A Scandinavian perspective on ecological gradients in north‐west European mires: reply to Wheeler and Proctor

In their recent review of gradient relationships in mires, Wheeler & Proctor (2000) question the central paradigm in Scandinavian mire ecology that there are three major gradients in mires, which relate to nutrient richness, water-table depth and proximity to the mire margin (see Sjörs 1948; Malmer 1962; Økland 1989a). Wheeler & Proctor (2000) propose that the simple ‘poor–rich’ gradient should be replaced by a dual system considering ‘…[first] variation in pH and concentrations of metallic cations [and, second] variation in the availability of limiting nutrients, primarily phosphorus (P) and nitrogen (N)’. They further suggest that the current practice of using the mineral soil water limit that distinguishes between ombrotrophic and minerotrophically influenced sites ‘as a general main division within mires’ should be discontinued because this limit ‘is not sharp, and cannot be related to consistent differences in either vegetation or water chemistry’, stating instead ‘That the most important natural division [of mires] is between “bog”, with pH generally < 5.0, … and “fen” with pH generally > 6.0’, ‘reflected in a bimodal distribution of pH’. The mire expanse–mire margin gradient is considered ‘not in itself clearly definable or ecologically useful’, while Wheeler & Proctor (2000) recognize that the water-table and minor gradients may be significant. Gradients have been extensively studied on many Scandinavian (Swedish and Norwegian) and Finnish mires during the 20th century (see Malmer 1986; Økland 1989a, 1992). Our intention is to discuss the results of these Fennoscandian studies, which were not taken into account by Wheeler & Proctor (2000) in terms of their proposals, in order to seek a consensus on ecological gradients in north-west European mires. From studies, mainly on British and Dutch mires, Wheeler & Proctor (2000) concluded that there are two major, more-or-less independent, gradients related to ‘richness’. (We use the term gradient as synonymous with ecocline: i.e. variation in species composition and variation in environmental factors that is correlated with variation in species composition.) Similar conclusions were reached by Bridgham et al. (1996), based upon studies in North Carolina peatlands. Fennoscandian studies do not, however, provide evidence for a ‘fertility’ gradient that is independent of the ‘poor–rich’ gradient. Studies in which multivariate methods have been applied to data sets comprising the full range from acid sites poor in Ca to alkaline, calcareous sites (Pakarinen & Ruuhijärvi 1978; Heikkilä 1987; Vorren et al. 1999) all reveal a ‘poor–rich’ gradient without separate gradients related to P- and/or N-availability. In a recent study of spruce swamp forests in south-east Norway by Økland et al. (2001), environmental variables associated with either the alkalinity or fertility gradient by Wheeler & Proctor (2000), join to form one complex gradient correlated with the ‘poor–rich’ gradient in vegetation. Furthermore, Fennoscandian studies of ‘rich’ fens typically show strong variation in extractable (Sepponen et al. 1978; Økland et al. 2000) and total (Starr & Westman 1978) P, and fail to reveal a consistent relationship between P and other elements in peat (compare Tyler [1979], Heikkilä[1987] and Økland et al. [2000]). Partial independence of the vegetation gradients related to P availability and alkalinity has been demonstrated only in ‘extremely rich’Schoenus-dominated calcareous fens in south-east Sweden by Tyler (1979). The difference between the English and central European studies, and those from Fennoscandia, is likely to be due to regional differences in the supply of nutrients to mires: C Europe, S England and S Sweden, but not C and N Fennoscandia, receive large amounts of N, to the extent that P has replaced N as the growth-limiting nutrient in mires (Aerts et al. 1992). As a result, a single ‘poor–rich’ gradient in mire vegetation from bog to moderately rich fen (correlated with several factors, including soil and water acidity [pH], electric conductivity, and concentrations of Ca, Mg, Mn, total N and even P) appears to be typical where deposition of airborne pollutants is low, although more-or-less independent gradients related to ‘fertility’ (P availability) and alkalinity occur elsewhere. In support of a ‘natural’ bipartition of mire sites into ‘poor’, buffered by humic material and with (water) pH < 5.0, and ‘rich’, buffered by the bicarbonate system and with pH > 6.0, Wheeler & Proctor (2000) cite three studies showing bimodal distributions of water pH. A ‘natural bipartition’ does, however, require that the distribution patterns of environmental factors and species composition coincide. Opinions on the distinctness of the ‘poor–rich’ vegetation border reflect the authors’ standpoint in the continuum controversy (see Økland 1990a). While Du Rietz (1949) stated that the limit between ‘poor’ and ‘rich’ sites is one of the sharpest borderlines in mire vegetation, Sjörs (1952) considered it indistinct. Thus, several ordination studies of mire vegetation from different parts of the world (Pakarinen & Ruuhijärvi 1978; Jeglum 1991; Prieditis 1999) demonstrate continuous distribution of plots along axes related to peat acidity vs. alkalinity. However, a study of 150 swamp-forest plots from south-east Norway by Økland et al. (2000) reveals a sparse region in detrended correspondence analysis (DCA) ordination, suggesting a clear split into ‘poor’ and ‘rich’ vegetation, and a bimodal distribution of water pH (Fig. 1), as proposed by Wheeler & Proctor (2000). However, the data are inconsistent, as the separating antimode is at water pH ≈ 4.3 rather than at the ≈ 5.5 predicted, and does not coincide with the separating antimode for the gradient in species composition; ‘poor’ and ‘rich’ plots overlap broadly over a range of water pH from 4.5 to 6.0 (Fig. 1; Økland et al. 2001). Frequency distribution for water pH (measured in situ in tube wells in August 1998) in 150 1-m2 plots from 11 spruce swamp forests in Østmarka, south-east Norway (Økland et al. 2000, in press). The mid-point of each of 10 equal pH intervals is indicated on the horizontal axis. Plots are divided into ‘poor’ and ‘rich’ on the basis on species composition (with scores < 2.5 SD and > 2.5 SD units, respectively, along the first axis of a detrended correspondence analysis ordination of the full species composition of 150 plots, 1 m2 each), as indicated by filled and unfilled bars, respectively. This division corresponds to a sparse region (relative discontinuity) in the ordination. There is a broad-scaled geographical pattern of variation in vegetation in relation to water acidity and associated factors. ‘Intermediate fens’, where there are species typical of both ‘rich’ and ‘poor’ sites, are widespread in the boreal zone of Scandinavia (Sjörs 1950; Fransson 1972). Chemically, such fens are characterized by a combination of high water pH (up to 6.5) and a relatively low base mineral content (Witting 1949; Sjörs 1952). Gradual transitions in both vegetation (Cooper & Andrus 1994) and chemical characteristics (Sjörs 1952) have been reported. In the southern, nemoral and boreo-nemoral zones, fen vegetation of ‘intermediate character’ is much less frequent (Sjörs 1950, 1952; Malmer 1962; Fransson 1972) and the pH of mire water more closely follows its base cation content (Sjörs 1952). The Østmarka study area, which is situated on the southern border of the boreal zone, supports intermediate fen vegetation where concentrations of several metallic cations are as low as in ‘poor’ swamp-forest plots (a typically boreal trait), while also having a distinct floristic split (a typical southern trait), although the split here corresponds neither to an antimode in water pH nor to water pH = 5.5. Other observations that do not fit into this geographical pattern have been provided from the northern hardwood region of the US by Vitt & Slack (1984) and from boreal Canada by Vitt et al. (1995). Available evidence thus supports Sjörs’s (1952) view that there is a strong degree of individuality among study areas with respect to both the sharpness of the ‘poor–rich transition’ and the way in which particular environmental variables contribute to the gradient from acid, base-poor sites to alkaline, base-rich sites. This is at odds with existence of a ‘natural’ limit, as claimed by Wheeler & Proctor (2000), and calls for pragmatism in classification. The hydrological distinction between a ‘bog’, exclusively fed by ombrogeneous water, and a ‘fen’, which also receives minerogeneous water, has been recognized as important in Scandinavia since the 1940s (e.g. Sjörs 1948; Malmer 1962; Fransson 1972). Wheeler & Proctor (2000) give five reasons for reducing the emphasis on this ‘limit’: (i) it is hydrologically indistinct; (ii) no discontinuity in vegetation coincides with it; (iii) it cannot be characterized by a regionally consistent set of ‘fen plants’ (species indicating supply of minerogeneous water); (iv) the chemical composition of rain water varies regionally; and (v) unverified sets of indicator species used practically to determine this limit make the argument circular. Detailed integrated hydrological and botanical investigations (Sjörs 1948; Malmer 1962; Økland 1989a) show that the absolute limit for supply by minerogeneous water, the hydrological limit between bog and fen, can actually be determined as a narrow borderline zone, even in cases where minerogeneous water spreads diffusely over a very gently sloping raised bog surface. Thus, this limit is one point on a gradient from truly ombrogeneous sites, via sites with weak and intermittent supply of minerogeneous water, to strongly minerogeneous sites. Furthermore, because species appear along this gradient in a site- or region-specific sequence, locally valid fen plants can be identified and the circular argumentation avoided (cf. point v discussed earlier). The absence of a ‘sparse region’ in the ordinations of Pakarinen & Ruuhijärvi (1978), Økland (1990b) and Singsaas (1990) suggest that point (ii) does apply to Fennoscandia, as do (iii) and (iv) (Økland 1990c). However, the regional variation in the relationship between species’ responses to the gradient from ombrogeneous to minerogeneous sites (and thus their suitability as indicators of the latter) reflects the distribution of habitable sites (Økland 1990c), and accords with the general rule in biogeography that species will generally become more demanding towards their distributional limits (Hengeveld & Haeck 1981). Available evidence indicates that the mineral soil water limit is hydrologically distinct and that it is characterized by at least a local set of indicator species. The distinction between mire sites supplied with minerogeneous water and those that are not is known to be important (e.g. Clymo 1984) and is no less precise than any along the ‘poor–rich’ gradient. Furthermore, regional variation in the distribution of species on habitats and, hence, in their value as ecological indicators, is typical of most (or all) environmental gradients (e.g. Boyko 1947; Økland 1996). We therefore recommend that use of the mineral soil water limit is retained. In the absence of any generally distinct limit along the ‘poor–rich gradient’, we suggest that the present usage of bog is retained for sites that are not fed by minerogeneous water and that the present usage of fen is retained for sites that are fed by minerogeneous water. This terminology is well established in the mire hydrological literature, and amended definitions based on vegetation characteristics as suggested by Damman (1995), Bridgham et al. (1996) and Wheeler & Proctor (2000) will cause extensive confusion. Furthermore, we suggest that the terms ‘poor’ (can be subdivided into extremely and moderately poor), intermediate and rich (can be subdivided into moderately and extremely rich) are retained for steps along the gradient (see Table 1). Ample evidence, such as lists of differential species (e.g. Sjörs 1948, 1952; Malmer 1962; Fransson 1972; Økland 1989a) and vegetation gradients identified by ordination (Pakarinen & Ruuhijärvi 1978; Økland 1990b; Vorren et al. 1999), demonstrate that Fennoscandian mires often show considerable shifts in species composition from the (mostly open) mire expanse to the mire margin (mostly with trees). Less gradual shifts in species composition occur from the mire margin (swamp forest) to forest on mineral soil (Økland et al. 2000; Økland et al. 2001), suggesting that this gradient is useful as a floristic concept. Wheeler & Proctor (2000), however, maintain that ‘… this “gradient” embraces a number of different effects, varying from site to site …’ and that it therefore lacks ecological usefulness. We admit that the ecological cause of this ‘direction of variation in vegetation’ remains unclear, with no agreed ecological explanation. Nevertheless, there are indications that, at least in Fennoscandia, this gradient is caused by a specific set of environmental factors. Malmström (1931), Økland (1989a) and Økland et al. (2000) provide empirical support that the annual range of water-table fluctuations is larger, and that well-aerated peat occurs over larger areas and longer time-periods, in mire margin than in mire expanse sites (Malmer 1962, 1986). Waterlogging constrains tree growth, and mire margin conditions therefore enable trees to grow to greater heights. Trees reduce radiation and increase litterfall to the understorey and are therefore likely to determine its species composition. The mire margin–mire expanse gradient therefore partly parallels the gradient in the understorey vegetation of forests on mineral soil, from gaps between trees to beneath trees and in dense tree stands (Økland & Eilertsen 1993; Rydgren 1996; Økland et al. 1999). Until further studies on the causal factors underlying this gradient in species composition have been published, we recommend that it is considered as regionally important. The term ‘mire margin’ is ambiguous in the sense that ‘vegetation of a mire margin type’ may also occur near the geographical centre of a mire and we therefore suggest that the somewhat ambiguous term ‘mire margin’ is replaced by swamp forest (Økland et al. 2000). Vegetation of a mire margin type (sensuSjörs 1948) is mostly, but not entirely (see Fransson 1972), developed on mire sites with trees, and the occurrence of trees appears to be both the cause and effect of variation along this gradient. We propose the term open mire as a replacement for ‘mire expanse’, as the latter term has failed to gain general acceptance. Wheeler & Proctor (2000) state that ‘there has been little attempt to categorize mires quantitatively into “wetness types”… in part because point-to-point variation of water level within a mire is often relatively small, and equalled or exceeded by temporal variation.’ However, the temporal and spatial patterns of variation in depth to the water-table and its relationship with variation in species composition has been thoroughly studied in Fennoscandia. At a given point on a mire, the water-table depends on the flux of water, determined primarily by precipitation and temperature (Malmer 1962; Økland 1989a). Strong correlations among depth-to-the-water-table characteristics (such as the minimum, median and maximum depths) over one or several growing seasons in spatially explicit points on one or several mires clearly demonstrate that the water-table fluctuates in a temporally and spatially consistent manner (Malmer 1962; Økland 1989a; Økland et al. 2000). The importance of depth to the water-table and the mechanisms by which it influences species abundances and composition are known for individual mires (Malmer 1962; Tyler 1981; Økland 1989a, 1990b; Nordbakken 1996b; also see Rydin 1993). The Pearson product–moment correlation coefficient, r = 0.920, found between species scores along the main DCA ordination axis and median depth to the water-table for 800 plots on the boreal mire N. Kisselbergmosen in south-east Norway by Økland (1990b), is among the strongest relationships between a vegetation gradient and an environmental variable reported in the ecological literature. Scandinavian mire ecologists after Sjörs (1948) have used the strong relationship between species composition and water-table characteristics for dividing mires into microtopographic or microstructural levels along the water-table gradient; each level having characteristic physiognomy, species composition and depth-to-water-table properties (Sjörs 1948; Malmer 1962; Økland 1989b). Sjörs (1948) and Fransson (1972) divide the gradient into carpet, lawn and hummock; Malmer (1962) and Økland (1989a) further divide lawns and hummocks into lower and upper levels. In sites with trees, species’ responses to the water-table gradients tend to become somewhat less clear-cut and even three floristically distinct levels may be difficult to recognize (Økland et al. 2000). Wheeler & Proctor’s (2000) statement that ‘it is surprisingly difficult to relate species distributions to water-tables in between-site comparisons’ strongly contrasts with the results of regional studies in Fennoscandia, where such geographical variation in species’ distributions is predictable (Malmer 1986; Økland 1990c; Rydin 1993). The Fennoscandian literature suggests three fine-scaled local gradients in addition to the ‘spring–flush–fen gradient’ and the ‘lithotrophic–thalassotrophic [fresh–salt] gradient’ mentioned by Wheeler & Proctor (2000). A vegetation gradient from strongly peat-producing, ‘progressive’, Sphagnum-dominated sites to weakly peat-producing, ‘regressive’ sites, dominated by hepatics or lichens or covered by bare peat, was recognized by Malmer (1962) and Økland (1989a), and has been substantiated by ordination studies (Økland 1990b; Nordbakken 1996a). No relationship has been found between this vegetation gradient and measurable environmental factors; it is most likely caused by natural vegetation dynamics (Malmer 1962; Økland 1989a, 1990b). Mire margin fens grade from continuous, flat, lawn-like, drier areas dominated by large bryophytes such as Polytrichum commune and Sphagnum girgensohnii to sites such as small vertical walls that occur within hummock-like drier areas (Økland et al. 2000) and are dominated by small mosses and hepatics (‘pocket species’; Økland & Bendiksen 1985). Within hummocks, sites range from those with a stable snow cover in winter, dominated by ‘normal [chionophilous] mire vegetation’ to others, dominated by chionophobous lichens such as Cetraria nivalis and C. cucullata (Sonesson 1969). The importance of this gradient increases from the middle boreal to the alpine vegetation zones. A fourth gradient, as yet unsubstantiated, is related to water flow rates and oxygenation (Havas 1961). We conclude that two major and generally important gradients exist in north-west European mires while two gradients can be considered as major and regionally important, and five additional gradients have been demonstrated locally to be important. Table 1 summarizes the main characteristics of these gradients and may be useful as an informal gradient reference frame, from which locally or regionally important gradients may be selected for use in practical description, classification and communication in general. We thank P.D. Moore and M.C.F. Proctor for constructive comments on an earlier version of the manuscript. Received 14 July 2000 revision accepted 26 January 2001