Estimation of lichen biomass with emphasis on reindeer winter pastures at Hardangervidda , S Norway

Quanti!cation of lichen abundance is important for management of reindeer populations. We measured dry lichen biomass in 876 microplots (16.5 cm × 16.5 cm) systematically sampled within 219 vegetation plots (2 m × 2 m) from 7 di#erent areas in S Norway. Lichen biomass was quanti!ed as: (a) dry weight in g m, (b) lichen height in cm, (c) lichen cover, and (d) lichen volume (lichen height × lichen cover). Lichen biomass decreased with increasing precipitation and increasing altitude. On local scale, the variation in lichen biomass varied strongly with snow conditions. $e grazed parts of Hardangervidda had in general a low average lichen biomass (an average mostly lower than 150 g m). Lichen biomass was much higher in area where reindeer migration was interferred by human activity. Lichen height and lichen volume were strongly linearily correlated with dry lichen biomass. $ese proxy methods may therefore be used to predict lichen biomass, but deviations from exact measurements should be expected.


Introduction
Lichen heaths are the most important natural winter pasture resources for both wild and semi-domesticated reindeer (Rangifer tarandus), and the most abundant generas are al .,1979),given su cient availability.Hardangervidda is a mountain plateau which has the largest population of wild reindeer in Europe and Norway has therefore an international obligation to manage this population.Estimation of total lichen abundance available for reindeer grazing during the winter is essential when the carrying capacity of an area is determined.e area used by the reindeer is approximately 8000 km 2 , but estimations of the total area considered to be available lichen heaths during the winter show highly diverging results, ranging from 431 to more than 2100 km 2 in di erent studies (Bjerketvedt et al., 2012).Previous measurements of lichen abundances based on lichen cover or lichen volume measurements in the Hardangervidda area show highly diverging results (Gaare et al.,2005;Falldorf et al., 2014).
e method used by Falldorf et al. 2014 was veri ed by in-situ measurements of lichen volume combined by robust statistics and should therefore be regarded as much more accurate than the estimates made by Gaare et al. (2005).
e reindeer population has varied strongly during the last decades, but a sustainable winter population has been assumed to be between 9000 -12000 depending on the quality of the winter pastures.
An estimate of the carrying capacity of Hardangervidda based on available winter pastures was rst described by A. Tveitnes (1980).His calculation was, however based on seven assumptions which have rarely been mentioned when referring his paper (c.fBjerketvedt, 2013).
Measurement of lichen biomass (LB) dry weight is therefore an essential component of ecological and reindeer management studies in alpine areas, but often avoided because it is destructive, laborious, and time-consuming (Moen et al.,2007).In a destructive sampling strategy, samples are collected within speci c vegetation types, weighted after drying and the lichen dry weight is mostly given as g m -2 .
Lichen dry biomass has previously only been estimated from one small, fenced area at Hardangervidda ca 40 years ago (Kjelvik & Kärenlampi, 1975).Here the average LB was 380 g m -2 , and the annual lichen production was estimated to 0.23 g lichen g -1 year -1 (Kjelvik, 1978).
Four di erent strategies for not-destructive sampling have previously been used (Kumpula et al.,2000;Moen et al.,2007;Kastdalen L., 2011, Falldorf et al., 2014): (a) Lichen cover estimated in percentage cover, (b) average lichen podetia height measured in cm, (c) volume estimation calculated as lichen cover x lichen height, and (d) cover estimation based on presence/absence data with a strict criterion of 100 % lichen cover.ese proxy methods can never give exact estimates of the actual lichen biomasses, and statistical testing of the relationships between proxy data and lichen biomass data are therefore necessary.
Di erent methods of estimating LB from ground cover and lichen thallus heights of four common lichen species have previously been compared (Moen et al.,2007).It was found that di erent methods gave mostly similar results with strong linear relationships between LB and mean thallus height, but average thallus heights within the plots were found to explain the variation in LB as well as lichen volume.In northern Finland, dry lichen biomass and other proxies were measured and it was found that LB was best explained by a quadratic relationship to lichen volume (Kumpula et al.,2000).Estimated LB was between 260-280 as g m -2 (max 700 g m -2 ) in N Finland.
It is well known that "snow conditions" (thickness or duration) is a major determinant for lichen heath development (Dahl, 1957;Walker et al.,2001;Vistnes & Nellemann, 2008;Odland & Munkejord, 2008a), however, quantitative relations between LB and snow conditions have not been investigated.In comparative studies it is essential that data from sites with approximately the same snow conditions are compared.e amount of snow may vary considerably between years, but due to snow drift the uneven distribution of snow is repeated every year and the snow melt pattern is about the same (Gjaerevoll, 1956).Relative estimates of snow layer duration for di erent plant communities may be calculated by the use of plants as indicators (Odland & Munkejord, 2008a) by the use of Weighted averages for plant communities or vegetation plots (WA Si ). is method takes into account both occurring species and their abundances.Species optima along gradients from chionophobous to chionophilous communities have been quantied by giving species with signi cant responses along the snow layer duration relative snow duration values ranging from 1 to 9. Strictly chionophobous communities will have a WA Si value close to 1, while strictly chionophilous communities have a WA Si value close to 9. In average, during a "normal year" the di erence between each step is a 11-12 days di erence in snow layer duration (Odland & Munkejord, 2008b).
e most exposed sites (with WA Si values between 1 and 2) will normally have no or a very thin snow cover, while the most chionophilous communities will not be melted out before late July.Consequently, the LB in particular sites (vegetation type) should be related to their snow conditions when di erent areas or sites are compared.
e present study aims at answering the following questions: sured by snow indicator values related to the variation in LB? grazed and not-grazed areas as measured on data sampled on sites with approximately the same snow conditions?proxies for LB estimation?-tween oceanic and continental study areas as measured by average annual precipitation?
increasing altitude as measured from sites with the same "snow conditions"?

Study areas and sampling methods
Nine study areas were selected (  (Jordhøy & Strand, 2009).Data from other areas were included to compare the data from Hardangervidda with LB data from other areas in S Norway.e investigation was based on oristic and environmental data from homogenous stands where lichen abundance data have been systematically sampled.A homogenous stand is de ned as an area of vegetation that shows no obvious variation in the spatial distribution or relative abundance of at least the major species present and that shows small substrate variations.Selection of homogenous (Gjaerevoll, 1956;Dahl, 1957;Diekmann, 1995) plots is essential when the distribution of plants is related to environmental variables sampled from the actual vegetation plots (Diekmann, 2003).
As the lichen biomass varies strongly with snow conditions, the sampling sites were selected along gradients from exposed ridges to snow-beds.In each area sites at di erent elevations were also studied.Emphasis has been on low alpine sites from the most exposed sites to snow demanding dwarf shrub heaths and early graminoid dominated early snow-beds.Wetland vegetation was omitted because the lichen cover is there generally sparse.
Total cover of all vascular plants and the cover of the most abundant mosses and lichens within the plots were estimated in percentage.Within the plots, four smaller plots (16.5 ×16.5 cm) 60 cm from the corners were systematically selected along the diagonals in the quadrat where total lichen cover estimated in percentage and height of the lichen podetia was measured in cm.All lichens were sampled and brought to the laboratory and the average lichen heights and volumes were calculated for each plot.Lichen samples from the four sample plots were mixed and brought to the laboratory to be carefully sorted, and dried for 24 h at 105 o C. en it was weighted and an average LB for the whole plot was calculated as g m -2 .In the thickest lichen mats, 0.5 -1.0 cm of the basal part of the podetia were often decaying, and only the living parts were measured.
Weighted averaged snow indexes (WA Si ) were calculated for each 2 × 2 m plot based on plants as snow indicators (Odland & Munkejord, 2008b) occurring in the plots.e weighted average method takes into account the abundances of the di erent species and calculated according to the following equation: Wa Si is the weighted average snow indicator value for the actual plot, X 1 -n are species abundance values, and SI 1-n are the actual snow indicator values.

Numerical analyses
e samples were classi ed by the TWINS-PAN program (Hill, 1979), where six pseudospecies cut-levels (1 -5 -10 -30 -50 -70) and ve indicator species were selected.Eleven types of vegetation were selected for further analyses, primarily based on their separation along the two main ordination axes.Detrended Correspondence Analysis (DCA) analyzed by the use of CANOCO 5 (ter Braak & Smilauer, 2012) was applied to the oristic data to es- timate the oristic turnover or compositional change along the main gradient as assessed by standard deviation (SD) units of turnover (detrending by segments, non-linear rescaling, and no down-weighting of rare species).Species abundance data measured as percent cover were square-root transformed.Lumped taxa (some bryophytes and lichens identi ed to genus only) were deleted in the statistical analyses.e environmental/explanatory variables were log-transformed in the DCA analyses.
Taxa occurrences and abundance (SOA) in the di erent TWINSPAN community groups are given in percentage, calculated according to the formula given in (Odland et al.,1990).Statistical analyses were performed by use of the MINITAB program.

Floristic and environmental gradients
e vegetation plots include a large oristic variation as shown in Figure 1.DCA axis 1 describes a gradient from exposed sites with species such as Alectoria spp., Flavocetraria spp., and Arctous alpinus in the left part and meadows with species such as Geranium sylvaticum and Phleum alpinum in the right part.DCA axis 2 represents a gradient from moist low alpine communities with Salix lapponum and S. glauca in the upper part, and snow-bed species mostly at high altitudes in the lower part (Salix herbacea, Diphasium alpinum, Harrimanella hypnoides, Luzula confusa and Juncus tri dus).Most of sampled explanatory variables were closely correlated with the oristic gradients as shown in Table 2. e eigenvalues/ gradient length of DCA axes 1, 2, and 3 were 0.457/4.35,0.326/3.83,and 0.21/2.74respectively.e main oristic gradient (DCA axis 1) was strongly correlated with snow indicator values and total cover of lichens.DCA axis 2 is best correlated with altitude, and DCA axis 3 with total bryophyte cover.Plots with high LB have a small distribution in the DCA ordina-99 tion diagram (approximately 2 SD units both on axis 1 and 2), and are primarily associated with low annual precipitation, low WA Si values, and relatively low altitudes (Table 2).Eleven separated types (clusters) were selected to be compared with previously described vegetation types.SOA values for the di erent species in the di erent types and also average values for WA Si , LB, CoL, CoB, LH and lichen volume were calculated for each type (Table 3).
LB was much higher in type B and D as compared to the other types.Type A represents highly exposed sites at high altitudes (mainly the Middle alpine zone).Type C represents less exposed sites but at relatively high altitudes (upper Low alpine and Middle alpine zone).Types E, F, G, H, I , and J represents lee-side communities which are dependent on a stable and relatively thick and longlasting snow cover, while type K represents an early snow-bed community.Within some of the vegetation types, plots were both grazed and not-grazed, and therefore average values were calculated for both groups.In type B and D no plots were grazed, while in type A and C both grazed and not grazed plots were grouped together.In type A and C, LB in the grazed areas were around 50 % of the LB in not grazed areas.e differences in LB between grazed and not grazed plots within type E and F were smaller.

Lichen abundance estimations
LB and di erent proxies in vegetation types with or without grazing and in di ererent areas are shown in Table 3 and 4. In all study areas the standard deviations of the means were high which is a result of variations in snow conditions between the plots.High LB were also found in dry, not grazed Pinus sylvestris vegetation in the Tunhovd area where Cladonia stellaris had developed thick mats.Relationships between measured LB and the di erent proxies for LB estimation are shown in Figure 3. e Lowess smoothers show that the relationship between CoL and LB is very di erent to the trends between LB, LH, and LH × CoL.All three proxies show signi cant linear relationships, but lichen volume (LH × CoL) was best correlated with measured LB.Lichen height (LH) also explained LB well, but there were several deviating plots.Lichen cover (CoL), however, was poorly linearily related to the LB measurements.Based on these data, LB may be estimated on the proxy data by the following equations: Eq.1 LB = -70.8+ 1.5 × (LH × CoL), (R 2 = 74.1% , P < 0.0001) Eq.2 LB = -170.7 + 113.4 × LH, (R 2 = 52.6%, p <0.0001) Figure 2. e relative di erences in lichen biomass according to the plot positions along DCA axes 1 and 2 (cf. Figure 1).

Variation in LB along snow duration gradients
Variations in LB and the three di erent proxies along gradients in the WA Si show poor linear trends (Figure 4).LB higher than 500 g m -2 were mainly found in plots with WA Si values lower than 3.0, and the highest LB values were found in plots with WA Si values around 2.0, decreasing toward the most exposed sites (WA Si values below 1.7).Similar patterns were also shown for lichen volume and lichen height.Lichen cover (mainly dominated by Cetraria islandica), however, could be higher than 75 % in plots with WA Si around 5.0 (i.e.snow-bed communities).
Estimation of available lichen resources for reindeer during the winter have often been based on average data from the most exposed sites (Figure 5).erefore, average values for LB from the di erent areas where WA Si was lower than 3.0 were calculated, and these results are given in Table 1.In continental areas without reindeer grazing (Vågå, Iming ell E, and Tron ell), the average LB was higher than 700 g m -2 , while in the grazed areas LB was mostly below 150 g m -2 .e oceanic areas, also with little or no grazing (Haukeli ell and Suleskard) have average LB below 250 g m -2 .A quadratic regression analyses on plots with WA Si values between 3.0 and 1.7 showed a trend in LB as measured on data from continental, ungrazed areas with a maximum LB at a WA Si value of 2.

Regional di erences in LB
ere was a strong signi cant general decrease in LB and CoL (lichen cover) from oceanic areas (1200 mm at Haukeli) to a strongly continental area (370 mm at Vågå), but in all areas there were major variations in LB according to snow conditions.Regression analyses gave the following equations: Eq. 5 LH × CoL = 554.5 -0.42 × Precipitation, (R 2 = 18.5, p < 0.0001) Eq. 6 LB = 98.3 -0.07 × Precipitation, (R 2 = 37.4, p < 0.0001) A regression analysis showed that there was a general decrease in LB with increasing altitude.All plots (both grazed and ungrazed) from the continental areas where WA Si < 3.0 gave the following equation: Eq. 7 LB = 2420 -1.604 × Altitude, (R 2 = 16.7%, p < 0.0001) e main results of the study showed that the variation in average LB between the areas was strongly in uenced both by altitude, snow condition, reindeer grazing, and annual precipitation.ree main areas may be separated: oceanic areas with annual precipitation higher than 1200 mm where LB generally was below 200 g m 2 even where reindeer grazing was low or absent; continental areas without grazing, where average LB (calculated on all plots where WA Si < 3.0) was higher than 650 g m -2 , and LB was about 100 g m -2 in grazed areas.

LB in di erent vegetation types and total lichen production
Di erent vegetation types include major variation in LB from exposed ridges to snow-bed veg- etation, as previously shown by vegetation transects analyses (Gjaerevoll, 1956;Dahl, 1957), it is therefore essential to relate the lichen measurements to actual vegetation types and their snow conditions.Types A-D are similar to previously described as oligotrophic chionophobous belonging to Loiseleurio-Arctostaphylion (Nordhagen, 1943) or Arctostaphyleto-Cetrarion nivalis (Dahl, 1957).It includes alpine dwarf-shrub-and lichen heaths, Juncus tri dus heaths as well as subalpine (northern boreal) chionophobous coniferous and Betula pubescens forests.
e most exposed (A and B) include several di erential species against other groups, e.g.Arctous alpinus, Alectoria nigricans, Bryocaulon divergens, Coelocaulon aculeatum, Flavocetraria nivalis, Gymnomitrium concinnatum and amnolia vermicularis.Similar communities are previously included in the A group (Odland, 2005).Such vegetation types appear to have been the basis for previous estimates of LB (Wielgolaski, 1975), Type E-I can be described as oligotrophic lee-side communities similar to previously allocated to Phyllodoco-Vaccinion myrtilli (Nordhagen, 1943;Odland, 2005).e limit towards the chionophobous community Arctostaphyleto-Cetrarion nivalis is drawn where Cladonia stellaris and Flavocetraria nivalis begin to dominate.Oligotrophic lee-side communities have not been included in previous LB-studies on Hardangervidda.
A vegetation map based on vegetation classi cations has been published where the area of the di erent vegetation types were calculated, and also their distribution in relation to alti-tude (Hesjedal, 1975).e two lichen communities available for reindeer grazing were there described as Loiseleuria procumbens heaths (1a) and Flavocetraria nivalis-Juncus tri dus heaths (1b).e rst covered 7.1 % of the total area (8310 km 2 ), while the second covered 3.0 %, but this was mainly con ned to high altitudes (>1350 m a.s.l.).

E ects of grazing on LB
Studied plots classi ed to the same vegetation types had highly di erent LB when grazed and not grazed areas were compared, with a 50 % reduction in the grazed areas.Types with a higher WA Si value had a lower (ca 30 %) reduction (Table 3).Average LB from the grazed areas at Hardangervidda were around 100 g m -2 , while not-grazed continental areas had LB higher than 700 g m -2 .During the period from 2001 to 2011 the reindeer population at Hardangervidda increased from ca 5200 to 11000 animals (Bjerketvedt et al., 2014), and the estimated average LB may be considered to reect this grazing pressure.is value indicates that LB at Hardangervidda lie close to a critical value adequate to ensure survival of the present reindeer population size.It is remarkable that Iming ell east in the outskirts of Hardangervidda situated only one km from the grazed Iming ell west appear to remain ungrazed.e main reason for this is presumably that Iming ell east is separated from Hardangervidda by human activities (roads and cabins).Migration of reindeer supplied with GPS collars (Jordhøy & Strand, 2009) showed that they hardly crossed the valley during winter.It has previously been found (Nelleman et al.,2000) that available LB was ca 1200 g m -2 0-5 km from a tourist resort decreasing to a low of ca.250 g m -2 at 15-25 km distance, a pattern that was assumed to re ect overgrazing as a result of avoiding a tourist resort in the Rondane national part, S Norway.
A reindeer consumes 70 % of the entire amount of lichens within a grazing area and can dig and graze 30 m 2 day -1 , and then on average 90-100 g m -2 of lichens should be available (Kumpula et al.,2000). is represents on average around 50 % lichen cover 3 cm thick which is equivalent to earlier presented lichen biomass data.A lichen cover of 50 % with a thickness of 3 cm will according to Eq.1 represent an estimated LB of 86 g m -2 .A general model for the dependence of lichen range condition on the mean density of semidomesticated reindeer in Finland indicates that a LB at such a condition level, the number of reindeer on lichen ranges in winter should not exceed 5-7 reindeer km -2 .It has also been found that LB lower than 200 g m -2 was associated with reindeer densities of 4-8 reindeer km -2 (Kumpula et al.,2000).
It has been estimated that the highest annual yield of lichens (120-160 kg ha -1 ) is achieved from lichen stands that contain LB of 600-1200 kg ha -1 (Helle et al.,1990).Studies of vegetation in protected ungrazed sites in the Finnish reindeer management area (Väre et al.,1996) estimated an average of nearly 8000 kg ha -1 of lichen (total amount) at ungrazed sites.eis value t quite well the average LB value found here from ungrazed areas in continental parts of S Norway.
In general, there was a strong increase in lichen abundance from oceanic to continental areas.As shown by Eq. 5 and 6, the lichen abundance (based on not-grazed plots with WA Si < 4) increased from approximately 20 % in an oceanic site to 80 % in a continental site.It has previously been estimated that exposed heaths have often a 75-85 % lichen cover in areas where annual precipitation is less than 400 mm, and 40 % or less where the annual precipitation is 1200 mm (Heggberget et al.,2002).

Lichen height, cover and volume as proxies for biomass
e present study shows that lichen height and lichen volume were well linearly correlated with LB. ere were, however frequently major discrepancies between LB and lichen volume (Fig. 3). is can partly be explained by the fact that volume measurements in the most exposed vegetation types may deviate strongly from the measured LB (cf.Eq. 4).Fig. 3 also show that LH (and thereby also the volume) is particularly a poor estimate for LB when the lichen mat is heigher then 10 cm. is is because the podetia density decrease strongly the the top to the bottom.Relationship between LB and CoL indicate that lichen cover was generally a poor estimate for LB.In plots where CoL was high (>75 %), LB varied from <100 g m -2 to >2000 g m -2 , and plots with a CoL of 90 % could have a variation in LB from less than 400 g m -2 to1800 g m -2 (Figure 4). is is particularly evident in grazed area where the lichen cover could be quite high, but where the lichen heights were very low.Consequently, we suggest that reindeer management evaluations should not be based on lichen cover data only.
It has also previously been found (Moen et al.,2007) that di erent not-destructive methods gave fairly similar results, except when cover estimation was based on presence/absence data with a strict criterion of 100 % lichen cover.Average thallus heights within the plot explained the variation in LB as well as lichen volume (estimated from cover and average heights).Relationships between biomass and volume or height were also very similar for four lichen species studied (Cladonia arbuscula, C. rangiferina, C. stellaris, and Cetraria islandica), and the separation into species thus did not seem necessary for practical purposes.
In a recent published study estimating reindeer pasture quality at Hardangervidda, lichen volumes measurements on the ground were performed and combined with Landsat image analyses (Falldorf et al.,2014).Sampling data over a large area, they found that the maximum lichen volume was 600 (6 cm height × 100 % cover), and the average value within alpine heaths was approximately 154.By the use of Eq. 1, this will translate to a maximum LB of 829 g m -2 , and a mean of ca 160 g m -2 which do not deviate much from results of the present study.We agree with Falldorf et al. (2014) who maintained that in future studies one should discriminate between dominant lichen species on alpine heaths, e.g. by the use of three lichen catogeries which have slightly di erent distribution optima: (i) Cladonia stellaris dominant heaths (primarily type D), (ii) C. rangiferina and C. arbuscula dominant heaths (primarily type C, D, and E), and (iii) Flavocetraria nivalis dominant ridges (cf.type A and B) (cf. Figure 5).
Figure 5. Exposed alpine lichen heaths have often no or a thin snow cover during the winter and are therefore available for reindeer grazing.ese areas are most distinctly separated from areas (vegetation types) with a thicker snow layer that has not melted during late winter-early spring.e exposed vegetation on the picture belongs to Type C in Table 3. Figure 6.In continental areas with a gently sloping terrain where wind blows the snow away, large stands with a vegetation cover similar to Type A and B (Table 3) may occur.e picture is from the Vågå area, ca 1250 m a.s.l.

Use of snow indicators to predict variation in LB
In comparative LB studies it is essential that sampling has been performed in the same environments, i.e., under similar snow conditions.One way to control that LB has been measured under the same snow conditions are to relate the biomass to average Snow index in the sampled plots.
Plants have for a long time been recognized as potential indicators of environmental conditions, and there are several reasons for, and advantages of using plant indicators instead of physical or chemical measurements (Diekmann, 1995).Studies have shown that vegetation, especially long-established vegetation, provides a sensitive integrated measure of the environment.
Species with an optimum on exposed sites with no or little snow during the winter have indicator value 1, and species with an indicator values 5 (lee-side species such as Vaccinium myrtillus) are normally melted out around medio May (Odland & Munkejord, 2008b).As compared with a previous study (Dahl, 1957), Caricetum nivalis communities (type A and B) had a snow-depths <50 cm, and the calculated Wa Si -value was 1.49 ± 0.30, Cladonietum alpestris betuletosum communities (type C and D) had snow-depths between 0.3 and 2.0 m (Wa Si = 2.16 ± 0.35), Phyllodoco-Vaccinium myrtilli communities (type E-I) had snowdepths between 1.6 and 4.5 m (Wa Si = 3.3 ± 0.31), and Deschampsieto-Dicranetum fuscae communities (type K) had depths between 2.0 and 4.2 m (Wa Si = 5.66 ± 0.35) (cf.15).Maximum snow depths and Wa Si -values are not always well correlated because the snowmelt rates may di er highly according to variation in altitude and aspect.
An essential question related to available grazing resources for the reindeer population in an area is the snow thickness and hardness.
e snow thickness is highly variable both during the year and between years, and evaluations must therfore be based on "normal values".Studies show that reindeer can dig craters 70-80 cm deep, but the depth depends on snow hardness and possible layers of ice (Helle, 1984;Nelleman, 1996;Heggberget et al., 2002).According to this, mainly the two most exposed types studied will normally be available for reindeer grazing.LB are estimated to 119 g m -2 (Wa Si = 1.8) in the middle alpine zone and 652 g m -2 (Wa Si = 2.0) in the low alpine zone, but the spatial distribution is relatively small.Vegetation types where LB is highest (type B and D, LB = 802 g m -2 and Wa Si = 2.7) would therfore probably not be available for wintergrazing in a normal year.

Conclusions
As to the main questions raised in the study, the following answers can be given: to average snow layer duration as quanti ed by average snow indicator values for the studied sites (Wa Si ).
e relationship was not linear and highest average LB (650 g m -2 ) was found in sites with an average WA Si value of 2.0 (i.e.snow layer lasts until medio April).Average LB decreased on sites where the snow duration was both shorter and longer.available for reindeer grazing at Hardangervidda had average LB of 87 and 118 g m -2 , representing an average decrease of LB of approximately 60 % compared to similar vegetation in ungrazed areas.
signi cantly linearily correlated with LB and are therefore useful proxies for LB.Lichen percentage cover was, however a poor proxy for LB.negatively correlated with average LB.In areas with annual precipitation higher than 900 mm, LB was generally low and rarely higher than 200 g m -2 .increasing altitude.On average, LB decreased with approximately 20% for each hundred m increase in elevation.
Classi cation of the studied vegetation plots e studied vegetation plots as classi ed by TWINSPAN are shown in Table 3. e rst division separated two clusters with Cladonia rangiferina, C. stellaris and Flavocetraria nivalis as the main indicators (Types A-F, n = 165) from the rest (Types G-K, n = 51) .e second division separated two clusters: (Type A-C, n=55) with high abundance of Flavocetraria nivalis, F. cucullata, and Alectoria ochroleuca and (Types D-T, n = 110) where 43 plots had high abundance of Cladonia stellaris.e cluster (Type G-K, n = 51) included a group of di erent vegetation types dominated by species such Salix glauca, S. lapponum, S. herbacea, Vaccinium myrtillus, Nardus stricta, Pleurozium schreberi, and low abundances of lichens.

Figure 3 .
Figure 3. Relationship between measured lichen cover (CoL), lichen height (LH), and lichen volume as estimated by LH × CoL.Lowess smoothers are drawn.Results of the linear regression analyses are shown in the text.

Figure 4 .
Figure 4. Relationships between average snow indicator value (WA Si ) for the plots and associated values for lichen biomass (LB), lichen cover (CoL), lichen height (LH), and lichen volume as estimated by LH × CoL.