Peatland vegetation organization and dynamics in the western subarctic, Northwest Territories, Canada

Transcription

1 Peatland vegetation organization and dynamics in the western subarctic, Northwest Territories, Canada M. A. JASIENIUK Department of Biology, McGill University, Montreal, P.Q., Canada H3A 1B1 AND E. A. JOHNSON Department of Biology, University of Calgary, Calgary, Alta., Canada E'N IN4 Received June 9, 1980 JASIENIUK, M. A., and E. A. JOHNSON Peatland vegetation organization and dynamics in the western subarctic, Northwest Territories, Canada. Can. J. Bot. 60: Species frequency abundance was determined in 77 peatland stands and the variance partitioned between the effects of fire and habitat. Each variance partition was then analyzed using principal component analysis. The resulting habitat ordination was interpreted in terms of a moisture-nutrient gradient and a sphagnum substrate gradient. The fire frequency ordination gave a gradient of plant adaptations to the interval between fires. Vegetation change after fire is predominantly change in species abundance. It consists of a regeneration to the predisturbance species composition which depends on the moisture-nutrient and sphagnum substrate conditions. Changes in species composition are suggested to be due to major changes in habitat characteristics usually the result of water level fluctuations. A discussion of the frequency of disturbance due to water level fluctuations and fire for different peatland communities is presented. JASIENIUK, M. A., et E. A. JOHNSON Peatland vegetation organization and dynamics in the western subarctic, Northwest Territories, Canada. Can. J. Bot. 60: La frkquence et l'abondance des esptces ont CtC dcterminces dans 77 peuplements de tourbitres et une partition des variances selon les effets des feux et les effets des habitats a ensuite CtC effectuce. Chaquepartition de variance a ensuite CtC analysce par les composantes principales. L'ordination de l'habitat resultant de cette analyse est interprctce selon un gradient d'humiditk - ClCments nutritifs et un gradient de substrat sphagnique. Quant h l'ordination de la frcquence des feux, elle rcvtle un gradient d'adaptations h l'intervalle entre les feux. Le changement de la vcgktation aprts un feu est surtout un changement dans l'abondance des esptces. I1 consiste en une regknkration vers la composition spkcifique antcrieure h la perturbation, cette composition dtpendant des conditions d'humiditt - ClCments nutritifs et des conditions de substrat sphagnique. Les changements de la composition spccifique semblent Ctre dus a des modifications importantes dans les caractkristiques du milieu, ces modifications rcsultant habituellement de fluctuations de la nappe phrkatique. Une discussion est consacrke a la frcquence des perturbations dues aux fluctuations de la nappe phreatique et au feu dans diverses communautcs des tourbitres. [Traduit par le journal] Introduction Studies of peatland vegetation dynamics have relied heavily on the identification of successional sequences. Two approaches have predominated: spatial-temporal conversion and stratigraphy. In the spatial-temporal conversion approach, peat was assumed to accumulate with time and induce environmental changes favoring the selection of successively less hydrophytic communities. The spatial arrangement of vegetation types around open water was considered to reflect this ordering in time. As a result this approach consisted of describing the vegetation associations around an open pool and placing them according to their spatial position and (or) composition in a successional sequence (Transeau 1903; Dachnowski 1912, 1924; Cooper 1913; Clements 1916; Tansley 1939; Gates 1942; Conway 1949). These hydroseres received most attention in early North American literature. However, long-term records of vegetation changes around some of these bog pools show that the spatial-temporal approach does not accurately describe the dynamics of these peatlands. In fact the stages of hydroseral succession are often disrupted as a result of fluctuations in water level (Schwintzer and Williams 1974). Furthermore, these water level fluctuations, significant with respect to their affect on vegetation, can occur as frequently as every 10 years (Schwintzer 1978). In the stratigraphic approach, the vegetation history as recorded in the sequence of macrofossils, pollen, and peat types was used to represent peatland succession (e.g., Granlund 1932; Godwin 1946; Conway 1948; Kulczynski 1949; Pearsall 1950). These early studies gave a much more deterministic interpretation than have recent ones which use more sophisticated techniques (e.g., Heinselman 1963, 1970; Janssen 1967, 1968; Griffin 1975, 1977). The latter have revealed that a /82/ $01.OO/O National Research Council of Canada/Conseil national de recherches du Canada

2 2582 CAN. J. BOT. VOL variety of transitions between vegetation types exist (Walker 1970) and that vegetation change in peatlands is a stochastic process occurring continuously and dependent on innumerable local and regional events (Heinselman 1970, 1975). From both long-term records of spatial-temporal studies and recent stratigraphic evidence, it is clear that the recurrence of numerous natural disturbances frequently alters the simple vegetation succession due to peat accumulation. In this study we assume that the variation in vegetation composition in a regional collection of stands reflects two environmental complexes. The first environmental complex is fire. It is generally of such frequency and intensity as to cause changes in species abundance but little long-term effect on the habitat. The second environmental complex is covered by the term habitat. The water regime (water level) and nutrient regime (access to mineral rich waters) are the important factors and only change significantly over longer periods. These changes however result in shifts in species composition not just abundance. The problem then is to partition the variation in species abundance into that correlated with time since fire and a residual which is largely correlated with habitat variables. To do this a multivariate regression is used to partition the variation in species abundance using the concomitant variable age since fire. Each of the resulting partitions are then subject to principal components analysis. The partitioning is necessary because the two environmental complexes act at different time scales. Study area The area studied is east of Great Slave Lake, Northwest Territories, extending from 60 to 62ON latitude and 104 to 112" W longitude. The climate is dry subhumid continental and at Fort Smith (60" N, 112" W) the mean total precipitation is 33 cm, mean January temperature is -26 C and mean July temperature is 17 C. Average temperature and precipitation decrease toward tree line. Permafrost is present in most raised and some forested nonraised bogs throughout July and August at a mean depth of 35 cm but is absent in wet peatlands. The region is located on the Precambrian Shield. The bedrock consists mainly of granites, gneissic granites, and granodiorites, overlain by noncalcareous glacial drift of varying thickness. The entire area was glaciated and glacial landforms dominate the landscape. East of 108" W long, the terrain is of reduced relief with thick deposits of drumlinized drift. West of 110" W long, the terrain becomes more rugged and bedrock controlled, and has a thinner covering of drift. The extent of peatlands decreases from east to west. In general, lakes in this region are low in dissolved ions as a consequence of the granitic substrates. The concentrations of dissolved ions in lake and first-order stream waters are only vegetation in the region is characteristically ombrotrophic with little fen development. Upland vegetation is mostly dominated by open black spruce (Picea mariana') and jack pine (Pinus banksiana) forests with lichen understories of Stereocaulon paschale, Cladonia mitis, and Cetraria nivalis (cf. Johnson 1981). Peatland vegetation is generally dominated by open forests of black spruce with infrequent occurrences of tamarack (Larix laricina). At the base of upland slopes and near runoff channels, the understory vegetation consists of Sphagnum fuscum and ericaceous shrubs (Ledum groenlandicum, Rubus chamaemorus, Oxycoccus microcarpus). Black spruce - lichen (Cladonia mitis, C. coccifera) stands occur in areas less topographically favourable for moisture input. Wet depressions and pools with aquatic sphagna (Sphagnum angustifolium, S. riparium) mats and (or) sedges (Carex limosa, C. paupercula, Eriophorum russeolum) are common within forested peatlands. Treeless lichen-covered (Cetraria nivalis, Cladonia mitis, C. coccijera) peat plateaus become more common and polygonally patterned towards tree line. For a flora of the region see Jasieniuk and Johnson (1979)., Lightning fires are a frequent and natural occurrence in the study area and follow a seasonal occurrence pattern advancing towards tree line in the summer and retreating in the fall. Because of this, more fires and larger areas burn in the southwest than in the northeast (Johnson and Rowe 1975). The expected recurrence time of fire on uplands is between 55 and 100 years based on studies of stand age and fire intervals (Johnson 1979). Ages of trees in peatlands are found to be considerably older, often greater than 200 years. Methods One hundred and three peatland stands (Fig. I) were sampled to give a representation of different peatland vegetation compositions, environments, and ages since fire. The criteria for selection of stands required that peat accumulation be equal to or greater than 25 cm, and that the vegetation, physical environment, and age since fire be homogeneous in a circular 100-m2 stand. The size of the stand was chosen to insure relatively homogeneous vegetation and physical environment. Density, diameter at breast height (DBH), and height (using an Abney level) were recorded for all trees (22 cm DBH) within a stand. Frequency of all species of understory shrubs, herbs, lichens, and bryophytes (i.e., everything <2 cm DBH) was estimated from 10 randomly placed 0.6-m2 quadrats in each stand. Initially 20 quadrats were sampled but frequency of occurrence of the major species differed by less than 10% from that measured for only 10 quadrats. As a result 10 quadrats were considered sufficiently representative of a stand's species composition and abundance. Sampling fewer quadrats speeded sampling and allowed for an increase in the number of stands sampled. Ages since fire were determined for each stand. For stands less than 10 years old, exact dates were available from Indian and Northern Affairs fire records. For older stands, ages were slightly higher than in the precipitation. Specific electrical '~omenclature of vascular plants follows Scoggan (1979, conductance (at 20 C) of precipitation is 9 pmho (I mho = 15) Sphagnum species follows Vitt and Andrus (1977), lichens and of surface lake water is 20 pmho at Porter Lake (61 42' N, follows Hale and Culberson (1970) except for Cladonia which 108" W). Owing to this low ion concentration, peatland follows Thomson (1967).

3 JASIENlUK AND JOHNSON 2583 with 10 ml 0.5 M CaC12 solution and allowed to equilibrate for 4 h with intermittent hand shaking. The suspensions were then centrifuged for 30 s and ph measurements were taken of the supernatant. Specific electrical conductance was measured in supernatants of slurries of surface peat and distilled water that were allowed to equilibrate for 4 h. Specific electrical conductance is a measure of the mobility of solute ions in a solution. Although it is not simply related to ionic concentration owing to differences in dissociation, proportions, and types of ions, it can be used to give a general indication of nutrient ion status. In bog waters much of the specific electrical conductance can be attributed to the high hydrogen ion concentrations. The standard procedure has been to subtract the contribution of the hydrogen ions from the measured conductance (Sjors 1950; Gorham 1956; Malmer 1963). Hydrogen ions have an ionic conductivity several times greater than that of other ions and, for the high concentrations (low ph) found in this study, the remaining or reduced conductivity is zero or even negative in most stands. Similar results have been shown by Malmer (1963) and Hofstetter (1969). No adequate method of solving this problem has been suggested. In this study phh2, measurements are uniformly low across all stands such that the differences that do occur in measurements of specific electrical conductance are considered to indicate relative differences in 6 m" nutrient status ' 106' 105 Maximum water retention capacity and ash content of FIG. 1. Map of the study area showing the general location and numbers of stands samples. Triangles represent approximate tree line. determined by tree ring counts of basal sections of several of the largest trees and fire-scarred trees found in the peatland type. Treeless peatlands were assumed to have a minimum age equivalent to that of the immediately surrounding bog forest. Physical measurements of peat depth and depths of water table, mineral soil and permafrost were recorded at the center of each stand. If permafrost was found no attempt was made to reach mineral soil owing to the difficulty in coring through permafrost. Cores of the top cm of peat were removed for laboratory analyses. The phh2, and specific electrical conductance of surface samples from the peat cores taken in the field were measured shortly after sampling. The cores were then stored at 0 C in plastic bags and analyzed within 3 months for maximum water retention capacity, ash content, ph in 0.5 M CaC12, and a relative measure of dissolved organic matter using optical density (Hofstetter 1969). The ph was measured in supernatants of slurries of fresh surface (0-2 cm) peat and distilled water. The ratio of fresh peat to water depended on the soil moisture content of the peat. The objective was to add the least amount of water possible but still obtain a supernatant in which to take a stable reading. Direct insertion bf the ph probe into the peat did not give a stable reading. The same procedure was used for the measurement of specific electrical conductance. The ph measurements were also taken in 0.5 M CaC12 to increase the exchange for H+ by Ca2+. This measurement gives a better indication of the hydrogen ion concentration and the base status of a peat substrate (Puustjwi 1956, 1957). Subsamples of 0.2 g of air-dried and ground peat were mixed surface peat samples (the rooting layer) were determined following the methods of Farnham and Finney (1965) and expressed as a percentage of ovendry weight. Maximum water retention capacity is defined as the amount of water that a soil retains against gravity (i.e., at zero suction). It is dependent on the plant composition but especially on the degree of decomposition of the peat substrate decreasing with an increase in decomposition as a result of a considerable decrease in the effective diameter of pores for moisture absorption (Boelter 1968). Ash content is a measure of the mineral content of the peat substrate and is primarily dependent on its plant composition and on the minerals in the water input into the peatland. The laboratory measurements were carried out on as many of the peat samples as possible. However, there were limitations as the peat was difficult to keep intact and often the surface rooting layer became mixed, through handling, with the subsequent layers. These were then not analyzed. This accounts for the variable number of values obtained for different measurements in the Results section. Gradient analysis procedures The gradient analysis is divided into three parts: (i) partitioning the total variation in each species' frequency over all sampled stands into that determined by fire and by the habitat, (ii) ordering species by their degree of selection (correlation) along each environmental complex, that is, ranking species by their adaptations relative to each other and the specific environmental gradient (species gradient analysis), and (iii) ranking stands by their abundances of ordered species along each gradient (stand gradient analysis). To accomplish these objectives, 24 species (of the total number encountered) occurring in 30% or more of the 103 sampled stands were used in the analysis. Species were not used that were shared by only few stands to reduce the

4 2584 CAN. 1. BOT. VOL. 60, 1982 distortion in multivariate ordination models due to zeros in the data matrix (Swan 1970; Jeglum et al. 1971; Gauch and Whittaker 1972) and to more closely satisfy the statistical assumption of a continuous distribution of abundances for each variable (species). Twenty-six stands were eliminated that contained fewer than 6 species of the selected group of 24. These 26 stands consisted of pools with aquatic spaghna (Sphagnum angustifolia and S. riparium and (or) sedges (Carex limosa, C. paupercula, and Eriophorum russeolum). The 24 species in these remaining 77 stands accounted for over 92% of the variation in the total species list. The temporal gradient, related to the age since last fire, must be removed by a between-group statistical technique. This is necessary so that the within-group technique of gradient analysis has a homogeneous (in the statistical sense) set of variables. If this is not done, multimodal species distributions will occur along the environmental gradients and severe clumping will occur in the stand ordination. This can be seen in the analysis of Maikawa and Kershaw (1976) and Black and Bliss (1978) in which the ages of stands are confounding the ordination. The variation due to fire was removed by using age since last fire as a concomitant variable in the equation: Y = XT + E, where Y is the n x s matrix of stands (n) by species (s), elements of the matrix being species frequency, X is a n x 1 vector of stand ages, T is a 1 x s vector of regression coefficients, and E is an n X s matrix of deviations from the regression. The residual matrix is R = Y'Y - Y'XT, where the original data matrix (Y) is subtracted from the predicted Y given in the stand ages X. The residual matrix is consequently the variation in the vegetation data that does not include that due to age since the last fire. This will be called habitat variation for reasons which will become clear later. Principal component analysis (Morrison 1967) was then performed on the residual matrix to give the habitat gradient analysis. The species loadings of the principal component express the correlation between the species and the gradient (component). Loadings can therefore be considered to order species by their adaptations relative to each other along the gradient. A stand's component score is the sum of the products of each species loading on the gradient and its abundance in the sampled stand. The fire frequency gradient analysis was determined by taking the difference between the principal component analysis of the residual matrix (habitat gradient analysis) and a principal component analysis of the unpartitioned data matrix (Y). Before this difference could be determined a linear transformation (rotation) of the components of the habitat gradient analysis onto those of the unpartitioned analysis was performed. This technique ensures the closest possible congruence between the components while keeping the order of species and stands rotationally and translationally invariant. Only the first two components are compared as they account for most of the variation in the vegetation. The rotated habitat species loadings and stand component scores minus the unpartitioned species loadings and stand component scores, respectively, gave the ordering of species and stands along the fire frequency gradient. A detailed discussion of these procedures can be found in Johnson (1981). Results Habitat gradient analysis Two major components are extracted by the habitat principal components analysis. Along the first component (Fig. 2) species are ordered from the high loadings of Cladonia mitis, C. rangiferina, C. amaurocraea, C. coccifera, Cetraria nivalis, Dicranum undulatum, and Vaccinium vitis-idaea to the low loadings of Ledum groenlandicum, Rubus chamaemorus, Oxycoccus microcarpus, Polytrichum juniperinum, Cephalozia spp., and Pohlia nutans. Stands are ordered along this gradient (Fig. 3) from treeless and treed (black spruce) lichen (Cladonia mitis, Cetraria nivalis, Cladonia rangiferina) peatlands to Sphagnum fuscum - Ericaceae stands and Sphagnum magellanicum runoff channels. The major change in vegetation along this gradient is a change in the abundance of Cetraria nivalis and the Cladoniae. Component 1 can be interpreted as a moisturenutrient gradient based on the following changes in moisture and nutrient conditions along the component. Differences in moisture conditions of peatland stands FIG. 2. The habitat species gradient analysis. Component I is interpreted as a moisture-nutrient gradient and accounts for 23% of the total variance. Component I1 is a sphagnum substrate gradient accounting for 15% of the variance. Numbers correspond to the species listed below: 1, Cetraria nivalis (L.) Ach.; 2, Cladonia alpestris (L.) Rabenh.; 3, Cladonia amaurocraea (Florke.) Schaer.; 4, Cladonia coccifera (L.) Willd.; 5, Cladonia deformis (L.) Hoffm.; 6, Cladonia gracilis (L.) Willd.; 7, Cladonia rangiferina (L.) G. H. Web.; 8, Ichmadophila ericetorum (L.) Zahlbr.; 9, Cladonia mitis Sandst.; 10, Cephalozia spp.; 11, Dicranum undulatum Brid.; 12, Pohlia nutans (Hedw.) Lindb.; 13, Polytrichum juniperinum Hedw.; 14, Sphagnum fuscum (Schimp.) Klinggr.; 15, Andromeda polifolia L.; 16, Empetrum nigrum L.; 17, Ledum decumbens (Ait.) Small.; 18, Ledum groenlandicum Oeder; 19, Oxycoccus microcarpus Turcz.; 20, Pinguicula villosa L.; 21, Rubus chamaemorus L.; 22, Vaccinium uliginosum L.; 23, Vaccinium vitis-idaea L.; 24, Mylia anomala (Hook.) Gray.

5 JASIENIUK AND JOHNSON FIG. 3. The habitat stand gradient analysis. Stands are ordered along the moisture-nutrient gradient (component I) and the sphagnum substrate regime (component 11) by their component scores. Component scores are the sum of the abundances of species weighted by their loadings from Fig. 2. Stands are grouped into communities based on the canopy cover and the abundance of species with high loadings on the two components. FIG. 4. Changes in moisture conditions of stands on the habitat gradient analysis given by the depth to water table (A, 0-10 cm; A, cm; A, >21 cm), topographic position for moisture input (0, stands along runoff channels; 0, stands at the base of upland slopes), and maximum water retention capacity of the peat substrate: 1, 0-10% dry weight; 2, 11-20%; 3, 21-30%; 4, 31-40%. are characterized by depth to water table, topographic position with respect to moisture input, and maximum water retention capacity of the peat substrate (Fig. 4). Depth to water table increases from Sphagnum magellanicum runoff channels to Sphagnum fuscum - Ericaceae peatlands (see Fig. 3). Permafrost is present at depths of cm throughout July and August in black spruce, black spruce - Sphagnum fuscum - Ericaceae and treed and treeless lichen-covered peatlands. In these stands no measure of depth to water table is recorded because of the presence of permafrost. Topographic position for moisture input and maximum water retention capacity are therefore used as indices of moisture conditions. Based on these properties, moisture status decreases from black spruce and black spruce - Sphagnumfuscum - Ericaceae to treed lichen to treeless lichen-covered peatlands. The nutrient input decreases, as indicated by the change in specific electrical conductance (Fig. 5), from Sphagnum magellanicum runoff channels and Sphagnum fuscum - Ericaceae peatlands to black spruce, black spruce - Sphagnum fuscum - Ericaceae, and

6 2586 CAN. J. BOT. VOL. 60, 1982 I Component I score FIG. 5. Specific electrical conductance (micrornhos per centimetre at 25OC) of slurries of surface peat and distilled water plotted against stand component scores on the moisturenutrient gradient (component I, Fig. 3). Highest values are obtained in Sphagnum magellanicum runoff channels and Sphagnum fuscum - Ericaceae peatlands (component scores between 13.9 and 5.0); lower values are obtained in black spruce, black spruce - Sphagnum fuscum - Ericaceae, black spruce - lichen - Sphagnum fuscum - Ericaceae, and both treed and treeless lichen peatlands (component scores between 2.3 and -13.1). lichen peatlands. The abundance of characteristically higher nutrient demanding species as Chamaedaphne calyculata and Sphagnum magellanicum (Sjors 1950; Segadas-Vienna 1955; Lems 1956; Crum 1973) for this set of data also decreases (from greater than 50% to less than 10%) with the decreasing specific electrical conductance. Higher electrical conductance (40-60 kmho) in those stands that have a higher water table (less than 20cm from the surface) suggests that some nutrient enrichment occurs from waters in contact with mineral substrates. In black spruce, black spruce - Sphagnum fuscum - Ericaceae and both treed and treeless lichen peatlands no water table is found owing to the presence of permafrost and specific electrical conductance is uniformly low (20-30 kmho). Other general measures of nutrient status such as ash content and phh,o remain consistently low and show no pattern over the gradient. The values range from 3.2 to 4.7 with most values between 3.4 and 4.0; ash content ranges from 2.25 to 5.03% dry weight. The ph in 0.5 M CaC12, a more sensitive indicator of base status in these low nutrient peats, however, did show somewhat of a pattern (Fig. 6). The highest values were in Sphagnumfuscum - Ericaceae peatlands where specific electrical conductance was also high. Values were lower in black spruce - Sphagnum fuscum - Ericaceae. It is unclear why phcacl, increases in lichen peatlands. Comparison with the literature on vegetation and moisture and nutrient conditions of other peatlands suggests that the gradient is at the extreme dry and nutrient poor end (e.g., Heinselman 1963; Malmer 1963; Hofstetter 1969; Jeglum 1971, 1972, 1973; Vitt and Slack 1975; Vitt et al. 1975; Johnson 1977). Mean for the gradient is 3.84; mean specific electrical conductance is 23.7 kmho; mean ash content is 3.59%. Under these conditions small changes in field moisture conditions would cause large changes in ion activity (because of changes in ion concentration). Thus even moderate increases in moisture would decrease the ion activity of the peat solution and increase the potential between it and the plant absorption structures. This would increase ion uptake by the vegetation. On the other hand, drier conditions would decrease ion uptake by the vegetation. This may partly explain the changes in vegetation composition and abundance on this gradient without corresponding changes in ph and conductance. Both ph and conductance are measured under standard moisture content which may be eliminating the ion activity differences affecting availability of nutrients. Similarly ash content measures only the absolute amount of minerals in the substrate not the differences in concentration and activities of cations that are important in cation exchange. Species are ordered along component I1 (Fig. 2) of the habitat principal components analysis from Cladonia coccifera, Cetraria nivalis, and Cladonia mitis characteristic of dead and (or) partially decomposed sphagnum to Pinguicula villosa and the ericaceous shrubs Oxycoccus microcarpus, Rubus chamaemorus, and Ledum groenlandicum, and the bryophytes Mylia anomala and Cephalozia spp. characteristic of living sphagnum. Stands are ordered along component 11 (Fig. 3) from lichen, black spruce, and burned (510 years old, 275% bum) peatland communities to stands of black spruce - lichen - Sphagnum fuscum and Sphagnum fuscum - Ericaceae peatlands. The major change in vegetation along this gradient is an increase in the abundance of black spruce and the frequency of the ericaceous shrubs Rubus chamaemorus, Oxycoccus microcarpus, and Ledum groenlandicum with increasing abundance of living sphagnum. The organization of species and stands along this component appears to be related to changes in the hydrophysical properties of the peat substrate as determined by the amount of living versus dead, partially decomposed sphagnum. Changes in frequency of living Sphagnum fuscum along the gradient are given in Fig. 7, Sphagnumfuscum forms the peat substrate in 93% of the

7 JASIENIUK AND JOHNSON 2587 I Component I s core FIG. 6. The ph of surface peat in 0.5 M CaC12 plotted against stand component scores on the moisture-nutrient gradient. Groupings correspond to communities delimited in Fig. 3 excluding stands that deviate markedly in ph from the main body of stands in a community. I - I Component i7 score MWRC f % dry velghl I FIG. 7. Frequency of living Sphagnum fuscum in a stand plotted against component scores on the sphagnum substrate gradient (component 11, Fig. 3). Highest values are obtained in Sphagnum fuscum - Ericaceae, black spruce - Sphagnum fuscum - Ericaceae, and black spruce - lichen - Sphagnum fuscum - Ericaceae communities (component scores between -4.0 and 0); lower values are obtained in lichen, black spruce, and burned peatlands (component scores between 0 and 6.0). total 77 stands. Living Sphagnum fuscum has a frequency >0.6 in 51% of these stands; dead S. fuscum, in varying degrees of decomposition, forms the major comdonent of the substrate in 42% of the stands. Sphagnum magellanicum and S. balticum are abundant in only six stands (the S. magellanicum runoff channels in Fig. 3). Figure 8 shows the relationship between frequency of living Sphagnum fuscum and maximum water retention capacity, and Fig. 9 the relationship of maximum water retention capacity and optical density at 320nm of FIG. 8. Frequency of living Sphagnum fuscurn in a stand plotted against maximum water retention capacity (MWRC percent dry weight) of the peat substrate. filtrates from slumes of peat substrates and distilled water. High absorbance of short wavelength light by bog and other natural waters has been shown to be related to an increase in the amount of dissolved organic carbon due to decomposition (Tolpa and Gorham 1961; Mackereth 1963; Hofstetter 1969). Thus from Figs. 7 to 9 moisture retention is seen to decrease as the percentage of dead and decomposed sphagnum increases in the substrate. The large water storage capacity of undecomposed sphagnum peats is primarily due to a high total porosity within the plants (and their remains) and between their stalks. With drying or decomposition there is a considerable decrease-in-the pore size distribution and the effective diameter of pores for moisture absorption

8 CAN. J. BOT. VOL. 60, 1982 snow abrasion. Some allelopathic effects by the lichens may also explain the low abundance of Ericaceae shrubs and black spruce. I Absorbance (OD units) FIG. 9. Maximum water retention capacity (MWRC, percent dry weight) of peat substrate plottedagainst absorbance of short wavelengths (320nm) of light by filtered slurries of sphagnum substrates. (Boelter 1964, 1965; Romanov 1968) such that moisture retention is reduced. Hydraulic conductivity through sphagnum peats is also primarily related to pore size distribution within the plant material and between stalks. Living sphagnum substrates have many large pores which allow rapid water movement and are easily drained; dead highly decomposed sphagnum peats permit little water movement, and have hydraulic conductivity rates that may be lower than many clays and glacial tills (Boelter 1965; Romanov 1968). Death of sphagnum can come about by surface drying due to decreases in precipitation, drainage, fire, or even as the consequence of the growth of Sphagnum too high above the water table (Conway 1948; Tallis 1964; Romanov 1968). The increase in oxygen availability as a result of drying then initiates decomposition by microbial activity (Given and Dickinson 1975). The implications of a decrease in water retention and movement with increasing decomposition are decreases in water availability and aeration. The decrease in moisture retention no doubt also affects cation exchange. Thus, the absence of trees and decreased abundance of ericaceous shrubs in treeless lichen peatlands (Fig. 3) are the result of the combined effects of moisturenutrient factors and sphagnum substrate factors. Other environmental changes associated with the decrease in tree and shrub cover from living sphagnum to dead, partially decomposed sphagnum substrates may include increases in wind speed and consequently lower winter temperatures with reduced snow cover and increased Firefrequency gradient analysis Figure 10 gives the species loadings associated with species recovery patterns after fire. Species at the top of the gradient such as Polytrichum juniperinum, Ledum groenlandicum, and Cladonia gracilis increase in abundance within a short interval after fire and decrease in abundance in older stands (see chronosequences in Fig. 10). Species at the bottom of the gradient such as Cladonia rangiferina, Cetraria nivalis, and Andromeda polifolia recover more slowly after fire and generally reach their highest abundance in old stands. Intermediate along the gradient are Vaccinium vitisidaea, Ledum decumbens, and Sphagnum fuscum which show a variable abundance in both recently burned and older stands. Stands are ordered along the fire gradient in Fig. 11. Generally, as might be expected, recently burned stands (<20 years old) have a large number of species that increase rapidly after fire although other species are also - LEOUM GROENLANOICUM <a 05 0?-T77-3, yl.iw 150 2W 250 AGE -- VACCINIUM - VITIS-IOAEb ClTRARlb NlVALlS AGE -. FULYTRICHUM..-- LUEE-Ug W 253 AGE W 250 AGE I50 2W 2Y1 AGE FIG. 10. The fire frequency species gradient analysis. The numbers beside the points correspond to the species in Fig. 2. The species gradient analysis results from the difference between the habitat principal component analysis in which variation due to age since fire has been removed (Fig. 2) and a principal component analysis of the unpartitioned vegetation data. See text for explanation of technique. The small figures adjacent to each species number give the chronosequences for the three habitats differentiated on the basis of moisture and nutrient conditions: 1 includes Sphagnum fuscum - Ericaceae stands and Sphagnum magellanicum runoff channels; 2 includes black spruce and black spruce - Sphagnum fuscum - Ericaceae peatlands; 3 includes lichen and black spruce - lichen - Sphagnum fuscum - Ericaceaepeatlands. The gradient orders species according to their response after fire as is demonstrated by the chronosequences.

9 ;;:I.,.: -:: - I-. JASIENIUK AND JOHNSON 2589 TABLE 1. The number of species for all 2 stands less than or equal to 10 years since 0 a SPECIES LOAD~NGS. -03 I fire. Division into completely burned (all aboveground parts killed) and incom-. i a o pletely burned show the difference in the a SPECIES LOADlHGS In.. - recovery of species richness. Number of,..,..:-... species means all species present in the --ol a 0 a SPECIES LOIOINGS 1: 100-m2 stand I Age (years since.'+:/p;k.:..; last fire) No. of species...-. I - m ). a SPECIES LOADINGS.I. Y Incompletely burned stands *::JAi].:- 1 " a SPECIES LoPinlNGs 6 37 * I : $! ;, /:.: =;. FIG. 11. The fire frequency stand gradient analysis. Stands are ordered along the gradient by their component scores (the sum of products of each species loading and its abundance in the stand). The small figures display a component score's "ingredients," i.e., the species loadings (loadings on the right are on the top of Fig. 10) versus the species abundance in a stand. Notice that in the small figures the greater abundance shifts from species loadings on the right to species loadings on the left as one moves down the gradient. Species with loadings on the right have rapid recovery strategies after fire, those on the left have slow recovery strategies (see Fig. 10). found in low abundance. Older stands consist of species which recover more slowly after fire. Discussion Fire frequency dynamics In this part of the Canadian central subarctic we have suggested that the predominant environmental gradients acting on peatland vegetation are moisture-nutrients, peat substrate, and the frequency of fire (i.e., interval between fires). In general, fire kills only the surface vegetation and is not severe enough to destroy the peat to any depth. In many stands fire selectively bums only the larger, erect woody shrubs and trees and has less effect on the Sphagnum and closely appressed vascular plants such as Oxycoccus microcarpus and the bryophytes Mylia anomala and Cephalozia spp. Differences in species composition between stands are primarily determined by habitat differences. Fire does not usually appear to change the habitat very much or significantly affect stand composition. However, fire does affect changes in species abundance within the stand. Stands in which only part of the vegetation is burned rapidly regenerate to approximately the predisturbance composition. Table 1 indicates that young stands in which all of the surface vegetation was destroyed show the differential responses of species reinvasion and Completely burned stands reestablishment after fire. The composition of these stands can be compared to the greater species richness of incompletely burned young peatlands. These latter stands include both species adapted to rapid recolonization of the partially fire opened habitat and the surviving predisturbance species. Unlike after upland fires (Johnson 1981), there appears to be no early invasion in peatlands of fugitive species such as Corydalis, Epilobium, and Calamogrostis. The very low nutrient status of the peat substrate seems to mitigate against this reproductive strategy. One of the adaptations of woody and herbaceous species in nutrient-deficient habitats is an extremely slow growth rate, thereby making low demands on the small nutrient reserve available (Small 1972a, 1972b; Grime 1977). The fast growth rates of fugitive species required to escape crowding and to insure reproduction may necessitate a larger nutrient supply than is available in these peatlands and thus prevent their invasion after fire. Because of their wider habitat tolerance and excellent regenerative ability from rhizomes, rhizoids, and gemmae, Ledum groenlandicum, Vaccinium vitis-idaea, Polytrichum juniperinum, Pohlia nutans, and Ceratodon purpureus seem to predominate in most habitats for a short period after fire. This explains why in many stands of years since fire Polytrichum, Pohlia, and Ceratondon are often found to have been overgrown by lichens or sphagnum. In the dry, low nutrient habitats of dead sphagnum

10 2590 CAN. J. BOT. VOL. 60, 1982 substrates, lichens show a continuum of recovery patterns (Fig. 10). Cladonia deformis and C. gracilis recover very rapidly and obtain a high abundance soon after fire. This is a result of a combination of rapid invasion by soredia, thallus fragments, or sexually produced propagules (in C. gracilis), basal regeneration, and rapid growth rates. Other lichens increase later in the order Cladonia mitis, C. coccifera, C. amaurocraea, C. rangiferina, C. alpestris, and Cetraria nivalis. Cladonia mitis regenerates more rapidly than other cladinae species owing in part to its faster growth rate (Scotter 1963). Cladonia rangiferina is mainly found on relatively high moisture and nutrient substrates. Cetraria nivalis has a very slow recovery after fire. It is primarily found in treeless lichen peatlands and has a range predominantly in the tundra (Ahti and Hepburn 1967). Its small stature closely appressed to the peat surface allows it to escape desiccating winds and snow abrasion in the treeless habitat in winter. Furthermore, its extremely slow response to fire may be an adaptation to reduce the rate of utilization of extremely limited resources. In habitats of higher moisture status and associated nutrient availability, regeneration is to the prefire composition of living sphagnum substrate with an ericaceous shrub cover. Species characteristic of these stands such as Sphagnum fuscum, Oxycoccus microcarpus, and Rubus chamaemorus return more rapidly following fire than many of the fruticose lichens (Cladonia rangiferina, C. alpestris, Cetraria nivalis) characteristic of lichen peatlands (Fig. 10). Sphagnum fuscum reproduces vegetatively by dichotomous branching (Lane 1977) and remnants have been observed to readily regenerate and invade burned-out patches in partially destroyed, moist habitats. Oxycoccus microcarpus and Rubus chamaemorus also reproduce vegetatively by creeping stems and rootstocks, respectively. Habitat dynamics Johnson (1981) has suggested that for uplands in this region the fire frequency and habitats are stationary for an average of about 400 years. Between these periods of relatively stable climate are transitional intervals which involve change in the water, nutrient, and heat budgets of upland habitats. The pattern of these transitions appears to be random in time. Consequently, for uplands, he proposed that his fire frequency gradient analysis represented a shorter term dynamics in the vegetation conforming to the regrowth of species after fire but with little change in either the species composition or the quality of the habitat. The habitat gradient analysis represented a longer term dynamics in the vegetation related to changes caused by major and infrequent environmental disturbances. Peatlands require some modification of this explana- tion of community organization and dynamics. Because of the lowland location and highly oxidizable nature of the substrate, peatlands are much more subject to change in their physical properties than are most upland mineral soil sites. Thus, while upland habitats are physically and compositionally stable for long periods this same generalization cannot be made of peatlands. In Table 2 we suggest a classification of communities along the moisture-nutrient gradient and the recurrence and severity of disturbance by fire and water level for each. Below we discuss our reasoning for the suggestions in Table 2. Sphagnum magellanicum stands are subject to large variations in water level during one season and from year to year. This is due to the generally near-littoral or stream-channel location of these stands. They are subject to this variation because the regional hydrology allows negligible groundwater storage due to the largely granitic substrates, and runoff is thus unbuffered (Kakela 1973). Water is stored primarily as surface water in poorly drained lakes, adjacent areas of streams and rivers. Figure 12 gives a rough idea of the variation in water level in terms of the flow. (Long-term records of water levels are very limited for any regions of the western subarctic.) Fires rarely bum in these extremely wet habitats so except for indirect effects, the importance of fire in shaping the communities is minor. The treed peatland communities in the middle of the moisture-nutrient gradient are situated so that they are less affected by changes in lake level and runoff fluctuations. However, they are not immune. The black spruce - Sphagnum fusclim - Ericaceae community seems to have a higher chance of being flooded or having their peat eroded by rising lake levels. Fires occur in these peatlands at what appears, from comparisons with adjacent uplands, to be once for every two fires on the uplands. From observations of over 30 fires in these communities, burning is spotty with only local areas around trees having much peat removed. Regeneration is usually rapid because of the relatively moderate and stable moisture conditions and sufficient nutrient availability. Treeless lichen-covered peatland communities represent the other extreme on the moisture-nutrient gradient. These communities no longer have water budgets which allow the growth of most peat-forming species. Despite their drier conditions compared with other peatland communities, they bum even less frequently and as sparsely as the treed peatlands. They do not bum well for the same reasons that tundra vegetation does not: lack of fuel, nonconducive fuel geometry, and small but effective microtopographic fire bamers (Wein 1976). Because of their generally upland locations and dry conditions, these treeless peatlands always have permafrost within 40cm of the surface and often

11 JASIENIUK AND JOHNSON TABLE 2. Suggested relationships between different disturbance frequencies and severities for peatland communities* Disturbance Recurrence causes habitat Community Disturbance interval changet Sphagnum magellanicurn runoff channels S. fuscum - Ericaceae Water table Fire Black spruce - S. fuscum - Ericaceae Water table Fire Black spruce - lichen Treeless - lichen Water table (increase or decrease) 5-10 years Yes Water table Fire Water table Fire *During climate stationary periods. THabitat change then causes permanent change in species composition Decades Hundreds of years >50 years Hundreds of years >200 years Hundreds of years Rare during stationary period Hundreds of years Yes No Yes No Yes Yes son, D. Johnson, A. Lincoln, and T. Trottier. Analysis of data was supported by the Biology Department, University of Saskatchewan, and manuscript preparation, by the Natural Sciences and Engineering Research Council of Canada. The senior author held scholarships from the Institute for Northern Studies and College of Graduate Studies, University of Saskatchewan, during the study. AHTI, T., and R. L. HEPBURN Preliminary studies on woodland caribou range, especially on lichen stands, in 101 B 3 r a lo PO itelurn Perlod I" years Ontario. Ontario Department of Lands and Forests Research FIG. 12. The maximum and minimum recurrence interval of Report, Wildlife ~ d 74.. high and low water levels for Great Slave Lake BENNET, R. M Great Slave Lake water levels. Water (from Bennett 1973). Survey of Canada. Environment Canada, Ottawa. BLACK, R. A., and L. C. BLISS Recovery sequence of polygonal surface patterns. In their dry and exposed Picea mariana-vaccinium uliginosurn forests-afte; burning location, negative heat budget (as typified by the near Inuvik, Northwest Territories, Canada. Can. J. Bot. permafrost) and species composition, these communi- 56: ties represent a tundra environment. BOELTER, D. H Water storage characteristics of several peats in situ. Soil Sci. Soc. Am. Proc. 28: Acknowledgments Hydraulic conductivity of peats. Soil Sci. 100: We thank J. W. Sheard, J. K. Jeglum, B. R. Neal, Important physical properties of peat materials. and J. W. Stewart for their help and criticism of the peat congr. 3rd. pp, research and drafts. The fieldwork was supported by a CLEMENTS, F. E Plant succession: an analysis of the Supply and Services Canada contract to J. S. Rowe and development of vegetation. Carnegie Inst. Washington field collections were assisted by R. Godwin, J. Harald- Publ No

12 2592 CAN. J. BOT. VOL. 60, 1982 CONWAY, V. M Von Post's work on climatic rhythms. New Phytol. 47: The bogs of central Minnesota. Ecol. Monogr. 19: COOPER, W. S The climax forest of Isle Royale, Lake Superior, and its development Bot. Gaz. (Chicago), 55: CRUM, H Mosses of the Great Lakes Forest. Contrib. Univ. Mich. Herb. No. 10. DACHNOWSKI, A. P Peat deposits of Ohio. Ohio Geographical Survey Bulletin The stratigraphic study of peat deposits. Soil Sci. 17: FARNHAM, R. S., and H. R. FINNEY Classification and properties of organic soils. Adv. Agron. 17: GATES, F. C The bogs of northern lower Michigan. Ecol. Monogr. 12: GAUCH, H. G., and R. H. WHITTAKER Comparison of ordination techniques. Ecology, 53: GIVEN, P. H., and C. H. DICKINSON Biochemistry and microbiology of peats. In Soil biochemistry. Vol. 3. Edited by E. A. Paul and A. B. McLaren. Marcel Dekker, New York. GODWIN, H The relationship of bog stratigraphy to climatic change and archaeology. Proceedings of the Prehis- toric Society, No. 1. GORHAM, E On the chemical composition of some waters from the Moor House Nature Reserve. J. Ecol. 44: GRANLUND, E De svenska hogmossarnas geologi. Sver. Geol. Unders. Ser. C, No GRIFFIN, K Vegetation studies and modem pollen spectra from the Red Lake peatland, northern Minnesota. Ecology, 56: Paleoecological aspects of the Red Lake peatland, northern Minnesota. Can. J. Bot. 55: GRIME, J. P Evidence for the existence of three primary strategies in plants and its relevance to ecological and evolutionary theory. Am. Nat. 111: HALE, M. E., and W. L. CULBERSON Afourth checklist of the lichens of the continental United States and Canada. Bryologist, 73: HEINSELMAN, M. L Forest sites, bog processes, and peatland types in the Glacial Lake Agassiz Region, Minnesota. Ecol. Monogr. 33: Landscape evolution, peatlands, and the environment in the Lake Agassiz Peatlands Natural Area, Minnesota. Ecol. Monogr. 40: Boreal peatlands in relation to environment. In Coupling of land and water systems. Edited by A. D. Hasler. Springer-Verlag New York, Inc., Secaucus, NJ. pp HOFSTETTER, R. H Floristic and ecological studies of wetlands in Minnesota. Ph.D. thesis, University of Minnesota, Minneapolis. JANSSEN, C. R Stevens Pond: a postglacial pollen diagram from a small Typha swamp in northwestern Minnesota, interpreted from pollen indicators and surface samples. Ecol. Monogr. 37: Myrtle Lake: a late- and post-glacial pollen diagram from northern Minnesota. Can. J. Bot. 46: JASIENIUK, M. A., and E. A. JOHNSON A vascular flora of the Caribou Range, Northwest Territories, Canada. Rhodora, 81: JEGLUM, J. K Plant indicators of ph and water level in peatlands at Candle Lake, Saskatchewan. Can. J. Bot. 49: Boreal forest wetlands near Candle Lake, Saskatchewan. I. Vegetation. Musk-Ox, 11: Boreal forest wetlands near Candle Lake, Saskatchewan. 11. Relationships of vegetational variation to major environmental gradients. Musk-Ox, 12: JEGLUM, J. K., C. F. WEHRHAHN, and J. M. A. SWAN Comparisons of environmental ordinations with principal component vegetational ordinations for sets of data having different degrees of complexity. Can. J. For. Res. 1: JOHNSON, E. A A multivariate analysis of the niche of plant populations in raised bogs. I. Niche dimensions. Can. J. Bot. 55: Fire recurrence in the subarctic and its implications for vegetation composition. Can. J. Bot. 57: Vegetation organization and dynamics of lichen woodland communities in the N.W.T., Canada. Ecology, 62: JOHNSON, E. A., and J. S. ROWE Fire in the wintering grounds of the Beverley caribou herd. Am. Midl. Nat. 94: KAKELA, P Thomwaite's climatic water balance: Evaluation of annual discharge estimates for two subarctic basins. Can. Geogr. 17: KULCZYNSKI, S Peat bogs of Polesie. Memoires de 1'Academie Polonaise des Sciences Cracovie B. pp LANE, D. M Extent of vegetative reproduction in eleven species of Sphagnum from northern Michigan. Mich. Bot. 16: LEMS, K Ecological study of the peat bogs of eastern North America Notes on the behaviour of Chamaedaphne calyculata. Can. J. Bot. 34: MACKERETH, F. H. H Some methods of water analysis for limnologists. Sci. Publ. Freshwater Biol. Assoc. No. 21. MAIKAWA, E., and K. A. KERSHAW Studies on lichen-dominated systems. XIX. The postfire recovery sequence of black spruce-lichen woodland in the Abitau Lake Region, N.W.T. Can. J. Bot. 54: MALMER, N Studies on mire vegetation in the Archaean area of SW Gotaland (S. Sweden) On the relation between specific conductivity and concentration of ions in the mire water. Bot. Not. 116: MORRISON, D. F Multivariate statistical methods. McGraw-Hill Publications, New York. PEARSALL, W. H Mountains and moorlands. William Collins, Sons & Co. Ltd., London. PUUSTJARVI, W On the cation exchange capacity of peats and on the factors of influence upon its formation. Acta Agric. Scand. 6: On the base status of peat soils. Acta Agric. Scand. 7: ROMANOV, V. V Hydrophysics of Bogs. (Gidrofizika balot.) Israel Program for Scientific Translations, Jerusalem.