18
http://hol.sagepub.com/ The Holocene http://hol.sagepub.com/content/24/8/985 The online version of this article can be found at: DOI: 10.1177/0959683614534745 2014 24: 985 originally published online 4 June 2014 The Holocene Vivian A Felde, Sylvia M Peglar, Anne E Bjune, John-Arvid Grytnes and H John B Birks Setesdal, southern Norway The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Published by: http://www.sagepublications.com can be found at: The Holocene Additional services and information for http://hol.sagepub.com/cgi/alerts Email Alerts: http://hol.sagepub.com/subscriptions Subscriptions: http://www.sagepub.com/journalsReprints.nav Reprints: http://www.sagepub.com/journalsPermissions.nav Permissions: http://hol.sagepub.com/content/24/8/985.refs.html Citations: What is This? - Jun 4, 2014 OnlineFirst Version of Record - Jul 15, 2014 Version of Record >> at TEXAS SOUTHERN UNIVERSITY on October 19, 2014 hol.sagepub.com Downloaded from at TEXAS SOUTHERN UNIVERSITY on October 19, 2014 hol.sagepub.com Downloaded from

The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

  • Upload
    h-j-b

  • View
    218

  • Download
    0

Embed Size (px)

Citation preview

Page 1: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

http://hol.sagepub.com/The Holocene

http://hol.sagepub.com/content/24/8/985The online version of this article can be found at:

 DOI: 10.1177/0959683614534745

2014 24: 985 originally published online 4 June 2014The HoloceneVivian A Felde, Sylvia M Peglar, Anne E Bjune, John-Arvid Grytnes and H John B Birks

Setesdal, southern NorwayThe relationship between vegetation composition, vegetation zones and modern pollen assemblages in

  

Published by:

http://www.sagepublications.com

can be found at:The HoloceneAdditional services and information for    

  http://hol.sagepub.com/cgi/alertsEmail Alerts:

 

http://hol.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

http://hol.sagepub.com/content/24/8/985.refs.htmlCitations:  

What is This? 

- Jun 4, 2014OnlineFirst Version of Record  

- Jul 15, 2014Version of Record >>

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 2: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

The Holocene2014, Vol. 24(8) 985 –1001© The Author(s) 2014Reprints and permissions:sagepub.co.uk/journalsPermissions.navDOI: 10.1177/0959683614534745hol.sagepub.com

IntroductionThe relationships between vegetational composition and pollen assemblages in surface-sediment samples provide a valuable guide to the interpretation of fossil pollen assemblages (e.g. Birks and Gordon, 1985). There have been many studies on modern pollen assemblages in relation to contemporary vegetation. How-ever, the focus of such studies varies from detecting how well modern pollen composition reflects contemporary vegetation types at a range of spatial scales (e.g. Birks, 1973a, 1973b; Birks et al., 1975; Fletcher and Thomas, 2007; Lichti-Federovich and Ritchie, 1968), finding pollen taxa that have a strong relationship with modern climate (e.g. Fall, 2012; Herzschuh and Birks, 2010; Ortuño et al., 2011; Wei et al., 2011; Zhao et al., 2012) or estimat-ing relative pollen-representation values (e.g. Andersen, 1967, 1970; Bradshaw, 1981; Hjelle, 1998) or pollen-productivity esti-mates (e.g. Broström et al., 2008; Bunting et al., 2005). Such stud-ies are invaluable in helping to reduce some of the inherent biases in pollen representation and in the reconstruction of past vegeta-tion (Davis, 2000; Sugita, 2007a, 2007b).

Recently, the focus of modern pollen–vegetation studies has shifted to examining pollen–plant richness relationships (e.g. Melt-sov et al., 2011; Odgaard, 1994). Drastic changes in global diversity are predicted as a result of changes in climate and human land-use changes (Bellard et al., 2012). The decrease in diversity observed

over the last century is comparable to mass extinction rates in the past, and more knowledge about species vulnerability to external changes is urgently needed (Bellard et al., 2012). Clearly, the palaeo-ecological record is a valuable source of long-term information about how organisms have responded to rapid climate or land-use changes in the past (e.g. Willis et al., 2010). However, to be able to study ter-restrial richness of plant species using pollen assemblages through time, it is important to establish whether there is a strong and robust relationship today between the composition and richness of vegeta-tion and the composition and richness of pollen assemblages.

The vegetation–pollen relationship is not a 1:1 relationship as pollen production and dispersal differ between species. Distance

The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

Vivian A Felde,1,2 Sylvia M Peglar,2 Anne E Bjune,1 John-Arvid Grytnes2 and H John B Birks2,3,4

AbstractReconstructing and interpreting past vegetation composition can be enhanced by studying modern pollen samples and contemporary vegetation. Here, we compare pollen in surface sediments from 52 medium-sized lakes with the surrounding vegetation along an elevational gradient covering six major vegetation zones in south-central Norway. The aims are to detect how well the vegetational composition and terrestrial pollen assemblages distinguish the major vegetation zones, whether the pollen composition in surface-sediment samples reflects the composition of the surrounding vegetation and whether aquatic pollen and spores reflect the major vegetation zones. We use multivariate classification trees, ordination and co-correspondence analysis to address these questions. We show that it is possible to separate the major zones using terrestrial pollen assemblages and using plant species in the vegetation reasonably well, whereas aquatic pollen and spores poorly reflect the zones. Surprisingly, the terrestrial pollen assemblages separate the zones better than vegetational composition does. The compositional match between the pollen assemblages and surrounding vegetation is consistent for sites along the elevational gradient within the forested zones, but deteriorates in increasingly open vegetation zones. Our results are consistent with other investigations of modern pollen–vegetation relationships. Careful interpretation of past vegetation from pollen assemblages is needed when the vegetation is treeless because of a larger potential pollen-source area and hence a higher proportion of long-distance dispersed pollen in open areas.

Keywordsco-correspondence analysis, indicator species analysis, modern pollen–vegetation relationships, multivariate classification trees, Procrustes analysis, Protest randomisation test

Received 17 January 2014; revised manuscript accepted 11 April 2014

1Uni Research Climate and Bjerknes Centre for Climate Research, Norway2University of Bergen, Norway3University College London, UK4University of Oxford, UK

Corresponding author:Vivian A Felde, Uni Research Climate and Bjerknes Centre for Climate Research, Allegatén 55, N-5007 Bergen, Norway. Email: [email protected]

534745 HOL0010.1177/0959683614534745The HoloceneFelde et al.research-article2014

Research paper

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 3: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

986 The Holocene 24(8)

from lake, lake size and morphometry, vegetation structure, wind direction and a range of taphonomic processes (e.g. differential preservation and sedimentation) can all affect the probability of pollen being represented in lake-sediment samples (e.g. Davis, 2000; Prentice, 1985; Sugita, 1993, 1994; Tauber, 1965). Plants with high pollen production and wind-dispersed pollen are over-represented, whereas species with poor pollen production and/or insect pollination are under-represented or not represented at all. Small hollows (10- to 20-m radius) receive pollen from local stand-scale vegetation (Janssen, 1973), whereas medium-sized lakes (200 to 300 m radius) derive pollen from a larger regional-scale vegetation area (Calcote, 1995; Davis, 2000; Jacobson and Bradshaw, 1981; Janssen, 1973). Pollen assemblages from open treeless vegetation may contain more extra-regional pollen (sensu Janssen, 1973) from plants that are not locally present in the area. Pollen spectra from treeless vegetation sites can thus be difficult to separate from spectra from forested sites (Birks, 1973a; Pelánková et al., 2008). Homogenous vegetation types may usually have a smaller relevant pollen-source area than het-erogeneous vegetation types, thus making it difficult to assess the relevant spatial scale. However, despite these biases, results from various studies show that the predictive abilities of pollen for discriminating major vegetation types is generally good, with, on average, about 20–30% prediction errors (e.g. Birks and Gordon, 1985; Fall, 2012; Felde et al., in press; Pelánková et al., 2008).

Vegetation–pollen relationships can also vary from site to site and in different vegetation types, emphasising the importance of studying such relationships in different areas, and preferably along major vegetation gradients. In general, investigations of the relationships between modern pollen and vegetation compo-sition have usually been conducted by comparing pollen spectra at different sites to major vegetation types (e.g. Lichti-Federovich and Ritchie, 1968), or by investigating the amounts of a pollen type and of the relevant plant species in the surrounding vegeta-tion (e.g. Andersen, 1970). Few studies (e.g. Birks, 1973a) have collected full plant species lists and pollen data, and compared two multivariate data-sets directly. This is necessary for investi-gating modern pollen–plant richness relationships (Birks, 1973a; Meltsov et al., 2011), and it can provide detailed information about the relationships between pollen and vegetation composi-tion and between pollen and vegetation richness. By sampling pollen assemblages and associated vegetation data along eleva-tional gradients, it is possible to investigate the pollen–vegetation relationships in different vegetation types and to compare these differences (e.g. Birks, 1973a, 1973b; Djamali et al., 2009; Wright et al., 1967).

In addition, when studying pollen assemblages from lakes, aquatic pollen and spores and terrestrial pollen and spore types are usually considered separately because aquatic taxa represent strictly local within-lake conditions. These conditions may be determined by climate, lake morphometry, geochemical and hydrological processes within the lake; input of nutrients from the catchment; and the vegetation and soils of the catchment (e.g. Birks et al., 2000; Väliranta et al., 2005). However, the relation-ship between terrestrial vegetation and aquatic pollen has not, as far as we know, been rigorously investigated. If catchment vege-tation is important in determining aquatic pollen assemblages, aquatic pollen should be able to predict major terrestrial vegeta-tion types. If other factors such as geochemical processes within the lake, lake-water temperature and lake morphometry are more important for determining the aquatic macrophyte flora, aquatic pollen may not satisfactorily discriminate between major terres-trial vegetation types.

In this study, we use the elevational gradient in the Setesdal valley in south-central Norway to investigate the following questions:

1. How well do the composition of vegetation and the com-position of terrestrial pollen assemblages distinguish the major vegetation zones?

2. How well do the pollen assemblages in surface-sediment samples of small-to-medium sized lakes represent the composition of the vegetation around the lakes?

3. How well do aquatic pollen and spores reflect the major terrestrial vegetation zones?

The terms local, extra-local, regional and extra-regional are used throughout in the sense of Janssen (1966, 1973, 1981). If used in a different sense, a relevant reference is given (e.g. Sugita, 2007a, 2007b).

Methods and materialStudy areaThe study area is the Setesdal valley, in south-central Norway. It stretches approximately 200 km from Kristiansand in the south to Haukelifjell in central Norway (58–60°N; Figure 1). The valley forms an elevational gradient ranging from 0–1318 m a.s.l. This gradient represents a climate gradient where the mean July tem-perature ranges from 15.6°C to 10.0°C, and the mean January temperature varies between −1.7°C and −8°C. Annual precipita-tion varies between 840 and 1300 mm (http://sharki.oslo.dnmi.no/portal/page). The bedrock is mainly gneiss (http://geo.ngu.no/kart/berggrunn/) with extensive overlying morainic and glacial outwash deposits in southern Setesdal. We studied 52 lakes along this transect. The lake locations and characteristics are sum-marised in Table 1.

The Setesdal valley provides an ideal setting for ecological and palaeoecological studies because it covers six major vegeta-tion zones – nemoral, boreonemoral, southern-boreal, middle-boreal, northern-boreal and low-alpine (Figure 1) – within 200 km. The definition of these vegetation zones follows Moen’s (1998) classification of broad-scale vegetation types within Nor-way. The different zones are defined by overall species composi-tion that is interpreted as reflecting different precipitation and temperature conditions (Moen, 1998). Some of our sites have been used in other palaeoecological studies (Birks, 2007; Brooks, 2003; Eide et al., 2006; Lüder, 2007; Panizzo et al., 2008).

The nemoral zone in Norway is a narrow belt in the southern-most part of Norway, representing the northernmost part of this vegetation type in the world (Moen, 1998). Deciduous forests, especially oak forests (Quercus spp.), are characteristic together with species whose northern extent is limited by low winter tem-peratures. Coniferous forests in this zone are planted. Moving northwards, the next zone is the boreonemoral zone. This is tran-sitional between the nemoral and the extensive boreal zones. It is much affected by agricultural land-use, and the vegetation is patchy consisting of pastures and a mosaic of deciduous forests with, for example, Quercus spp., Tilia cordata, Fraxinus excel-sior, Corylus avellana, Betula pendula and Alnus glutinosa, and boreal forests with Picea abies and Pinus sylvestris. The next zone, the southern-boreal zone, can be difficult to distinguish from the boreonemoral zone because it is also much affected by land-use and contains localised patches with warmth-demanding deciduous trees. However, boreal forests are more frequent, and trees and shrubs such as Betula pubescens and Salix spp. are more common than in the boreonemoral zone. The middle-boreal zone is dominated by coniferous forests and ombrotrophic bogs. Deciduous trees such as B. pubescens, Alnus incana, Sorbus aucuparia and Populus tremula are also common, but warmth-demanding tree species are rarer and only occur in locally suitable sites such as on south-facing slopes or cliffs. In the main Setesdal valley, the middle-boreal zone can be difficult to distinguish from

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 4: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

Felde et al. 987

Table 1. Sites, site codes, coordinates, elevation, vegetation zone and size and depth of each of the lakes ordered from north to south.

Locality Code Latitude (°N) Longitude (°E) Elevation Vegetation zone Lake size (ha) Lake depth (m)

Reinstjørna REIN 59.85081 7.02168 1318 LA 84.56 15.5Holebudalen HOLE 59.84186 6.99235 1144 LA 30.20 7.9Hermodholtjønni HERM 59.80733 7.25255 1107 LA 48.01 6.5Godtholtjønni GODT 59.81409 7.20968 981 LA 46.07 10.5Lomtjønn LOMT 59.80777 7.41007 907 LA 36.49 3.2Skreutjønn SKRE 59.79463 7.41571 901 LA 8.31 6.5Lille Kjelavatn LILL 59.79603 7.25229 960 NB 4.16 3.3Kyrkjestøyltjønni KYRK 59.76019 7.65303 955 NB 49.37 16.0Sæsvasstjønn SAESV 59.66277 7.51264 899 NB 69.67 8.6Stokketjønn STOK 59.66338 7.53651 898 NB 59.52 4.5Flotatjønn FLOT 59.66972 7.54042 889 NB 62.86 8.1vBotnetjønni vBOT 59.77002 7.41534 813 NB 1.25 7.2Isbenttjønn ISBE 59.76471 7.43588 785 NB 44.50 10.0Hovden1 HOV1 59.55834 7.34626 780 NB 26.21 4.0Hovden2 HOV2 59.55645 7.35119 780 NB 4.41 2.4Gravdalstjønn GRAV 59.76439 7.57624 726 NB 4.34 5.2Troddetjønn TROD 59.68142 8.05793 729 MB 86.64 3.3Øygardstjønn OYGA 59.6258 7.98732 676 MB 28.92 8.1Gregarstjønn GREG 59.57468 8.02657 660 MB 53.86 8.1Steintjønn STEI 59.60554 7.80756 574 MB 11.38 4.1Bjørntjønn BJOR 59.54524 8.01851 530 MB 2.20 7.8Lisletjønn LISL 59.34353 7.30609 518 MB 50.31 7.1Øytjønni OYTJ 59.31834 7.341 531 MB 26.54 18.5Ormetjønn ORME 59.56966 7.98317 453 MB 107.78 25Furuteig FURU 59.17324 7.52834 300 MB 2.74 11.7Tjovetjørn TJOV 58.76308 7.81138 274 SB 19.48 3.3Moseid MOSE 58.64434 7.7974 205 SB 4.97 5.5Vassendkilen VASS 58.66641 7.81727 203 SB 145.33 2.5Pederstjønn PEDE 58.45575 7.79042 200 SB 28.56 21.0Fisketjønn FISK 58.62294 7.80195 187 SB 13.36 3.6Grostjørna GROS 58.53828 7.73457 192 SB 22.88 9.5Bærvannet BAER 58.33601 7.78476 315 BN 26.06 17.0Tovetjønna TOVE 58.30039 7.74892 295 BN 27.92 9.5Reiersdalsvann REIE 58.32648 7.78701 244 BN 48.47 9.2Espeland ESPE 58.36136 7.79715 242 BN 42.03 10.0Nordtjønn NORD 58.3161 7.78931 215 BN 13.43 7.8Poddetjønn PODD 58.26124 7.7847 206 BN 27.99 7.0Skarpenglandsvann SKAR 58.29488 7.84713 180 BN 82.48 17.0Gjeningstjønn GJEN 58.27131 7.73883 165 BN 26.79 6.3Store Gangtjønn STOR 58.27176 8.05405 135 BN 23.01 8.9Åmdalstjønna AMDA 58.32449 8.00136 70 BN 24.74 12.0Årstølvannet ARST 58.13665 7.72172 197 N 19.07 12.5Skogevann SKOG 58.24657 7.8239 152 N 97.55 22.0Rypestøl RYPE 58.18499 7.86905 123 N 27.80 11.0Tvitjønn TVIT 58.14421 7.91406 80 N 6.00 7.8Dalanstjønn DALA 58.24374 8.00565 40 N 17.56 9.2Stemtjørn STEM 58.17404 8.19653 25 N 38.04 10.0Heimretjønn HEIM 58.24758 8.28647 30 N 11.79 19.5Hommerkleivtjønn HOMM 58.22078 8.24782 25 N 10.43 4.5Kostøltjønn KOST 58.22268 8.02307 18 N 12.85 6.5Oftenestjønna OFTE 58.07579 7.74401 17 N 13.08 13.0Kjøstveittjønna KJOS 58.18838 8.2403 10 N 38.04 4.7

LA: low-alpine; NB: northern-boreal; MB: middle-boreal; SB: southern-boreal; BN: boreonemoral; N: nemoral.

the southern-boreal zone because of this. Scattered individuals of Ulmus glabra are observed as far north as Bykle (c. 500 m a.s.l.), and C. avellana and B. pendula are often observed along the val-ley, on south- or west-facing slopes. However, most sites from this zone are from the eastern part which is mainly dominated by P. abies (Figure 1). The northern-boreal zone grades into low-alpine vegetation and is dominated by mountain-birch forest (B. pubescens) and conifers with increasingly scattered individuals of

P. abies and P. sylvestris as elevation increases. Tree growth ceases at the tree-line. The low-alpine zone consists mainly of shrubs (Juniperus communis, Salix spp.), dwarf-shrubs (Empe-trum nigrum, Calluna vulgaris, Vaccinium spp.), graminoids and herbs.

The Setesdal valley has a long cultural history. Archaeological and palaeoecological evidences show that human activity dates back to c. 8000 cal. BP (Selsing, 2010). These prehistoric people

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 5: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

988 The Holocene 24(8)

were most likely hunter–gatherers who depended on reindeer populations in the mountains. During the Holocene, this area has been used with varying intensity, and humans have probably affected the vegetation to different degrees over millennia (Sels-ing, 2010). When humans settled and started agriculture, transhu-mance practices became an important lifestyle where people lived in the mountains with their animals during the summer and moved down to the lower parts of the valley during the winter. In the last century, there have been changes in this traditional land-use, and transhumance practices in the mountains have gradually shifted to tourism and recreational activities such as hiking and skiing.

Sediment sampling and pollen analysisSediment sampling was conducted during the summers of 1996 and 1997. The 52 lakes were selected to represent, as far as pos-sible, natural, non-disturbed vegetation (low human impact), and with a lake catchment of approximately similar size. The top sedi-ments were sampled using a HTH gravity corer (Renberg and Hansson, 2008) with the surface sample being the uppermost 0.5 cm of sediment at the sediment/lake-water interface taken at the deepest part of each lake. Pollen samples were prepared follow-ing standard methods (Method B of Berglund and Ralska-Jasiewiczowa, 1986) and mounted in silicone oil. Identifications were made to the lowest possible taxonomic level using keys (Birks, 1973b; Faegri et al., 1989; Punt et al., 1976–1995) and modern reference material in the Department of Biology, Univer-sity of Bergen. The pollen sum of terrestrial pollen and spores counted ranged from 554 to 896. The pollen counts and associated meta-data are stored in the European Modern Pollen Database (Davis et al., 2013).

Vegetation samplingVegetation sampling was done during the summers of 2011 and 2012. We compiled vascular plant species lists for each lake, where all observed plant species along a chosen path were

recorded. Vegetation sampling in rugged terrain such as in parts of the Setesdal valley is time-consuming and challenging, and to balance time and effort, we developed a standardised sampling scheme for recording plant species around the lakes that attempts to match the likely relevant spatial scale of the pollen composition in the lake surface sediments. Based on knowledge from previous published studies on pollen deposition, it is well known that the probability of pollen coming from a particular plant decreases with increasing distance from the lake (Davis, 2000). Since the vegetation around each lake was relatively homogenous and some of the lakes had high near-vertical cliffs that restricted the vegeta-tional survey, we walked half-way around the lake, and then at c. 500 m radii from the edge of the lake covering all the different visible vegetation types, and recording all the plant species we observed. We considered that when the vegetation is spatially homogenous (as it commonly is in boreal and low-alpine vegeta-tion), half-way around the lake and a 500 m sampling radius is an adequate compromise between increased sampling effort and the number of additional new species. Each species was given an abundance based on how many times it was observed on a fre-quency scale from 1 to 5. This is a subjective estimate where the most frequent species were assigned a value of 5, very frequent species scored 4, frequent species 3, occasional species 2 and rare species 1.

Data preparationFour data-sets were prepared for numerical data analysis: (1) ter-restrial pollen and spores from the surface-sediment samples, (2) aquatic pollen and spores from the surface-sediment samples, (3) terrestrial plant species recorded from the vegetation around each lake and (4) terrestrial plant species from the vegetation trans-formed into pollen and spore types. To transform the plant species into pollen and spore types, we compiled two translation tables of native and non-native plant species of the Norwegian flora (Lid and Lid, 2005). The plant species were then assigned to their rel-evant pollen or spore types based on five sources: Beug (2004),

Figure 1. Overview map of the Setesdal valley in south-central Norway, with pictures from sites to illustrate the major vegetation zones described by Moen (1998). The stars represent the position of the 52 lakes.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 6: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

Felde et al. 989

Faegri and Iversen (1964), Faegri et al. (1989), Moore et al. (1991) or Birks (1973b) and Birks and Peglar (unpublished data; Felde et al., 2012). In this study, the pollen and spore types are based on the Birks and Peglar categories in Felde et al. (2012), and nomenclature of plant species follows Lid and Lid (2005). To avoid the problem that some plant species may produce two or more morphologically similar pollen or spore types, the pollen types Rumex acetosa and Rumex acetosella-type were merged into R. acetosa/acetosella-type; Blechnum spicant, Dryopteris filix-mas-type and Dryopteris carthusiana spores were merged into Dryopteris-type; and species within the Rosaceae family were merged into Rosaceae undiff. except in a few cases where the pollen types could be confidently and consistently identified to genus. Pollen percentages were calculated on the basis of the sum of total terrestrial pollen and spores, while aquatic pollen were calculated as the sum of aquatic + terrestrial pollen and spores. Both the modern plant data and the plant data converted into pollen types were transformed into percentages as follows: 1 = 0.5%, 2 = 12.5%, 3 = 37%, 4 = 62.5% and 5 = 87.5%. Diagrams of selected plant species, plants as pollen and spores and terres-trial and aquatic pollen assemblages along the Setesdal vegeta-tional gradient were prepared using Tilia and TGView (Grimm, 1990, 2004) (Figures 2–5).

Numerical analysisTo detect how well the composition of vegetation, of plants in the vegetation transformed to pollen and spore types, terrestrial pol-len and spores from surface sediments and aquatic pollen and spores are able to differentiate Moen’s (1998) six vegetation zones in the Setesdal valley, multivariate classification trees (MCTs) were used. MCTs are useful and robust tools to explore, describe and predict relationships between complex multivariate data-sets and associated environmental or predictor variables (De’ath, 2002; Simpson and Birks, 2012). Hellinger-transformed percentage data were used as input for the four data-sets, and Moen’s vegetation zones were the only constraining factor to compare the amount of variation captured by the data, the predic-tion accuracy and the statistically important indicator taxa for each vegetation zone. Because there were so many response vari-ables (pollen and plant taxa), the most significant indicator taxa were assessed by combining MCTs with indicator species analy-sis (IndVal; Borcard et al., 2011; Legendre and Legendre, 2012). Only taxa with an IndVal higher than 20% and with a p-value lower than 0.01 based on IndVal randomisation tests were considered.

Detrended correspondence analysis (DCA; Hill and Gauch, 1980) was used to estimate the amount of compositional change in standard deviation (SD) units of turnover and to assess whether linear- or unimodal-based ordination methods were appropriate for these data (Ter Braak and Prentice, 1988). Since the compositional turnover was larger than 2.5 SD units for the vegetation data, DCA was conducted on Hellinger-transformed percentage data for all four data-sets. Generalised Procrustes analysis (Digby and Kempton, 1987; Gower, 1975; Peres-Neto and Jackson, 2001) and associated Protest significance tests (Jackson, 1995) were used to compare the similarities between the different DCA ordinations and to assess how well the differ-ent data-sets match each other in terms of compositional gradi-ents. By using similar ordination methods for all the data-sets, we avoid the problem that differences detected by Procrustes analysis may result from using different ordination methods. The Procrustes sum of squares (m2) is 0 when the two ordina-tions are identical, and rises when the distances between pairs of sites increase. The Protest randomisation test is a one-tailed test that accounts for the probability that the random m2 value is equal to or less than the observed m2 (Jackson, 1995).

The comparisons of ordinations were done for (1) plants in the vegetation versus terrestrial pollen and spores, (2) plants in the vegetation transformed to pollen and spores versus terrestrial pollen and spores, (3) plants in the vegetation versus aquatic pollen and spores, (4) plants in the vegetation as pollen and spores versus aquatic pollen and spores and (5) terrestrial pollen and spores versus aquatic pollen and spores. The results are dis-played in DCA diagrams with Moen’s vegetation zones fitted to the ordination space as passive objects, and Procrustean plots illustrating the similarities between paired samples.

To detect co-variation between the pollen assemblages and vegetational composition from our sites in Setesdal, co-corre-spondence analysis (Co-CA) was used. It allows the direct com-parison of two different multivariate compositional data-sets (Ter Braak and Schaffers, 2004; Schaffers et al., 2008). In palaeoecol-ogy, it has been used to investigate relationships between long-term changes in environment and diatom and chironomid assemblages (Bitušík et al., 2009), and environment and cladocer-ans (Davidson et al., 2011). It can be used in both a predictive and a descriptive mode, but only the descriptive mode is used here because the data-sets are not fully independent as pollen assem-blages are derived from the surrounding vegetation. Co-CA attempts to identity the underlying pattern that is common in both data-sets by maximising the covariance between two assemblage data-sets (Ter Braak and Schaffers, 2004). Finally, the percentage match in the pollen sum between the pollen and spore types in the vegetation and the pollen assemblages were compared along the elevational gradient.

All analyses were conducted with R (R Core Team, 2012) using the vegan package (Oksanen et al., 2013), the mvpart pack-age (De’ath, 2013), the MVPARTwrap package (Ouellette and Legendre, 2013) and the cocorresp package (Simpson, 2009).

ResultsMoen’s vegetation zonesIn total, 406 terrestrial plant species, 125 terrestrial pollen and spore types and 11 obligate aquatic pollen and spore types were recorded. After transforming the plant species into pollen and spore types, there were 180 plants as pollen and spore types. Rare taxa with a low frequency (<12.5%) for plants and plants as pollen and spore types, and pollen and spore types in the sedi-ment samples with percentages less than 5%, were deleted from the MCT analyses because such taxa have little or no effect on the results (Felde, unpublished results). The remaining taxa for these analyses were 280 plant species, 144 plants as pollen and spore types and 37 pollen and spore types from the sediment samples.

The MCT results show that the vegetation and plants trans-formed into pollen and spore types are similar (Figure 6(a) and (b)). The transformation was done to permit a more realistic comparison with the modern pollen assemblages. Moen’s vege-tation zones capture 35.4% of the variation in the vegetational composition (Figure 6(a)) and 34% in the transformed vegetation data (Figure 6(b)). The adjusted r2 for vegetation is 26.8%, while it is slightly higher for the transformed vegetation data (27.8%). The prediction accuracies of 16.2% (vegetation) and 13.7% (transformed vegetation data) are surprisingly low. Basically, the MCT results show that the main split between the low-alpine and northern-boreal and the lower elevation zones to the right explain 17.4% or 16.43% of the variation. The second split on the right branch separates the middle-boreal and southern-boreal from the boreonemoral and nemoral zones, and captures an additional 6.7% or 6.58% of the variation. The terminal nodes show that variations between the middle-boreal and southern-boreal, and the boreonemoral and nemoral zones, are relatively low (4.4% or 3.86% and 3.2% or 3.45%, respectively), which is similar to the

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 7: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

990 The Holocene 24(8)

Fig

ure

2. D

iagr

am o

f pla

nt s

peci

es s

elec

ted

to s

how

the

mai

n va

riat

ion

in t

he d

istr

ibut

ion

and

abun

danc

e of

pla

nts

obse

rved

in t

he v

eget

atio

n ar

ound

the

52

lake

s in

the

Set

esda

l val

ley.

The

spe

cies

in t

he d

iagr

am

are

orde

red

acco

rdin

g to

the

ir p

hysi

ogno

mic

gro

ups,

that

is, t

rees

and

shr

ubs;

gras

ses,

sedg

es a

nd r

ushe

s; he

rbs

and

fern

s an

d fe

rn a

llies

, and

arr

ange

d ac

cord

ing

to t

heir

occ

urre

nces

alo

ng t

he t

rans

ect

from

the

ne

mor

al t

o th

e lo

w-a

lpin

e zo

ne. T

he s

ites

are

arra

nged

acc

ordi

ng t

o th

e si

x m

ajor

veg

etat

ion

zone

s of

Moe

n (1

998)

. The

bar

s re

pres

ent

the

1–5

freq

uenc

y sc

ale.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 8: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

Felde et al. 991

Fig

ure

3. D

iagr

am o

f pla

nt s

peci

es in

the

veg

etat

ion

(Fig

ure

2) t

rans

form

ed in

to p

olle

n an

d sp

ore

type

s. T

he o

rder

of t

axa

and

follo

ws

the

sam

e sy

stem

as

for

the

plan

t sp

ecie

s (F

igur

e 2)

. It

is p

ossi

ble

to s

ee t

hat

som

e in

form

atio

n is

lost

abo

ut t

he o

ccur

renc

es o

f cer

tain

pla

nt s

peci

es w

ithin

pol

len

and

spor

e ty

pes

that

con

tain

ver

y m

any

diffe

rent

spe

cies

. Eve

n th

ough

the

re is

tax

onom

ic r

educ

tion

and

info

rmat

ion

abou

t th

e fr

eque

ncie

s fo

r ce

rtai

n ta

xono

mic

gro

ups

has

chan

ged

in t

he s

ense

tha

t th

ey h

ave

beco

me

com

mon

eve

ryw

here

(e.

g. Sa

lix, P

oace

ae, C

arex

-typ

e, B

etul

a), t

here

are

stil

l maj

or v

aria

tions

in t

he d

istr

ibut

ion

and

abun

danc

e of

mos

t of

the

pol

len

and

spor

e ty

pes

repr

esen

ting

the

plan

t sp

ecie

s in

the

veg

etat

ion.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 9: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

992 The Holocene 24(8)

Fig

ure

4. D

iagr

am o

f the

per

cent

age

abun

danc

e of

sel

ecte

d po

llen

and

spor

e ty

pes,

and

the

polle

n su

m c

ount

ed in

the

sur

face

-sed

imen

t sa

mpl

es fr

om 5

2 la

kes

in t

he S

etes

dal v

alle

y. T

he o

rder

of t

axa

follo

ws

the

sam

e sy

stem

as

for

the

plan

t sp

ecie

s an

d po

llen

and

spor

es o

f pla

nt s

peci

es (

Figu

res

2 an

d 3)

.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 10: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

Felde et al. 993

variation between the low-alpine and the northern-boreal on the other branch (3.7% or 3.67%).

The important statistically significant indicator taxa accord-ing to the IndVal analyses are identified for each zone (Table 2), and the occurrences of selected plant species within the different zones are included in Figure 2. The main difference between plants and plants as pollen and spore types is that the latter group has more significant indicator taxa for some of the zones. Indica-tor species analysis shows that the nemoral zone is discriminated by Quercus robur, B. pendula, Frangula alnus, P. tremula and the herb Convallaria majalis, while for the plants transformed into pollen and spore types, more indicator taxa are added including Acer, Alnus, C. avellana, F. alnus, F. excelsior, P. tremula, Quer-cus, T. cordata, Lonicera periclymenum, Viburnum opulus, Lysi-machia vulgaris-type, Oxalis acetosella, C. majalis, Plantago lanceolata, Polypodium and Athyrium filix-femina. Indicator taxa for the boreonemoral zone are Erica tetralix, Myrica gale, Pteridium aquilinum and Trifolium pratense, and the southern-boreal zone is considered as the limit of many warmth-demand-ing species restricted by cold winter temperatures (Moen, 1998). Many species have a high occurrence in this zone (Figure 2 or 3), but only P. sylvestris is an important indicator species. These results are the same for plants transformed into pollen and spore types. In the middle-boreal zone, the coniferous forest shifts into a dominance of P. abies, where Rubus idaeus is also an indicator species. Then, the coniferous forest grades into mountain-birch forest and shrub-dominated vegetation in the northern-boreal and low-alpine zones. Indicator species in the northern-boreal zone are B. pubescens and Luzula multiflora. A few additional herb taxa are included for pollen of plants, including Aconi-tum, Paris-type, Pedicularis, Silene-type, Pinguicula and Rhinanthus-type. The transition from the northern-boreal to

Table 2. The results of the IndVal analyses showing which taxa are important indicators using plants, plants as pollen and spores, terrestrial pollen and spores, and aquatic pollen and spores with a threshold above 20% IndVal and a significant p-value above 0.01. The IndVal is in parentheses.

Low-alpine Northern-boreal Middle-boreal Southern-boreal Boreonemoral Nemoral

Plants Empetrum nigrum (0.35), Eriophorum angustifolium (0.33), Rubus chamaemorus (0.37)

Betula pubescens (0.34), Luzula multi-flora (0.26)

Picea abies (0.29), Rubus idaeus (0.35)

Pinus sylvestris (0.32)

Erica tetralix (0.42), Myrica gale (0.36), Pteridium aquilinum (0.33), Trifolium pratense (0.30)

Betula pendula (0.35), Convallaria majalis (0.45), Frangula alnus (0.47), Populus tremula (0.33), Quercus robur (0.51)

Plants as pollen

Diphasiastrum (0.71), E. nigrum (0.35), Gentiana cf. purpurea (0.44), Huperzia selago (0.57), Juncaceae (0.25), Salix herbacea-type (0.71), Saxifraga stellaris (0.59)

Betula (0.27), Pinguicula (0.44), Rhinanthus-type (0.40)

Aconitum (0.40), Paris-type (0.40), Pedicularis (0.53), Picea (0.29), Silene-type (0.47)

Pinus (0.31) M. gale (0.38), Narthe-cium ossifragum (0.54), P. aquilinum (0.34), T. cf. pratense (0.31)

Acer (0.51), Alnus (0.45), Athyrium filix-femina (0.58), cf. Phragmites (0.53), Convallaria-type (0.43), Corylus avellana (0.51), F. alnus (0.45), Fraxinus excelsior (0.61), Lonicera periclymenum-type (0.80), Lysimachia vulgaris-type (0.52), Oxalis acetosella (0.43), Plantago lanceolata (0.45), Polypodium (0.56), P. tremula (0.32), Quercus (0.50), Tilia cordata (0.49), Viburnum opulus (0.49)

Ter-restric pollen

Artemisia (0.27), Carex-type (0.27), Cyperaceae undiff. (0.33), Dryopteris-type (0.25), Ericaceae-type (0.32), Poaceae (0.23), Salix (0.34), Solidago-type (0.33)

Diphasiastrum (0.34), Juniperus (0.22), Melampyrum (0.26), Selaginella selaginoides (0.42), Vaccinium-type (0.31)

Picea (0.28) Calluna vulgaris (0.24), Pinus (0.20)

– Alnus (0.25), Betula/Cory-lus/Myrica (0.28), cf. Phrag-mites (0.56), C. avellana (0.26), Filipendula (0.32), F. excelsior (0.27), M. gale (0.35), Quercus (0.36), T. cordata (0.35)

Aquatic pollen

– – – – – Nymphaea (0.39)

Figure 5. Diagram of the percentage abundance of aquatic pollen and spores in the surface-sediment samples from 52 lakes in the Setesdal valley. The rare pollen and spore types are multiplied by a factor of 16 to emphasise their presence. Site order follows Figures 2–4.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 11: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

994 The Holocene 24(8)

the low-alpine is mostly indicated by an increased frequency of alpine species such as Salix herbacea, E. nigrum, Carex bigelowii, Carex lachenalii, Poa alpina, Saxifraga stellaris, Diphasiastrum alpinum and Huperzia selago. E. nigrum, Eri-ophorum angustifolium and Rubus chamaemorus are important indicator species, while the pollen of plant species include taxa such as Gentiana cf. purpurea, Juncaceae and S. stellaris.

MCT results for the terrestrial pollen and spores from the surface-sediment samples show that the classification tree is slightly different than the classification trees for vegetation and for plants as pollen and spore types (Figure 6(c)). The first node shows the distinction between the low-alpine, middle-boreal and northern-boreal zones to the left; and the boreonemoral, nemoral and southern-boreal zones to the right. The total per-centage of variation captured by Moen’s vegetation zones is 47.1%, and the adjusted r2 is 41.5%. The prediction accuracy is 31%, which is better than for vegetation and for plants as pollen and spore types. The nemoral zone is relatively easily distin-guished palynologically by deciduous trees and warmth-demanding dwarf-shrubs and herbs such as Alnus, Betula/Corylus/Myrica, cf. Phragmites, C. avellana, Filipendula, F.

excelsior, M. gale, Quercus and T. cordata. Most match the plants as pollen and spore types observed in the vegetation. Taxa frequently observed in the nemoral zone are also present in the boreonemoral zone, and there are thus no important pol-len indicators for the boreonemoral zone. The southern-boreal zone is distinguished by a dominance of C. vulgaris and P. syl-vestris, and the middle-boreal zone is dominated by P. abies. The northern-boreal sites are dominated by shrubs and low-growing taxa such as Diphasiastrum, Juniperus, Melampyrum, Selaginella selaginoides and Vaccinium-type. The low-alpine zone has Artemisia, Carex-type, Cyperaceae undiff., Dryopteris-type, Ericaceae-type, Poaceae undiff., Salix and Solidago-type as indicators. Overall, the pollen indicators cor-respond well with the indicators of plant species for the differ-ent vegetation zones.

In contrast to the results based on vegetation and terrestrial pollen and spores, the MCT results for aquatic pollen and spores show no consistent patterns (Figure 6(d)). This indicates that it is not possible to distinguish Moen’s (1998) vegetation zones based on the composition of the aquatic pollen and spores in our 52 lake-samples.

17.43%LA,NB BN,MB,N, SB

3.69%NB

6.70%MB, SB BN,N

4.40%SB MB

3.20%N BN

adjusted R2: 26.77 % R2 : 35.4 % Error : 0.646 CV Error : 0.838 SE : 0.039

2.5 : n=103.82 : n=112.57 : n=91.72 : n=63.27 : n=111.76 : n=5

LA

(a)

24.99%

LA, MB, NB BN, N, SB

7.45%MB, NB LA

5.78%NB

6.40%SB BN, N

2.52%BN N

adjusted R2: 41.45 % R2 : 47.1 % Error : 0.529 CV Error : 0.69 SE : 0.0552

0.642 : n=9 0.644 : n=11 0.334 : n=5 0.293 : n=6 0.453 : n=100.57 : n=11

(c)

MB

BN, MB, NB, SB LA, N

9.98 : n=36 5.76 : n=16

Error : 0.935 CV Error : 1.09 SE : 0.107

(d)

16.43%

LA, NB BN, MB, N, SB

3.67%LA NB

6.58%MB, SB BN,N

3.86%SB MB

3.45%BN N

adjusted R2: 27.81 % R2 : 34 % Error : 0.66 CV Error : 0.863 SE : 0.0442

1.57 : n=5 2.57 : n=11 1.43 : n=6 2.29 : n=9 2.3 : n=10 2.9 : n=11

(b)

Figure 6. Multivariate classification tree of the vegetation and Moen’s (1998) vegetation zones for (a) plants, (b) plants transformed into pollen and spore types, (c) terrestrial pollen and spores and (d) aquatic pollen and spores. The vegetation zones and r2 for each split are included. The information under each terminal node is the number of sites at each leaf (n), and the relative error (RE). The trees show the hierarchical division of the different data-sets individually. The results also show that it is not possible to separate the vegetation zones on the basis of the composition of the aquatic pollen and spore taxa in the pollen assemblages. BN: boreonemoral; LA: low-alpine; MB: middle-boreal; N: nemoral; NB: northern-boreal; SB: southern-boreal.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 12: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

Felde et al. 995

Comparison of vegetation and pollen compositionThe DCA results show that the vegetation data have a composi-tional turnover of 3.78 SD units, plants as pollen and spore types 2.98 SD, terrestrial pollen and spores 1.34 SD and aquatic pollen and spores 0.82 SD. The DCA plots display the two-dimensional distribution of individual sites in relation to the each other based on the composition for vegetation, plants as pollen and spores, terrestrial pollen and spores and aquatic pollen and spores (Figure 7(a)–(d)). The total inertia is for (1) 3.06, (2) 2.04, (3) 1.10 and (4) 1.55. The first DCA axis in Figure 7(a)–(c) distinguishes sites that are dominated by either deciduous or coniferous forest, while the second DCA axis may be related to vegetation structure as this gradient separates the upper low-alpine and northern-boreal sites from the vegetation zones at lower elevation. The underlying gra-dients for the axes in Figure 7(d) are not known. The results of the aquatic pollen and spores support the MCT results that Moen’s vegetation zones are not reflected by the composition of aquatic pollen and spores (Figure 7(d)).

The results for the comparison of the DCA results based on the Procrustes and Protest analyses of vegetation and terrestrial pol-len and spores, and pollen and spores of plants and terrestrial pol-len and spores show consistent results (Figure 8(a) and (b)). The Procrustes m2 is 0.57 for vegetation versus pollen, and 0.56 for plants as pollen and spore types versus pollen and spore types. The Protest correlations are 0.66 and 0.67, both highly statisti-cally significant (p = 0.001). In general, distances between paired sites are smaller in the nemoral, boreonemoral, southern-boreal

and middle-boreal zones, while the distances increase between paired sites in the northern-boreal and low-alpine zones. Some outliers are observed, but the overall pattern is consistent. In con-trast, comparison between vegetation and aquatic pollen and spores, and pollen and spores of plant species and aquatic pollen and spores show that there is no consistent relationship between the vegetation surrounding the lakes and the aquatic assemblages. The Procrustes m2 values are 0.92 and 0.94, and Protest r values are 0.28 and 0.24. These are not statistically significant (p = 0.071, p = 0.225). When comparing terrestrial pollen and spore composition with aquatic pollen and spores, the results are slightly different. The distances between paired sites are still relatively large, and Procrustes m2 is 0.87 and Protest r is 0.36. However, the Protest test suggests that these relationships are statistically significant (p = 0.003). This contrasts with the result based on the comparison of vegetation composition and aquatic pollen and spores, and suggests that there may be a very weak link between the catchment vegetation as represented by the modern pollen assemblages and the aquatic pollen and spore assemblages.

Co-correspondence between vegetation and pollen compositional patternsCo-CA shows that there is a good match in the compositional pat-tern between vegetation and plants as pollen and spores in the vegetation and terrestrial pollen and spores in the sediment sam-ples from lakes at low and mid-elevations (i.e. nemoral,

−1 0 1 2

−2

−1

01

DC

A 2

BNLA

MBN NB

SB

(a)

−1.0 −0.5 0.0 0.5 1.0 1.5

−1.0 −0.5 0.0 0.5 1.0 1.5

2.0

−1.

0−

0.5

0.0

0.5

1.0

1.5

BNLA

MB

N

NB

SB

(b)

0.60.40.2–0.2–0.6 0.0

0.6

0.4

0.2

–0.2

–0.6

0.0

1.0

0.5

–0.5

–1.0

0.0

DCA 1

DC

A 2

BN

LA

MB

N

NBSB

(c)

DCA 1

BNLA

MBNNB

SB

(d)

Figure 7. Detrended correspondence analysis (DCA) of four compositional data-sets – (a) vegetation, (b) plants in the vegetation converted into pollen and spore types, (c) terrestrial pollen and spore assemblages and (d) aquatic pollen and spore assemblages in the Setesdal valley. Moen’s (1998) vegetation zones are passively displayed in the ordination diagrams and marked with different colours along with their centroids.LA: low-alpine; NB: northern-boreal; MB: middle-boreal; SB: southern-boreal; BN: boreonemoral; N: nemoral.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 13: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

996 The Holocene 24(8)

boreonemoral, southern-boreal and middle-boreal zones), whereas this pattern starts to deviate at higher elevations (i.e. northern-boreal and low-alpine zones; Figure 9(a) and (b)). Since there is a prominent arch effect along the second correspondence analysis axis in the vegetation data, only the first Co-CA axes are used for comparisons. The Co-CA axis 1 for vegetation and ter-restrial pollen and spores captures 43.3% of the variance, while the Co-CA axis 1 for pollen and spores of plants and terrestrial pollen and spores captures 46.4%. The observed pattern also matches the results of the percentage match in the pollen sum for the pollen and spores in the pollen assemblages and the pollen and spores in the surrounding vegetation (Figure 9(c)). The percent-age correspondence in the pollen sum for pollen and spore types in the surface sediments and plants as pollen and spores in the surrounding vegetation is 80% or more in the forested areas, whereas it decreases sharply to 20–50% at the upper northern-boreal sites and low-alpine sites.

The resulting Co-CA species scores are ordered for plants, pollen and spores of plants and terrestrial pollen and spores along the first Co-CA axis (Figure 10). Figure 10 shows that the taxon order is consistent with expectations based on the distribution of

the vegetation zones. Only the most common taxa are included to permit an overview and ease of comparison. The taxa that are considered statistically significant indicators in the indicator spe-cies analysis for certain vegetation zones are marked with differ-ent symbols. Taxa closely related to high elevations have the highest positive values, whereas taxa show a gradual turnover from low-stature vegetation to coniferous forest in the middle, to a dominance of deciduous forest taxa in the lower part of the Co-CA axis.

DiscussionOur results show that it is possible to distinguish the major veg-etation zones (Moen, 1998) in the Setesdal valley based on the vegetational composition around our 52 lakes, on pollen and spores of these plant species around the lakes and on terrestrial pollen and spore assemblages from surface-sediment samples from the lakes. However, the prediction accuracy using our veg-etation data is surprisingly low. The MCT shows that the vegeta-tion zones capture a higher proportion of the variation in the pollen assemblages than in the vegetation data. The prediction

−0.1 0.0 0.1 0.2 0.3

−0.

15−

0.05

0.05

0.15

Dim

ensi

on 2

(a)

−0.1 0.0 0.1 0.2 0.3

−0.15

−0.05

0.05

0.15

(b)

−0.2 −0.1 0.0 0.1 0.2 0.3

−0.

15−

0.05

0.05

0.15

Dim

ensi

on 2

(c)

−0.2 −0.1 0.0 0.1 0.2 0.3−

0.15

−0.

050.

05

(d)

−0.2 −0.1 0.0 0.1 0.2

−0.

100.

000.

10

Dimension 1

Dim

ensi

on 2

(e)

−0.2 −0.1 0.0 0.1 0.2 0.3

−0.

15−

0.05

0.05

0.15

Dimension 1

(f)

Figure 8. Procrustes analysis errors in two-dimensional ordination space between paired sites for the comparison of DCA ordinations (Figure 7) of (a) vegetation and terrestrial pollen and spore assemblages in surface-sediment samples, (b) vegetation vs plants in the vegetation transformed into pollen and spores, (c) plants in the vegetation transformed into pollen and spores vs terrestrial pollen and spores, (d) vegetation vs aquatic pollen and spores, (e) plants in the vegetation transformed into pollen and spores vs. aquatic pollen and spore types, and (f) terrestrial pollen and spore assemblages vs aquatic pollen and spores. The length of the arrows indicate the magnitude of similarity between paired sites with short arrows suggesting high similarity and long arrows suggesting low similarity.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 14: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

Felde et al. 997

accuracy is also higher for the pollen assemblages than for the vegetation data, indicating that the pollen assemblages delimit the different vegetation zones better than the vegetational com-position around our study lakes. When investigating the distribu-tion of sites in the DCA results, both the pollen and vegetation data show that there is a gradual change in composition along the first DCA axis from the nemoral to the low-alpine zones. This can be a plausible explanation for the low prediction accuracy of the vegetational composition, where individuals or patches of warmth-demanding tree species can be observed quite far north

in the valley. Even though a gradual change in composition along the gradient is also observed for the pollen assemblages, the pol-len assemblages from sites within the same vegetation zone are more similar to each other than for the vegetation data.

The most prominent distinction in both the vegetation and pollen data is between sites in the upper and lower part of the valley, observed in the MCT and DCA results. According to Moen (1998), the middle-boreal zone is distinguished from the southern-boreal zone because it contains boreal and alpine spe-cies that tolerate low winter temperatures. This is not strictly true in our case, because the MCT shows that for vegetation, the mid-dle-boreal sites are more similar to the nemoral, boreonemoral and southern-boreal sites, whereas in terms of the pollen assem-blages, the middle-boreal sites are closest to the northern-boreal and low-alpine sites. In general, comparison of the DCA results by Procrustes analysis and Protest randomisation tests confirms that there is a very strong relationship between modern vegeta-tional composition and the modern pollen assemblages. How-ever, the DCA results also show a higher amount of turnover in the vegetational data than in the pollen data, and the match in composition between vegetation and pollen assemblages, and between pollen and spores of plants and pollen assemblages starts to deviate at higher elevations. A few outliers or individual sites with larger Procrustes sum-of-squared distances are observed in the lower part of the gradient. It is possible that there have been changes in the vegetation around some of the lakes in the last 15 years. However, this seems to have little effect on the overall results. The largest differences between the composition of the pollen assemblages and the vegetation corresponds with where the vegetation becomes more open and consists of less dense mountain-birch forest and dwarf-shrub heath and grass-land vegetation. The Co-CA results (Figure 9(a) and (b)) also reveal that the pollen data have an almost non-linear pattern in composition along the elevational gradient, indicating that the pollen assemblages from the low-alpine zone have many taxa in common with vegetation zones at lower elevations. This is because of the influence of long-distance (‘extra-regional’) dis-persal of pollen from lower down the valley. Different tree pollen taxa, for example, Ulmus, Alnus and Fraxinus, are observed in low-alpine surface samples, even though it is many kilometres to the nearest trees. These false presences of tree pollen taxa lead to a lower apparent turnover in the pollen data in the upper vegeta-tion zones because the pollen composition is more similar to assemblages from the lower zones than the actual vegetation. Similar observations have been found using pollen traps in simi-lar vegetation zones to investigate annual pollen and macrofossil influx across the tree-line (Birks and Bjune, 2010). This raises concerns about the reliability of using pollen assemblages to study tree-line dynamics. These concerns are also supported by the low percentage match in the pollen sums between pollen and spore types in the vegetation and pollen and spore types in the surface-sediment samples for the northern-boreal and low-alpine sites (Figure 9(c)).

Even though the low-alpine and northern-boreal sites seem to contain more extra-regional pollen and there is less compositional turnover in the pollen data than in the vegetation data, pollen spectra from the low-alpine zone can still be distinguished rea-sonably well from pollen assemblages from other zones. This is because of the variation in the proportions among the different taxa. The low-alpine pollen assemblages and vegetation are domi-nated by grasses, low-growing herbs, heaths and ferns. Even though pollen of tree taxa are present, their frequencies are gener-ally lower in these sites. Spruce (P. abies) forest is dominant at middle-boreal sites, while pine (P. sylvestris) is the most domi-nant tree in the southern-boreal sites. The boreonemoral and nemoral sites are mainly dominated by warmth-demanding decid-uous trees, shrubs and herbs. This general trend is also observed

(a)

(b)

(c)

−1

0

1

2plantspollen

0 200 400 600 800 1000 1200

−1

0

1

2plants as pollenpollen

Elevation (m)

40

60

80

% m

atch

in p

olle

nsum

LANBMBSBBNN

Co-CA1

Co-CA1

Figure 9. Axis 1 of the co-correspondence analysis (Co-CA) for (a) vegetational composition (plants) vs terrestrial pollen composition, (b) plants as pollen and spores in the vegetation vs pollen and spore composition in the surface-sediment samples, and (c) the percentage match in the pollen sum between the presence of plants as pollen and spore types in the surrounding vegetation and pollen and spore types in the surface-sediment samples along the elevation gradient. Blue squares represent the Co-CA1 of the pollen assemblages in the sediment samples, and green triangles represent the composition of plants or plants as pollen and spores in the vegetation. Co-CA axis 1 for pollen assemblages and vegetation captures 43.3% of the total variance in the data, while Co-CA axis 1 captures 46.4% of the variance for pollen assemblages and pollen and spores of plants in the vegetation. The different symbols in (c) represent the different vegetation zones. A smoothing spline with 10 degrees of freedom is fitted to plot (c).LA: low-alpine; NB: northern-boreal; MB: middle-boreal; SB: southern-boreal; BN: boreonemoral; N: nemoral.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 15: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

998 The Holocene 24(8)

in the vegetation data and shows how well the pollen composition reflects the underlying vegetation pattern.

Differences in composition between the vegetation and the pollen assemblages are also affected by factors such as differen-tial pollen production and dispersal, differential preservation, growing distance from the lake, lake size and vegetation structure in the pollen-source area (Bunting et al., 2004; Davis, 2000;

Jackson, 1990). The extent of the pollen-source area is largely determined by lake size (Jackson, 1990; Janssen, 1966, 1981). In this study, the lakes are considered to be small-to-medium sized (1.25–97.55 ha) and each lake should primarily reflect a regional pollen-source area (sensu Janssen, 1973, 1981). Since only the local and extra-local vegetation (sensu Janssen, 1966, 1973, 1981) around each site was recorded, the spatial scales of the

Figure 10. Species scores on the co-correspondence analysis axis 1 for vegetation (left), terrestrial pollen assemblages from surface-sediment samples (middle), and plants in the vegetation transformed to pollen and spore types (right). Taxa that are also statistically significant indicators (IndVal) for particular vegetation zones are indicated by a small symbol – see the legend for the six vegetation zones.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 16: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

Felde et al. 999

vegetation data and the pollen data may differ to some degree. This raises the questions of whether the vegetation data really rep-resent the regional vegetation, and whether our vegetation sam-pling is adequate. Based on the results of this study, it seems that our sampling effort for the vegetation is adequate overall, particu-larly when the vegetation is homogenous. In such vegetation, the probability of adding new species decreases markedly with dis-tance from the lake and with sampling effort. The consistent results between the vegetation composition and the pollen assem-blages indicate that the pollen and spores in the sediment samples reflect the vegetation in the area sampled. However, the relevant pollen-source area can also be affected by vegetation structure (Bunting et al., 2004). Dense forested sites are probably less affected by distant extra-regional dispersal of pollen types, and it is natural to assume that a higher proportion of the pollen sum represents the local and extra-local vegetation (sensu Janssen, 1966, 1981) at the site, because the canopy and trunk space in the forest will function as a filter for pollen coming from increasing distances from the lake (Jacobson and Bradshaw, 1981; Tauber, 1965). When the forest is less dense, this may increase the poten-tial pollen-source area, and more pollen from the regional vegeta-tion (sensu Janssen, 1966) and beyond can be deposited in the lake. This explanation is also consistent with the close match in composition between vegetation and pollen assemblages from sites surrounded by forest. This match is worse for sites with open vegetation.

Our results also show that it is not possible to distinguish the different vegetation zones on the basis of the composition of aquatic pollen and spores, and there is no relationship between vegetation composition around the lakes and the aquatic pollen and spore assemblages. The exception is the weak statistical link between the terrestrial pollen assemblages and the aquatic pollen and spores. When investigating past vegetation changes, aquatic pollen richness usually shows an increase when there are changes in forest composition (e.g. mixed deciduous forest with Ulmus, Fraxinus and Corylus (Eide et al., 2006)) that may facilitate soil development in response to warmer temperatures. However, our results suggest that the aquatic macrophyte composition is prob-ably more affected by local factors at each site such as water tem-perature, pH, water level and geochemical conditions within the lake rather than by the surrounding vegetation. It is also important to bear in mind that aquatic pollen is often less well preserved than plant macrofossils (Zhao et al., 2006), and macrofossils were not analysed in these surface samples. Hence, the aquatic pollen and spore data in this study may be insufficient to draw any reli-able conclusions about the relationships between aquatic plant composition and major vegetation zones. More studies are clearly needed on aquatic–terrestrial biotic relationships.

ConclusionMoen’s (1998) major vegetation zones in Setesdal can be distin-guished by the vegetational composition around our lakes and by terrestrial pollen assemblages from these lakes. Surprisingly, pol-len composition reflects the major zones better than the vegeta-tional composition. A plausible explanation is that pollen reflects a regional vegetation scale, while our vegetation data may record more local and extra-local variation. The direct comparison of paired sites shows that the pattern in composition between pollen assemblages and vegetation matches well for the nemoral, boreonemoral, southern-boreal and middle-boreal zones, but starts to deviate when the vegetation becomes more open in the northern-boreal and low-alpine zones. Increases in potential pol-len-source area and a resulting increase in long-distance extra-regional dispersed pollen reaching the lakes occur when the vegetation becomes more open. However, despite these biases, it is possible to distinguish the northern-boreal and low-alpine

zones based on modern pollen assemblages. In contrast, the com-position of aquatic pollen and spore types does not reflect the major vegetation zones, which indicates that the surrounding regional vegetation may have little or no direct effect on the aquatic flora of the lakes.

AcknowledgementsWe thank Tone Martinessen, Siri Skoglund and Ildikó Orbán for help with the botanical surveys, and Ewan Shilland, Nancy Bi-gelow, Hilary Birks, Annette Stavseth Furnes, Tim Allott, Cath-erine Dalton and Lindsey Allott for help with lake sampling. We are grateful to Cathy Jenks for editorial help. We also thank Thomas Giesecke and an anonymous reviewer for helpful com-ments on an earlier version of the manuscript.

FundingThis work has been supported by Miljø 2015 LAND: ‘Terrestrial biodiversity through time – novel methods and their applications’ (203804/E40) funded by NFR, and from the Olaf Grolle Olsen’s Legacy to the University of Bergen with the addition of the request of Miranda Bødtker. The collection of the pollen data was supported by the NFR grant to ‘Holocene climatic history and ecological impacts in Setesdal, southern Norway: a quantitative pollen-analytical study’ (110486/720).

ReferencesAndersen ST (1967) Tree-pollen rain in a mixed deciduous for-

est in south Jutland (Denmark). Review of Palaeobotany and Palynology 3: 267–275.

Andersen ST (1970) The relative pollen productivity and pollen representation of north European trees, and correction factors for tree pollen spectra. Danmarks Geologiske Undersoegelse, Raekke 2 (Afhandlinger) 96: 1–99.

Bellard C, Bertelsmeier C, Leadley P et al. (2012) Impacts of cli-mate change on the future of biodiversity. Ecology Letters 15: 365–377.

Berglund BE and Ralska-Jasiewiczowa M (1986) Pollen analy-sis and pollen diagrams. In: Berglund BE (ed.) Handbook of Holocene Palaeoecology and Palaeohydrology. Chichester: Wiley, pp. 455–484.

Beug H-J (2004) Leitfaden der Pollenbestimmung für Mitteleu-ropa and angrezende Gebiete. München: Verlag Dr. Friedrich Pfeil.

Birks HH and Bjune AE (2010) Can we detect a west Norwegian tree line from modern samples of plant remains and pollen? Results from the DOORMAT project. Vegetation History and Archaeobotany 19: 325–340.

Birks HH, Battarbee RW and Birks HJB (2000) The development of the aquatic ecosystem at Kråkenes Lake, western Norway, during the late glacial and early Holocene – A synthesis. Jour-nal of Paleolimnology 23: 91–114.

Birks HJB (1973a) Modern pollen rain studies in some arctic and alpine environments. In: Birks HJB and West RG (eds) Qua-ternary Plant Ecology. Oxford: Blackwell Scientific Publica-tions, pp. 143-168.

Birks HJB (1973b) Past and Present Vegetation of the Isle of Skye – A Palaeoecological Study. Cambridge: Cambridge University Press.

Birks HJB (2007) Estimating the amount of compositional change in late-Quaternary pollen-stratigraphical data. Vegetation His-tory and Archaeobotany 16: 197–202.

Birks HJB and Gordon AD (1985) Numerical Methods in Quater-nary Pollen Analysis. London: Academic Press.

Birks HJB, Webb T and Berti AA (1975) Numerical analysis of pollen samples from central Canada: A comparison of meth-ods. Review of Palaeobotany and Palynology 20: 133–169.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 17: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

1000 The Holocene 24(8)

Bitušík P, Kubovcik V, Stefkova E et al. (2009) Subfossil diatoms and chironomids along an altitudinal gradient in the High Tatra Mountain lakes: A multi-proxy record of past environ-mental trends. Hydrobiologia 631: 65–85.

Borcard D, Gillet F and Legendre P (2011) Numerical Ecology with R. New York: Springer.

Bradshaw RHW (1981) Modern pollen-representation factors for woods in south-east England. Journal of Ecology 69: 45–70.

Brooks SJ (2003) Chironomid analysis to interpret and quantify Holocene climate change. In Mackay A, Battarbee RW, Birks HJB et al. (eds) Global Change in the Holocene. London: Arnold, pp. 328–341.

Broström A, Nielsen A, Gaillard M-J et al. (2008) Pollen produc-tivity estimates of key European plant taxa for quantitative reconstruction of past vegetation: A review. Vegetation His-tory and Archaeobotany 17: 461–478.

Bunting MJ, Armitage R, Binney HA et al. (2005) Estimates of ‘relative pollen productivity’ and ‘relevant source area of pol-len’ for major tree taxa in two Norfolk (UK) woodlands. The Holocene 15: 459–465.

Bunting MJ, Gaillard M-J, Sugita S et al. (2004) Vegetation struc-ture and pollen source area. The Holocene 14: 651–660.

Calcote R (1995) Pollen source area and pollen productivity – Evidence from forest hollows. Journal of Ecology 83: 591–602.

Davidson TA, Bennion H, Jeppesen E et al. (2011) The role of cla-docerans in tracking long-term change in shallow lake trophic status. Hydrobiologia 676: 299–315.

Davis BAS, Zanon M, Collins P et al. (2013) The European Mod-ern Pollen Database (EMPD) project. Vegetation History and Archaeobotany 22: 521–530.

Davis MB (2000) Palynology after Y2K – Understanding the source area of pollen in sediments. Annual Review of Earth and Planetary Sciences 28: 1–18.

De’ath G (2002) Multivariate regression trees: A new technique for modeling species-environment relationships. Ecology 83: 1105–1117.

De’ath G (2013) mvpart: Multivariate partitioning (R package version 1.6–1). Available at: http://CRAN.R-project.org/package=mvpart.

Digby PGN and Kempton RA (1987) Multivariate Analysis of Ecological Communities. New York and London: Chapman & Hall.

Djamali M, de Beaulieu JL, Campagne P et al. (2009) Modern pollen rain-vegetation relationships along a forest-steppe transect in the Golestan National Park, NE Iran. Review of Palaeobotany and Palynology 153: 272–281.

Eide W, Birks HH, Bigelow N et al. (2006) Holocene forest devel-opment along the Setesdal valley, southern Norway, recon-structed from macrofossil and pollen evidence. Vegetation History and Archaeobotany 15: 65–85.

Faegri K and Iversen I (1964) Textbook of Pollen Analysis. 2nd Edition. Oxford: Blackwell Scientific Publications.

Faegri K, Kaland PE and Krzywinski K (1989) Textbook of Pollen Analysis. 4th Revised Edition. Chichester: John Wiley & Sons.

Fall PL (2012) Modern vegetation, pollen and climate relation-ships on the Mediterranean island of Cyprus. Review of Pal-aeobotany and Palynology 185: 79–92.

Felde VA, Birks HJB, Peglar S et al. (2012) Vascular plants and their pollen- or spore-type in Norway. Available at: http://www.uib.no/rg/EECRG/artikler/2012/03/vascular-plants-and-their-pollen-or-spore-type-in-norway.

Felde VA, Bjune AE, Grytnes JA et al. (in press) Comparing mod-ern pollen assemblages with major vegetation-landform types using novel numerical methods. Review of Paleobotany and Palynology.

Fletcher M-S and Thomas I (2007) Modern pollen–vegetation relationships in western Tasmania, Australia. Review of Pal-aeobotany and Palynology 146: 146–168.

Gower JC (1975) Generalized Procrustes analysis. Psychometrika 40: 33–51.

Grimm EC (1990) TILIA and TILIA GRAPH: PC spreadsheet and graphics software for pollen data. INQUA Commission for the Study of the Holocene: Working group on data-handling meth-ods. Newsletter 4: 5–7.

Grimm EC (2004) TGView Version 2.0.2. Springfield, IL: Research and Collection Center, Illinois State Museum.

Herzschuh U and Birks HJB (2010) Evaluating the indicator value of Tibetan pollen taxa for modern vegetation and climate. Review of Palaeobotany and Palynology 160: 197–208.

Hill MO and Gauch HG Jr (1980) Detrended correspondence analysis: An improved ordination technique. Vegetatio 42: 47–58.

Hjelle KL (1998) Herb pollen representation in surface moss sam-ples from mown meadows and pastures in western Norway. Vegetation History and Archaeobotany 7: 79–96.

Jackson DA (1995) PROTEST: A PROcrustean randomization TEST of community environment concordance. Ecoscience 2: 297–303.

Jackson ST (1990) Pollen source area and representation in small lakes of the northeastern United States. Review of Palaeobot-any and Palynology 63: 53–76.

Jacobson GL Jr and Bradshaw RHW (1981) The selection of sites for paleovegetational studies. Quaternary Research 16: 80–96.

Janssen CR (1966) Recent pollen spectra from the deciduous and coniferous-deciduous forests of northeastern Minnesota: A study in pollen dispersal. Ecology 47: 804–825.

Janssen CR (1973) Local and regional pollen deposition. In: Birks HJB and West RG (eds) Quaternary Plant Ecology. Oxford: Blackwell Scientific Publications, pp. 31–42.

Janssen CR (1981) On the reconstruction of past vegetation by pollen analysis: A review. Proceedings of the IVth Interna-tional Palynological Conference, Lucknow (1976–1977) 3: 163–172.

Legendre P and Legendre L (2012) Numerical Ecology. 3rd Edi-tion. Amsterdam: Elsevier.

Lichti-Federovich S and Ritchie JC (1968) Recent pollen assem-blages from the western interior of Canada. Review of Palaeo-botany and Palynology 7: 297–344.

Lid J and Lid DT (2005) Norsk Flora. Oslo: Det Norske Sam-laget.

Lüder B (2007) The potential of non-alpine lakes for quantita-tive palaeotemperature reconstructions based on subfossil chironomids: A comparative palaeolimnological study from southern Norway. PhD Thesis, University of Bremen.

Meltsov V, Poska A, Odgaard BV et al. (2011) Palynological rich-ness and pollen sample evenness in relation to local floristic diversity in southern Estonia. Review of Palaeobotany and Palynology 166: 344–351.

Moen A (1998) Vegetasjonsatlas for Norge. Hønefoss: Statens kartverk.

Moore PD, Webb JA and Collinson ME (1991) Pollen Analysis. 2nd Edition. Oxford: Blackwell Scientific Publications.

Odgaard BV (1994) The Holocene vegetation history of northern West Jutland, Denmark. Opera Botanica 123: 1–171.

Oksanen J, Blanchet GF, Kindt R et al. (2013) vegan: Commu-nity ecology package (R package version 2.0–9). Available at: http://vegan.r-forge.r-project.org/.

Ortuño T, Ledru M-P, Cheddadi R et al. (2011) Modern pollen rain, vegetation and climate in Bolivian ecoregions. Review of Palaeobotany and Palynology 165: 61–74.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from

Page 18: The relationship between vegetation composition, vegetation zones and modern pollen assemblages in Setesdal, southern Norway

Felde et al. 1001

Ouellette M-H and Legendre P (2013) MVPARTwrap: Addi-tional features for package mvpart (R package ver-sion 0.1–9.2). Available at: http://CRAN.R-project.org/package=MVPARTwrap.

Panizzo VN, Jones VJ, Birks HJB et al. (2008) A multiproxy palaeolimnological investigation of Holocene environmental change, between c. 10700 and 7200 years BP, at Holebudalen, southern Norway. The Holocene 18: 805–817.

Pelánková B, Kuneš P, Chytrý M et al. (2008) The relationships of modern pollen spectra to vegetation and climate along a steppe–forest–tundra transition in southern Siberia, explored by decision trees. The Holocene 18: 1259–1271.

Peres-Neto P and Jackson D (2001) How well do multivariate data sets match? The advantages of a Procrustean superimposition approach over the Mantel test. Oecologia 129: 169–178.

Prentice IC (1985) Pollen representation, source area, and basin size: Toward a unified theory of pollen analysis. Quaternary Research 23: 76–86.

Punt W, Blackmore S, Clarke GCS et al. (1976–1995) The North-west European Pollen Flora. Vol. I (1976); Vol. 2(1980); Vol. 3 (1981); Vol. 4 (1984); Vol. 5 (1988); Vol. 6 (1991); Vol. 7 (1995). Amsterdam: Elsevier.

R Core Team (2012) R: A Language and Environment for Statisti-cal Computing. Vienna: R Foundation for Statistical Comput-ing. Available at: http://www.R-project.org/.

Renberg I and Hansson H (2008) The HTH sediment corer. Jour-nal of Paleolimnology 40: 655–659.

Schaffers AP, Raemakers IP, Sykora KV et al. (2008) Arthropod assemblages are best predicted by plant species composition. Ecology 89: 782–794.

Selsing L (2010) Mennesker og natur i fjellet i Sør-Norge etter siste istid med hovedvekt på mesolitikum. Stavanger: Arkeolo-gisk museum, Universitetet i Stavanger.

Simpson GL (2009) cocorresp: Co-correspondence analysis ordi-nation methods. R package version 0.2–0. Available at: http://cran.r-project.org/package=cocorresp.

Simpson GL and Birks HJB (2012) Statistical learning in palaeo-limnology. In: Birks HJB, Lotter AF, Juggins S et al. (eds) Tracking Environmental Change Using Lake Sediments, Vol-ume 5: Data Handling and Numerical Techniques. Dordrecht: Springer, pp. 249–327.

Sugita S (1993) A model of pollen source area for an entire lake surface. Quaternary Research 39: 239–244.

Sugita S (1994) Pollen representation of vegetation in Quaternary sediments: Theory and method in patchy vegetation. Journal of Ecology 82: 881–897.

Sugita S (2007a) Theory of quantitative reconstruction of vegeta-tion. I. Pollen from large sites REVEALS regional vegetation. The Holocene 17: 229–241.

Sugita S (2007b) Theory of quantitative reconstruction of vegeta-tion. II. All you need is LOVE. The Holocene 17: 243–257.

Tauber H (1965) Differential pollen dispersion and the interpre-tation of pollen diagrams. Danmarks Geologiske Undersoe-gelse, Raekke 2 (Afhandlinger) 89: 1–69.

Ter Braak CJF and Prentice IC (1988) A theory of gradient analy-sis. Advances in Ecology Research 18: 271–313.

Ter Braak CJF and Schaffers AP (2004) Co-correspondence anal-ysis: A new ordination method to relate two community com-positions. Ecology 85: 834–846.

Väliranta M, Kultti S, Nyman M et al. (2005) Holocene develop-ment of aquatic vegetation in shallow Lake Njargajavri, Finn-ish Lapland, with evidence of water-level fluctuations and drying. Journal of Paleolimnology 34: 203–215.

Wei H-C, Ma H-Z, Zheng Z et al. (2011) Modern pollen assem-blages of surface samples and their relationships to vegeta-tion and climate in the northeastern Qinghai-Tibetan Plateau, China. Review of Palaeobotany and Palynology 163: 237–246.

Willis KJ, Bailey RM, Bhagwat SA et al. (2010) Biodiversity baselines, thresholds and resilience: Testing predictions and assumptions using palaeoecological data. Trends in Ecology and Evolution 25: 583–591.

Wright HE, McAndrews JH and van Zeist W (1967) Modern pol-len rain in western Iran, and its relation to plant geography and Quaternary vegetational history. Journal of Ecology 55: 415–443.

Zhao Y, Li F, Hou Y et al. (2012) Surface pollen and its rela-tionships with modern vegetation and climate on the Loess Plateau and surrounding deserts in China. Review of Palaeo-botany and Palynology 181: 47–53.

Zhao Y, Sayer C, Birks HH et al. (2006) Spatial representation of aquatic vegetation by macrofossils and pollen in a small and shallow lake. Journal of Paleolimnology 35: 335–350.

at TEXAS SOUTHERN UNIVERSITY on October 19, 2014hol.sagepub.comDownloaded from