Dynamics of plant-mediated organic matter and nutrient cycling following water-level drawdown in boreal peatlands

Abstract

[1] If boreal peatlands face drought more often due to climatic warming, the responses of vegetation may drastically change the functions of the ecosystem. We assessed the effects of water-level drawdown on plant-mediated organic matter (OM) and nutrient fluxes in a chronosequence of undrained and drained, originally sparsely treed fens. The chronosequence mimicked the reduced growing season moisture predicted by current climate change scenarios. In a pristine state, OM and nutrient fluxes were characterized by annual cycling through graminoids and mosses. Water-level drawdown initiated a “forest succession,” in which the OM and nutrient cycles shifted from being dominated by graminoids and mosses to dominance by arboreal vegetation in two decades. Simultaneously, the quantity and tissue type composition of annual litterfall, as well as the quantity and allocation of annual nutrient uptake, changed radically. The changes may have contrasting but as yet unexplored implications for the carbon and nutrient balances of these sites.

1. Introduction

[2] Water level, or soil moisture, is the major factor controlling plant community composition [Glaser et al., 1990; Jeglum and He, 1995], carbon (C) allocation [Weltzin et al., 2000], organic matter (OM) decomposition [Bridgham et al., 1991; Szumigalski and Bayley, 1996], and C fluxes [Moore and Dalva, 1993; Funk et al., 1994; Kettunen et al., 1999] in peatlands. Consequently, the function and element balances of peatland ecosystems may drastically change in response to alterations in water level. Climate change scenarios predict warmer temperatures and reduced growing season rainfall in parts of the boreal zone [Gitay et al., 2001]. These changes could cause a drawdown of the water level of peatlands [Roulet et al., 1992] which, although largely an indirect influence of climate change, may be the main agent of ecosystem change in the boreal and subarctic peatlands, outweighing the impacts of rising temperature and longer growing seasons [Gorham, 1991].

[3] In wet pristine peatlands, annual cycling of OM and nutrients through the ground vegetation, mainly graminoids and mosses, is important for all element cycles [Bernard and Hankinson, 1979; Urban and Eisenreich, 1988; Verry and Urban, 1992]. The poor aeration of the soil impedes the growth of unspecialized vascular plants, including most tree species [Jeglum, 1974; Macdonald and Yin, 1999], and the tree stand, when present, is usually sparse and has a minor role in the ecosystem function. Vegetation physiognomy usually clearly depends on the soil moisture regime [Verry, 1997].

[4] After water-level drawdown, the soil, and consequently, the whole ecosystem enter a dynamic state of change: a secondary succession that may last for decades [Laine et al., 1995a, 1995b; Vasander et al., 1997]. Because vegetation is an important mediator in the element balances [Hobbie, 1996; Aerts et al., 1999], changes in the C and nutrient cycling through vegetation have to be understood and taken into account when estimating ecosystem responses to climate change. Short measurement periods fail to capture the effects of the new succession on vegetation dynamics with their consequent implications for element balances, however.

[5] In parts of the boreal zone, large areas of peatlands have been drained to improve forest growth. These peatlands were usually sparsely forested (tree-covered) in their pristine state. An important point here is that ditching has commonly been the only operation done: the original tree stand (albeit often sparse) has formed the basis for the production forest stand. These drained peatlands give us an excellent opportunity to examine the long-term effects of water-level drawdown on peatland ecosystem function [e.g., Lieffers, 1988; Glenn et al., 1993; Martikainen et al., 1993; Roulet et al., 1993; Laine et al., 1995a; Nykänen et al., 1998; Minkkinen et al., 1999; Silvan et al., 2000]. In addition to climate change scenarios, the results of drainage studies are currently needed for estimating the effects of land use changes on the C balance.

[6] The aim of this study was to analyze long-term (several decades) ecosystem responses to water table lowering in terms of changes in OM and nutrient cycling. Our method was to combine results from chronosequence and repeated measures studies with those of dendrochronological tree growth analysis. More specifically, we wanted to evaluate the dynamics following water-level drawdown of (1) plant biomass, (2) annual litterfall, (3) plant biomass nutrient pool, (4) annual nutrient cycle through the vegetation, and (5) total nutrient uptake by vegetation and its allocation to aboveground and belowground biomass. We will use nitrogen (N) and potassium (K), two elements with differing cycling and deposition patterns, as examples.

2. Materials and Methods

2.1. Study Area

[7] The peatlands chosen for this study were situated in southern Finland (61°35′–61°52′N, 24°05′–24°25′E). The area is part of the southern boreal coniferous forest zone and the eccentric raised bog region. The mean annual temperature in the region is +3°C and that of July is +16°C. The mean annual temperature sum (accumulated mean daily temperatures ≥+5°C) varies between 1150 and 1250 degree-days. The annual precipitation is ca. 710 mm, of which ca. 240 mm is snowfall.

2.2. Site Description

[8] Six sites forming a drainage succession chronosequence were chosen for detailed measurements (Table 1). Two of the sites were undrained, while the rest had been drained 8, 22, 30, and 55 years before the beginning of the investigation period (summer 1991 for sites 2 and 5, 1992 for other sites). The drained sites had not been further managed after ditching. The initial lowering in the water level in our study sites, caused by ditching, was 15–20 cm (Table 1). Later on in the forest succession, increased evapotranspiration has induced a further lowering of the water levels down to 50–70 cm during the growing seasons in our older drained sites.

Table 1. General Features of the Study Sites at the Beginning of the Investigation Period^aa For details of vegetation characteristics, see the work of Laiho [1996], Laiho and Finér [1996], and Laiho and Laine [1997].

	Site
	1	2	3	4	5	6
Year of drainage	…	…	1984	1970	1961	1937
Stand volume, m^3 ha^−1	6	35	10	76	102	150
Shrub biomass,^bb Aboveground, dry mass (105°C), standard error of mean (20 subsamples) given in parentheses. Different superscripts indicate differences that are significant at the 0.05 level (one-way ANOVA and Tamhane's T2 for pairwise comparisons). g m^−2	41 (8)^A	76 (11)^A	288 (30)^B	103 (32)^AC	54 (20)^A	186 (27)^BC
Graminoid biomass,^bb Aboveground, dry mass (105°C), standard error of mean (20 subsamples) given in parentheses. Different superscripts indicate differences that are significant at the 0.05 level (one-way ANOVA and Tamhane's T2 for pairwise comparisons). g m^−2	58 (8)^A	72 (6)^A	60 (10)^A	22 (5)^B	15 (3)^B	5 (1)^B
Moss biomass,^bb Aboveground, dry mass (105°C), standard error of mean (20 subsamples) given in parentheses. Different superscripts indicate differences that are significant at the 0.05 level (one-way ANOVA and Tamhane's T2 for pairwise comparisons). g m^−2	534 (20)^A	349 (15)^B	644 (21)^C	137 (40)^D	193 (28)^D	452 (53)^AB
WT,^cc Average water table depth during growing season. cm	10	10	20	50	50	60
Peat N,^dd Total concentration in the 0–20 cm peat layer. % of dry mass	1.5	2.2	1.7	2.2	2.4	2.0
Peat K,^dd Total concentration in the 0–20 cm peat layer. % of dry mass	0.042	0.052	0.044	0.028	0.026	0.043

• a For details of vegetation characteristics, see the work of Laiho [1996], Laiho and Finér [1996], and Laiho and Laine [1997].
• b Aboveground, dry mass (105°C), standard error of mean (20 subsamples) given in parentheses. Different superscripts indicate differences that are significant at the 0.05 level (one-way ANOVA and Tamhane's T2 for pairwise comparisons).
• c Average water table depth during growing season.
• d Total concentration in the 0–20 cm peat layer.

[9] All sites represented tall-sedge pine fen communities before drainage [in the sense of Laine and Vasander, 1996]. The sites were subjectively selected from the material (84 sites) of an earlier study dealing with the effect of drainage on peat properties and vegetation [Laiho and Laine, 1994a, 1995; Laine et al., 1995a; Weiss et al., 1998; Laiho et al., 1999]. The original material had been carefully selected to allow comparison between various sites at different stages of the postdrainage succession. For details of the selection, see the work of Laiho and Laine [1994a]. The sites of the present study were chosen to obtain as wide a drainage age variation as possible within the chosen site type, and so that the tree stand volumes on the drained sites would be close to the average of that drainage age. The two undrained sites represented the average and the higher end of tree stand volume variation [Heikurainen, 1971].

[10] In their natural, undisturbed state, tall-sedge pine fens are wet sites with stunted Scots pine (Pinus sylvestris L.) growing on low hummocks, often with some birch (Betula pubescens Ehrh.). The lawn between the hummocks is dominated and characterized by sedge species, such as Carex lasiocarpa Ehrh and/or Carex rostrata Stokes. Also typical is the dwarf birch Betula nana L. The moss layer is continuous and dominated by one or the other of the Sphagnum recurvum collective species, most often Sphagnum fallax (Klinggr.) Klinggr., or Sphagnum angustifolium (Russ.) C. Jens. These sites are minerotrophic and are commonly found at the slopes or margins of raised bogs, as well as in aapa mires. With its counterparts, tall-sedge pine fen is a very common mire site type throughout the boreal zone [e.g., Pyavchenko, 1963]. In North America, ecologically equivalent sedge fens usually have Larix laricina (Du Roi) C. Koch as the major tree species [e.g., Harris et al., 1996].

[11] After drainage of these sites, sedges and mire herbs disappear fairly soon, and dwarf shrubs such as B. nana and Ledum palustre L. gain dominance. Mire dwarf shrubs give way to Vaccinium myrtillus L. and Vaccinium vitis-idaea L. as the shading from the tree stand canopy increases. Cottongrass (Eriophorum vaginatum L.) may remain as the only relatively abundant graminoid. Sphagnum mosses are gradually replaced by such forest species as Pleurozium schreberi (Brid.) Mitt., even though Sphagnum russowii Warnst. and other Sphagnum species may still be abundant on old drained sites. Changes in plant species composition after drainage are described in more detail by Laine et al. [1995a]. The detailed lists of plant species with coverages of the study sites in the beginning of the investigation period have been published by Laiho [1996].

[12] On undrained sites, sample plots were laid out in the middle of an area where vegetation uniformly represented the chosen site type. Plot sizes were 600 and 900 m². On drained sites, the sample plots covered a whole strip between ditches, with one ditch counted in the plot area. Plot size varied from 1158 to 1328 m² depending on the ditch spacing, which varied from 30 to 37 m.

2.3. Plant Biomass and Nutrient Pools

[13] Twenty equally spaced subplots were sampled at each site for ground vegetation biomass. Shrubs were harvested on an area of 0.5 m² per subplot, graminoids and mosses on an area of 0.19 m². Zero level for sampling was the upper level of the rooting zone, i.e., the layer where the topmost fine roots could be found. This corresponded approximately to the bottom of the green part of the moss layer. Sampling was done in late June, at the time of maximum annual aboveground biomass. Sampling was repeated at site 2 to check the stability of the nutrient pool values, which turned out to be within ±6–12% units. Ground vegetation biomass at the beginning of the investigation period was reported by Laiho [1996]. We used the mean value from two undrained sites to represent the predrainage situation.

[14] Nutrient concentrations were measured for the shrub, the graminoid, and the moss components separately. Total concentrations of P, K, Ca, and Mg were measured using an ICP analyzer (ARL 3580) after nitric acid-perchloric acid digestion. Nitrogen concentrations were measured using a Leco CHN 600 analyzer.

[15] All trees in the sample plots were measured for diameter at breast height (DBH) (diameter at 1.3 m), bark thickness, total height, and height to the base of the crown. Six to 14 biomass sample trees per plot were chosen on the basis of stratified random sampling to cover the whole range of diameters, and tree species (pine and birch). Tree biomass measurements have been described by Laiho and Laine [1997]. Element concentrations were measured for each biomass component (wood, bark, live branches, dead branches, and needles/leaves) in each sample tree, using the same methods as for ground vegetation. The total nutrient content in each sample tree was calculated as the sum of the component pools. Equations were developed to estimate the total biomass and nutrient content of a single tree using DBH as the independent variable (Table 2).

Table 2. Equations for Estimating Aboveground (Dry) Biomass and Nutrient Contents of Single Trees^aa C_i, kg for biomass, g for nutrients; DBH, diameter at 1.3 m (cm); A, dummy variable adjusting the estimate for recently drained sites (A = 1 if time since water-level drawdown <10 years, otherwise A = 0); s.e.e., standard error of estimate. Asymptotic standard errors of the estimated parameters p1–p3 are given in parentheses.

Ci	p1	p2	p3	R^2	s.e.e.	n
Pine: Undrained Sites (Ci = p1 DBH^p2)
Biomass	0.094 (0.043)	2.255 (0.170)		0.986	3.2	14
N	0.079 (0.055)	2.714 (0.253)		0.996	8.4	5
P	0.017 (0.009)	2.237 (0.198)		0.987	0.6	9
K	0.158 (0.095)	2.033 (0.224)		0.980	4.0	9
Ca	0.090 (0.045)	2.603 (0.186)		0.991	7.0	9
Mg	0.057 (0.033)	2.008 (0.217)		0.982	1.3	9
Pine: Drained Sites (Ci = p1 DBH^p2 + A DBH^p3)
Biomass	0.106 (0.028)	2.369 (0.083)		0.987	14.8	38
N	0.921 (0.513)	1.988 (1.173)		0.979	63.6	18
P	0.202 (0.120)	1.547 (0.191)		0.931	5.4	33
K	0.168 (0.055)	2.100 (0.103)	1.047 (0.127)	0.970	10.5	33
Ca	1.040 (0.503)	1.692 (0.154)		0.957	34.3	33
Mg	0.062 (0.021)	2.116 (0.107)		0.982	5.0	33
Birch (Ci = p1 DBH^p2)
Biomass	0.155 (0.071)	2.226 (0.191)		0.983	2.9	15
N	0.295 (0.230)	2.534 (0.288)		0.990	17.9	5
P	0.151 (0.126)	1.677 (0.332)		0.922	2.2	11
K	0.023 (0.012)	2.974 (0.195)		0.987	4.0	11
Ca	0.415 (0.219)	2.025 (0.225)		0.984	5.2	10
Mg	0.122 (0.055)	1.870 (0.177)		0.979	1.5	11

• a C_i, kg for biomass, g for nutrients; DBH, diameter at 1.3 m (cm); A, dummy variable adjusting the estimate for recently drained sites (A = 1 if time since water-level drawdown <10 years, otherwise A = 0); s.e.e., standard error of estimate. Asymptotic standard errors of the estimated parameters p1–p3 are given in parentheses.

[16] We estimated the change in aboveground tree biomass over time for each plot using dendrochronological growth analysis. The past annual DBH was reconstructed for each sample tree back to 5 years before drainage using their annual ring data. The previous year's diameter increment at breast height was then regressed against the current DBH, using KPL software [Heinonen, 1994]. Applying the regressions to the recorded diameters at breast height produced the DBH of each tree a year earlier. This procedure was repeated either until the fifth predrainage year or until the tree's DBH became zero. Thus estimates of the annual DBH distributions of the tree stands were obtained for as far back as the fifth predrainage year. The reconstructed annual DBH data were divided by tree species into 1-cm classes. The mean DBH of each class were fitted into the biomass and nutrient content equations (Table 2). The biomasses and nutrient contents thus obtained for each DBH class were multiplied by the number of trees in that class. These values were then summed up to give the total biomasses and nutrient pools of the tree stands each year.

[17] Fine root (Ø ≤ 10 mm) biomass at the beginning of the investigation period was measured and reported by Laiho and Finér [1996]. Sixteen to 20 peat cores were systematically taken from each plot from 0–30 cm depth in late June. Zero level was defined the same as for the ground vegetation sampling. The cores were divided into three 10-cm-deep subsamples, starting from the surface. Roots were extracted by hand. Living roots were separated into three species groups: Scots pine, shrubs and birch, and Cyperaceae (Carex L. spp. sedges and cottongrass, E. vaginatum L.), and four diameter classes:<1, 1–2, 2–5, and 5–10 mm, except for Cyperaceae roots almost all of which had a diameter smaller than 1 mm. Element concentrations were measured for each species group with the same methods as for the ground vegetation.

[18] We estimated coarse root (Ø > 10 mm) biomass for trees using equations developed by Laiho and Finér [1996] for Scots pine, and by Finér [1989, Appendix 19; with the correction that +a₂d should be +a₂ ln d] for birch. These equations estimate the total dry mass of stump and coarse roots in single trees, using DBH as an independent value. The element contents of coarse roots were estimated using the element concentrations data of Finér [1989] measured for the same tree species and site type as in this study. Both Finér's and our data included nutrient concentration values measured from the stump (cutting level) of the sample trees. Thus we were able to adjust Finér's concentrations for coarse roots by a stump-level concentration ratio in our versus her sample trees. Accumulation of biomass and nutrients in coarse roots was estimated using the reconstructed annual DBH data.

2.4. Litterfall

[19] Aboveground tree litterfall was measured for 5 consecutive years (1 June 1993–31 May 1998) at sites 2, 4, and 5, and for 2 years at site 6, which was later harvested. Fine woody debris (FWD; twigs and branches) was collected in twenty 0.5-m² traps (wooden frames, 0.5 m × 1 m) per site, laid systematically across the plots on top of the forest floor. Nonwoody debris was collected in nine 0.5-m² traps (polyethene funnels, diameter 0.8 m) per site, suspended 1 m above ground. Litter from these traps was sorted into needles, leaves, cones, and “other nonwoody litter.” Twigs and branches were excluded from these samples. Traps were emptied once a month during the snowless period (usually from May to November); wintertime litterfall was collected in the first sampling point in the spring. The sorted samples were dried at 70°C. Their moisture content was measured by drying a subsample at 105°C. Nutrient concentrations in the litter materials were measured using the same methods as for plant biomass.

[20] We estimated the annual amount and nutrient content of nonwoody litterfall from ericaceous shrubs using litterfall to biomass ratios obtained by Håland [1994] for similar shrub cover. The amount of litterfall in his data was 6.2% of the maximum standing stock biomass, the ratios for nutrients varying between 2.6% for K and 14.7% for Ca. Twig litterfall from shrubs was obtained from the FWD traps, and is included in FWD in this paper.

[21] The annual aboveground litterfall from graminoids was estimated to be 86% of the annual maximum biomass of shoots, as measured by Saarinen [1998] for C. rostrata in a mesotrophic fen in the same region as our sites. The nutrient content of this litter was estimated based on the results of Bernard et al. [1988] and Bernard and Hankinson [1979] as a proportion of the nutrient content of the maximum standing stock biomass. Bernard and Hankinson [1979] measured decreases of 73–79% in the nutrient content in C. rostrata between maximum (summer) and minimum (winter) biomass. Of these decreases, we assumed 50% to be caused by litterfall and 50% by translocation for N, P, and K, and 100% by litterfall for Ca and Mg, based on the work of Auclair [1982] and Bernard et al. [1988].

[22] The annual moss litterfall was estimated based on the results of Vasander [1982], who measured moss biomass and production in the drainage succession of a bog. In all his sites (n = 14), which were undrained and drained with various microsites, moss turnover rates of the mixed species compositions similar to ours followed:

with R² = 0.924 and standard error of 0.0403. This implies that moss biomass is a good indicator of the productivity of the moss community in the drainage succession. Where moisture conditions are favorable for moss production, moss biomass is also high and vice versa. For estimating the nutrient content in moss litter, we used the concentration gradient between the living and dead moss segments observed by Pakarinen [1978]. We used living-to-dead ratios of 3 for K, 1.5 for N, 1.4 for P, and 1.0 for Ca and Mg. These are averages of Pakarinen's values for hummock and hollow species. The ratios follow the generally observed pattern of nutrient mobility.

[23] The estimation of fine root litter production was related to species composition and successional stage. For sedges in the undrained sites, we applied turnover of 1.2 (based on late summer biomass) from Bernard [1974] who presented data for C. rostrata in Minnesota with a four-seasons climate similar to ours. In the youngest drained site, practically all graminoids were E. vaginatum, which is known to renew practically all its fine roots annually [e.g., Chapin et al., 1979], i.e., the minimum turnover would be 1. N. Silvan et al. (unpublished data) suggest that the turnover of E. vaginatum roots would fall between 1 and 1.5, so we applied 1.25. Fine root turnover for pine (0.65), shrubs (0.56), and graminoids (1.27) were measured by Finér and Laine [1998] in our 30-year-old drained site during 1991–1993. These values were also applied to the 22- and 55-year-old drained sites based on rather similar hydrology and species composition. For pine and shrubs, litterfall was estimated only for the <2 mm fraction, whose annual mortality is higher than that of thicker roots [Persson, 1980; Laiho and Finér, 1996]. The nutrient content of fine root litterfall was estimated assuming the same element concentration for dying roots as was measured for our living fine roots. According to Bernard [1974] and Bernard and Hankinson [1979], the minimum belowground biomass and nutrient content in a standing crop of sedges occurs in late June (our sampling time), the time when maximum translocation of nutrients has occurred for growth aboveground and when many old rhizomes die. We measured the nutrient concentrations of dead fine roots in our sites, and they did not suggest the occurrence of further translocation from sedge roots before senescence. Neither did notable translocation seem to take place from tree and shrub roots before senescence [see also the work of Nambiar, 1987].

2.5. Nutrient Fluxes

[24] We used the nutrient return in annual litterfall as the measure of the annual cycle of nutrients through vegetation. With perennial plants, such as trees and shrubs, that retain a major biomass and nutrient reserve over winter, we face a conceptual problem of timing. The nutrients returned to soil in litterfall in a certain year may actually have been taken up in earlier years. We assumed that this would be of no significant consequence in our analysis, where the focus is on longer-term trends and most annual variation is averaged out on purpose.

[25] The annual changes in nutrient pools in ground vegetation biomass (shrubs, graminoids, mosses) were calculated as the difference of their (annual maximum) biomass nutrient pools between two points of time in the chronosequence, divided by the number of years between the time points. For the trees, the annual changes were calculated as differences between the reconstructed annual nutrient contents of the stands using the dendrochronology data. Thus the tree stands have a higher estimation frequency than ground vegetation.

[26] Total nutrient uptake from soil per year was estimated by adding the nutrients annually cycled through vegetation to the annual changes in the plant biomass nutrient pools. Moss litter was excluded from the total uptake estimates, however, because (1) moss turnover was not measured in our sites and thus moss litter estimates involved the greatest uncertainty in our analysis and (2) difficulties in separating moss litter from “peat proper.” The total uptake was further partitioned to aboveground and belowground components to estimate the allocation of nutrient uptake to aboveground and belowground biomass. A separate total uptake estimate was calculated for the mosses.

3. Results

3.1. Plant Biomass and Litterfall Dynamics After Water-Level Drawdown

[27] In the undrained sites, the total (aboveground and belowground) plant biomass was about 1.7 kg m⁻², two thirds of which was found in the ground vegetation (Figure 1). Within 50 years following water-level drawdown, the total biomass had increased sevenfold, and more than 90% of it was tree biomass.

Figure 1

(top) Aboveground and (bottom) belowground plant biomass (dry mass, g m⁻²) in the drainage succession chronosequence. FR, fine roots (diameter ≤ 10 mm).

[28] A closer examination of the vascular ground vegetation biomass (Figure 2) showed that the shrub and sedge biomasses were originally about equal. However, while the graminoid biomass steadily declined over time, shrub biomass increased dramatically at first, peaking 10–20 years after water-level drawdown. After that, mire shrubs declined while forest shrubs gained dominance.

Figure 2

Aboveground biomass of vascular ground vegetation (dry mass, g m⁻²) by species groups: graminoids, mostly C. lasiocarpa Ehrh. (before water-level drawdown) and E. vaginatum L. (after water-level drawdown); mire shrubs, B. nana L., Empetrum nigrum L., L. palustre L., Vaccinium oxycoccos L., Vaccinium uliginosum L.; forest shrubs, V. myrtillus L., V. vitis-idaea L.

[29] Estimated total annual litterfall from vascular plants slightly increased over time after water-level drawdown, from ca. 550 g m⁻² yr⁻¹ in the undrained state and during two decades following water-level drawdown, to ca. 600 g m⁻² yr⁻¹ when 30–50 years had passed (Figure 3). The total annual litterfall estimate including moss litter peaked soon after water-level drawdown at 960 g m⁻² yr⁻¹, and thereafter ranged from ca. 600 to 850 g m⁻² yr⁻¹.

Figure 3

Annual (top) aboveground and (bottom) belowground litterfall (dry mass, g m⁻²) by litter type. Leaves, shrub leaves + birch leaves; other nonw., unidentified nonwoody tree litter (male flowers, bark fragments etc.); FWD, fine woody debris (twigs and branches). Belowground litter is fine root litter only.

[30] Changes in the vegetation composition were clearly reflected in the litterfall composition (Figure 3). On undrained sites, about 80% of the annual aboveground litterfall was estimated to derive from mosses and sedges. Belowground, an estimated 80% of litter derived from sedges. Within 20 years after water-level drawdown, however, shrubs and trees had become the major litter source, producing 90% of the total vascular plant litter, and 75% of the total litter estimate including moss litter. Over time, the composition of litterfall from arboreal plants changed gradually: the proportions of slowly decomposing components, such as woody debris and cones, increased.

3.2. Plant Biomass Nutrient Pools and Nutrient Cycling Through Vegetation

[31] The patterns of change in the nutrient pools in the vegetation standing stock generally followed the corresponding changes in biomass. Accordingly, aboveground, arboreal biomass nutrient pools increased rapidly after water-level drawdown, and maintained a rather consistent increase over the 55-year study period (Figure 4). Graminoids showed a steady decline in the biomass nutrient pools, while for mosses, N and K showed slightly different initial patterns. Belowground, water-level drawdown initiated a decline in graminoid fine root nutrient pools (Figure 4). Tree and shrub fine root nutrient pools showed a fluctuation with two periods of increase and an intervening recession during the third decade following water-level drawdown.

Figure 4

Total (left) N and (right) K pools in plant biomass (g m⁻² in dry mass). Trees aboveground and coarse root values are from the dendrochronological growth analysis of the study site drained 55 years before measurements. Ground vegetation and fine root (FR) values are from the chronosequence data.

[32] Before water-level drawdown, most of the annual nutrient cycling (measured as nutrient return in litterfall) was accounted for by graminoids (Figure 5) and mosses (Figure 6), which shed a considerable part of their biomass and nutrients each year. Arboreal plants accounted for less than 25% of the annual N cycle, and less than 10% of the annual K cycle through vegetation. After two decades, however, arboreal plants had a major role in the annual nutrient cycle. They accounted for 90–98% of the annual N and K cycles through vascular plants, and 60–90% of the total annual cycles including mosses.

Figure 5

Amounts of (left) N and (right) K cycled annually through vascular plant biomass (g m⁻² in dry mass) measured as the total nutrient content in annual litterfall.

Figure 6

Estimated total annual (left) N and (right) K uptake by mosses and its components, i.e., changes in the N and K pools of moss biomass, and annual N and K cycle through mosses.

[33] After the initial peak, the total annual N uptake by vascular plants that was allocated to the aboveground parts remained permanently on a higher level than before water-level drawdown (Figure 7). The total annual N uptake allocated to the belowground parts, in contrast, remained fairly constant at first, and even seemed to temporarily decrease after some decades.

Figure 7

Estimated total annual (top) N and (bottom) K uptake by plants, and its allocation aboveground and belowground.

[34] Comparing the uptake of K by vascular plants allocated aboveground and belowground gives us an interesting picture of the total K dynamics. The rapid decline in sedge fine roots after water-level drawdown was enough to cause a decrease even in the total K uptake (Figure 7). Later on, the highs and lows of aboveground versus belowground uptakes more or less cancelled each other out, and the total uptake remained reduced for as long as some decades.

4. Discussion

4.1. Peatland Forest Succession: Changes in the OM and Nutrient Fluxes

[35] The succession induced by water-level drawdown that we have described may be called a “peatland forest succession.” Species adapted to wet conditions declined while shrubs and trees proliferated. Along with the forest succession, the OM and nutrient fluxes shifted from ground vegetation dominance to tree dominance in about 20 years in our sites, which were originally very sparsely stocked with trees.

[36] Simultaneously, with the change in growth form dominance, the quantity and tissue type composition of annual litterfall, as well as the quantity and allocation of total annual nutrient uptake, changed. On one hand, there was a shift from annual cycling of OM and nutrients through the ground vegetation to a long-term cycle through the cumulatively growing tree stand, but simultaneously, the volume of the annual OM and nutrient cycles increased as well. These changes may affect the OM and nutrient balances of peatlands in several ways.

[37] Graminoid and herbaceous litters, which were the major vascular plant litter types in undrained sedge-pine fens, as in many other minerotrophic mire types, decompose faster than foliage litter from shrubs and trees [Hobbie, 1996; Szumigalski and Bayley, 1996], which became the major litter types after water-level drawdown. Woody materials, branches, and cones, decompose especially slowly [Taylor et al., 1991; Hobbie, 1996]. Both Sphagnum mosses and forest species that increase after water-level drawdown, such as P. screberi and Hylocomium splendens (Hedw.) B.S.G., generally decompose slowly [Hobbie, 1996; Karsisto et al., 1996]. With the forest succession, the ample moss and woody litter input, decreasing soil temperature [Minkkinen et al., 1999] and increasing acidity of the surface soil [Laine et al., 1995a, 1995b] may lead to surface accumulation of OM [Vompersky et al., 1992; Laiho and Laine, 1994b].

[38] When evaluating the changes in nutrient cycles, it is clearly essential to include all vegetation layers, and both aboveground and belowground parts of vascular plants. Nutrient uptake in vascular plants showed partly contrasting response patterns to water-level drawdown aboveground and belowground (Figure 8). These patterns were largely caused by the changes in species and growth form composition, and tree stand development [e.g., Helmisaari, 1995]. Initially, they may also reflect changes in the shoot:root ratios with changing soil moisture conditions [Weltzin et al., 2000].

Figure 8

Total (top) N and (bottom) K concentrations (mg g⁻¹ dry mass) and amounts per unit area (g m⁻²) in a 0–30 cm peat layer in the chronosequence material of Laiho and Laine [1994a, 1995]. Only tall-sedge pine fen sites are included, and values have been calculated for the 30-cm layer that forms the major rooting zone in these sites [Laiho and Finér, 1996]. The data have been smoothed with a robust locally weighted regression, where the local flex of the curve was restricted by including 80% of the data points in the running window.

[39] The trends for both N and K may be attributed to the vegetation dynamics (Figures 1 and 3) [also Laine et al., 1995b]. The uptake of both elements into the aboveground biomass of vascular plants increased soon after drainage, due to a buildup of both the relatively nutrient-rich foliage and the accumulating woody biomass of shrubs and trees. The N uptake allocated to belowground biomass (Figure 7) more or less directly reflected the patterns of change in the total belowground biomass and litterfall (Figures 1 and 3). The pattern of K uptake was strikingly different. The disappearance of sedges, which cycle a lot of K through their fine roots (Figure 5), led to a decrease in the allocation of K uptake to belowground biomass. The increasing shrub and tree roots did not reverse this decrease until after a decade. The second period of decrease in belowground K allocation was largely caused by the decrease of shrub and birch roots during the time of the major species change from mire shrubs to forest shrubs (Figure 2).

[40] The total uptake by vascular plants of N, which is abundant in the soil (Figures 4 and 8), increased and remained at a higher level than in the pristine state. One might end up drawing the same conclusion for K if observations were restricted to the aboveground biomass only. Including belowground biomass reveals, however, that the total uptake of K did not change essentially, or rather decreased, after drainage. The soil pool of K is small (Figure 8), not much higher than the plant biomass pool (Figure 4), thus total uptake obviously cannot increase: when it increases in one vegetation component, it simultaneously decreases in another (Figure 7). The total uptake of K reached the predrainage level only after about 40 years following drainage. At this stage, the tree stand was already well developed and could thus itself affect the element balance of the site through the increased interception of (dry) deposition [Schauffler et al., 1996]. This interception may be crucial for the K regime of forested peatland sites [Laiho et al., 1999].

[41] Despite the changing nutrient requirements of the vegetation, the nutrient pools in the top 30 cm of peat (the major rooting zone, Laiho and Finér [1996]) have been found to remain rather constant over several decades after water-level drawdown, in the absence of other disturbances (Figure 8) [Laiho and Laine, 1994a, 1995; Westman and Laiho, 2003]. Subsidence of the mire surface and compaction of the surface peat following water-level drawdown [Rothwell et al., 1996; Minkkinen and Laine, 1998] compensate for the decrease observed in base cation concentrations [Laiho and Laine, 1995; Laiho et al., 1999]. Thus on a nutritional basis, the forest succession may proceed and forest ecosystem functioning may be sustained.

[42] The soil pool of N is so large that vegetation uptake may alter it only slowly. Notable leaching of the largely organic soil N has not been observed from drained peatland sites [Sallantaus, 1992]. Thus the stability of the soil N pool may be expected. The stability of the small, largely water-soluble soil K pool, however, is interesting as it proves the highly efficient retention of K in the nutrient cycle [Stone and Kszystyniak, 1977; Damman, 1978; Miller et al., 1979]. Net output of K from the soil, likely through leaching [Kaunisto, 1997], seems to take place only during the period of lowered total plant uptake immediately following drainage (Figure 7), when the K concentration (mg g⁻¹) in the soil decreases (Figure 8). This net output is obviously mediated by the disturbance in the biological cycle. However, we cannot deduce from our material whether it is a cause for or a consequence of the lowered total uptake by plants.

4.2. Is Peatland Forest Succession Feasible in the Changing Climate?

[43] Tree and shrub remains observed in peat stratigraphic studies suggest that forest successions have been initiated in peatlands during the Holocene, caused by changes either in local conditions or in the climate leading to lowered water levels [e.g., Bridge et al., 1990].

[44] Ditching causes relatively rapid lowering of the water level [Hillman, 1992; Paavilainen and Päivänen, 1995]. To what extent can we assume the responses of a gradual climate change to follow those induced by artificial drainage? Just one exceptionally dry summer has been found to cause a similar change in the water level [Alm et al., 1999] as ditching during the first decade (Table 1, site 3), when there has not yet been such profound change in evapotranspiration that the tree stand later causes. For tree growth, a deep water level during the growing season, especially in summer, is much more essential than the dormant season or early summer water levels [Päivänen, 1984]. Thus even if wintertime rainfall increased, creating high late fall to early summer water levels, persistent low summer water levels might induce release growth in the tree stand. Tree stand productivity may reach its potential maximum when the water level remains below 30–40 cm during 60% (pine) to 80–90% (spruce) of the growing season [Vompersky and Sirin, 1997]. However, sufficiently long dry periods are more easily experienced on drained sites, where after intermittent flooding events the “excess” water is drained via the ditch network, unlike in undrained sites where lateral water movement is usually slow [Hillman, 1992; Mugasha et al., 1993; Verry, 1997; Vompersky and Sirin, 1997].

[45] With several consecutive dry summers, a forest succession might begin more or less similarly as after artificial drainage. How rapidly trees may respond to the reduced water level depends on tree species, size, and age [Seppälä, 1969; Macdonald and Yin, 1999]. A response may usually be seen already during the growing season following water-level drawdown, first in fine roots and needles, then in radial growth [Seppälä, 1969; Mugasha et al., 1993]. It may take 5–30 years for the release growth to reach its maximum, depending on tree species and size/age, and site nutrient regime [Seppälä, 1969]. Once the water level has been down long enough for a tree stand to become well established, the stand itself has a major effect on the water balance and water level of the peatland through greatly increased interception of rainfall, and evapotranspiration [Heikurainen and Päivänen, 1970; Päivänen, 1980]. If intervening wet summers occurred frequently enough, however, they might lead to the deterioration, even the death, of the tree stand. This might leave mire dwarf shrubs as the dominant vegetation, as is typical of early stages of the succession (Figure 2).

4.3. Sensitivity of Our Estimates

[46] We want to emphasize that the focus in interpreting our results should be on the pattern of change rather than the actual values. Due to the higher-frequency measurements, the biomass and nutrient estimates of trees may be considered the most reliable.

[47] Moss and fine root litterfall were the components of OM and nutrient fluxes for which we had the least intensive measurements. They are difficult to measure, and there are relatively little literature data to be used either. However, net primary production and biomass generally have a close correlation within physiognomic species groups [e.g., Moore et al., 2002] (Table 2). Consequently, we estimated moss and fine root litterfall by applying turnover ratios to the measured biomass data, accounting for the hydrological status (undrained versus drained).

[48] Concerning moss litter, equation (1), which leads to a maximum turnover of 0.66, may be considered to produce minimum estimates for our fen sites, as higher turnover rates have been observed in more nutrient-rich sites than the bog sites from which equation (1) was derived. Moore et al. [2002] reported a turnover as high as 1.4 for a poor fen. Still assuming the turnover:biomass relation to follow equation (1) in the drainage succession, applying this higher turnover rate would result in about doubled amounts of moss litter, which may be considered the maximum estimates. The pattern of change would remain the same, however, and the higher estimates would not change our conclusions but would further emphasize the role of mosses.

[49] Based on biomass, sedge roots are the major belowground components in OM and nutrient cycling in the pristine state, and tree and shrub roots following water-level drawdown. Fine root turnover of graminoids and herbs, related to late summer biomass (our biomass measurement time) seems to generally vary between 1 and 1.5 [e.g., Bernard, 1974; Bernard and Hankinson, 1979; Finér and Laine, 1998]. The values that we applied fall in the middle of this range. The turnover of tree and shrub roots was measured in our 30-year-old drained site. We see no reason to assume that the turnover rates would have been essentially different at the 22- and 55-year-old drained sites.

4.4. Conclusions

[50] (1) During their Holocene-long persistence, boreal peatlands have developed an inherent autogenic changeability, and react fast to changing hydrological conditions. It is important to account for the changes in plant-mediated OM and nutrient fluxes when evaluating ecosystem response to changing environmental conditions.

[51] (2) The main adaptation mechanism to long-term alteration of water level is change in species composition and relative abundance. Long-term water-level drawdown induces a decline in graminoids and proliferation of shrubs and/or trees. Mosses retain their functional importance after adaptation at the community level.

[52] (3) Consequently, the tissue type composition of litterfall changes. The litter material of dried-out sites is largely slowly decomposable, which will counteract the enhancing effect of water-level drawdown on decomposition processes.

[53] (4) Nutrient uptake by plants shows partly contrasting response patterns aboveground and belowground, and between elements abundant or sparse in the soil. On a soil nutritional basis, however, the developing peatland forest ecosystem may be sustained.

Acknowledgments

[54] The assistance of Veli-Matti Komulainen, Jouni Meronen, Kari Minkkinen, Mikko Tiirola, Luo Jun, Leena Finér, Seija Repo, Kirsti Derome, Matti Siipola, Mikko Kukkola, and laboratory staff members of the Finnish Forest Research Institute in Joensuu, Rovaniemi, and Salla at various stages of the study is gratefully acknowledged. Line Rochefort, Jukka Alm, and the referees made valuable comments on the manuscript. This study was funded by the Academy of Finland and the Finnish Ministry of Agriculture and Forestry.