values, the sample-by-year matrix ofcovariables, and sample-by-treatment in yearmatrix of explanatory variables. The canonicalcoefficients of partial RDA (Cdt) quantify thecompositional differences between the controland other treatments (d) at each sampling date(t): temporal trends can be visualised by plottingCdt against time. The species weight bk indicatesthe affinity that each individual taxon has with theoverall community response as displayed in the PRCdiagram: species with high positive values followthe overall community response as indicated by the

eatments: each treatment has been replicated at least in

# Patch # Plot

ut 2 5s were cut 2 6t 3 13where shrubs were not cut 5 16

0:05 significance.

ARTICLE IN PRESS

S. Maccherini et al.336

PRC, while for negative values the communityresponse is the opposite.

After extraction of the first PRC, which shows thedominant response present in the data set, asecond, third, fourth, etc. PRC can be extracted(van den Brink & ter Braak, 1998). The second PRCshows the main deviations from the first PRC andmay be important if not only a single dominantresponse but also several sub-dominant responsesare present (van den Brink, van den Brink, & terBraak, 2003). Adding the second PRC, if significant,to the interpretation of data set can enhancevisualisation of the responses (van den Brink & terBraak, 1998). The change in log-abundance ascalculated per PRC must be summed across compo-nents to yield the joint fit of the change. The fittedchange can be plotted against time to yield fittedresponse curves for individual species or speciesgroups. For instance, if a species has an equalweight with both PRC, they can be summed to yieldthe fitted change for that species. If species weightsare different, the PRC should be weighted by thesespecies weights to obtain the fitted response.

Before multivariate analysis, the abundance ofcommunities was ln(20x+1)-transformed, to down-weight the influence of dominant species (van denBrink, Hattink, Bransen, van Donk, & Brock, 2000).

The significance of the PRC diagrams was testedby Monte Carlo permutation tests (999 permuta-tions), permuting whole time series in the partialRDA from which the PRC were obtained. MonteCarlo permutation tests were also performed for allsampling data (199 permutations), by testing eachtreatment against all other treatments. Statisticalanalysis was performed using the CANOCO 4 soft-ware package (ter Braak & Smilauer, 1998).

Results

General results

One hundred and fifty-three vascular specieswere found in the 4 years of observations. Mean

Table 2. Species richness and year of observation (mean7

Speciesm�2 (mean7SE)

Year 2000

Control scrubland 971.3Restored ex-arable 1471.5Restored grassland 2570.4Reference grassland 2570.3

species number in control sites was low whilerestored grassland, with a maximum of37 speciesm�2 recorded in one plot in 2002,showed higher values (Table 2). In 2000, the pre-treatment year, reference grassland was dominatedby Bromus erectus while the other treatments weredominated by Prunus spinosa, though with differentpercentages of cover (control scrubland 44–100%;restored ex-arable 28–100%; restored grassland3–100%).

Community changes

Differences between the treatments in timeaccounted for 21.9% of the between-plot variationin species composition, which was highly significant(Table 3). The two restored treatments showed thehighest variation in composition and cover (Fig. 2and 3). Fig. 2 shows the first Principal ResponseCurve (PRC 1), indicating a similar communitycomposition for the control scrubland and therestored ex-arable land for the pre-treatment year.In the same year of observation, the controlscrubland was very different from the restoredgrassland and especially from the reference grass-land.

The restored ex-arable plots had strong similaritywith the control plots: in 2002 and 2003, however,they were significantly different from the controlðpo0:01Þ, while remaining distinct from the re-stored and reference grassland. In 2000, therestored grassland was clearly different from thereference grassland, towards which it evolved aftershrub cutting. The differences between restoredgrassland, control scrubland and reference re-mained significant throughout the observationperiod ðpo0:01Þ.

In the second PRC (Fig. 3) the curves of therestored ex-arable and reference grassland stayedfairly close to the line of the control scrublandduring the whole period, while the restored grass-land showed a clear separation from the controlscrubland from 2000 to 2003.

SE; speciesm�2) for each treatment

2001 2002 2003

971.3 1071.4 1071.11571.6 2271.5 2371.42670.5 3070.4 2470.32770.3 2570.3 2270.3

ARTICLE IN PRESS

Table 3. Percentages of the total variance that can be attributed to time and treatment regime for the data set; thetable indicates the fraction of variance captured by the first and second principal components response (PRC)

% Variance accounted for by % Variance explained by treatment regime captured by p(999 permutations)

Time Treatment First PRC Second PRC First PRC Second PRC

2.1 21.9 58.4 23.8 0.001 0.001

The treatment component includes the interaction between treatment and time. The remaining fraction of variance is residual.

Figure 2. First Principal Response Curves diagram, showing the effects over time of shrub cutting on restored sites.Species weight on the right (bk) can be read as the affinity of every species with the shown diagram.

Assessing grassland restoration success 337

Species response

Bupleurum baldense, Festuca brevipila, Petror-hagia prolifera and Poa annua, showing highspecies weight (bk) only for the PRC 1 (Fig. 2) andlow for PRC 2, strongly changed in cover values inrestored grasslands approaching the cover observedin the reference grassland, especially in 2002 and2003. Additionally, although Galium corrudifoliumand Thymus longicaulis showed an increase withrespect to 2000, their cover remained lower than inthe reference grassland. Rubus ulmifolius andGalium album with negative bk values, were moreabundant in the restored ex-arable areas than inthe control and other treatments.

In the second PRC (Fig. 3), Phleum pratense, Airaelegans, Anthoxanthum odoratum and Tordyliummaximum (high bk values, low values for PRC 1)showed resilience to shrub cutting in the restoredgrassland.

Fig. 4 shows species weights for PRC 1 and PRC 2.Species with the same weights in the two PRC,except for sign, showed a pattern similar to that ofthe PRC at the corners of Fig. 4. The response of aspecies indicated by the first two PRC diagrams canbe calculated by adding up the PRC using thespecies weights as the weighing factor. Thediagrams in the corner indicate the response ifthe weight in each diagram is the same, i.e. that lieon the lines that have a 451 angle with the first andsecond axes. For example, Prunus spinosa hadsimilar abundance values in the control andrestored plots in 2000, showing a decrease sub-sequent to shrub cutting (negative bk for the PRC 1and positive bk for PRC 2, upper left quadrant, Ps).In the following years, P. spinosa cover in restoredex-arable sites returned to values similar to thoseof control scrubland, while in restored grassland itremained distinct from both control and referencesites during the whole period.

ARTICLE IN PRESS

Figure 3. Second Principal Response Curves diagram, showing the effects over time of shrub cutting on restored sites.

S. Maccherini et al.338

Many species, such as Cynosurus echinatum,Crepis neglecta, Eryngium campestre and Orlayagrandiflora had lower cover in the control plotscompared to all treatments (see upper rightquadrant of Fig. 4). In particular, Bromus erectusand Trifolium incarnatum ssp. molinerii, corre-sponded to the species with higher cover in therestored grassland than the reference grassland.

In contrast Anthyllis vulneraria ssp. praepropera,Sedum album, Centaurea deusta, Phleum ambi-guum and Thymus longicaulis (see lower rightquadrant of Fig. 4) had higher cover in thereference than in the restored grassland, repre-senting the guiding species for the desired commu-nity (reference grassland).

Discussion

Community changes

Although more years of study are necessary todefinitively declare the outcome of the restorationactions, from the present result we can assert thatin the short-term shrub cutting and a low grazingpressure did not suffice to restore the communitycomposition of semi-natural calcareous grasslands.

As in similar experiments, the greatest changeswere recorded in the second year after manage-ment practices (Dzwonko & Loster, 1998). Therestored ex-arable areas maintained a similar

species composition to the control scrubland,although in the last 2 years of observation theydiffered significantly. Conversely, although theystill differed significantly, the restored grasslanddiverged from the control and slightly convergedto the reference grassland. Even the first yearafter cutting, the dominant scrub species, Prunusspinosa, did not reach cover values similar to thereference grassland in any of the managed plots.This may be related to the mode of reproduction ofPrunus spinosa which has vigorous vegetativesuckering (Snow & Snow, 1986) with resproutingafter cutting.

In the restored grassland, the annuals Bupleurumbaldense, Petrorhagia prolifera and Poa annua,together with Festuca brevipila showed a positiveresponse to the management, with cover valuescloser to the reference grassland.

Significantly, the most abundant species inreference grassland are typical of xeric and/orcalcareous grassland, such as Inula montana,Sanguisorba minor, Teucrium chamaedrys and Thy-mus longicaulis and the species of high conserva-tion value Phleum ambiguum and Centaurea deusta(Italian peninsula endemism). Arable and/or waste-land species Tordylium maximum, Orlaya grand-iflora and Plantago lanceolata were more frequentin the restored than in the reference grassland.

Gaps created by shrub cutting in the twomanaged communities were colonised mostly byTrifolium incarnatum ssp. molinerii (with anincrease of 17.5% and 7.9% in restored grassland

ARTICLE IN PRESS

0

0.5

1

1.5C

dt

0

0.5

1

1.5d

t

Restored ex-arableRestored grassland

Year

-1.5

-1

-0.5

0

2000 2001 2002 2003

Cd

t

Cd

t

2000 2001 2002 200320030

0.5

1

1.5

2000 2001 2002 20030

0.5

1

1.5

2000 2002 2003

Be

Cd

Ce

Fb

Ga

Og

Ppr

Ru

Ti

Tl

Ae

Aela

Ag

Am

Ama

Ao

As

Av

Bb

Br

Cc

Cl

Cn

Cne

Dh Ec

Gc

Gco

Im

Ls

PaPan

Pl

Pp

Ps

Sa

Sh

SmTc

Tm

Vs

Xc

PRC 2

PRC 1- PRC 1

- PRC 2

Year

2000 2001 2002 20032000 2002 2003

Year

Reference grasslandControl scrubland

Figure 4. Two-dimensional plot of the species weights on the first and second PRC. The three diagrams in the corners,apply to species that have similar weights on the two PRC (except for the sign). If species weights are different, the PRCshould be weighted by these species weights to obtain the fitted response. Species are labelled as follows: Ae: Airaelegans, Am: Achillea millefolium, Ama: Armeria majellensis ssp. ausonia, Ag: Aegilops geniculata, Ao: Anthoxanthumodoratum, As: Arenaria serpyllifolia, Aela: Arrhenatherum elatius, Av: Anthyllis vulneraria ssp. praepropera, Bb:Bupleurum baldense, Be: Bromus erectus, Br: Brachypodium rupestre, Cc: Cynosurus cristatus, Cd: Centaurea deusta,Ce: Cynosurus echinatum, Cl: Cerastium luridum, Cn: Carduus nutans, Cne: Crepis neglecta, Dh: Dactylis hispanica, Ec:Eryngium campestre, Fb: Festuca brevipila, Ga: Galium album, Gc: Galium corrudifolium, Gco: Geranium columbinum,Im: Inula montana, Ls: Lathyrus sphaericus, Og: Orlaya grandiflora, Pa: Phleum ambiguum, Pan: Poa annua, Pl:Plantago lanceolata, Pp: Phleum pratense, Ppr: Petrorhagia prolifera, Ps: Prunus spinosa, Ru: Rubus ulmifolius, Sa:Sedum album, Sh: Stachys heraclea, Sm: Sanguisorba minor, Tc: Teucrium chamaedrys, Ti: Trifolium incarnatum ssp.molinerii, Tl: Thymus longicaulis, Tm: Tordylium maximum, Vs: Vicia sativa, Xc: Xeranthemum cylindraceum.

Assessing grassland restoration success 339

and restored ex-arable respectively from 2000, theyear before management, to 2002) and Dactylishispanica which are also present in the seed bank ofopen grassland and prairies afforested with conifers(Maccherini & De Dominicis, 2003). In ex-arableplots, the gaps were also colonised by Brachypodiumrupestre and in restored grassland by Bromus erectusas well. These species, are almost absent in thegerminable soil seed bank of these calcareousgrasslands (Maccherini & De Dominicis, 2003). Hemi-cryptophytes with regenerative strategy principally

by seeds (Grime, 1979) prevail while annual speciessuch as Linum bienne, Campanula rapunculus andVeronica agrestis, massively present in the soil seedbank of open and afforested grassland, were scarcelypresent or absent from restored patches.

Methodological considerations

This study has demonstrated that the BARCIapproach is a powerful tool to evaluate restoration

ARTICLE IN PRESS

S. Maccherini et al.340

efforts. Without control sites, it can be demon-strated that a managed site changed in thepredicted manner, but the transformation cannotbe unambiguously attributed to the restorativetreatment; further, without reference sites, theextent to which the restored sites follow atrajectory toward some specified target statecannot be assessed (Chapman, 1999). Resemblanceof control and restored sites in terms of communitycomposition can only be verified after statisticalcomparison since control sites should be ‘‘yassimilar as possible to the restored site’’ (Blocket al., 2001). Even if we tested that the controlscrubland was statistically different from therestored grassland before the intervention oc-curred, the control scrubland still remained a real‘‘control’’ to which the restored sites can becompared (Chapman, 1999) in time. In fact, theadoption of the BARCI approach allowed to effec-tively assess how and when the restored commu-nities, evaluated by means of species composition,followed in the years a trajectory from the controlscrubland toward the desired reference grasslandcondition.

The PRC method provided an overview of theresponse of the 153 species involved in thetreatment. Normally only the results of the firstPRC are presented: if only the first PRC had beenconsidered in this paper, the emphasis would havebeen put on the pre-treatment differences be-tween the control and restored grassland plots.When pooled with the second PRC these differencesstill hold, but the combination of the two providesa clear treatment axis from the upper left quadrantto the lower right quadrant (Fig. 4). Although this isone of the first ecological studies to include thesecond PRC, its results clearly indicate that, whensignificant, the second PRC should be considered,otherwise important information is left out of theassessment.

Implications for the conservation ofcalcareous grasslands

Different opinions have been expressed regardingthe restoration of grassland overgrown by trees orshrubs: Dzwonko & Loster (1998), for example,suggest that cutting alone is not sufficient tomaintain grassland communities. However, en-couraging results have been obtained in therestoration of overgrown grassland by clearingshrubs, mowing or grazing (Barbaro et al., 2001;Bobbink & Willems, 1993; Willems & Bik, 1998;Zobel, Suurkask, Rosen, Partel, 1996). On the otherhand, the restoration of formerly arable land seems

to be a slow and difficult process if target speciesare not sown, and if soil fertility and the establish-ment rate of competitive, early successionalspecies is not artificially reduced (Bissels, Holzel,Donath, Otte, 2004; Hutchings & Booth 1996; Kleijn2003; Walker et al. 2004).

Our study showed that after shrub clearing neitherthe overgrown grassland nor the scrubland on formerarable land reached a community compositionsimilar to the reference grasslands. The scarcity ofseeds of target species in the soil seed bank (Bisselset al. 2004; Maccherini & De Dominicis 2003)together with the documented high soil fertilitytypical of overgrown areas (Berendse 1998; Marrs1993) might be responsible for the negative resultobtained. Despite the presence of grazing livestock,the grazing pressure appeared too low to determinea consistent progressive shifting to the referencegrassland after shrubs cutting. Rigorously testing theefficacy of the adopted intervention is hence crucialto correctly address the scheduling and managementfor future intervention.

For these reasons, other factors not considered inthe present management project, such as increasedgrazing pressure, the introduction of target spe-cies, the reduction of nutrient status and pre-scribed burning used in the past to control shrubencroaching, should be considered for the restora-tion of these grasslands.

Acknowledgements

The authors thank Professor C.J.F. ter Braak forhis helpful suggestions, F. Casini, E. Santi and V.Zoni for their help with sampling, and the person-nel of the Parco Faunistico del Monte Labbro fortheir technical support, especially the Director, N.Cini. We acknowledge two referees for improving aprevious version of the manuscript. This study wasfunded by the European Commission LIFE-NatureProject NAT/IT/99/6229 ‘‘Monte Labbro and UpperAlbegna Valley, conservation and management’’,and by the Comunita Montana del Monte AmiataZona ‘‘I1’’, Area Grossetana.

References

Barbaro, L., Dutoit, T., & Cozic, P. (2001). A six-yearexperimental restoration of biodiversity by shrub-clearing and grazing in calcareous grasslands of FrenchPrealps. Biodiversity and Conservation, 10, 119–135.

Bechi, N. (1998). Ricerche citogeografiche in Cerastiumarvense L. s.l. (Caryophyllaceae) in Italia. Webbia, 53,1–12.

ARTICLE IN PRESS

Assessing grassland restoration success 341

Berendse, F. (1998). Effects of dominant plant species onsoils during succession in nutrient-poor ecosystems.Biogeochemistry, 42, 73–88.

Bissels, S., Holzel, N., Donath, T. W., & Otte, A. (2004).Evaluation of restoration success in alluvial grasslandsunder contrasting flooding regimes. Biological Con-servation, 118, 641–650.

Block, W. M., Franklin, A. B., Ward, J. P., Ganey, J. L., &White, G. C. (2001). Design and implementation ofmonitoring studies to evaluate the success of ecolo-gical restoration on wildlife. Restoration Ecology, 9,293–303.

Bobbink, R., & Willems, J. H. (1993). Restorationmanagement of abandoned chalk grassland in theNetherlands. Biodiversity and Conservation, 2,616–626.

Chapman, M. G. (1999). Improving sampling designs formeasuring restoration in aquatic habitats. Journal ofAquatic Ecosystem Stress and Recovery, 6, 235–251.

Chapman, M. G., & Underwood, A. J. (2000). The needfor a practical scientific protocol to measure success-ful restoration. Wetlands, 19, 28–49.

Debussche, M., Lepart, J., & Dervieux, A. (1999).Mediterranean landscape change: Evidence from oldpostcards. Global Ecology and Biogeography, 8, 3–15.

Dutilleul, P. (1993). Spatial heterogeneity and the designof ecological field experiments. Ecology, 74,1646–1658.

Dzwonko, Z., & Loster, S. (1998). Dynamics of speciesrichness and composition in a limestone grasslandrestored after tree cutting. Journal of VegetationScience, 9, 387–394.

Elzinga, C. L., Salzer, D. W., Willoughby, J. W., & Gibbs,J. P. (2001). Monitoring plant and animal population.Oxford: Blackwell Science.

Frampton, G. K., Van den Brink, P. J., & Gould, P. J. L.(2000). Effects of spring precipitation on a temperatearable collembolan community analysed using Princi-pal Response Curves. Applied Soil Ecology, 14,231–248.

Gallego Fernandez, J. B., Garcıa Mora, M. R., & GarcıaNovo, F. (2004). Vegetation dynamics of Mediterra-nean shrublands in former cultural landscape atGrazalema Mountains, South Spain. Plant Ecology,172, 83–94.

Grime, J. P. (1979). Plant strategies and vegetationprocesses. Chichester: Wiley.

Heegaard, E., & Vandvik, V. (2004). Climate changeaffects the outcome of competitive interactions—Anapplication of principal response curves. Oecologia,139, 459–466.

Hutchings, M. J., & Booth, K. D. (1996). Studies on thefeasibility of re-creating chalk grassland vegetation onex-arable land. I. The potential roles of seed bank andseed rain. Journal of Applied Ecology, 33, 1171–1181.

Johnson, D. H. (2002). The importance of replication inwildlife research. Journal of Wildlife Management,66, 919–932.

Kleijn, D. (2003). Can establishment characteristicsexplain the poor colonization success of late succes-

sional grassland species on ex-arable land? RestorationEcology, 11, 131–138.

Lake, P. S. (2001). On the maturing of restoration: Linkingecological research and restoration. Ecological Man-agement & Restoration, 2, 110–115.

Lazzarotto, A. (1993). Elementi di geografia e geomorfo-logia. In F. Giusti (Ed.), La Storia naturale dellaToscana meridionale (pp. 11–17). Cinisiello Balsamo(Milano): Monte dei Paschi di Siena.

Leps, J., & Smilauer, P. (2003). Multivariate analysis ofecological data using CANOCO. Cambridge: UniversityPress.

LIFE Projects Database, 2005. http://europa.eu.int/comm/environment/life/project/Projects/index.cfm

Maccherini, S., & De Dominicis, V. (2003). Germinable soilseed-bank of former grassland converted to coniferousplantation. Ecological Research, 18, 739–751.

Marrs, R. H. (1993). Soil fertility and nature conservationin Europe—Theoretical considerations and practicalmanagement solutions. Advances in Ecological Re-search, 24, 241–300.

Michener, M. K. (1997). Quantitatively evaluating re-storation experiments: Research design, statisticalanalysis, and data management considerations. Re-storation Ecology, 5, 324–337.

Moore, P. D., & Chapman, S. B. (1986). Methods in plantecology. Oxford: Blackwell Scientific Publications.

Muller, S., Dutoit, T., Alard, D., & Grevilliot, F. (1998).Restoration and rehabilitation of species-rich grass-land ecosystem in France: A review. RestorationEcology, 6, 94–101.

Ormerod, S. J. (Ed.). (2003). Restoration in appliedecology: Editor’s introduction. Journal of AppliedEcology, 40, 44–50.

Palmer, M. A., Bernhardt, E. S., Allen, J. D., Lake, P. S.,Alexander, G., Brooks, S., et al. (2005). Standards forecologically successful river restoration. Journal ofApplied Ecology, 42, 208–217.

Pignatti, S. (1982). Flora d’Italia. Bologna: Edagricole.Pistoi, M. (1989). Guida archeologica del Monte Amiata.

Siena: Nuova Immagine Editrice.Poschlod, P., Bakker, J. P., & Kahmen, S. (2005). Changing

land use and its impact on biodiversity. Basic andApplied Ecology, 6, 93–98.

Pykala, J. (2005). Plant species response to cattle grazingin mesic semi-natural grassland. Agriculture, Ecosys-tems and Environment, 108, 109–117.

Pywell, R. F., Bullock, J. M., Hopkins, A., Walker, K. J.,Sparks, T. H., Burke, M. J. W., et al. (2002).Restoration of species-rich grassland on arable land:Assessing the limiting processes using a multi-siteexperiment. Journal of Applied Ecology, 39, 249–309.

Ruiz-Jaen, M. C., & Aide, T. M. (2005). Restorationsuccess: How is it being measured? RestorationEcology, 13, 569–577.

Snow, D., & Snow, B. (1986). Some aspects of avianfrugivory in a north temperate area relevant totropical forest. In A. Estrada, & T. H. Fleming (Eds.),Frugivores and seed dispersal (pp. 45–54). Dordrecht:Junk.

ARTICLE IN PRESS

S. Maccherini et al.342

Stoate, C., Boatman, N. D., Borralho, R. J., Rio Carvalho,C., de Snno, G. R., & Eden, P. (2001). Ecologicalimpacts of arable intensification in Europe. Journal ofEnvironmental Management, 63, 337–365.

ter Braak, C. J. F., & Smilauer, P. (1998). CANOCOReference manual and user’s guide to Canoco forWindows: Software for canonical community ordina-tion (version 4). Ithaca, New York: MicrocomputerPower.

van den Brink, P. J., & ter Braak, C. J. F. (1997).Ordination of responses to toxic stress in experimentalecosystems. Toxicology and Ecotoxicology News, 4,174–178.

van den Brink, P. J., & ter Braak, C. J. F. (1998).Multivariate analysis of stress in experimental ecosys-tems by principal response curves and similarityanalysis. Aquatic Ecology, 32, 163–178.

van den Brink, P. J., & ter Braak, C. J. F. (1999). Principalresponse curves: Analysis of time-dependent multi-variate responses of biological community to stress.Environmental Toxicology and Chemistry, 18,138–148.

van den Brink, P. J., Hattink, J., Bransen, F., van Donk,E., & Brock, T. C. M. (2000). Impact of the fungicidecarbendazim in freshwater microcosms. II. Zooplank-ton, primary procedures and final conclusions. AquaticToxicology, 48, 251–264.

van den Brink, P. J., van den Brink, N. W., & ter Braak, C.J. F. (2003). Multivariate analysis of ecotoxicologicaldata using ordination: Demonstrations of utility on the

basis of various examples. Australasian Journal ofEcotoxicology, 9, 141–146.

Vandvik, V. (2004). Gap dynamics in perennial subalpinegrasslands: Trends and processes change duringsecondary succession. Journal of Ecology, 92, 86–96.

Walker, K. J., Stevens, P. A., Stevens, D. P., Mountford,O., Manchester, S. J., & Pywell, R. F. (2004). Therestoration and re-creation of species-rich lowlandgrassland on land formerly managed for intensiveagriculture in the UK. Biological Conservation, 119,1–18.

White, P. S., & Walker, J. L. (1997). Approximatingnature’s variation: Selecting and using referenceinformation in restoration ecology. Restoration Ecol-ogy, 5, 338–349.

Willems, J. H. (2001). Problems, approaches, and resultsin restoration of Dutch calcareous grasslands duringthe last 30 years. Restoration Ecology, 9, 147–154.

Willems, J. H., & Bik, L. P. M. (1998). Restoration of highspecies density in calcareous grassland: The role ofseed rain and soil seed bank. Applied VegetationScience, 1, 91–100.

Young, T. P., Petersen, D. A., & Clary, J. J. (2005). Theecology of restoration: Historical links, emergingissues and unexplored realms. Ecology Letters, 8,662–673.

Zobel, M., Suurkask, M., Rosen, E., & Partel, M. (1996).The dynamics of species richness in an experimentallyrestored calcareous grassland. Journal of VegetationScience, 7, 203–210.