Recent spatial development of Svalbard strandflat vegetation over a period of 31 years

  • Published on

  • View

  • Download


Recent spatial development of Svalbard strandat vegetationover a period of 31 yearspor_119 364..375Myrtille Moreau,1 Dominique Lafy2 & Thierry Brossard31 Laboratoire de Gographie Physique et Environnementale, 4 rue Ledru, FR-63057 Clermont-Ferrand, France2 Laboratoire Socit, Environnement, Territoire, Domaine Universitaire, FR-64000 Pau, France3 Laboratoire Thoriser et Modliser pour Amnager, 32 rue Mgevand, FR-25030 Besanon, FranceAbstractVegetation succession was analysed at the forefields of two glaciers in Svalbardover an interval of 31 years (19752006). In 1975, 85 sampling sites werepositioned along transects extending from the coastline to the glacier fronts:botanical observations were made at each sampling site. This protocol allowedus to carry out new observations in 2006 under the same conditions. Thus, itwas possible to undertake a botanical assessment of species and taxa, and to seehow the vegetation has changed with reference to a typology established bycoupling correspondence analysis and ascending hierarchical classification.Vegetation succession at the sampling sites was also measured by using vectorspositioned in the multidimensional space of correspondence analysis. In thisway, the changes over the interval between seven vegetation types wereplotted and mapped, and the colonization process was calibrated and datedagainst a series of reference stages, mainly since the end of the Little Ice Age.KeywordsCorrespondance analysis; plant dynamics;releves; Spitsbergen.CorrespondenceMyrtille Moreau, Laboratoire de GographiePhysique et Environnementale, 4 rue Ledru,FR-63057 Clermont-Ferrand, France. E-mail:myrtille.moreau@free.frdoi:10.1111/j.1751-8369.2009.00119.xSince the end of the Little Ice Age in Svalbard, dated thereto around 1880 by Lefauconnier (1990), glaciers havereceded, thereby opening up new areas for plant coloni-zation. By referring to aerial photographs that showdifferent stages of ice retreat, the different phases ofplant colonization of moraines can be dated. Kuc (1964)observed the vegetation dynamics on the Treskelenglacier foreland at Hornsund in the southern part of theisland of Spitsbergen. The foreland of the glacier MidreLovnbreen, on the Brggerhalvya peninsula in north-west Spitsbergen, has been the subject of several studies.Brossard (1985) described the species colonization ofsome of its terminal moraines. Hodkinson et al. (2003)observed plant colonization along a single transect acrossthe foreland (which did not cover the entire area).Moreau et al. (2005) and Moreau (2005) conducted avegetation survey of 300 plots to study the influence ofmicrotopography on vegetation cover. Elsewhere onBrggerhalvya, the foreland of the glacier AustreBrggerbreen has been the subject of several studies.Nakatsubo et al. (1998) observed the respiration of thebelow-ground parts of vascular plants. Kume et al. (2003)assessed the growth forms of Saxifraga oppositifolia in dif-ferent successional stages. Minami et al. (1997) sampledthree transects to analyse the relationship between plantspecies and microtopographic forms. At Isfjord, in thecentral part of Spitsbergen, Ziaja & Dubiel (1996) studiedthe colonization of the moraines of several mountainglaciers by vascular species.The direct monitoring of vegetation succession ontopographic features is less common. Klanderud & Birks(2003) compared vegetation sampling sites in the Jotun-heimen Mountains, Norway, in 193031 and 1998. Ourapproach is similar, and is based on field sessions in 1975and 2006. The study area is a complex strandflat, featur-ing moraine belts and outwash plains ahead of twoglaciers, as well as raised marine landforms. In 1975,Brossard (1977) positioned 85 sampling sites alongtransects crossing the main landscape units. Woodenposts were used to mark these sampling sites. The plotswere photographed and their plant species weresystematically counted. Many of these markers survived,making it possible to identify most of the sampling sites in2006, and to carry out a fresh survey based on the sameprotocol. The sampling sites were accurately positionedusing a differential GPS set. Accordingly, we have twohomogeneous series of vegetation sample data separatedby a time interval of 31 years. To our knowledge, suchdata are unique in Svalbard. The aims of the study are: (1)to examine the differences between the two series ofPolar Research 28 2009 364375 2009 The Authors364vegetation sample data; (2) to define the vegetationdynamics, in accordance with the different environmen-tal contexts; and (3) to provide a spatial assessment of thevegetation succession.Study areaThe study was conducted 6 km east of Ny-lesund, in thevicinity of the J. Corbel French research base. This area islocated between the Kongsfjorden fjord and the two frontsof the Midre Lovnbreen and Austre Lovnbreen glaciers.These glaciers reached theirmaximumadvancemore thana century ago, at the height of the Little Ice Age (Hagenet al. 1993). At that time, the ice tongues extended wellover the piedmont. In the course of their retreat, whichis still continuing, the glaciers have left several morainebelts that are major features of the present-day landscape(Lefauconnier 1990). Ahead of the moraines, a doublelimestone outcrop creates anchoring lines for severalraised glaciomarine terraces (Forman 1990; Fig. 1), whichare related to the post-Weichselian isostatic uplift. Maturetundra of the Northern Arctic Tundra Type defined byElvebakk (1997) covers these glaciomarine terraces. Theoldest terrace dates to around 9000 B.P. (Corbel 1966).Streams of water from the upper glacial basins have builtan outwash plain of detrital cones (sandurs). As the iceretreated, the water flow became concentrated in just afew very dynamic and shifting water courses, whereas thesecondary channels and cones were abandoned and thencolonized by vegetation (Mercier 1998).Materials and methodThe observation protocol began in 1975, and was basedon field sampling along transects extending from thecoast to the glacier fronts (Fig. 1). The transects crossed allof the different landscape units in the area: marine ter-races, active and inactive cones, and moraines. A total of85 sampling sites were set up at intervals of roughly200 m. A 4-m2 area was delimited at each sampling site,providing a frame of reference for the observation ses-sions in 1975 and 2006. Other studies have shown that,as a rule, a 4-m2 plot is large enough to cover theminimum area occupied by plant communities, except forextreme pioneering contexts, for which the minimumarea is greater (Moreau 2005).A portable 1-m2 quadrat, subdivided into a 10 10 cmgrid, was used to observe the plant species, and eachsampling site was photographed. It was possible, then, tolocate the exact position of each sampling site with theminimum of error. During the 2006 session, 76 of theinitial 85 sampling sites were identified. The missing plotshad been located on active sandurs, and had thereforebeen washed away during the intervening period. Forthe observation protocol itself, our main concern was toavoid any artefact that might distort the information col-lected on the two dates. Species abundance was thereforeomitted from the analysis, even though it was recorded inboth surveys. Only presence/absence was considered,because species abundance estimates are somewhat sub-jective, and are unlikely to be stable over a 31-yearFig. 1 Study area with transect positions andstages of the Midre Lovnbreen glacier retreatsince the end of the Little Ice Age. The imagebackground is a mosaic of two aerial photo-graphs (provided by the the Norwegian PolarInstitute) from 1966 and 1990. The photo-graphs were digitized at 2-m resolution. Themain part of the area was covered by an infra-red photograph from 1990 (Norwegian PolarInstitute reference number S 90 5788). Wehave used a section dating from 1966 (Norwe-gian Polar Institute reference number 66 4455)to complete the coverage of the area. The twoimages were simply assembled after geomet-ric correction, as they are used here merely toillustrate the sampling sites in their geographi-cal context.Vegetation evolution (19752006) in SvalbardM. Moreau et al.Polar Research 28 2009 364375 2009 The Authors 365interval. Hence, the survey data took into account thecomplete species list established for this sector by Bros-sard et al. (1984). In addition to the vegetation cover, wealso noted the presence of any biotic crust. This takesthe form of a black or grey pellicle, which is a combina-tion of cyanobacteria, cyanophytes, cyanophylous lichens(Ochrolechia frigida, Lecanora epibryon and Lecophysma fin-markicum), and some small liverworts belonging to theCephalozia and Cephaloziella genera (Nilsen et al. 1999). Bygrouping certain taxa in this manner, information is lost,but observations are made fully comparable both in space(from one sampling site to another) and in time (fromone date to another). This constraint takes precedencein the context of a study designed to provide a spatialassessment of the vegetation succession. Figure 2 lists thevegetation characteristics that were processed for a com-parative analysis between the years 1975 and 2006.The data collected were arranged in two tables, withthe rows corresponding to the georeferenced samplingsites, and the columns corresponding to the vegetationitems observed and selected. The information wasanalysed in several steps. First, a tally was made of thecharacteristics (species or others) that had appeared ordisappeared. This yielded an initial indication of thegeneral pattern of change in the study area. The datawere then analysed quantitatively using STATISTICA 7. Bycoupling the correspondence analysis and ascendinghierarchical classification (Benezecri 1992; Le Roux &Rouanet 2004), all of the elements (vegetation items, andsampling sites for 1975 and 2006) were positioned in afactorial space. This plot was used firstly as a system ofreference for defining the trajectories, corresponding tovectors, followed by the sampling sites over time, andthen to see how these trajectories corresponded to tran-sitions between the various vegetation types, defined byascending hierarchical classification. The geographicallocation of the sampling sites meant that the results couldbe mapped in such a way as to show whether the move-ments observed in the factorial space corresponded to thespatial logic of vegetation dynamics across the differentenvironmental contexts. Before analysing the results,some practical questions about how the method wasimplemented require clarification.The data were collated in a binary table, and werecoded as either present (1) or absent (0). For species thatwere present at more than 50% of the sampling sites, acolumn was added in which their absence was coded as(1) and their presence was coded as (0). This was becausethe absence of a species that is found at a majority oflocations is a discriminating feature worth highlighting.In this way, the absence of Salix polaris, Saxifraga oppositi-folia, Sterocaulon alpinum and mosses was noted. Bycontrast, some species were excluded from processingwhen they were inventoried only once or twice, as wasFig. 2 Frequency of species occurrences at all 85 sampling sites in 1975 and 2006.Vegetation evolution (19752006) in Svalbard M. Moreau et al.Polar Research 28 2009 364375 2009 The Authors366the case with Saxifraga aizodes, Cassiope tetragona, Papaverdahlianum and Saxifraga nivalis.The method used is based on correspondence analysis,which makes it possible to process tables of binary data,and to define the resulting space structured by the axes.Another useful feature of this method is that the spacesfor variables (vegetation species) and for individuals(sampling sites) are merged into the same ordinationspace. This property is known as distributional equiva-lence (Escofier 1978). The relative positions of individualsand variables are therefore helpful in giving a thematicmeaning to the different factorial axes, as it is easier tomake a connection between the logical structure arisingfrom the analysis (relative position of the species in thefactorial space) and the geographical structure of thespace (relative position of the sampling sites on the map).Three axes of the correspondence analysis were sufficientto explain more than 48% of the inertia: this is a highpercentage, as binary coding, which generates a largenumber of empty cells in the data matrix, leads to a dropin the levels of inertia. It must be noted here that thefactorial space was calculated first using the 1975 dataalone. The sampling sites for 2006 were then projectedinto this space by the supplementary elements tech-nique that is possible in correspondence analysis(Greenacre 1984). As a result of successional change(i.e., the appearance or disappearance of species), eachsampling site occupies a different factorialspace positionin 2006 compared with 1975, which can be expressedgraphically by a vector. This application made it possibleto show successional trends, and to identify explanatoryfactors related to the environment. We coupled factoranalysis with a method of ascending hierarchical classifi-cation (AHC, using Wards metric): seven vegetationtypes were defined. In the final step, vegetation types forthe two dates were mapped against a background madeup of vertical aerial photographs. Each vegetation typewas identified by a separate colour. This made it possibleto see which sampling sites had changed type, and wherethey were located. The results furnished by the map werethen interpreted to show the successional trends (or sta-bilities) observed.ResultsGeneral assessmentFigure 3 shows the frequency of occurrence of the taxa(species or type) from all of the sampling sites on the twodates. No new species appeared. Pedicularis dasyantha nolonger occurred in 2006. Taxa can be classified into threeFig. 3 Correspondence analysis: species loca-tions in factorial space and groupings resultingfrom the classication.Vegetation evolution (19752006) in SvalbardM. Moreau et al.Polar Research 28 2009 364375 2009 The Authors 367groups depending on whether their frequency of occur-rence increased, decreased or remained the same.Vegetation maturation. Certain vascular species thatare recognized as pioneers became less frequent. This istrue of Braya purpurascens, the frequency of which fellfrom 28 to 5%, and of Cerastium arcticum and of Saxifragacespitosa. By contrast, species that are ubiquitous or arecharacteristic of mature vegetation groups either main-tained (Dryas octopetala, Carex nardina and Carex rupestris)or increased their frequency (Salix polaris and Polygonumviviparum). The increased occurrence of bryophytes, suchas Drepanocladus sp. and Polytrichum sp., from 30 to 40%and from 18 to 25%, respectively, can be explained onthe same grounds, as these species are associated withmature tundra formations. Lichens of the genus Cetrariaincreased in frequency from 30 to 40%, and Sterocaulonalpinum increased in frequency from 60 to 80%.It is noteworthy that although Cetraria sp. occur at alarger number of sampling sites, the cover of these lichenshas fallen drastically on mature tundra, because of theovergrazing by reindeer that were reintroduced in 1978(Elvebakk 1997). Cetraria sp. exhibit increased dispersionbut decreased biomass. Despite the uncertainties aboutthe qualitative change in cover since 1975, the lossesobserved in the stable tundra environment are too size-able and too widespread to be dismissed as differencesin observational calibration between the two dates.However, apart from the effect on Cetraria sp., ourinvestigation detects no other impact of reindeer onvegetation. Indeed species that are particularly prone tograzing, such as Dryas octopetala, Silene acaulis (Wada1999), Saxifraga oppositifolia and Salix polaris (Rnning1996), do not appear to have been affected, at least interms of their frequency.The increased frequency and the cover of Stereocaulonalpinum in stabilized moraine and sandur habitats arevery significant: this species is an indicator of a transitionfrom pioneer to mature stages, with a ratio of mineral/vegetation cover that tends to be balanced.The drying of the environment. A decrease in thefrequency of species that are characteristic of wet habitats(Rnning 1996), such as Cerastium regelii (from 15 to 4%),Juncus biglumis (from 22 to 12%), Cochlearia officinalisand Deschampsia alpina, was observed over the interval.Although they favour less water-logged habitats, Luzulasp. and Silene uralensis also declined in frequency. Con-versely, the vascular species peculiar to dry settings, suchas Polygonum viviparum and Minuartia rubella (Rnning1996; Aiken et al. 1999) increased in frequency. Thisobservation may be related to the depth of the seasonalthaw and soil drainage. Soils are well drained overall, assilt- and clay-sized particles make up only a small fractionof the substratum almost everywhere in the study area.The soils have a tendency to dry out earlier in thegrowing season, which is a constraint for most species.Many of the instances of decline in taxa frequency can beinterpreted as a consequence of drying.Changes from pioneer to mature vegetation typesTo make it readable, only the plant species names and theellipsoids grouping sampling sites of the same type areincluded in Fig. 2. Even though they have not beenexplicitly indicated here, the distributional equivalencedoes allow the individual positions of the sampling sitesto be inferred. As the environmental conditions charac-terizing each sampling site (water flow, age and edaphicconditions) are known from different sources (additionalfield observations, literature and maps), it is possible,when interpreting the graph, to determine how they arerelated to vegetation types, and to their changes. Thevegetation types are ranked from 1 to 7 along a paraboliccrescent. This geometric shape reveals a logical structurethat also organizes the plant species distribution. Thischaracteristic shape is a sign that the data obey anordered relationship, sometimes referred to as theGutman effect (Escoffier 1979). Such a structure indi-cates that the interpretations of axes 1 and 2 cannot bedistinguished, and that they obey the same logic. In ourcase, this structure can be interpreted as a temporal gra-dient extending from the initial mineral stages to maturestages, via the pioneer and intermediary stages thatpunctuate the colonization process. Axis 3, which is notshown on the graph, defines a secondary wetdry gradi-ent. It distinguishes species that favour wet habitats, suchas Equisetum sp., Deschampsia sp., Juncus biglumis andCochlearia officinalis, from species that prefer dry habitats,such as Cetraria nivalis, Carex nardina, Carex rupestris andPedicularis dasyantha.Community types associated with stages invegetation developmentOn the basis of distributional equivalence, each plantspecies is positioned at the barycentre of the differentsampling sites at which it is observed. In this way, certaintaxa are seen to be poorly dispersed, and strongly char-acterize a given type. For example, Cetraria nivalis, Carexnardina, Carex rupestris and Pedicularis dasyantha areobserved exclusively, or almost exclusively, at type-7 (drymature) sampling sites; Saxifraga tenuis, Deschampsia sp.and Equisetum sp. are characteristic of type-6 (mature)sampling sites. Other taxa, such as Dryas octopetala, LuzulaVegetation evolution (19752006) in Svalbard M. Moreau et al.Polar Research 28 2009 364375 2009 The Authors368sp. and Poa sp., are slightly more dispersed across thetypes, but they remain within a limited sector of thefactorial space. On the other hand, certain ubiquitoustaxa, situated in the middle of the graph, are dispersedover a large number of types, and are thus not verydiscriminating: Bryum sp., Saxifraga oppositifolia, bioticcrust and, to a lesser extent, Stereocaulon alpinum fall intothis category. By analysing the correlations between taxaand sampling site types in this way, it is possible to specifythe content and the significance of the different vegeta-tion types recognized. Because the ordered relationship,brought out by correspondence analysis, is defined as atemporal gradient, we have used the term stage here todefine the vegetation types. Moreover, as we also havecertain reference points from old aerial photographs(Fig. 1), or from radiocarbon dating (Corbel 1966), thedifferent stages can also be ascribed to particular agebrackets (Moreau 2005).Vegetation type 1: initial mineral stage. This refersto the initial stage of colonization, when the samplingsites, marked exclusively by characters of absence, aredevoid of vegetation, and are located on recentlydefrosted fringes or active transit areas subject to flu-vioglacial dynamics. Less often, vegetation type 1 is foundin stable areas, where succession has been impeded byparticularly limiting factors, such as a coarse-grainedsubstratum or by late snowmelt in some moraine depres-sions. Judging from our sampling sites, the response timefor the advent of the first species is 510 years.Vegetation type 2: first pioneer stage (1030years). The landscape remains very mineral, and char-acters of absence are still dominant at the sampling sites.However, several pioneer taxa are observed initiatingthe colonization process, such as mosses (Bryum sp.),biotic crust, Braya purpurascens, Sagina nivalis and Saxi-fraga oppositifolia. Cochlearia officinalis and Juncus biglumismay occur where a fine-grained substratum and humid-ity are combined. None of the species observed at thisstage are exclusive to it: they all occur in the followingstage too.Vegetation type 3: second pioneer stage (30100years). Eleven new species were counted, includingSaxifraga cespitosa, Draba sp., Cerastium arcticum, Cerastiumregelii, Poa sp., Minuartia rubella, Oxyria digyna and Salixpolaris. This group, which is more variable from onesampling site to another, occupies the bulk of the areasliberated since the end of the Little Ice Age: moraine beltsand stabilizing sandur strata. Plant cover values remainmodest, at less than 10%.Vegetation type 4: first intermediate stage (100150 years). At this stage the number of plant speciesstabilizes, although Stereocaulon alpinum appears for thefirst time in this stage, and is immediately very frequent.Braya purpurascens is still present, although it is rarer,thereby displaying its pioneer credentials. Silene uralensisis also observed. This phase of stabilization is expressedby cover values of 2030%, and, at a large majority ofsampling sites, by the presence of mosses, biotic crust,Saxifraga oppositifolia, Cerastium arcticum, Draba sp., Saginanivalis and Salix polaris. This type occupies the spaces thatwere established at around the maximum of the Little IceAge: i.e., terminal moraine ridges and defunct proglacialoutwash cones (Mercier 1998). Secondly, Cochlearia offi-cinalis, Poa sp., Juncus biglumis and Cerastium regelii occurat sampling sites in the wetter settings of intramorainedepressions, or in sedimentation areas on sandurs.Vegetation type 5: second intermediate stage. Thisstage characterizes the long transition leading to maturetundra vegetation. Distinctive species appear, such asPedicularis hirsuta, Polygonum viviparum, Cetraria sp., Poly-tricum alpinum, Luzula sp. and Drepanocladus sp., andSterocaulon alpinum is typically abundant. Plant coverattains 50%, giving the landscape its dominant phys-ionomy. The time bracket in which these environmentsevolved is unknown, because precise temporal indicatorsare lacking. These sectors are older than the Little Ice Age,but are younger than the raised glaciomarine terraces,three samples of which were radiocarbon dated, andyielded inclusive dates of between 9260 and 9650 B.P.(Corbel 1966). These local measurements are consistentwith the results of other work carried out in this part ofSpitsbergen (Salvigsen 1977; Forman & Miller 1984).Vegetation type 6: mature stage (mesic andwet). Vegetation types 6 and 7 dominate the glacioma-rine terraces: these may be raised marine landforms orold, long-stabilized sandurs. We have no indicators todate the time boundary between the intermediate andmature stages. Vegetation type 6 includes the species thatare typical of the mature tundra already observed invegetation type 5, but they tend to be found at all of thesampling sites, which display greater floristic richness.The fruticose lichens Cetraria delisei and Cetraria islandica,the abundant large thalli of which gave this tundra itscharacteristic appearance in 1975, have been largelyeradicated by the recent reintroduction of reindeer. Thisvegetation group has a preference for well-sorted sedi-ments (sands and granules), and has brought about adegree of edaphic development (brown soils). A wetsubtype, which the classification did not distinguish, isVegetation evolution (19752006) in SvalbardM. Moreau et al.Polar Research 28 2009 364375 2009 The Authors 369marked by the presence of Equisetum sp., Juncus biglumisand Deschampsia alpina. Where the soil is waterlogged andits grain-size distribution is heterogeneous, metric-sizedstone rings form.Vegetation type 7: mature dry stage. Vegetationtype 7 is distinguished from type 6 by its environmentalconditions, rather than by any age criterion. The lime-stone bedrock outcrops from place to place on the marineterraces, and is subjected to frost action, thereby produc-ing a gelifraction layer, where patterns such as stripes andsmall stone rings develop. The vegetation cover rate dropsand the number of species is reduced: Dryas octopetala,Carex misandra, Carex nardina, Carex rupestris, Pedicularisdasyantha and Cetraria nivalis are characteristic of thisgroup. This type has two subtypes, corresponding to themicrohabitats that our sampling procedure did notdifferentiate between. Sites facing south-west that aresheltered from the wind are characterized by Cassiopetetragona, Lycopodium selago, Tomentypnum nitens and Ptil-idium ciliare. Exposed sites situated along the limestonebars, and used as look-out posts by certain predatorybirds, are distinguished by the presence of Xanthoriaelegans, Hypnum revolutum, Rhacomitrium lanuginosum andAlectoria ochroleuca.Vegetation change between 1975 and 2006Figure 4 shows how the relative percentage of each of theseven vegetation types changed in the 31-year interval.Paradoxically, the initial mineral stage (vegetation type 1)occurred at a slightly increased number of sampling sites,because although some sampling sites evolved beyondthis type by integrating new species, others regressed andreturned to this initial stage as a result of hydrolic erosionremoving the vegetation. The number of sampling siteswith pioneer vegetation types 2 and 3 declined as thecover there matured into intermediate vegetation types4 and 5, which became more common. The number ofsampling sites with mature vegetation types 6 and 7remained largely stable, even though the slightly highernumber of locations of the dry mature type tends toconfirm the drying effect that we observed in the assess-ment by species.Changes between vegetation types. In 1975 and2006, 56% of the sampling sites were unchanged underour coding system, with these being mostly the maturestages. By projecting the sampling sites for 2006 in thefactorial space, the changes observed can be expressedby vectors. The original position of these vectors ismarked by black dots in Fig. 5. Figure 5a shows onlyintertype movements, so as to avoid overcrowding thegraph. Intratype movements are shown in Fig. 5b. Typechanges were recorded for 31 of the 85 sampling sites.Most changes were between neighbouring vegetationtypes, and were part of a general pattern of maturation.The number of regressions was logically higher for thepioneer stages, and lower for the more mature stages.However, progressions were very common in themedian section of the graph, with a particularly strongtransition between vegetation types 3 (second pioneerstage) and 4 (first intermediary stage). Some samplingsites that were at the mineral stage in 1975 hadmatured to the second pioneer stage by 2006, whichconforms to the timescale established elsewhere(Moreau 2005) to calibrate the process of vegetationcolonization. Certain major regressive phenomena weremarked by a return from the mature stages to themineral stage: here, we found the effect of vegetationdestruction by proglacial waters eroding the older levelsof the tundra. As the glacial front recedes, the successivemoraine ridges dam the different water streams, whichare diverted and flow into a few main courses. Thecones through which all of the water is discharged fromthe moraines are then undersized, and tend to spreadoutwards, encroaching on older stable levels: the sam-pling sites that markedly regressed were victims of thisFig. 4 Number of sampling sites correspond-ing to each vegetation type in 1975 and 2006.Vegetation evolution (19752006) in Svalbard M. Moreau et al.Polar Research 28 2009 364375 2009 The Authors370scenario. Two other more sustained regressive shiftswere observed between vegetation types 6 and 5, result-ing from a decrease in the number of species observedat the corresponding sampling sites. In all likelihood thisis further evidence of drying, because the species thatvanished are associated with wet environments.Changes within vegetation types. These changes areconcentrated on the right-hand section of Fig. 5b, andaffect vegetation types 37, indicating that readjustmentsexist in most settings. Here again, the direction of thevectors conforms to the ordered relationship: regressivemovements are rare, and analysis of the correspondingsampling sites shows that they are linked to the inexpli-cable erratic disappearance of a species.Mapping the vegetation succession. A spatial visu-alization (Fig. 6) was produced by creating a mosaic ofaerial photographs: two symbols have been added foreach plot, the ones on the left apply to 1975 and the oneson the right to apply to 2006. The colours refer to theseven vegetation types identified by the classification. Inaddition to the successions observed, a logical distributionof vegetation types appeared. In 1975, the mineral stageswere found at the glacier fronts, or on active sandurs thatwere flooded each year. To a large extent, the pioneerstages occupied the core of the moraine amphitheatresand the sandurs undergoing stabilization. The intermedi-ate stages, which were less frequent, were concentratedon the terminal moraine arcs or the stabilized sandurs.Vegetation type 6 occupied the older sections of thesandurs and the marine accumulation terraces, whereasvegetation type 7 was strung along the linear limestoneoutcrop, easily seen on the photograph, crossing the areafrom the north-west to the south-east. Comparisonbetween 1975 and 2006 data shows that few samplingsites with vegetation of types 6 and 7 underwent change.However, the drying phenomenon was in evidence: thedisappearance of hydrophilous species from vegetationtype 6 brought about a regression to type 5, whereas theappearance of xerophilous species in vegetation type 6Fig. 5 (a) Changes between vegetation types.The graph corresponds to the correspondenceanalysis with the sampling sites that changedvegetation type between 1975 and 2006. (b)Changes within vegetation types. The graphcorresponds to the correspondence analysiswith the sampling sites that did not changevegetation type between 1975 and 2006. Inboth plots, the black dots show the positionsof the sampling sites in factorial space in 1975,and the lines show the paths of change from1975 to 2006.Vegetation evolution (19752006) in SvalbardM. Moreau et al.Polar Research 28 2009 364375 2009 The Authors 371marked a transition towards type 7. The large number ofsampling sites that remained at the mineral stage isanother stable feature in the overall pattern. Thisobservation is mainly explained by proglacial activitypreventing colonization from getting underway. Thevariation in water level in the lagoon in the north-west isanother explanatory factor. As for the rest, it appears thatthe most intensive succession was concentrated in themoraines, with a marked progression from vegetationtype 3 to type 4. Succession of the same type was alsoobserved on the external sandurs, but was caused bymuch more localized phenomena when certain channelswere no longer functional. Instances of regression are alsoobserved in this context, as channels that have dried upand are undergoing colonization may be flooded whenthere is a rise in the water level, or during a spring thaw.The advanced phases of maturation from vegetation type4 to type 5 and beyond are less frequently observed. It canbe surmised that when the process of vegetation coloni-zation reaches these phases of development, it occurs ona longer timescale than our 31-year observation intervalwas able to capture.DiscussionRegular sampling along transects is a commonlyemployed method for investigating spatial variations ofplant species and vegetation types. Repeating this type ofprotocol after a 31-year interval meant making adjust-ments to ensure the two series of observations werecomparable. The precautions taken at the time of theinitial surveyground marking, and the precise photog-raphy of the sampling sites and their settingsmade iteasier to identify the sampling sites and accurately repo-sition the sampling quadrant. The sampling site locationshave now been recorded in a GPS system for future work.A further consideration was the level of precision ofthe information collected. This had to remain consistentin space, from one sampling site to another, and in time,from one survey date to another. Accordingly, somespecies that were too rare were either omitted or weregrouped with others. The vegetation types we have iden-tified do not correspond to plant communities in thephytosociological sense of the term, but to biogeographi-cal entities that we use as markers to calibrate changesin vegetation. Moreover, by coupling correspondenceanalysis with ascending hierarchical classification, wehave been able to measure these changes using vectors(factorial distance), and to represent them in two ways:first, by plotting them on a graph to reveal their logicalstructure, which is essentially time-related; and second,by mapping them to reveal their spatial pattern. However,the scope of some observations is limited because speciesabundance evaluations are subjective and had to beomitted. For instance, the quantitative impact of over-grazing by reindeer could not be measured, because thepresence/absence criterion fails to provide an accuratemeasure of the substantial decline in the cover of speciessuch as Cetraria delisei or Cetraria islandica. On the scale ofthe study zone as a whole, the species that were missingfrom the second survey were mostly species characteristicof wet habitats. This was interpreted as a sign that soilshave dried out because the permafrost level has fallen,although this is merely a hypothesis. A measurement ofFig. 6 Map of vegetation succession stagesfor 1975 (squares on the left) and 2006(squares on the right).Vegetation evolution (19752006) in Svalbard M. Moreau et al.Polar Research 28 2009 364375 2009 The Authors372species abundance would have enabled a better diagnosisof the phenomenon. The observation protocol needs to beimproved to overcome this difficulty. The solution envis-aged consists of measuring abundance from frequencies.To this end, the same 1-m2 quadrat divided into10 10 cm squares can be used. Species are recorded interms of their presence or absence in each square. Thenumber of squares where a species is observed gives itsabundance. This mode of observation is more reliablebecause it does away with the subjective character ofthe evaluation, but it is more time-consuming. This con-straint can be limited by observing just 16 squares takenat random from the 100 squares of the quadrat. Testshave shown that abundance differences between theexhaustive method and the sampling method are insig-nificant (Moreau 2005). This, then, is a potential solutionfor ensuring that abundance measurements remain com-parable over long intervals.Despite these limitations, the method used was preciseenough to reveal the major changes that have affectedthe vegetation in the study area as a whole. The essentialcontribution was to show how, over a given interval oftime, the sampling sites have shifted within a typologicalreference space. Many studies of vegetation dynamicsconcentrate on stabilized areas where dating features canbe found, say from photographs in which the successivestages of glacial retreat can be identified. In this way,the spatial position provides a sort of time reference.Hodkinson et al. (2003), who also worked on the MidreLovnbreen glacier, opted to study unreworked morainezones only. Plant colonization is considered, then, as astraightforward ordered process that passes through aseries of progressive stages (Kuc 1964; Ziaja & Dubiel1996; Nakatsubo et al. 1998; Kume et al. 2003). Thedirect survey reported here supplements that approach.The succession of stages of colonization shown on thefactorial graphs (Fig. 5) relates to a timescale that can becalibrated from Fig. 7. The first three stages of the suc-cession can be completed in less than 100 years. The timevectors clearly show the continuity of the process thatleads from the mineral stage to the mature stages, butthey also show that transitions from one type to the nextdid not occur everywhere with the same intensity. Overthe 31-year interval, all sampling sites recorded changes,and adjustments are underway everywhere, even in themature tundra. Analysis of these changes reveals thatthey are not all consistent with the temporal logic under-pinning plant colonization, marked by the succession ofthe various stages. The impact of certain processes thatinterfere with vegetation dynamics can be appraised. Themajor role played by run-off can be noted, as 13 sam-pling sites were reworked: this erosional processcontributes to maintaining large areas at the mineralstage, both in the zone within the moraine amphitheatreand in the outlying sandurs. Run-off also accounts formost of the regressive phenomena observed: theseregressions may be partial, where run-off is occasionaland slight, or complete, with a return to the initial stagewhere water flow is intense; even sections of maturetundra were affected in the observation interval. Otherprocesses have far less effect than run-off. In someinstances, plant colonization is impeded by the grain-sizecharacteristics of the substratum (natural paving), or bysite exposition (such as a hollow where snow is trappedand only melts late in the growing season). Solifluxion isalso a factor of constraint observed at 11 sampling sites,although its effect is most marked on the axial part of thelobes, whereas the edges remain stable: colonization wasable to continue at the sampling sites affected by solif-luxion, as seven of them changed type, with six shiftingfrom type 3 to type 4, and with one shifting from type 4to type 5.Fig. 7 Diagram of the different stages of vegetation succession on glacial forelands in Svalbard. The line shows the trajectory of vegetation over timerelative to the various vegetation stages and habitat types.Vegetation evolution (19752006) in SvalbardM. Moreau et al.Polar Research 28 2009 364375 2009 The Authors 373A final issue is the representative character of the workcarried out. Most of the studies cited here relate to zonesthat have recently been opened to colonization (morainesand recent sandurs) because of glacial melting, and soomit the transition stages leading to mature tundra. Thecharacteristics of our study site, where ancient glacioma-rine terraces occur alongside more recent sandur areas,pre-dating the Little Ice Age, provide insight into howthese biogeographical units with their very differenttimescales, and which are widespread in polar regions,are related to one another by vegetation dynamics. Thetime interval under study is long enough to give someidea of this process.ConclusionThe method is an important aspect of this paper. Ourstudy specifies, for an applied case, how changes invegetation can be observed at relatively long intervalsby comparing repeated samplings from sites that areregularly spaced along transects. Although localizationproblems have been quite satisfactorily resolved, moredifficulty was found in deciding what informationalcontent could be exploited while ensuring the two dataseries remained comparable. We opted to sacrifice preci-sion for the sake of consistency. In this way, we could besure the changes observed were real, and were not mereartefacts generated by the protocol. The downside is thatwe detected major changes only. The discussion sectionsuggests how this limitation might be overcome, and howobservations might be made more precisely, especiallyby measuring species abundance from frequencycalculations.As the data collected were referenced in space andtime, a processing procedure was required to cover bothof these aspects. Correspondence analysis coupled withclassification revealed the time pathways of the samplingsites, whereas the mapping of types at the two datesbrought out the spatial pattern of vegetation dynamics.Correlating the graphs and maps provided an interestingbasis for interpreting the results.The results are consistent with many earlier studies.Time is the main factor differentiating vegetation typesfrom the mineral and pioneer stages through to maturetypes, where moisture and aridity appear to be the dis-tinguishing features. The changes affecting vegetationwere measured against these reference types. Most of thetrajectories followed by the sampling sites are consistentwith this general pattern, and reveal the continuity of thepattern as the vectors follow on from one vegetation typeto the next. The vector length provides a ready measureof change, and is an indicator of the dynamics of thesampling sites. Change and continuity are found in alltypes of settings, although with different characteristics.Vectors for the recent moraine and sandur sectors showthat plant colonization may be rapid, as some samplingsites have completed several stages in the successionwithin the 31-year interval. Conversely, vectors for themature tundra sectors are shorter, and indicate only slightadjustments in sampling site composition, the meaning ofwhich is less obvious; however, the hypothesis that thesoils have dried out is advanced, as wet-habitat specieshave been lost from several sampling sites. The impact ofrun-off in explaining regression has been shown, as hasthe effect of certain limiting factors such as grain size orthe late melting of snow cover.AcknowledgementsThe fieldwork for this study was carried out with logisticalresources provided by the Jean Corbel research base,operated by the French Polar Institute. We are grateful tothe anonymous reviewers for their advice on the paper.ReferencesAiken S.G., Dallwitz M.J., Consaul L.L., McJannet C.L.,Gillespie L.J., Boles R.L., Argus G.W., Gillett J.M., ScottP.J., Elven R., LeBlanc M.C., Brysting A.K. & Solstad H.1999. Flora of the Canadian Arctic Archipelago. Accessed onthe internet at on3 April 2009.Benzecri J.-P. 1992. Correspondence analysis handbook. NewYork: Dekker.Brossard T. 1977. Les quilibres biogographiques lis auxquilibres dynamiques de versants dans les paysages du Svalbard(Ny-lesund). (Relation between biogeography and slopedynamics in Svalbard landscapes [Ny-lesund].) PhD thesis,Ecole Pratique des Hautes Etudes en Sciences Sociales,Paris.Brossard T. 1985. volution biogographique dun secteurde moraine au Svalbard (glacier Loven central) bilanquantitatif 19781981. (Biogeographical evolution of asector of moraine in Svalbard [central Lovn glacier]:quantitative assessment 19781981.) Inter-Nord 17, 3346.Brossard T., Druelle S., Nimis P.L. & Petit P. 1984. Aninterdisciplinary approach to vegetation mapping in HighArctic Environments, Ny-lesund, Svalbard. Phytocoenolgia12, 433453.Corbel J. 1966. Datation au carbone 14 des terrasses marinesde la Baie du Roi. (Radiocarbon dating of marine terracesin the Kongsfjorden area.) In: Spitsberg 1964 et premiresobservations 1965. (Spitsbergen 1964 and first observations1965.) Pp. 309314. Lyon: Audin ditions.Elvebakk A. 1997. Tundra diversity and ecologicalcharacteristics of Svalbard. In F.E. Wielgolaski (ed.):Ecosystems of the world. Vol. 3. Polar and alpine tundra.Pp. 347359. Amsterdam: Elsevier.Vegetation evolution (19752006) in Svalbard M. Moreau et al.Polar Research 28 2009 364375 2009 The Authors374Escofier B. 1978. Analyse factorielle et distances rpondantau principe dquivalence distributionnelle. (Factoranalysis and distances in accordance with thedistributional equivalence principle.) Revue de StatistiqueApplique 26, 2937.Escofier B. 1979. Une reprsentation des variables danslanalyse des correspondances multiples. (A way ofrepresenting variables in multiple correspondenceanalysis.) Revue de Statistique Applique 27, 3747.Forman S.L. 1990. Post-glacial relative sea-level history ofnorthwestern Spitsbergen, Svalbard. Geological Society ofAmerica Bulletin 102, 15801590.Forman S.L. & Miller G.H. 1984. Time-dependent soilmorphologies and pedogenetic processes on raisedbeaches, Brggerhalvhya, Spitsbergen, SvalbardArchipelago. Arctic and Alpine Research 16,381394.Greenacre M.J. 1984. Theory and applications of correspondenceanalysis. London: Academic Press.Hagen J.O., Liestl O., Roland E. & Jrgensen T. 1993.Glacier atlas of Svalbard and Jan Mayen. Norsk PolarinstituttMeddelelser 129. Oslo: Norwegian Polar Institute.Hodkinson I.D., Coulson S.J. & Webb N.R. 2003. Communityalong proglacial chronosequences in the High Arctic:vegetation and soil development in north-west Svalbard.Journal of Ecology 91, 651663.Klanderud K. & Birks H.J.B. 2003. Recent increases inspecies richness and shifts in altitudinal distributionsof Norwegian mountain plants. The Holocene 13,16.Kuc M. 1964. Deglaciation of Treskelen-Treskelodden inHornsund, Vestspitsbergen, as shown by vegetation. StudiaGeologica Polonica 11, 197206.Kume A., Bekku Y.S., Hamba Y.T. & Kanda H., 2003. Carbonisotope discrimination in diverging growth forms ofSaxifraga oppositifolia in different successional stages in aHigh Arctic glacier foreland. Arctic, Antarctic, and AlpineResearch 35, 377383.Lefauconnier B. 1990. Fluctuations rcentes des glaciers dansle Kongsfjord, 79N Spitsberg. (Recent fluctuations ofglacier fronts in Kongsfjorden, 79N Spitsbergen.)Inter-Nord 19, 449453.Le Roux B. & Rouanet H. 2004. Geometric data analysis: fromcorrespondence analysis to structured data. Dordrecht: KluwerAcademic Publishers.Mercier D. 1998. Le ruissellement au Spitsberg. Limpact dunprocessus azonal sur les paysages dun milieu polaire,presqule de Brgger. (Runoff in Svalbard. The polar worldfacing climate changes.) PhD thesis, Universit BlaisePascal Clermont-Ferrand II.Minami Y., Okitsu S. & Kanda H. 1997. Relationshipbetween plant community and topographic factors onthe moraine at deglaciated Arctic terrain in Ny-lesund,Svalbard. Bulletin of the Faculty of Agriculture TamagawaUniversity 37, 2130.Moreau M. 2005. Dynamique des paysages vgtaux depuis lafin du Petit ge Glaciaire au Spitsberg. Analyse intgre de lareconqute vgtale des marges proglaciaires. (Dynamics ofvegetated landscapes since the end of the Little Ice Age: analysis ofplant colonization in proglacial areas.) PhD thesis, UniversitBlaise Pascal Clermont-Ferrand II.Moreau M., Laffly D., Joly D. & Brossard T. 2005. Analysis ofplant colonization on an Arctic moraine since the end ofthe Little Ice Age using remotely sensed data and aBayesian approach. Remote Sensing of Environment 99,244253.Nakatsubo T., Bekku Y., Kume A. & Koizumi H. 1998.Respiration of the belowground parts of vascular plants:its contribution to total soil respiration on a successionalglacier foreland in Ny-lesund, Svalbard. Polar Research 17,5359.Nilsen L., Elvebakk A., Brossard T. & Joly D. 1999. Mappingand analysing Arctic vegetation: evaluating a methodcoupling numerical classification of vegetation data withSPOT satellite data in a probability model. InternationalJournal of Remote Sensing 20, 29472977.Rnning O.I. 1996. The flora of Svalbard. Oslo: NorwegianPolar Institute.Salvigsen O. 1977. Radiocarbon dating and the extension ofthe Weichselian ice sheet in Svalbard. Norsk Polarinstituttrbok 1976, 209224.Wada N. 1999. Factors affecting the seed-setting success ofDryas octopetala in front of Brggerbreen (Brgger Glacier)in the High Arctic, Ny-lesund, Svalbard. Polar Research 18,261268.Ziaja W. & Dubiel E. 1996. Vascular plant succession duringcontemporary deglaciation in the mountains ofNordenskild Land, Spitsbergen. In W.E. Kawczyk (ed.):XXIII Sympozjum Polarne. (23rd Polar Symposium.)Pp. 99104. Sosnowiec: University of Silesia.Vegetation evolution (19752006) in SvalbardM. Moreau et al.Polar Research 28 2009 364375 2009 The Authors 375


View more >