Allometric relationships and dendroecology of thedwarf shrub Dryas integrifolia near Churchill,subarctic Manitoba

R. Au and J. C. Tardif

Abstract: Shrubs have generally been overlooked in dendrochronology, and little information exists on allometric relation-ships in dwarf shrubs. Dryas integrifolia M. Vahl. has been recognized as an important species in revegetation of disturbedgravel pits in northern areas. In this study, the dimensions (mat, stem, and root) of D. integrifolia shrubs were measured infour populations having recolonized disturbed areas, and anatomical slides containing growth rings were prepared for eachindividual. The first objective was to compare allometric relationships among descriptors of shrub dimension and betweenthese descriptors and shrub age. Secondary objectives included documenting maximum age and evaluation of the potentialfor cross-dating among shrubs. Strong, consistent allometric relationships between aboveground dimensions were observedamong individuals from all sites, although growth rates varied within and between sites. This indicated that even thoughaboveground shrub dimensions grow proportionally to one another, these measurements cannot be used to infer age, be-cause of differences in growth rates among D. integrifolia shrubs. However, numerous cross-sections from D. integrifoliashrubs could be successfully cross-dated, and a short chronology was developed. The radial growth – climate associationwas found to be similar to that of dominant tree species of the region regarding the impact of October conditions. In-creased snowfall in October prior to and warm May temperature during the year of ring formation appear to restrict growthof these shrubs by altering the onset of the growing season. This study has demonstrated that it is possible to accuratelydate D. integrifolia shrubs, and this ability may be applied to future studies involving population dynamics and remedia-tion of open gravel sites. Growth rings also have the potential to be used in other Arctic shrubs.

Key words: Entire-Leaved Mountain Avens, allometric relationships, growth rates, dendrochronology, microtoming, age.

Resume : On a generalement neglige les arbustes en dendrochronologie, et il y a peu d’information sur les relations allo-metriques chez les arbrisseaux. On reconnaıt le Dryas integrifolia M. Vahl. comme espece importante pour la restaurationdes bancs de gravier perturbes, dans les regions nordiques. Les auteurs ont mesure les dimensions (matte, tige, racine)d’individus du D. integrifolia, dans quatre populations ayant recolonise des surfaces perturbees, et ils ont prepare des la-mes anatomiques portant des anneaux de croissance de chaque individu. On cherchait d’abord a comparer les relations al-lometriques entre les donnees sur la dimension de l’arbuste, et entre ces donnees et l’age de l’arbuste. On cherchait deplus a determiner l’age maximum et a evaluer le potentiel pour obtenir des superpositions d’ages entre les arbrisseaux. Lesauteurs ont observe des relations robustes et congrues entre les dimensions epigees, parmi les individus de tous les sites,bien que les taux de croissance varient sur et entre les sites. Consequemment, meme si les dimensions epigees des arbustess’accroissent proportionnellement les unes par rapport aux autres, ces mesures ne peuvent pas servir a inferer l’age d’apresles taux de croissance, chez le D. integrifolia. Tout de meme, les auteurs ont pu croiser de nombreuses sections transversesde plants du D. integrifolia et obtenir une courte chronologie. On constate que la relation entre croissance radiale et climatressemble a celle des essences dominantes de la region, d’apres les impacts des conditions d’octobre. Des conditions d’en-neigement accru en octobre prealablement a une temperature elevee en mai, au cours de l’annee de formation des anneaux,semblent restreindre la croissance de ces arbrisseaux, en alterant le debut de la saison de croissance. On demontre donc lapossibilite de determiner l’age des individus du D. integrifolia avec precision, et que ceci peut etre applique a des etudesfutures portant sur la dynamique des populations, aux fins de la restauration de depots de gravier ouverts. Les anneaux decroissance ont egalement du potentiel pour l’etude de d’autres arbrisseaux arctiques.

Mots-cles : dryade a feuille entiere, relations allometriques, taux de croissance, dendrochronologie, microtomie, age.

[Traduit par la Redaction]

IntroductionDendrochronology has been successfully applied to those

plants capable of secondary growth and production of an-nual growth layers. Although these include trees, shrubs,and dwarf shrubs (Schweingruber 1996), dendrochronologyhas been mainly confined to trees. However, tree-ring chro-nologies cannot be developed north of the boreal tree linebecause of the absence of erect trees (Johnstone and Henry1997). Instead, dwarf shrubs producing a small amount of

Received 11 May 2007. Published on the NRC Research PressWeb site at canjbot.nrc.ca on 27 July 2007.

R. Au1 and J.C. Tardif. Centre for Forest InterdisciniplinaryResearch (C-FIR), University of Winnipeg, 515 Portage Avenue,Winnipeg, MB R3B 2E9, Canada.

1Corresponding author (e-mail: robau@mts.net).

585

Can. J. Bot. 85: 585–597 (2007) doi:10.1139/B07-055 # 2007 NRC Canada

growth every year are often found in these extreme environ-ments (Wilson 1964; Kolishchuk 1990; Woodcock andBradley 1994). Given that there are more than 1200 speciesof prostrate woody plants in the world (Kolishchuk 1990),the majority of shrubs and dwarf shrubs have generallybeen overlooked in dendrochronology (Beschel and Webb1963; Schweingruber 1996; Schweingruber and Dietz2001a).

Schweingruber and Dietz (2001a) indicated that woodanatomy in dwarf shrubs does not greatly deviate from thatof trees. In many cases, the anatomical pattern of growthrings was consistent among species within each of the Car-yophyllaceae, Lamiaceae, and Rosaceae. Moreover, dendro-chronological techniques have successfully been applied toperennial herbs (Dietz and Schweingruber 2002) as well asto shrubs and dwarf shrubs (Gordon 1991; Milton et al.1997; Schweingruber and Dietz 2001a; Lageard et al.2005). Schweingruber and Poschlod (2005) conducted anextensive review of the maximum ages, age structure, andanatomy of 800 species of herbs and dwarf shrubs fromCentral Europe. A major finding was that there was no dif-ference in anatomical response (wedging rings, reactionwood, compartmentalization, growth reductions, and struc-tural changes) between dwarf shrubs and trees to mechanicalstress initiated by crown asymmetry, wounding, cell col-lapse, and damage of root systems. Chronologies of an arcticdwarf shrub based on morphological characteristics fromstems have been used to reconstruct mean summer air tem-perature (Rayback and Henry 2006). Other studies have alsodescribed dwarf shrubs as having good potential for dendro-climatological work (Kolishchuk 1990; Woodcock andBradley 1994; Schweingruber and Dietz 2001a; Bar 2005).

Entire-Leaved Mountain Avens (Dryas integrifoliaM. Vahl.) in the Rosaceae is a low-growing, mat-formingdwarf shrub widely distributed from the north of EllesmereIsland down throughout the Canadian Arctic Archipelago tosouthern Nunavut and the Labrador-Ungava Peninsula aswell as on the coasts of Greenland, James Bay, HudsonBay, Yukon, and Alaska (Scoggan 1978; Porsild and Cody1980; Johnson 1987a). Dryas integrifolia is a calciphilouspioneer shrub commonly occurring on barren, rocky, orgravelly substrates (Scott 1995). Hart and Svoboda (1994)reported median mat sizes to range from 208 to 717 cm2 atAlexandra Fiord, Ellesmere Island, NWT, and from 105 to2592 cm2 at Churchill, Manitoba (Hart 1988). Philipp et al.(1990) found that D. integrifolia seedling recruitment wasepisodic, and that population turnover was very slow withperiods during which no recruitment occurred. However, lat-eral rooting of branches in D. integrifolia can assist in itsspread (Porsild 1964), which may lead to vegetative repro-duction through clonal fragmentation (Hart 1988), especiallyin mechanically damaged shrubs whose centres have diedoff (Svoboda 1974).

Dryas integrifolia has been described as the dominantearly colonizer of lime-rich gravel pits in the Churchill re-gion (Scott 1995) as well as one of the most important sta-bilizers of marine beaches and gravel regions within thearctic (Johnson 1987a). Northern gravel sites usually havelimited moisture and nutrient holding capacities and there-fore are difficult to revegetate (Johnson 1987b). It has beensuggested that D. integrifolia was a suitable species to reve-

getate recently exposed gravel pits, because of its ability togrow under adverse conditions and to facilitate growth ofless stress-tolerant species (Hart and Svoboda 1994; Firlotte1998). A better understanding of early-colonizing shrubs andtheir allometric relationships in the subarctic is of particularrelevance for the field of restoration ecology. For example,it would be of interest to discover whether allometric rela-tionships of D. integrifolia shrubs remain constant acrossthe region. Knowledge of growth rates also provides infor-mation to aid in site remediation. More specifically, betterestimates of the time needed by D. integrifolia shrubs to re-vegetate a gravel site could be obtained with the develop-ment of a valid age–size model. In addition, the analysis ofa plant population’s age structure and growth variabilitycould be of value for management of the species through re-construction of historical disturbances (Schweingruber andPoschlod 2005). Given that crown diameters of some arcticshrubs have been successfully used to date when substratesbecame available for colonization (Beschel and Webb 1963;Benedict 1989), it may also be possible to determine if a re-lationship exists between time since disturbance (gravel ex-traction) and shrub size in northern gravel sites.

To provide answers to some of these questions, the firstobjective of this study was to compare allometric relationshipsalong with growth rates in four populations of D. integrifoliashrubs growing in previously disturbed areas in theChurchill region, subarctic Manitoba. Secondary objectivesincluded the evaluation of the potential for cross-datingamong individual shrubs and documentation of maximumshrub age. An understanding of which climatic factors influ-ence shrub annual growth increment would contribute toknowledge on how the radial growth of shrubs and treescompare at the tree line.

Materials and methods

Study areaThe Churchill region is located in the northeastern corner

of Manitoba on the western shore of Hudson Bay (Fig. 1).The study area is located southeast of the town of Churchill(58844’N and 94804’W) within the Hudson Bay LowlandEcoregion and York Factory Ecodistrict (Smith et al. 1998).The High Sub-arctic Ecoclimatic region in Churchill (Scott1995) is largely influenced by cold air masses from HudsonBay (Rouse 1991). The Churchill meteorological station(Fig. 1, 58844’N and 94803’W) situated at 28.7 m a.s.l.showed a mean annual daily average temperature of –6.9 8Cfor the period 1971–2000 (Environment Canada 2004). InChurchill, 61.3% of the total annual precipitation (431.6 mm)falls as rain.

The coastal calcareous substrates east of Churchill haveless than 2000 years of profile development because of theisostatic rebound of Hudson Bay (Scott 1995) and they cur-rently rise at approximately 0.5 m per century, leaving anextensive pattern of beach ridges (Dredge and Nixon 1992).At these beach ridges, limited water-holding capacities and alow store of nutrients (Smith et al. 1998) severely restrictthe vegetation able to inhabit these areas. Sandy, imperfectlyto well-drained Regosols are found on beach ridges, whilepoorly drained Rego Gleysols are associated with clayeybrackish tidal flats (Smith et al. 1998). Precambrian, Ordovi-

586 Can. J. Bot. Vol. 85, 2007

# 2007 NRC Canada

cian, and Silurian bedrock underlie the region (Dredge andNixon 1992), which is presently dominated by black spruce(Picea mariana (Mill.) BSP.), white spruce (Picea glauca(Moench) Voss), and tamarack (Larix laricina (Du Roi)K. Koch).

Sampling sitesIn August 2004, four anthropogenically disturbed sites do-

minated by D. integrifolia shrubs (A–D) were selected alongroads or gravel openings (Fig. 1). These four sites were se-lected for accessibility and on the criteria that individualD. integrifolia mats were distinguishable from one another.In addition, an effort was made to select sites characterizedby well-drained, Regosolic soils typical of northern gravelsites in need of revegetation. The substrate was relativelyconsistent across these sites and consisted of small- to me-dium-sized gravel with little organic material. Site A was lo-cated near the edge of a recent road with scattered P. glaucaand Salix spp. found in adjacent locations. Site B was lo-cated between an all-terrain vehicle road and a forest ofL. laricina and P. glauca. Site C was centred on a pastgravel extraction pit with some legumes mixed withD. integrifolia shrubs. Site D was located north of Twin

Lakes on a shallow gravel pit. Mat sizes varied from smallto continuous in places. Picea glauca and L. laricina borderedthe edge of the pit.

Aerial photographs from the Canadian Department of Ener-gies, Mines and Resources were consulted to determine theapproximate date at which these four sites were disturbed. Allavailable photos capturing sites A–D were examined for evi-dence of road construction. Aerial photos showed no road de-velopment at sites C and D in August 1947; at site B in June1948; and at site A in August 1991. Conversely, the earliestphotos displaying road development for sites B, C, and Dwere taken during the summer of 1956, and the road at site Awas developed in 1994 (C. Paddack, personal communication,2004). In addition to the four disturbed sites, a total of 26D. integrifolia shrubs were collected from a series of undis-turbed sites to assess the maximum age of the species. Indi-vidual D. integrifolia mats at these sites came into contactforming a continuous cover of vegetation.

Data collectionAt each disturbed site (A–D), D. integrifolia shrubs were

chosen according to nearest neighbor protocol. Sampleswere consecutively selected along two parallel transects at

Fig. 1. Map of the Churchill region, northeastern Manitoba, showing locations of sites where Dryas integrifolia shrubs were sampled inAugust 2004. (A), site A; (B), site B; (C), site C.

Au and Tardif 587

# 2007 NRC Canada

each site based on mat diameter classes of 10 cm increments.The mat axis with the longest diameter (MD1) was used toselect and assign each shrub to a particular diameter class.An attempt was made to collect at least two replicates foreach diameter class. In some instances, this was not possiblebecause of an under-representation of larger individuals atthe sites. Consequently, few large samples within diameterclasses approaching 100 cm or greater were collected. Atotal of 25, 19, 24, and 30 D. integrifolia shrubs weresampled at sites A, B, C, and D, respectively.

The following measurements were directly taken in thefield for each D. integrifolia: the longest mat diameter(MD1) and a second diameter (MD2) taken perpendicular toMD1. Mat symmetry was determined by dividing MD2 overMD1. Mat dieback was estimated as the percentage of matarea not currently occupied by the living shrub. This includedthe leafless area of the mat occupied by leaf litter, rock,and bare ground in addition to dead or discolored portionsof mat. The number of flowers and presence of lateralrooting branches (LRB) was also determined for each indi-vidual.

Before taproot extraction, a measuring tape was placed ad-jacent to each D. integrifolia, and a digital photo was taken tolater derive other measurements (see laboratory section).After shrub extraction, taproot length (TRL) was measuredfor each shrub. Because of the difficultly of extraction in thegravelly substrate, entire taproots were collected only forD. integrifolia shrubs with a MD1 of up to 25 cm with the ex-ception of those at site A. The TRL measurements pertainingto D. integrifolia shrubs from site A with a MD1 exceeding25 cm were later removed to avoid bias in site comparisons.The excavated taproots were retained for laboratory analysis.Lateral roots and belowground biomass were neither col-lected nor estimated.

Laboratory analyses

Image analysisIn the laboratory, each digital photograph of D. integrifolia

was analyzed to derive additional measurements. First, the per-centage mat dieback for each D. integrifolia was re-estimatedto compare with that estimated in the field, and a mean diebackvalue was calculated using both estimations. Second, the soft-ware WinFolia Pro1 2005b (Regent Instruments Inc., 2005)was used to calculate parameters related to mat dimension:mat area (MA), mat perimeter (MP), longest mat diameter(D1), and mat diameter perpendicular to D1 (D2). Each picturewas individually calibrated using both vertical and horizontalgraduations, in centimetres, from the measurement tape.

Sectioning of Dryas integrifolia stemsTo determine the age of each sample and to assess if a

chronology could be developed, each shrub specimen wasprepared for each microscopic examination . Before section-ing, the diameter (or radius if the root was split) of each sam-ple was measured to the nearest 0.10 mm. The longest stemdiameter (SD1) and the diameter perpendicular to it (SD2)were measured near the stem–root interface using a digitalcaliper. The mean of these two measurements produced amean stem diameter (MSD) value.

To ease the process of making anatomical slides,

D. integrifolia samples were kept refrigerated until sectioning,so that the wood remained soft. The stems were transverselysectioned as close to their stem–root interfaces as possible.Previous studies on shrubs and dwarf shrubs extracted stemsections just above the root collar (Woodcock and Bradley1994) or at soil level (Milton et al. 1997; Zalatan and Gajewski2006) with acceptable results. Branches of shrubs from undis-turbed sites and occasionally from disturbed sites (A–D) werepreferably sectioned when stem and (or) root portions werehighly degraded. In each case, the basal portion of largestbranches was selected for sectioning. Stems of woody plantsgrowing in extreme environments often split longitudinally(Milton et al. 1997) or have rotten centres (Kuivinen and Law-son 1982).

A sliding microtome with Feather S35 blades was usedfor microsectioning stems. The blade was angled between28 and 158 depending on the particular piece of wood. Thesections taken from each specimen were mounted onto aslide, each of which displayed 8–10 sections varying be-tween 6 and 16 mm in thickness to offer a range for analysis.All sections were stained with 1% Safranin O solution andthen cleared, while still on the slides, with 50% ethanol(two baths), 95% ethanol (three or four baths), reagent alco-hol, and D-limonene, respectively. An additional fourth bathwas necessary to clear those sections more prone to cloud-ing. On average, each bath would last from 2 to 5 min foreach chemical. After clearing, a few drops of Permount1(Fisher Scientific, Fairlawn, N.J.) were applied on the la-beled slides to secure the cover slips. These were laterplaced on metal trays and weighted down with flat magnetsto discourage bubbling underneath the cover slip. The metaltrays were placed in a drying oven at 40 8C for at least 2 d.All slides were then stored on trays for at least 1 month untilthe mounting medium was properly set.

Age determination and growth-ring measurementThe annual growth rings contained in transverse sections

of each D. integrifolia (Fig. 2) were counted by two peopleusing a Nikon E200 Eclipse microscope to determine theage for each shrub. Each person counted the ages of threesections per slide (shrub) containing the most clearly definedrings and highest quality stem sections. If a discrepancy inage was found, the slide was reexamined by both people untila consensus was reached.

A Velmex slide stage micrometer interfaced with a DOS-based computer was used to record the ring widths of 24D. integrifolia shrubs (above a minimum age of 15 years)with a precision of 0.001 mm. A single radius was measuredfor each shrub. The sections were cross-dated by an experi-enced technician (France Conciatori) using the skeleton plotmethod to determine pointer years. Cross-dating ensured thatpatterns of ring widths were matched among several series toidentify the correct calendar date of formation for every ringin all the samples (Schweingruber 1996). Cross-dating andring-width measurements were validated using COFECHA(Holmes 1983).

Statistical analyses

Population comparisons and allometric relationshipsSeveral measurements were taken in the field and labora-

588 Can. J. Bot. Vol. 85, 2007

# 2007 NRC Canada

tory to determine mat diameter, stem diameter, and percentdieback in each D. integrifolia. Field and laboratory meas-urements describing a given variable were compared usingPearson correlation analysis. Laboratory measurements fromphotos were found to be highly correlated to field measure-ments, with the weakest correlation coefficient being 0.973(data not shown). The field variable MD1 was thus used inall subsequent analyses. As a way to reduce the number ofmeasurements recorded in the field or laboratory for a givenvariable, the one most highly correlated for each was sys-tematically retained for subsequent analysis.

The characteristics of the D. integrifolia populations atthe four disturbed sites were compared using the nonpara-metric Kruskal–Wallis analysis of variance. The data didnot conform to a normal distribution and had greater weight-ing toward individuals with smaller dimensions. Therefore,the Kruskal–Wallis single factor analysis of variance byranks was used to test for the presence of a statistical differ-ence among the four populations. The Mann–Whitney U testwas subsequently employed for those variables shown to bestatistically significant by the Kruskal–Wallis test, to deter-mine which specific pairwise comparisons were statisticallysignificant (Sheskin 1997).

Growth rates and their relationships with dimensionsAnnual growth rate of D. integrifolia individuals was cal-

culated by dividing mat dimension descriptors (MD1, MA,MP, MSD, TRL) by age. Status variables (mean dieback,flower abundance, and mat symmetry) were divided by ageto produce mean values corresponding to the annual pro-gression of the associated status. The variable MD1 wasalso divided by MSD, TRL, MA, and MP to produce allo-

metric ratios. Finally, the ratios flower abundance over MAas well as over live area (LA) were calculated. The latter ra-tio represented the number of flowers in proportion to LAonly, excluding dieback. The variable LA was calculatedfrom the equation: LA = MA – ((mean dieback/100) �MA). Following the same approach as the comparisons be-tween D. integrifolia dimensions among the four popula-tions, the Kruskal–Wallis analysis of variance was alsoapplied to the series of derived growth ratios, and theMann–Whitney U test was employed to discern differencesin the pairwise comparisons.

Allometric relationships were compared among sites. Therelationships between age and dimensions and status wereinvestigated as well. Linear and nonlinear regression analy-ses were used to model the relationships between selectedvariables. However, an obvious outlier sampled near theedge of site A was excluded from regression analysis. Stat-istical analyses were conducted with Systat (version 11) forWindows (SYSTAT 2004a) and SigmaPlot (version 9.01)for Windows (SYSTAT 2004b).

Chronology development and its association with climateA D. integrifolia chronology was developed from the 24

cross-dated growth-ring measurement series after standard-ization using the program ARSTAN. Standardization trans-forms measurement series into tree-ring indices to providecomparable means and variances for all indices. A cubic-spline function of 100% of the series length and a frequencyresponse of 50% was applied to each measurement series toremove unwanted signals related to individual D. integrifoliagrowth and to maximize the common variation given theshort length of the measurement series. As a standard prac-

Fig. 2. A cross-section of a Dryas integrifolia stem showing annual growth increments. Outer stem segments often became isolated fromone another in older individuals as a result of cambial division directed at only select locations. This particular individual was determined tobe 42 years of age.

Au and Tardif 589

# 2007 NRC Canada

tice in dendrochronology, a residual chronology was thendeveloped by removing temporal autocorrelation from eachstandardized measurement series using autoregressive mod-eling, and a biweight robust mean was used to enhance thecommon signal of the residual chronology (Cook andHolmes 1986; Girardin et al. 2005).

To assess the association between the D. integrifolia re-sidual chronology and climate, Pearson correlations werecalculated between the residual chronology and monthly cli-mate data obtained from the Churchill meteorological stationfor the period from 1933 to 1999 (Girardin et al. 2005). Inour study, correlations between D. integrifolia shrubs andclimate were calculated for the period 1978–2004, corre-sponding to the establishment of a minimum sampling depthof 12 series. The weather variables included total monthlysnowfall, total monthly precipitation, mean monthly mini-mum temperature, and mean monthly maximum temperaturefor a 17-month period from May of the year prior to (t – 1)ring formation to September of the year (t) during ring for-mation. Two significance levels were used P < 0.01 and P <0.05. Some missing meteorological data occurred between2000 and 2004, and these were not estimated.

Results

General characteristics of Dryas integrifoliaFive variables (MSD, TRL, mean dieback, presence of

LRB, and age) were significantly different among the fourpopulations (Table 1). Dryas integrifolia shrubs from site Aconsistently had the lowest mean values for these variables.In site A, shrubs were significantly younger, with shortertaproots and less dieback than the other populations. Site Acontained individuals with a mean age of less than one-quar-ter that of other sites, even though mat dimensions did notdiffer significantly from those of the other disturbed sites(Table 1). Although the characteristics of shrubs from undis-turbed sites were not included in the Kruskal–Wallis tests ofsignificance, their mean values were often extreme as in thecases of MSD and age. The oldest shrub was found amongundisturbed sites and was recorded to be 68 years of age(data not shown).

The comparison of growth rates and allometric ratiosamong the four populations indicated that at least one popu-lation differed from the others for these variables: MD1/age,MP/age, MSD/age, TRL/age, mean dieback/age, MS/age,MD1/TRL, and MD1/MP (Table 2). Ratios from site Awere higher than other sites for the majority of significantlydifferent cases. Dryas integrifolia shrubs from site A werealso found to hold the highest mean values among sites ex-cept for mean dieback/age and MD1/MP. This indicated thatshrubs from this population had much faster growth rates(i.e., mat size growth rates) relative to the others.

Allometric relationshipsThe relationships between variables describing mat di-

mension in D. integrifolia shrubs indicated that a consistentrelationship was maintained among populations (Figs. 3A,3B). The relationship between MD1 and MP was best ap-proximated by a linear regression (y = 13.406 + 5.474x; adj.r2, 0.933; P < 0.001; n = 97), whereas the association be-tween MD1 and MA was best illustrated by a quadratic pol-

ynomial equation (y = 141.829 + –15.965x + 0.782x2; adj.r2, 0.979; P < 0.001; n = 97). In contrast, the relationshipbetween TRL and either MD1 or MSD was found to bemore variable among sites (Figs. 3C and 3D). These rela-tionships of D. integrifolia shrubs were best approximatedby a linear relationship at site A (TRL and MD1: y =4.262 + 1.491x; adj. r2, 0.882; P < 0.001; n = 23; and TRLand MSD: y = 0.622 + 8.386x; adj. r2, 0.890; P < 0.001; n =23) indicating that a larger mat size is supported by shorterTRL compared with other sites (Figs. 3C, 3D). The relation-ship between MD1 and MSD was found to be consistentacross all four populations, although residuals become in-creasingly important among shrubs with MD1s greater than50 cm (Fig. 3E). This relationship was best approximated bya two-parameter exponential model (y = 14.685(1 – e–0.019x):adj. r2, 0.848; P < 0.001; n = 99).

Age-related relationshipsThe relationships between D. integrifolia age and various

descriptors of mat dimension and status indicate that indi-viduals from site A have higher growth rates than in allother sites (Figs. 4A–4C). It should be noted that at site A,there was one outlier 20 years of age that exceeded the agesof all the other D. integrifolia shrubs at this site and wasthus excluded from the analyses. In contrast with the othersites, shrubs from site A showed a linear increase in ageand MD1 (y = –12.639 + 6.649x: adj. r2, 0.821; P < 0.001;n = 22), MSD (y = –2.084 + 1.260x: adj. r2, 0.829; P < 0.001;n = 22), and TRL (y = –18.613 + 10.852x: adj. r2, 0.826;P < 0.001, n = 22), whereas this relationship becomes morevariable within and between sites B, C, and D. A largerange of shrub dimensions was observed at any given ageat these three sites as illustrated by the departures from thelinear regression models between age and MD1 (y =–20.734 + 1.899x: adj. r2, 0.398; P < 0.001; n = 73),MSD (y = –1.864 + 0.278x: adj. r2, 0.547; P < 0.001; n =73), and TRL (y = –11.189 + 3.082x: adj. r2, 0.409; P <0.01; n = 20). It should be noted that after log transform-ing MD1, the linear regression with age increased to anadj. r2 of 0.619 (P < 0.001; n = 73; not shown). MSDand age increased consistently across all sites, althoughthose at undisturbed sites showed more similarity to thoseat sites B, C, and D than to those at site A (Fig. 4B). Inaddition, the relationship between mean dieback and age(Fig. 4D) was best described by a two-parameter exponen-tial equation (y = 80.794(1 – e–0.028x): adj. r2, 0.690; P <0.001; n = 96). Mean dieback values for D. integrifoliashrubs at undisturbed sites were relatively lower for anygiven age than for those at disturbed sites. Finally, theyoung shrubs from site A showed a vigourous stem diame-ter growth rate (MSD/age) compared with those at theother three disturbed sites (Fig. 4E). A nonlinear decreasein MSD/age with increasing age, however, described theindividuals from the undisturbed sites, which contained theoldest D. integrifolia shrubs (y = 0.095 + 6.895/x: adj. r2,0.580; P < 0.001; n = 25) indicating reduced stem diametergrowth with age.

Dryas integrifolia chronology and its association withclimate

The residual chronology developed covered the period

590 Can. J. Bot. Vol. 85, 2007

# 2007 NRC Canada

from 1962 to 2004 and was well replicated during the last26 years (Fig. 5A). Favorable years of radial growth wereobserved in 1983, 1995, and 1999 and poor growth was ob-served in 1980, 1992, and 1998. Dryas integrifolia shrubshad a mean ring width of 0.133 mm and mean standard de-viation of 0.061 mm. Over the common period 1982–2004,the residual series had a mean sensitivity value of 0.332,the mean correlation among all series was 0.228, and the ex-pressed population signal value reached was 0.876 (24 stemsincluded).

The Pearson correlation analysis from 1978 to 2004 indi-cated that the residual chronology of D. integrifolia was sig-nificantly and positively correlated with Octobert–1minimum temperature (r, 0.471; P < 0.05; n = 27) and neg-atively with Octobert–1 snowfall (r, –0.524; P < 0.01; n =26) (Figs. 5B and 5C). During the year of ring formation,Mayt minimum temperature (r, –0.434; P < 0.05; n = 26)was significantly and negatively correlated with radialgrowth. Although nonsignificant, radial growth also showednegative trends (P < 0.1) with Junet snowfall and Septembertsnowfall and a positive trend with Augustt total precipitation(Fig. 5B).

Discussion

Allometric relationshipsThe first objective of this study was to compare allometric

relationships among shrub dimensions and to compare shrubdimensions with age in four populations of D. integrifoliashrubs. All allometric relationships were consistent acrosssites except with taproot length. The relationship betweenMD1 and MSD in all sites followed a two-parameter expo-nential model. At first, increases in mat diameter arefollowed by proportional increases in stem diameter, sug-gesting that the taproot is the main supplier of water andnutrients to the small expanding mat. The increase in stemdiameter, however, slowly decreases with increasing matsize. This result suggests that the role of rooting branchesmay increase in importance at greater mat diameters.Specifically, larger individuals may increasingly utilize ad-ventitious roots for resource acquisition, thereby reducingthe demand for greater taproot girth. In larger shrubs, thetaproot may also no longer be able to entirely support thewater and nutrient requirements of larger mat sizes andmay, therefore, inhibit further growth. In support of this,Svoboda (1974) observed that there was a potential forvegetative reproduction in older D. integrifolia shrubs whosemats had begun to degrade at their centres.

Another contributing factor to the weakening of the rela-tionship between mat and stem diameter at mat diametersgreater than 50 cm could relate to dieback within mats,since mean dieback initially increases with age. At disturbedsites, dieback area along with live mat area would both con-tribute to the length of shrub MD1 measurements. Giventhis, there would be no requirement for any increase ofstem girth for individuals containing significant dieback, be-cause their stems would not be providing water and nutrientsfor the entire mat area but rather, only for the functional matarea. In support of this idea, a higher correlation (r = 0.730;P < 0.001; n = 97) was found between stem diameter andT

able

1.D

escr

iptiv

est

atis

tics

for

each

ofth

efo

urdi

stur

bed

and

pool

edun

dist

urbe

dD

ryas

inte

grif

olia

popu

latio

ns(t

otal

n=

124)

.

AB

CD

KW

Und

istu

rbed

*

Var

iabl

en

Mea

nSD

nM

ean

SDn

Mea

nSD

nM

ean

SDP

nM

ean

SD

MD

1(c

m)

2517

.915

.419

39.5

41.0

2428

.821

.830

25.8

25.1

0.14

83

71.5

33.8

MA

(cm

2 )21

361.

952

7.3

1920

06.6

4087

.224

646.

811

52.8

3075

5.4

1669

.90.

363

331

58.0

2514

.2M

P(c

m)

2111

0.6

77.7

1920

6.0

201.

224

201.

815

3.4

3016

0.7

165.

00.

205

333

2.4

184.

3M

SD(m

m)

233.

8a2.

819

6.5b

4.6

246.

2b3.

430

4.6a

b3.

10.

031{

2610

.94.

5T

RL

(cm

)23

32.1

a25

.110

41.7

ab23

.410

65.1

b36

.20

——

0.02

8{0

——

TR

L(c

m){

1720

.3a

14.6

1041

.7b

23.4

1065

.1b

36.2

0—

—0.

002{

0—

—M

ean

dieb

ack

%23

3.5a

7.5

1938

.4b

16.8

2443

.0b

17.3

3044

.4b

15.4

0.00

0{6

12.5

14.0

Flow

erab

unda

nce

2523

.451

.719

7.8

13.3

248.

112

.130

17.7

31.7

0.49

83

239.

312

7.6

Pres

ence

ofL

RB

230.

4a0.

519

0.8b

0.4

240.

6ab

0.5

300.

7b0.

50.

011{

31.

00.

0A

ge23

5.2a

3.8

1927

.2bc

8.3

2430

.0c

11.5

3024

.3b

8.8

0.00

0{25

41.0

14.4

MS

250.

90.

119

0.8

0.1

240.

80.

130

0.8

0.1

0.26

03

0.9

0.0

Not

e:T

hech

arac

teri

stic

sof

undi

stur

bed

site

sw

ere

not

incl

uded

inth

eK

rusk

al–W

allis

test

sof

sign

ific

ance

.Sa

mpl

ing

ofun

dist

urbe

dsi

tes

was

mai

nly

mea

ntto

iden

tify

the

max

imum

age

ofsh

rubs

.Su

mm

ing

the

high

est

num

ber

ofca

ses

atea

chsi

tean

dpo

oled

undi

stur

bed

site

sw

illre

sult

in12

4in

divi

dual

s.M

D1,

long

est

mat

diam

eter

;M

A,m

atar

ea;

MP,

mat

peri

met

er;

MSD

,mea

nst

emdi

amet

er;

TR

L,

tapr

oot

leng

th;

LR

B,

late

ral

root

ing

bran

ches

;an

dM

S(m

atsy

mm

etry

),M

D2/

MD

1;K

W,

Kru

skal

–Wal

lis;

SD,

stan

dard

devi

atio

n;P

,pr

obab

ility

valu

e.Pa

irw

ise

com

pari

sons

betw

een

popu

latio

nsca

lcul

ated

from

the

Man

n–W

hitn

eyU

test

havi

ngdi

ffer

ent

lette

rsar

est

atis

tical

lysi

gnif

ican

t(P

£0.

05).

*A

llth

eun

dist

urbe

dsi

tes

wer

epo

oled

toge

ther

form

ing

the

colu

mn

‘‘Und

istu

rbed

’’.

{ Mar

ksva

riab

les

that

wer

esi

gnif

ican

t(P

£0.05).

{ For

this

vari

able

,on

lysa

mpl

esw

ithin

the

rang

eof

aM

D1

ofup

to25

cmw

ere

kept

(six

case

sw

ere

dele

ted

from

site

Ato

even

out

sam

plin

gsi

zes

betw

een

grou

ps).

Au and Tardif 591

# 2007 NRC Canada

live mat area than between stem diameter and entire matarea (r = 0.675; P < 0.001; n = 97).

Shrub dimensions and ageIn Churchill, the maximum age (68 years) was observed

for a shrub sampled at an undisturbed site. This is compara-ble to that of 80 years of age recorded for Dryas octopetalaL. in alpine conditions (Schweingruber and Dietz 2001b),although Schweingruber and Poschlod (2005) have madereference to a 108-year-old individual found at extreme con-ditions (Schroter 1926). Alternatively, Svoboda (1974) hadestimated many D. integrifolia shrubs at Devon Island to be80–120 years old based upon mean shoot increment per yearand mean shoot lengths. At disturbed sites in Churchill, themaximum shrub age observed was 47 years. Considering theestablishment years of the oldest D. integrifolia shrubs atthese sites: site D, 1957; site C, 1959; site B, 1962; and siteA, 1994 (we disregarded the 20-year-old outlier, which webelieve was collected at the roadside adjacent to the nearbyscattered network of trees), all dates agreed with those de-rived from available aerial photos (see section on Samplingsites in Materials and methods). The short temporal lag be-tween road construction estimated from aerial photos andD. integrifolia establishment may simply reflect the possibil-ity that the oldest individuals at each site were not sampled,or that shrub establishment had been delayed. In all cases,the close agreement between these dates provides furtherconfidence in the annual character of the growth ring inD. integrifolia shrubs.

At site A, shrubs required much less time to develop sim-ilar dimensions to those at other sites, and the relationshipbetween shrub dimension and age was best approximatedby a linear regression. This relationship was much weakerand irregular when considering individuals from sites B, C,and D, and undisturbed ones. Other studies have also re-ported that shrubs and small trees of similar size varied

widely with growth rate (Ward 1982; Zammit and Zedler1992; Grice et al. 1994; Martin and Moss 1997). Less com-paction of the substrate at the more recently disturbed site Amay be one explanation. Firlotte (1998) indicated that thecompaction of gravel pads in the Churchill region has cre-ated conditions unsuitable for root penetration into the sub-strate, and Babb (1977) observed that D. integrifolia shrubsexposed to damage by vehicle movement would unlikely re-cover for a considerable period. Similarly, Woodcock andBradley (1994) attributed differences in microsite sensitivityas one of the factors causing the low correlation betweenshrub age and stem growth rate in arctic willow (Salix arc-tica Pall.). It should also be noted that the lower growthrates of shrubs at disturbed sites other than site A may be,in part, due to older shrubs producing smaller annual radialincrements, although little overlap in age exists between siteA and the other disturbed sites. Interestingly, the radialgrowth of shrubs at undisturbed sites decreased with age,suggesting that competition could become an important fac-tor for shrubs in areas where a continuous cover of vegeta-tion is formed.

Another explanation for this inconsistency may be topo-graphic aspect. Shrubs on the crests and slopes of Devon Is-land, NWT, developed more rapidly following snowmelt(Svoboda 1974). Beck et al. (2005) also found that the oc-currence of D. octopetala was positively related to a highermean June temperature and exposure to a south-facing as-pect. However, some other factors potentially affectinggrowth rate can be disregarded. McGraw (1985) concludedfrom experimentation with differential manipulations onecotypes of D. octopetala in Alaska, that variability in eithernutrient or moisture availability had no direct significant ef-fect on the growth of D. octopetala. Dryas integrifolia wasalso reported to have a relatively high tolerance to soil mois-ture regimes (Svoboda 1974). Shrub dimension, therefore,cannot be used as a surrogate for age data as a result of

Table 2. Selected Dryas integrifolia growth rates, allometric proportions and status ratios calculated for disturbed sites (total n =96).

A B C D KW

Variable n Mean SD n Mean SD n Mean SD n Mean SD P

MD1/ age 23 3.26b 1.75 19 1.22a 0.95 24 0.88a 0.50 30 0.99a 0.84 0.000*MA/age 21 46.35 52.77 19 52.94 98.00 24 17.51 27.06 30 25.75 52.15 0.096MP/age 21 19.91b 10.34 19 6.46a 4.77 24 6.13a 3.86 30 6.19a 5.90 0.000*MSD/age 23 0.69b 0.35 19 0.22a 0.10 24 0.21a 0.08 30 0.18a 0.09 0.000*TRL/age{ 17 5.17b 3.15 10 1.86a 0.88 10 3.07b 1.03 0 — — 0.002*Mean dieback/age 23 0.54a 1.15 19 1.42b 0.51 24 1.48b 0.49 30 2.06c 1.16 0.000*Flower abundance /age 23 3.05 6.29 19 0.21 0.32 24 0.25 0.39 30 0.65 1.11 0.692MS/age 23 0.14b 0.07 19 0.02a 0.01 24 0.02a 0.03 30 0.02a 0.02 0.000*MD1/ MSD 23 4.83 0.80 19 5.02 2.09 24 4.30 2.02 30 4.81 1.91 0.245MD1/ TRL{ 17 0.56c 0.17 10 0.36b 0.26 10 0.19a 0.06 0 — — 0.000*MD1/ MA 21 0.35 0.53 19 0.16 0.23 24 0.22 0.39 30 0.29 0.49 0.328MD1/ MP 21 0.18bc 0.03 19 0.19c 0.02 24 0.16a 0.05 30 0.16ab 0.02 0.001*Flower abundance/area 21 0.03 0.05 19 0.01 0.01 24 0.01 0.02 30 0.03 0.03 0.186Flower abundance/LA 21 0.03 0.05 19 0.01 0.02 24 0.03 0.03 30 0.06 0.06 0.142

Note: Summing the highest number of cases at each site will result in 96 individuals. MD1, longest mat diameter; MA, mat area; MP, mat peri-meter; MSD, mean stem diameter; TRL, taproot length; MS, mat symmetry; and LA, live area; KW, Kruskal–Wallis; SD, standard deviation; P,probability value. Pairwise comparisons calculated from the Mann–Whitney U test having different letters are statistically significant (P £ 0.05).*Marks variables that were significant (P £ 0.05).{Only samples within the range of a MD1 of up to 25 cm were kept (six cases were deleted from site A to even out sampling sizes between groups).

592 Can. J. Bot. Vol. 85, 2007

# 2007 NRC Canada

highly variable growth rates among D. integrifolia shrubs.Alternatively, the accumulation of detrimental effects (meandieback) was generally shown to follow a two-parameter ex-ponential model with age at disturbed sites. However, lowdieback values of some shrubs at undisturbed sites may re-flect a lack of damage from disturbance.

Dryas integrifolia chronology and its association withclimate

The secondary objective of the study was to evaluatewhether D. integrifolia shrubs could be cross-dated andtherefore have a potential use in dendrochronology. Discon-tinuous and missing rings were common, and this was alsoobserved in Salix spp. (Beschel and Webb 1963; Woodcockand Bradley 1994; Zalatan and Gajewski 2006), in Ceano-thus crassifolius Torr. (Gordon 1991), in Pteronia pallensL. f. (Milton et al. 1997), and in Empetrum hermaphroditum

Hagerup and other dwarf alpine shrubs (Schweingruber andDietz 2001a; Bar 2005). Moreover, boundaries betweenrings were occasionally difficult to recognize, and recent an-nual increments were often present only at certain locationsalong the stem transverse section in older individuals. Thus,many stem sections had to be rejected, and only a selectnumber were successfully cross-dated.

Despite these difficulties, the D. integrifolia residual chro-nology was found to correlate significantly with both Octo-bert–1 minimum temperature and snowfall (positive andnegative correlation, respectively) of the year prior to ringformation. Elevated temperature for Octobert–1 in theChurchill region has also been shown to be positively asso-ciated to radial growth in P. glauca (Jacoby and Ulan 1982;Girardin et al. 2005) as well as in P. mariana andL. laricina (Girardin et al. 2005). October is usually thetime when snowfall begins in Churchill. It is speculated that

MD1 (cm)

0 40 80 120 160

MS

D(m

m)

0

4

8

12

16

20

AAAAAAA

AAA

AAAAA

AAA

AA

AA

A

BBBBBBBB

BBB

B

BB

B B

B

BB

CC

CCCCCCC

CC

C

C

C

C

CC

CC

C

C

C

C

C

DDDDDDDD

DDDDDDDD

DDD

DDD

DDD

D

D D

D

D

MD1 (cm)

0 40 80 120 160

MA

(cm

2x10

2)

0

40

80

120

160

200

AAAAAAAAAAAAA AAA

AA AAA

BBBBBBBB BBB BBBBB

B

B

B

CCCCCCCCCCCCCCC CCCCCCC C

C

DDDDDDDDDDDDDDDDDDDDDDDDDDD

DD

D

MD1 (cm)

0 40 80 120 160

MP

(cm

)

0

200

400

600

800

1000

AAAAAAAA

AAAAA A

AA

AAAA

A

BBBBBBBB

BBB B

BBB

B B

B B

CCCCCCCCCC

CC

CC

C

CCCCCC

CC

C

DDDDDDDDDDDDDDDDDDDDDDD

DDDD

DD

D

MD1 (cm)

0 15 30 45 60

TR

L(c

m)

0

30

60

90

120

150

AAAAAAAAAA

AA A

A AA

AA

AA

A

A

A

B

B

B

B

B

BBB

BB

CC

CC

C

C

C

C

C C

(A) (B)

(C)

(E)

MSD (mm)

0 2 4 6 8 10

TR

L(c

m)

0

30

60

90

120

150

AAA AAAAAAA

AA A

A AA

A A

AA

A

A

A

B

B

B

B

B

BB

B

BB

CC

CC

C

C

C

C

C C

(D)

Fig. 3. Allometric relationships for Dryas integrifolia shrubs (A–E). MD1, longest mat diameter; MSD, mean stem diameter; MA, mat area;MP, mat perimeter; and TRL, taproot length. Individuals from both disturbed and undisturbed sites are indicated. Linear regressions (darkgray) in subfigures (C) and (D) used data from site A only to illustrate a departure from those of other sites. (A), site A; (B), site B; (C), siteC; *, undisturbed site.

Au and Tardif 593

# 2007 NRC Canada

warm October temperatures coupled with low snow accumu-lation in the autumn may delay the onset of ground snowcover and further extend the length of the growing seasonduring ring formation. In the future, it would be desirable toproduce longer chronologies with more site replication tofurther validate these correlations and extend them at the re-gional level. However, the likeness of both erect trees andD. integrifolia shrubs in response to climate reaffirms thevalidity of the climate signal in the D. integrifolia chronol-ogy and demonstrates that climate is a limiting factor ofsimilar importance to different growth forms at the tree linein the subarctic region of Churchill. This response to climateis probably limited in scope and may not encompass higherlatitude regions, where climate places different growth re-strictions on the species. But given that spruce trees sampled

from both western Alaska and the western shore of HudsonBay have been shown to cross-date and respond similarly tosummer temperature (Giddings 1954), it is hypothesized thatD. integrifolia shrubs along the tree line may also share acommon climatic signal. This remains to be tested, however,and would require further dendrochronological studies ofthis species elsewhere along the tree line.

Moreover, the results indicated that warmer than averageminimum Mayt temperatures during the year of ring forma-tion had a negative association with radial growth. Girardinet al. (2005) have also observed this relationship forP. mariana and L. laricina, which they associated with anearly snowmelt at a time when the ground has not yetthawed. This may later reduce the infiltration of water intothe substrate by increasing evaporation and (or) runoff

0 10 20 30 40 50

D1

(cm

)M

0

40

80

120

160

AAAAAAAAAAAAAA

AA

A

AAAA

A

B BBB BBBBB BB

BBBB

B

B

BB

CCCC CCC

C CCC CCC

C

CCC CCC CC

C

DDDD DDDDDDD DD DD D

DDD DDDDDD

D D

DD

D

A

0 20 40 60 80

MS

D(m

m)

0

5

10

15

20

25

AAAAAAAAAA

AAAAAAA

AAAA

A

B BBBBBBBBBBBB

BBB

B

BB

CC

CCCCCC CCC

CC

C

C

CC

CCC

CC

C

C

DDDDDDDD

D DDDD DD D

DD

D

DDDDDDD

DD

D

D

A

0 10 20 30 40 50

TR

L(c

m)

0

40

80

120

160

AAAAAAAAAA

AAAAA

AA

AAA

A

A

B

B

B

BB

BBB

BB

CC

CC

C

C

C

CCC

A

(A)1

(C)1

(B)1

Age

AgeAge

Age0 20 40 60 80

Mean

die

back

(%)

0

15

30

45

60

75

90

A

AAAAAAAAAAAAAA

A

A

AAAAA A

B

B

B

B

B

B

BB

BB

B

B

B

B

B

B

B

B

B

CC

C

C

C

CCC

CC

C

C

C

CCC

CC

CC

CC

C

C

DDDD

D

D

DD

DDD

D

D

D

D

D

D

D

DD

DD

D

DD

D

D

DD

D

(D)2

Age0 20 40 60 80

MS

D/A

ge

(mm

/year)

0.0

0.5

1.0

1.5

AAA

A

AA

A

AA

A

A

AAAAA

A

A

AA

A

A

A

B BBB BBBBBBBBB

BBB

BBBCC CCCCCC CCC CC

C

CCC

CCC

C

C

C

CDDDD DDDDD

DD

DD DD D DDD DDDDDDD

D

D D D

(E)3

Fig. 4. Relationships between age of Dryas integrifolia shrubs and (A) longest mat diameter (MD1), (B) mean stem diameter (MSD), (C)taproot length (TRL), (D) mean dieback, and (E) MSD/age. Individuals from both disturbed and undisturbed sites are indicated. The linearregressions of data from site A excluded the 20-year-old outlier. (A)1–(C)1 The linear regressions in these subfigures were calculated ex-cluding shrubs from undisturbed sites. A separate regression was calculated for site A. (D)2 This two-parameter exponential regression wascalculated with data from all disturbed sites excluding undisturbed sites. (E)3 This inverse first-order regression was calculated with datafrom undisturbed sites only. (A), site A; (B), site B; (C), site C; *, undisturbed site.

594 Can. J. Bot. Vol. 85, 2007

# 2007 NRC Canada

(Corbet 1972; Scott et al. 1993). At Victoria Island, North-west Territories, Canada, Zalatan and Gajewski (2006) alsosuggested that spring snowmelt limited radial growth of theshrub, Salix alaxensis (Andersson) Cov., in xeric conditionsby the amount of recharged soil moisture. Alternatively, alate snowmelt was reported to delay the onset of mat growthin D. integrifolia shrubs at one site and lower fruiting suc-cess as a result of the short growing season (Hart and Svo-boda 1994). This may be illustrated in our results by thenegative trend between radial growth and Junet snowfall.An exceptionally early or late snowmelt may therefore havenegative consequences to the growth of these shrubs.

Conclusion

It was expected that knowledge of D. integrifolia growth

rates could aid restoration ecologists to predict the durationrequired to revegetate a disturbed site. However, the ten-dency of D. integrifolia shrubs and other shrub species toexhibit inconsistent growth rates among sites strongly limitthe use of shrub size as an indicator of age. We speculatethat microsite condition (e.g., topography, substrate compac-tion) and recurrent disturances may be at the origin of thesedifferences in size/age ratios, and this should be addressedin future work on the species. Sites with undisturbed condi-tions where shrubs were shown to be the oldest, albeit notcross-dated, also merit further study.

It was found that the D. integrifolia shrubs near Churchill,Manitoba, which were cross-dated, displayed a radialgrowth – climate association similar to that of P. glauca,P. mariana, and L. laricina in the area. This finding furtherstrengthens the fact that climate is recorded in the ring

Fig. 5. The residual chronology (black line) and corresponding sampling depth (broken line) extend from 1962 to 2004 (A). Pearson’s cor-relation coefficients between the Dryas integrifolia residual chronology and climate variables for the period 1978–2004 (B and C). Correla-tion coefficients with monthly snowfall and total precipitation (B), and monthly minimum and maximum temperatures (C) are indicated.Probability values are indicated by the letters (X: P < 0.01 and Y: P < 0.05). Missing climate data were not estimated, and correlationanalyses were performed with n ranging from 22 to 27.

Au and Tardif 595

# 2007 NRC Canada

widths of D. integrifolia shrubs and opens up potential forfuture work regarding whether the species would display asimilar or different response to climate elsewhere along orbeyond the tree line. Moreover, longer ring-width series andbetter site replication would be needed to strengthen ourunderstanding of the radial growth – climate association ofthis species. The results of this study suggest that, like trees,endogenous (microsite) and exogenous (climatic) effects arealso reflected in the growth of D. integrifolia shrubs in sub-arctic Manitoba.

AcknowledgementsWe thank D. Ko Heinricks for his field participation and

assistance, B. Jones for preparing all microtome cross-sections, and B. Epp who provided technical assistance.Thanks are especially due to F. Conciatori, dendroecologytechnician, for having cross-dated all specimens. We ac-knowledge Dr. G. Scott and Dr. R. Staniforth for their pre-vious comments on this project. We also acknowledge thecontribution of the anonymous reviewers and associate editorin making this manuscript much better. This research wasundertaken, in part, thanks to funding from the CanadaResearch Chairs Program, the Natural Sciences and Engi-neering Research Council of Canada (NSERC), the ChurchillNorthern Research Centre, the Northern Scientific TrainingProgram, as well as the University of Winnipeg.

ReferencesBabb, T.A. 1977. High arctic disturbance studies associated with

the Devon Island project. In Truelove Lowland, Devon Island,Canada: a high arctic ecosystem. Edited by L.C. Bliss. Univer-sity of Alberta Press, Edmonton, Alta. pp. 647–654.

Bar, A. 2005. Citing online sources: annual conference for the as-sociation of tree-ring research, dendroecological investigationswith dwarf shrubs and their potential for ecosystem reconstruc-tions in high mountain ecosystems [online]. Available fromwww.unifr.ch/geosciences/geographie/Trace/Abstracts.html [ac-cessed 25 March 2006].

Beck, P.S.A., Kalmbach, E., Joly, D., Stien, A., and Nilsen, L.2005. Modeling local distribution of an Arctic dwarf shrub indi-cates an important role for remote sensing of snow cover. Re-mote Sens. Environ. 98: 110–121. doi:10.1016/j.rse.2005.07.002.

Benedict, J.B. 1989. Use of Silene acaulis for dating: the relation-ship of cushion diameter to age. Arct. Alp. Res. 21: 91–96.doi:10.2307/1551520.

Beschel, R.E., and Webb, D. 1963. Growth ring studies on arcticwillows. In Axel Heiberg Island research reports. Preliminary re-port 1961–1962. Edited by F. Muller. McGill University, Mon-treal, Que. pp. 189–197.

Cook, E.R., and Holmes, R. 1986. Guide for computer programARSTAN. Laboratory of Tree-Ring Research, The Universityof Arizona, Tucson, Ariz.

Corbet, P.S. 1972. The microclimate of arctic plants and animals,on land and in fresh water. Acta Arct. 18: 1–43.

Dietz, H., and Schweingruber, F.H. 2002. Annual rings in nativeand introduced forbs of lower Michigan. USA. Can. J. Bot. 80:642–649. doi:10.1139/b02-048.

Dredge, L.A., and Nixon, F. 1992. Glacial and environmental geol-ogy of northeastern Manitoba. Geological. Survey of CanadaMemoirs 432. Canada Communication Group, Ottawa, Ont.

Environment Canada. 2004. Canadian Climate Normals 1971–2000[online]. Available from climate.weatheroffice.ec.gc.ca/

climate_normals/index_e.html [updated 25 February 2004; ac-cessed 10 March 2006].

Firlotte, N. 1998. The revegetation of disturbed dry tundra areasnear Churchill, Manitoba. M.Sc. thesis, Department of Botany,The University of Manitoba, Winnipeg, Man.

Giddings, J.L., Jr. 1954. Tree-ring dating in the American arctic.Tree-Ring Bull. 20: 23–25.

Girardin, M.P., Berglund, E., Tardif, J., and Monson, K. 2005. Radialgrowth of tamarack (Larix laricina) in the Churchill area(Manitoba) in relation to climate and larch sawfly (Pristiphoraerichsonii) herbivory. Arct. Antarct. Alp. Res. 37: 206–217.doi:10.1657/1523-0430(2005)037[0206:RGOTLL]2.0.CO;2.

Gordon, H.J. 1991. Dendrochronological methods applied to theshrub Ceanothus crassifolius in southern California: precipitationas a factor in ring width variability. M.Sc. thesis, Departmentof Environmental Studies, California State University, Fullerton,Calif.

Grice, A.C., Westoby, M., and Torpy, C. 1994. Dynamics and po-pulation structure of Acacia victoriae Benth. Austral Ecol. 19:10–16. doi:10.1111/j.1442-9993.1994.tb01537.x.

Hart, G.T. 1988. Dryas integrifolia and its habitat diversity in theCanadian high and low arctic. M.Sc. thesis, Department of Bot-any, The University of Toronto, Toronto, Ont.

Hart, G.T., and Svoboda, J. 1994. Autecology of Dryas integrifoliain Alexandra Fiord lowland habitats. In Ecology of a polar oasis:Alexandra Fiord, Ellesmere Island, Canada. Edited by J. Svobodaand B. Freedman. Captus University Publications, Toronto,Ont. pp. 145–156.

Holmes, R.L. 1983. Computer assisted quality control in tree-ringdating and measurement. Tree-Ring Bull. 43: 69–78.

Jacoby, G.C., and Ulan, L.D. 1982. Reconstruction of past ice con-ditions in a Hudson Bay estuary using tree rings. Nature(London), 298: 637–639. doi:10.1038/298637a0.

Johnson, K.L. 1987a. Wildflowers of Churchill and the HudsonBay Region. Manitoba Museum of Man and Nature, Winnipeg,Man.

Johnson, L.A. 1987b. Management of northern gravel sites for suc-cessful reclamation: a review. Arct. Alp. Res. 19: 530–536.doi:10.2307/1551421.

Johnstone, J.F., and Henry, G.H.R. 1997. Retrospective analysis ofgrowth and reproduction in Cassiope tetragona and relations toclimate in the Canadian High Arctic. Arct. Alp. Res. 29: 459–469. doi:10.2307/1551993.

Kolishchuk, V.G. 1990. Dendroclimatological study of prostratewoody plants. In Methods of dendrochronology. Edited byE.R. Cook and L.A. Kairiukstis. Kluwer Academic Publishers,Dordrecht, the Netherlands. pp. 51–55.

Kuivinen, K.C., and Lawson, M.P. 1982. Dendroclimatic analysisof birch in South Greenland. Arct. Alp. Res. 14: 243–250.doi:10.2307/1551156.

Lageard, J.G.A., Wilson, D.B., Cresswell, N., Cawley, L.E., Jones,H.E., and Caporn, S.J.M. 2005. Wood growth response of Cal-luna vulgaris (L.) Hull to elevated N deposition and drought.Dendrochronologia, 23: 75–81. doi:10.1016/j.dendro.2005.08.001.

Martin, D., and Moss, J. 1997. Age determination of Acacia tortilis(Forsk.) Hayne from northern Kenya. Afr. J. Ecol. 35: 266–277.doi:10.1111/j.1365-2028.1997.067-89067.x.

McGraw, J.B. 1985. Experimental ecology of Dryas octopetalaecotypes. III. Environmental factors and plant growth. Arct.Alp. Res. 17: 229–239. doi:10.2307/1551013.

Milton, S.J., Gourlay, I.D., and Dean, W.R. 1997. Shrub growthand demography in arid Karoo, South Africa: inference fromwood rings. J. Arid Environ. 37: 487–496. doi:10.1006/jare.1996.0292.

596 Can. J. Bot. Vol. 85, 2007

# 2007 NRC Canada

Philipp, M., Bocher, J., Mattsson, O. and Woodell, S.R.J. 1990. Aquantitative approach to the sexual reproductive biology and po-pulation structure in some arctic flowering plants: Dryas integri-folia, Silene acaulis and Ranunculus nivalis. Medd. Grønl. 34:3–60.

Porsild, A.E. 1964. Illustrated flora of the Canadian Arctic Archi-pelago. 2nd ed. National Museum of Canada, Bulletin 146, Ot-tawa, Ont.

Porsild, A.E., and Cody, W.J. 1980. Vascular plants of continentalNorthwest Territories, Canada. Natural Museum of NaturalSciences, Ottawa, Ont.

Rayback, S.A., and Henry, G.H.R. 2006. Reconstruction of summertemperature for a Canadian high arctic site from retrospectiveanalysis of the dwarf shrub, Cassiope tetragona. Arct. Antarct.Alp. Res. 38: 228–238. doi:10.1657/1523-0430(2006)38[228:ROSTFA]2.0.CO;2.

Regent Instruments, Inc. 2005. WinFolia 2005b: image analysissoftware [computer program]. Quebec, Que.

Rouse, W.R. 1991. Impacts of Hudson Bay on the terrestrial cli-mate of the Hudson Bay Lowlands. Arct. Alp. Res. 20: 56–66.

Schroter, C. 1926. Das Pflanzenleben der Alpen. 2ed ed. AlbertBaustein, Zurich.

Schweingruber, F.H. 1996. Tree rings and environment dendroecol-ogy. Paul Haupt Publishers, Bern, Switzerland.

Schweingruber, F.H., and Dietz, H.J. 2001a. Annual rings in thexylem of dwarf shrubs and perennial dicotyledonous herbs. Den-drochronologia, 19: 115–129.

Schweingruber, F.H., and Dietz, H.J. 2001b. Jahrringe in Krautp-flanzen und Zwergstrauchern. Inf. Bl. Forsch. Bereich. Landsch.49: 1–2.

Schweingruber, F.H., and Poschlod, P. 2005. Growth rings in herbsand shrubs: life span, age determination and stem anatomy. For.Snow Landsc. Res. 79: 195–415.

Scoggan, H.J. 1978. The flora of Canada, part 3 – Dicotyledoneae(Saururaceae to Violaceae). National Museum of NaturalSciences, National Museums of Canada, Ottawa, Ont.

Scott, G.A.J. 1995. Canada’s vegetation: a world perspective.McGill-Queen’s University Press, Montreal, Quebec.

Scott, P.A., Hansell, R.I.C., and Erickson, W.R. 1993. Influences ofwind and snow on northern tree-line environments at Churchill,Manitoba, Canada. Arctic, 46: 316–323.

Sheskin, D.J. 1997. Handbook of parametric and nonparametric sta-tistical procedures. CRC Press LLC, Boca Raton, Fla.

Smith, R.E., Veldhuis, H., Mills, G.F., Eilers, R.G., Fraser, W.R.,and Lelyk, G.W. 1998. Terrestrial ecozones, ecoregions, andecodistricts of Manitoba: An ecological stratification of Manito-ba’s natural landscapes. Tech. Bull. 1998–9E. Land ResourceUnit, Brandon Research Centre, Research Branch, Agricultureand Agric.-Food Canada, Winnipeg, Man.

Svoboda, J. 1974. Primary production processes within polar semi-desert vegetation, Truelove Lowland, Devon Island, N.W.T., Ca-nada. Ph.D. thesis, Department of Botany, The University of Al-berta, Edmonton, Alta.

SYSTAT. 2004a. SYSTAT 11: statistics I. Systat Software Incor-porated, Richmond, Calif.

SYSTAT. 2004b. SigmaPlot 9.01. Systat Software Incorporated,Richmond, Calif.

Ward, L.K. 1982. The conservation of juniper: longevity and oldage. J. Appl. Ecol. 19: 917–928.

Wilson, J.W. 1964. Annual growth of Salix arctica in the High-Arctic. Ann. Bot. (Lond.), 28: 71–76.

Woodcock, H., and Bradley, R.S. 1994. Salix arctica (Pall.): its po-tential for dendrochronological studies in the High Arctic. Den-drochronologia, 12: 11–22.

Zalatan, R., and Gajewski, K. 2006. Dendrochronological potentialof Salix alaxensis from the Kuujjua river area, Western Cana-dian arctic. Tree-Ring Res. 62: 75–82.

Zammit, C.A., and Zedler, P.H. 1992. Size structure and seed pro-duction in even-aged populations of Ceanothus greggii in mixedchaparral. J. Ecol. 81: 499–511.

Au and Tardif 597

# 2007 NRC Canada