key: cord-0010780-xb2qp0uw authors: Carleton, T.J.; Maycock, P.F.; Arnup, R.; Gordon, A.M. title: In situ regeneration of Pinus strobus and P. resinosa in the Great Lakes forest communities of Canada date: 2009-02-24 journal: J Veg Sci DOI: 10.2307/3236287 sha: 92587c3400c7a2bb9c7a9fa9ffccb76ee6ebd5fc doc_id: 10780 cord_uid: xb2qp0uw Abstract. Two extensive forest vegetation survey datasets are explored, using ordination and classification, for evidence of in situ regeneration by Pinus strobus (Eastern white pine) and P. resinosa (Red pine). Ordination of tree species contributions to total basal area in 320 upland northern hardwood‐ conifer stands produced distinct stand groups for P.banksiana, P. resinosa, P. strobus and mesic hardwoods in an ascending sequence along the first axis. Quercus rubra (red oak), Q. alba (white oak) and tolerant conifer groups formed segregates from the hardwood complex along the second axis. P. strobus mixes with all other forest types, but P. resinosa is restricted to its own group. Seedlings and trees of P. strobus are more abundant than saplings, which are restricted to the pine and oak forests. Therefore, seed production, dispersal and seedling establishment seem to be less of a barrier to in situ regeneration by P. strobus than subsequent survival and growth. Canonical correspondence analysis of 170 pine‐dominated stands from the Canadian Shield of Ontario, in which tree species variables are segmented into height‐class pseudo‐ species, yielded no linear relationship between environmental features or stand structure and seedling densities of P. strobus. However, total tree basal area appears to impose an upper limit to seedling density on the forest floor. Strong correlations emerged between pine seedling density and understorey vegetation. Stand classification of the understorey vegetation, using constrained indicator species analysis, yielded distinct high and low seedling groups. Low pine seedling density was associated with abundant broadleaved shrubs, herbs and seedlings as well as feathermosses and tolerant conifers. High seedling density could not be ascribed to the presence of seedbed taxa, such as Polytrichum, but is ascribed to the absence of competition and other forms of inhibition in the understorey vegetation and down through the canopy profile. In situ regeneration of P. strobus does, therefore, occur but conditions over the forest landscape are largely restrictive. Pinus strobus (Eastern white pine) and P. resinosa (Red pine) each form monospecific stands and may also form a tall, emergent supercanopy in many mixed forests of the Great Lakes region, the Canadian Maritimes and throughout the Appalachian mountains (Burns & Honkala 1990) . Forests dominated by these pines have received attention in eastern Canada in recent years because of a decline in the abundance of large trees through heavy harvesting in the past and probably also through forest fire suppression activities. One aspect of such concern is a debate over the extent to which P. strobus can establish and regenerate through gap-phase replacement on the forest floor of an uneven-aged forest of which it is a component species. The prevalent view is that catastrophic clearance or partial canopy removal by forest fire is essential for effective pine regeneration and that little occurs by other means (Ahlgren 1976; Ahlgren & Ahlgren 1981; Day & Carter 1991; Frehlich 1992; Heinselman 1981; Maissurow 1941; Methven 1973) . Bergeron & Brisson (1990) demonstrate how recurrent forest fire leads to multiple cohorts of P. resinosa in a stand. Similar situations have been demonstrated for mixed P. strobus and P. resinosa (Cwynar 1977) and for Picea mariana beneath a Pinus banksiana canopy (Carleton et al. 1996) . Unlike P. banksiana, P. rigida and Picea mariana, neither P. strobus nor P. resinosa produce serotinous cones, the seed being dispersed at cone maturity. A countervailing view is that gap-phase replacement is possible in very old white pine forest (Gilbert 1978; Holla & Knowles 1988; Quinby 1991) . Others have referred to the critical nature of the interaction between site factors, light levels and the nature of hardwood competition in determining the success or failure of such in situ white pine regeneration (Hibbs 1982) . Most of the conflicting evidence derives from intensive case studies which focus largely on the detailed reconstruction of a few stands (Hibbs 1982; Day & Carter 1991) . Little attempt has been made to integrate this intensive approach with more extensive studies, as has been done for Picea mariana (black spruce) (Payette 1992) , and for tolerant northern hardwoods (Frehlich & Lorimer 1991) , into an overall perspective of the way in which environmental variation over the regional landscape affects forest structure, composition and succession. In order to address such an objective it is necessary to examine stand structure on an extensive basis in relation to composition and site characteristics. The purpose of this paper is to explore the evidence for in situ white and red pine regeneration in two, previously unreported, forest vegetation datasets from central eastern Canada. One survey encompasses the full range of forest canopy composition in the upland northern hardwood-conifer/Great Lakes-St. Lawrence lowlands region (Curtis 1959; Rowe 1972) . The second dataset derives from a more detailed structural, compositional and forest site survey of P. strobus and P. resinosa dominated forests located solely on the Canadian Shield of Ontario. The aim is to use the first dataset to examine the compositional relations of pine dominated stands in the context of the upland northern hardwood-conifer forest as a whole. With the pine forest survey on the Canadian Shield, the aim is to explore environmental and temporal correlates of stand structure, composition and regeneration in order to provide a perspective on intensive studies. P. strobus and P. resinosa occur naturally throughout the full geographical extent of stands in both surveys. To characterize the conditions under which white and red pine regeneration are likely to occur in the northern hardwood-conifer forest complex, in the absence of catastrophic disturbance, the following questions are posed: (1) Over what range of forest compositional variation do P. strobus and P. resinosa occur as trees, saplings and seedlings in central eastern Canada?; (2) is there a relationship between site condition, pine size and age structure?; (3) are differences in pine seedling and sapling abundance among stands linked to site conditions, are they linked to conditions of tree size and age structure or a combination of the two?; (4) Does the understorey vegetation appear to influence in situ pine regeneration? The conifer-hardwood survey encompasses the full spectrum of upland forest composition in the Great Lakes-St. Lawrence Forest Region of eastern Ontario and western Quebec (Rowe 1972) , traversing both the Paleozoic deposits of Southern Ontario and the Cana-dian Shield. The geographical area encompasses all of southern Ontario north of a line from Grand Bend, on the Lake Huron shoreline, eastwards to Toronto (Fig. 1) . The northern boundary corresponds to the line drawn by Rowe (1972) for the limit of the Great Lakes -St. Lawrence lowlands forest as far as the eastern limit of the Eastern Townships of Quebec. Data were collected during the period 1964 to 1972. 320 stands were included in the upland northern conifer-hardwood analysis. Stands on wet sites were omitted. Saplings are defined as any woody stem between 2.54 cm and 10 cm diameter at breast height (dbh) and a tree as any woody stem greater than 10 cm DBH. The data form part of the database of Dr. P.F. Maycock, Erindale College, University of Toronto, and derive from point-quarter, plotless sampling following methods of the Wisconsin school (Curtis 1959; Maycock 1963) . The second, more intensive survey of undisturbed pine forests was conducted during the summers of 1991 and 1992. It includes 170 stands with at least 10 % of the basal area composed of Pinus strobus and/or P. resinosa, located throughout eastern and north western Ontario on the Canadian Shield. The survey was stratified with a design of five age classes: 50 -80 yr, 81 -100 yr, 101 -120 yr, 121 -180 yr, and 180+ yr, in each of three main site quality categories, poor, medium and rich. In each stand a 20 m × 20 m (0.04 ha) plot was established within which the position of each woody stem >10 cm DBH was mapped and species identity, total height, height to the lowest live branch, dbh and crown radius were recorded. For stems <10 cm DBH, the species and numbers within a 5 m wide strip, along the centre line of the plot, were recorded in each of three height classes: 0.1 -1.5 m, 1.5 -3.0 m, 3.0 -10.0 m. Pine seedlings only (i.e. < 10 cm height) were counted in the same 5 m × 20 m strip. The cover of each understorey plant species was recorded in 20 contiguous 1 m × 1 m quadrats, forming a belt transect, along the centre line of the plot. In addition, the cover of various forest floor materials, including mineral soil, exposed rock, needle litter, broadleaved litter, tree bases, fine and coarse woody debris, were recorded. Site features recorded for each stand include angle of slope (inclinometer), aspect (compass degrees from 0 azimuth), position on the slope, volume of standing dead timber, and volume of coarse woody debris on the forest floor. In addition, average cover among the quadrats for each of the forest floor materials was summarized per stand. Two site-related indices were derived, a microclimatic index MICRO, and a white pine site index PWSI. For MICRO the compass direction of slope aspect was adjusted to within the range -90° to +90° at either side of a line from northwest (315° azimuth) to southeast (135 o azimuth) as 0° . This follows the scheme of Hills (1959) for irradiance distribution with landscape in Ontario. The calculation: MICRO = adjusted aspect %slope angle 1000 * ( ) yields an index in the range + 8.1 for which positive values are progressively warmer than normal and negative values are colder than normal, where a normal site is one with no slope. PWSI was based on the empirical construction of height/age curves in which the increment core ages and heights of five pine trees, representing the height class distribution in each stand, were fitted to harmonic normal yield tables for the region. Site index age was fixed at 120 yr. For stands with little or no white pine (P. strobus), an equivalent height/age relationship was derived for red pine (P. resinosa), with site index registered at 100y. Conversion from a red pine site index PRSI to a PWSI equivalent was done by regressing one set on the other in those stands containing both species (n = 89; PWSI = 11.0 + 0.61 * PRSI; R 2 = 38 %). The conversion was necessary for only 29 of the 170 stands in the dataset. Data on soil horizon thickness, colour, texture, mottling were collected from a 1 m × 1 m × 1.2 m deep soil pit and three point auger samples around the perimeter of each plot (Anon. 1985) . Air-dried soil samples, collected from the upper B horizon of the pit, were analyzed for pH, P, K, Ca, Mg, and particle size distribution (Anon. 1985) . Organic, F and H, soil layers were analyzed for total C and N using a Leco CHN-600 determinator. For the northern hardwood-conifer data, one data table, comprising relative dominance (i.e. percentage contribution to total basal area) for each tree species in each stand, was produced. For the pine survey data, individual trees were assigned to height classes. The seven height classes were < 5 m, 5. Three variable × sample matrices were generated: (1) tree height class pseudospecies × stands; (2) understorey species (average cover) × stands; and (3) environmental variables × stands. Data were analyzed with the ordination techniques, Detrended Correspondence Analysis (DCA: Hill & Gauch 1980) and Canonical Correspondence Analysis (CCA: ter Braak 1986), using the program CANOCO (ter Braak 1990). Stand classification was performed with the program COINSPAN (Constrained Indicator Species Analysis: Carleton et al. 1996) . This method is an extension of the popular correspondence analysis (CA) based technique, two way indicator species analysis (TWINSPAN: Hill 1979) . Whereas TWINSPAN involves the bisection of unconstrained CA axes, COINSPAN involves the constraint, or regression, of one or more external variables onto the axes such that they become canonical correspondence axes. The method is akin to a hybrid between canonical correlation analysis, in that external variables and vegetation ordination axes are first maximally correlated (rotated to congruence), and a polythetic-divisive cluster analysis, in that an hierarchic stand classification is created by successive axis bisections. Where the external variables are categorical, rather than continuous, there exists a close analogy with canonical discriminant analysis, such that indicator species which emerge, typify the various categories. Multiple regression was conducted using SAS (Anon. 1989) . The location of 320 northern hardwood-conifer stands and 170 pine dominated stands from each of the two surveys in Ontario illustrate a discontinuity due to the northern shore of Lake Superior extending into the southern boreal forest zone. Neither P. strobus nor P. resinosa occurs here. Only 27 stands from the northern hardwood-conifer survey are located in northwestern Ontario whereas 60 stands from the pine forest survey occur in this region (Fig. 1) . Table 1 shows Acer saccharum (sugar maple) to be the most abundant species in both the tree and sapling strata in the hardwood-conifer survey and this species accounts for almost one quarter of total tree basal area in the data. Acer rubrum is also widespread but contributes far less to total basal area. However, P. strobus accounts for an equally high proportion of the total basal area even though it is a less frequent species. In contrast, P. resinosa contributes considerably less to the total basal area and is found in fewer than 10 % of the stands in the survey. In the case of neither pine species are saplings particularly frequent or abundant compared with shade tolerant species such as A. saccharum, Fagus grandifolia, Table 1 . Percentage of plots in which each tree species occurs (%P), average proportional basal area (%BA), sapling presence, and average proportional sapling density (%D) for species among 320 northern hardwood-conifer stands in Ontario. In order to examine the compositional position of the pine species in relation to all other trees, the data for percentage contribution to total basal area (relative dominance) were ordinated using detrended correspondence analysis (DCA). Axes 1 and 2 respectively accounted for 8.3 % and 5.6 % of the total trace. The species ordination illustrates the pine species toward the lower (left) end of the first axis grading to the higher scores for tolerant hardwoods at the extreme right (Fig. 2) . The second axis grades from species of oak at the low end to tolerant conifers, including A. balsamea, Thuja occidentalis and T. canadensis at the high end. The third ordination axis (5.2 %: not shown) is characterized by Betula papyrifera and P. tremuloides at the high end with shade tolerant species scoring at the low end. This is, therefore, best interpreted in terms of a hardwood disturbance/succession trend. On the ordination of axis 1 vs 2, P. banksiana, P. resinosa and P. strobus form a sequence from the extreme left toward the centre (Fig. 2). Fig. 3 . a. Relative dominance (1 = 1 -5 %; 2 = 6 -20 %; 3 = 21 -50 %; 4 = > 50 %) per stand for trees of Pinus strobus plotted on the stand ordination by DCA (Fig. 1) . b. Ibid. for P. resinosa. c. Percentage frequency of occurrence for seedlings of Pinus strobus plotted in the same ordination diagram. d. Ibid. for P. resinosa. Fig. 3a illustrates the relative dominance (in four classes) of P. strobus, on the corresponding stand ordination. While P. strobus attains maximum dominance in the centre of the ordination, the species clearly is associated with P. resinosa and P. banksiana further to the left, but also extends into the hardwood complex to the right. In addition, P. strobus is associated both with the oak complex toward the lower end of axis 2 and with the tolerant conifers at the upper end of axis 2. The spread of P. resinosa on the stand ordination, in contrast, is limited to the left side and does not extend into the (hardwood sections at all (Figs. 3 b) . P. banksiana is even more limited to the extreme left hand cluster (not shown). The abundance of P. strobus saplings, as a percentage contribution to the total number of stems per stand (relative density), differs from that of the trees in being largely limited to pine-dominated forests with some extension into the oak-dominated region of the ordination. There is a notable absence of pine saplings in the tolerant conifer and tolerant hardwood regions of the ordination (not shown). Seedlings of P. strobus in the understorey are more widespread in the ordination diagram than the saplings and shows something of a concentration in the tolerant hardwood region (Fig. 3c) . Saplings of P. resinosa and P. banksiana were almost wholly restricted to the stand clusters dominated by their respective tree species. Seedlings of P. resinosa were somewhat more widespread than the trees and saplings but they appear sporadic in occurrence (Fig. 3d ). The data summary by ordination, therefore, locates the optimum occurrence of each of the pine species as being rather distinct compared with most other tree species. P. strobus shows more intermingling with the hardwoods and tolerant conifers than P. resinosa. The comparatively restricted range of P. strobus sapling occurrence indicates that seedling establishment is less of a barrier to in situ regeneration than subsequent survival and growth. The 170 stand dataset consisting of white and red pine dominated forests on the Canadian Shield contain higher total basal area, by an order of magnitude, for the pine species than any others. Other species that are prominent include Thuja occidentalis, Abies balsamea, Populus tremuloides and Betula papyrifera (Table 2) . P. strobus is the most frequent throughout stands in the data but P. resinosa is comparable to A. balsamea and B. papyrifera. Compared with the 320 stand hardwoodconifer dataset, therefore, the shade-tolerant hardwood species, including A. saccharum and Fagus grandifolia, are poorly represented in the pine dataset. The pine stand tree species basal area data were divided into pseudospecies, representing up to seven height classes for each true species. These pseudospecies data were then ordinated by DCA and CCA in order to explore the relations with postdisturbance stand age, stand productivity, as assessed by site index, and by the more directly measured site related variables. The percentage of total compositional variance in the respective ordination axes drops more quickly in the CCA than the corresponding DCA. However, the first CCA axis accounts for approximately 90 % of the variance in the first DCA axis (Table 3 ). The species-environment biplot of CCA axes 1 and 2 indicates a strong primary trend in tree species composition linked to white pine site index (PWSI) and postdisturbance stand age (Fig. 4) . The second axis appears, from the biplot diagram, to be linked with a soil depth gradient which is inversely related to slope angle. The third axis (not shown) corresponds best with the microclimate index. Because each tree species has been split into height classes a through to g, the shift in associates and environmental features is evident with increasing size. Mid-height stems of P. strobus are abundant throughout the data under a wide range of conditions. Height classes b, c, d and e are, therefore, seen located at their average position near the centre of the ordination. Short P. strobus stems appear mostly associated with oak species and others on shallow sites and steep slopes. To a certain extent, these stands represent the youngest in the data and offer comparatively limited growth potential. However, the stands also include the most uneven size structures for pine in the 170-stand dataset. The tallest white pine, classes f and g, appear towards the right associated with Acer saccharum, short Thuja occidentalis, old stands and high PWSI (Fig. 4) . Therefore, large white pine trees derive from a combination of stand longevity and productive soils but with a slight bias toward the latter. The bias is more apparent when ordinated in three dimensions on the first three axes (not shown). The pattern for P. resinosa is similar to that for P. strobus but with a general shift toward warmer and drier microclimates for small height classes and a stronger bias toward stands of advanced age, versus high productivity, for the tallest stem classes. Although comparable interpretations could be applied for all other species shown, the data do not necessarily include their full ecological range in the region. Therefore, these species are not considered. The stand ordination by CCA, corresponding to that for the species-environment biplot, of the height class data demonstrates a clear segregation along the second axis between P. resinosa dominated stands at the upper end and P. strobus dominated stands at the lower extreme (Fig. 5) . The first axis places stands of increasing age and productivity from left to right. Plotting the seedling density for each species on this diagram illustrates that seedlings of P. strobus are abundant but those of P. resinosa are not. Significantly, the oldest stands, toward the bottom right of the ordination, support relatively few or no pine seedlings whereas Table 5 . DRAIN = drainage class; AGE = estimated stand age (yr) since last disturbance. younger and mid-aged stands support a wide range of pine seedling densities. To examine links between pine regeneration potential and stand structure, a scatterplot of total woody stem density versus the density of well established P. strobus seedlings (i.e. 11-150 cm tall) illustrates that an upper bound of tree density appears to limit the number of seedlings (Fig. 6) . However, many points lie below and to the left of this upper bound line indicating that other factors also operate to limit pine seedling density. A conclusion from this phase of the work is that the potential for in situ regeneration of P. strobus, as indicated by the abundance of seedlings, is quite strong but that the nature of stand structure and the phase of stand development only play a partial role in promoting or restricting seedling establishment and persistence. To further explore correlations between pine seedling abundance and environmental factors, a multiple regression/correlation was performed between P. strobus seedling density as the dependent variable and 22 site related variables. These included moisture regime, microclimate index, slope angle, soil texture class, depth of organic matter, depth to bedrock, mineral soil pH, available nutrients (P, K, Ca, and Mg), C/N ratio of the F-layer, average cover of unvegetated parts for forest floor substrata (logs, needle litter, exposed rock, broadleaf litter), white pine site index (PWSI), total basal area, percentage conifer basal area, percentage hardwood basal area, DCA on understorey vegetation (axes 1 and 2). Of these variables PWSI, broadleaf litter, pH and Ca group 3. Although at considerably lower abundance, both seedling height classes of P. resinosa concentrate in group 4. The four stand groups are quite uneven in size, respectively accounting for 9 %, 58 %, 27 %, and 6 % of the total. Table 5 , showing average basal area of tree species and site characteristics in each of the four stand groups, indicates that the small groups 1 and 4 support far less tree biomass than groups 2 and 3. Indeed, tree density in group four, comprising a few very old P. resinosa stands is exceedingly low. These Classification was conducted on the understorey vegetation data of the 170 pine stand dataset using four seedling density variables as constraints. These were (1) P. strobus seedlings < 10 cm; (2) P. strobus stems 11 cm to 150 cm; (3) P. resinosa seedlings < 10 cm; (4) P. resinosa stems 11 cm to 150 cm. COINSPAN division was terminated at the four group level and Table 4 illustrates the concentration of P. strobus seedlings < 10 cm in stand groups 3 and 4. However, a large concentration of larger seedlings (11 -150 cm) is seen only in Table 4 . Average density, ± standard error (s.e.) of Pinus resinosa and P. strobus seedlings in two height classes, among the four stand groups derived from a constrained indicator species analysis (COINSPAN) of understorey vegetation, using the pine seedling density data as constraining variables. Group Table 5 . Average basal area (dm 2 /ha) and average site-related variables in each of the four understorey stand groups. Moisture score runs from 1 (fry) to 5 (wet); Family particle size class runs from 1 (coarse) to 8 (fine). stands are also the least productive, on the criteria of PWSI and total basal area. In contrast, the most productive stands in the data belong to group 2 and support, on average, over twice the biomass of the next most productive group (3: Table 5 ). The distinctive feature of Group 1 is that Picea mariana is prominent and, on examining the stand structure profiles (not shown), has clearly invaded the understorey. Group 1 stands are the northernmost in the data and here white pine regeneration may be influenced by climate. Group 2 stands only include hardwoods, Thuja occidentalis (eastern white cedar) and Tsuga canadensis (eastern hemlock) in any number. In addition, Abies balsamea is prominent in this group as an invasive understorey tree. Few distinguishing features emerge for stands in group 3, except that they lack the hardwoods and competitive understorey conifers and support a moderate to low density of pine stems. Table 6 summarizes the mean cover of the dominant understorey species in each of the four stand groups. Species dominant in stand group 3, in which P. strobus seedlings are abundant, include Pteridium aquilinum, Pleurozium schreberi, Dicranum polysetum, Vaccinium angustifolium and Maianthemum canadense. Group 1, that with the minimum number of P. strobus seedlings, is typified by abundant Pleurozium schreberi, Corylus cornuta, Abies balsamea, Acer spicatum, Ptilium cristacastrensis and Diervilla lonicera. The very large group 2 is typified by Corylus cornuta, Abies balsamea, Aralia nudicaulis, Maianthemum canadense, Aster macrophyllus, and Acer spicatum. The small group 4 is characterised by abundant Vaccinium angustifolium, Picea mariana, and Pteridium aquilinum. The recurrence of the same dominants among these groups reflects the broad range of the main understorey species throughout all such pine stands. However, the shifts in peak abundance for each species among the four groups indicates more telling differences (Table 5) . On this criterion, species which show such optima in group 1 include Hylocomium splendens, Ptilium crista-castrensis, Coptis trifolia, Pleurozium schreberi and Clintonia borealis. This assemblage typifies a feathermoss carpeted boreal Table 6 . Percentage presence (%P), average cover among all 170 stands (%C all), and average cover (%C) in each of the four understorey stand groups derived by COINSPAN (see Table 4 ), for those understorey plant species occurring in 10 or more of the 170 stands. The comparatively widespread occurrence of Pinus strobus throughout the northern hardwood-conifer forest complex indicates that the species is generally successful in achieving a significant presence notably on well-drained and somewhat, but not extremely, nutrient poor sites. In contrast, Pinus resinosa is quite restricted to dry, shallow soil outcrops and to well-drained sand plains. Trees and seedlings of P. strobus are more widespread than saplings, which occcur only in pine and oak dominated stands. One interpretation would be that an episodic wave of pine recruitment is occurring through-forest floor beneath a P. mariana canopy. Distinctive group 2 species include Aster macrophyllus, Acer saccharum and Dryopteris intermedia. In addition there are many minor species which typify a herb-rich, deciduous forest floor. Group 3 species include Polytrichum juniperinum, Carex communis, Dicranum polysetum, Gaultheria procumbens, Juniperus communis, Deschampsia flexuosa and Cladina rangiferina. These species indicate a heterogeneous forest floor of illuminated gaps, shallow soil, rocky outcrops interspersed with deeper soil pockets. Distinctive group 4 species are: Vaccinium angustifolium, Dicranum ontariense, Picea mariana seedlings and Viburnum cassinoides. out the northern hardwood complex. However, the northern hardwood -conifer data were accumulated between two and three decades ago and no corresponding waves of pine recruitment to the sapling and canopy strata are evident in the more recent pine forest survey data. Another interpretation is that white pine trees in mixed forests, on relatively productive soils, produce seed which germinate and establish but which tend not to survive into larger seedlings and small saplings. Faster growth and associated competitive suppression by the hardwoods is the most likely mechanism (Hibbs 1982) . Some of the oak dominated stands in the data support the lowest density of trees and are more akin to savannas than closed forests. Such stands include moderate white pine seedling frequencies and a notable pine sapling component. In these habitats it seems probable that a proportion of establishing pine seedlings will survive and eventually form an upper, discontinuous canopy or emergent supercanopy (Whitney 1986) . The distribution of trees and saplings of P. resinosa, with a very limited presence of seedlings throughout the northern hardwood-conifer complex, indicates that no continuous regeneration is occurring for this species. Indeed, saplings, as recorded here may simply represent small, suppressed trees of the same age as those stems recorded as trees in a stand. Some stands may support several cohorts, including saplings, because of surface fire recurrence and the stimulating effect this has on red pine seedling establishment (Burns & Honkala 1990; Ahlgren 1976; Cwynar 1977) . The restricted number of seedlings also implies that P. resinosa is not particularly adventive into new habitats as seedlings, in contrast with P. strobus. Analysis of the 170-stand, pine-dominated dataset is largely consistent with conclusions drawn from the broader stand base of upland northern hardwood-conifer forest data. Comparable separations between P. resinosa and P. strobus dominated stands are evident in the ordination results. Similarly, the comparative abundance of P. strobus seedlings is consistent between the two surveys. The environmental features, emphasized by the CCA ordination (Fig. 4) , lend more detail to the site related factors linked to the distribution of each species. Nevertheless, none of those environmental features recorded were strongly linked to abundant pine seedling densities, although some were significantly associated (PWSI, broadleaved litter, pH, Ca). An unexpected finding from the stand structure data is the comparatively poor regeneration of P. strobus in old-growth pine stands with the largest trees (Fig. 5) . Shade tolerant conifers, notably Abies balsamea and Thuja occidentalis, cast heavy shade, deposit abundant litter and greatly reduce throughfall in many such stands in our data. Seedling establishment and mortality may be short-lived events under such conditions. The strongest finding is the tight relationship between established pine seedling abundance and the understorey vegetation composition. Alternative explanations for a certain understorey vegetation type being associated with high white pine seedling density are threefold: (1) the understorey species are more sensitive indicators of appropriate site conditions than the single point-in-time measures and estimates of site and microsite that were made in the course of the vegetation surveys; (2) certain species, most abundant in understorey group 3, act as a favourable seedbed for pine; (3) species in the other three understorey groups directly inhibit pine seedling establishment. Stands on sites with comparatively shallow soil overlying cracked but otherwise impervious bedrock, promote the survival and persistence of P. strobus seedlings (cf. Ahlgren 1976) . Associated with such forest floor conditions, the tree canopy cover must be discontinuous or there must exist abundant lateral illumination, as on hill slope sites. However, the site related measures do effectively capture many such features. Furthermore, the understorey vegetation may differ considerably on otherwise comparable sites (Table 5 ) and this may relate to past opportunities for establishment and spread. Consequently, it seems unlikely that the vegetation is acting as an indicator of appropriate contemporary site conditions for pine seedling establishment, which cannot otherwise be detected, to any significant degree, by survey measurements. Moss species of the genus Polytrichum offer favourable seedbed conditions for shade intolerant conifers (Burns & Honkala 1990 ) and these show a peak in understorey stand group 3. However, there was no simple correlation in our data between pine seedling abundance and Polytrichum cover, nor for any other understorey species or life-form group. One explanation may be that it is the combined seedbed effect of all plants in a certain understorey community which promotes pine establishment. This seems unlikely among a suite of species in understorey group 3 with growth forms which vary from moss mats to coniferous shrubs and which are not necessarily unique to that stand group. Inhibitory conditions correspond, on one hand, to those stands in which a herb-rich and deciduous shrub understorey dominates, as in understorey group 2. On the other hand a feathermoss dominated forest floor appears equally inhibitory, as in understorey group 1. Herb rich stands characterize sites of relatively high site index (PWSI: Table 5 ). Competitive suppression of pine seedlings by a comparatively fast growing, broadleaved flora and by tolerant conifers, down through the attenuated canopy profile to the forest floor, is probable. At the other extreme, stands with abundant feathermoss cover invaded the understorey of a thinning white pine stand on such productive sites and have subsequently grown to produce the main canopy. White pine survives in these stands as large, very old emergent trees at low density. Thus, at the landscape level, large gaps on productive sites appear to be readily colonised by the shade tolerant tree species which dominate the forest complex (Frehlich & Lorimer 1991) . Heinselman (1981) identified two classes of fire origin pine forest in the pre-settlement era: (1) pure P. resinosa and P. strobus needle mat stands growing on fire prone sites with light surface fire cycles of 5 to 50 yr and severe fires at longer intervals; (2) mixed pine with hardwoods and/or shade tolerant conifers on relatively productive sites which experience only very severe fires in the frequency range of 150 to 300 yr. Both types are evident in the pine survey reported here. Whereas recruitment of pine to the upper canopy is unlikely in the mixed type (2) forests, it is probable that some limited second and even third generation establishment will occur in type (1) stands under an active fire prevention regime. Where canopy cover is incomplete, subcanopy tolerant conifers are at low density, broadleaved tree, shrub and herb species are limited, and feathermosses are relatively low in abundance, then white pine seedling establishment and subsequent growth seem probable. offer a very poor seedbed for conifer seedlings. Feathermosses dry out completely near the surface of the moss mat during dry spells (Jessee 1990) . Whereas the poikilohydric mosses tolerate such episodes of moderate local drought, vascular plant seedlings, with shallow rooting in the moss carpet, cannot avoid permanent wilting. Alternatively, where frequent wetting occurs, the aerial growth rate of feathermosses can outstrip that of pine, smothering the seedlings. Competitive and inhibitory processes, therefore, appear to primarily govern in situ pine seedling establishment. Where conditions cannot sustain the productive broadleaved species and tolerant conifers, yet also where the moss carpet is poorly developed, then white pine seedling establishment seems probable given sufficient seed rain and acceptably low levels of predation and parasitism (Abbott & Quink 1970) . The two stand surveys reported here were conducted on well established, closed canopy forest vegetation. Sites of recent episodic disturbances by extensive blow-down and fire were not included, largely because of the non-availability of such stands under prevailing forest landscape management regimes. The cohort origin of most stands indicates that such episodic disturbances determine most P. strobus regeneration and nearly all of that by P. resinosa. Most of the pineries of the Great Lakes region established following extensive surface fires associated with both fire usage by hunting tribes (Curtis 1959 ) and slash burning following land clearance (Heinselman 1973) . Pine seed for colonizing such disturbed areas appears to have originated from surviving individual trees, from those on adjacent ridges and from along the margins of wetlands. Such widespread fires promoted pine establishment over a broad range of soil, site and climatic conditions by delaying competition from broadleaved species on the relatively productive sites (Kershaw 1993; Maissurow 1935 Maissurow , 1941 van Wagner 1971) . However, it appears from our results that it was not only broadleaved tree species that acted as potential competitors but also other components of the forest ecosystem including tolerant conifers, broadleaved shrubs, herbs and feathermosses. Little evidence emerges in our data for the scenario in which a cluster of pine trees establish in a large forest gap on a productive site, followed by encroaching competitive exclusion by hardwood species from the exterior of the cluster to finally leave a single surviving pine tree at the centre (Hibbs 1982) . Our results indicate that the understorey vegetation is likely to outcompete pine seedlings under such circumstances. However, monospecific tree clusters or patches do appear to govern the dynamics of early, postfire succession between white pine and aspen (Squiers & Klosterman 1981) . More in evidence here is the situation where hardwoods have Field Manual for Describing Soils SAS user's guide: statistics. Version 6. SAS Institute Ecology of eastern white pine seed caches made by small forest mammals Regeneration of red and white pine following wildfire and logging in northeastern Minnesota Ecological effects of forest fires Fire regime in red pine stands at the northern limit of the species range Constrained Indicator Species Analysis (COINSPAN): A strategy for the classification of ecological community survey data A new list of mosses of North America north of Mexico The Vegetation of Wisconsin Recent fire history of Barron Township Stand structure and successional development of the white and red pine communities in the Temagami forest. Ontario Ministry of Natural Resources The relationship of natural disturbances to white pine stand development Natural disturbance regimes in hemlock-hardwood forests of the upper Great Lakes region Growth and development of white pine (Pinus strobus L.) at Lake Temagami The New Britton and Brown Illustrated Flora A fourth checklist of the lichens of the continental United States and Canada Fire in the virgin forests of the Boundary Waters Canoe Area Fire and succession in the conifer forests of northern North America White pine in the transition hardwood forest TWINSPAN. A FORTRAN program for twoway indicator species analysis. Section of Ecology and Systematics Detrended correspondence analysis: an improved ordination technique A ready reference to the description of the land in Ontario and its productivity. Preliminary Report Ontario Department of Lands and Forests Age structure analysis of a virgin white pine, Pinus strobus, population. Can. Field Natur Water retention and drainage in the forest floor organic layer under upland black spruce Early successional processes of eastern white pine and red pine in the Great Lakes-St. Lawrence forest. Forest Fragmentation and Biodiveristy Project, Rept Fire as a necessary factor in the perpetuation of white pine The role of fire in the perpetuation of virgin forests of northern Wisconsin The phytosociology of the deciduous forests of extreme southern Ontario Fire succession and community structure in red and white pine stands Fire as a controlling process in the North American boreal forest Self-replacement in old-growth white pine forests of Temagami Forest Regions of Canada Spatial patterning and competition in an aspen-white pine successional system Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis CANOCO -a FORTRAN program for canonical community ordination by correspondence analysis, principal components analysis, and redundancy analysis Fire and red pine Relation of Michigan's presettlement pine forests to substrate and disturbance history Acknowledgements. Financial assistance derived from NSERC operating grants (to P.F.M. & T.J.C.) and the Ontario Ministry of Natural Resources (OMNR: forest fragmentation and biodiversity project), for field data acquisition. OMNR staff were extremely helpful with advice and logistic support for the pine forest survey.