1 Nutrient dynamics and growth of the feathermoss Hylocomium splendens were studied in a subarctic birch woodland. My aims were to estimate the species’ nutrient use strategy (and its determinants: mean residence time (MRT) and annual nutrient productivity (aNP)) and to unravel possible traits related to nutrient conservation and their implications for ecosystem nitrogen flux. Three methods to estimate nutrient losses in bryophytes were evaluated: a conventional growth analysis technique, a retrospective analysis and a 15N-tracer approach.
2 Estimates for nitrogen retention varied between 3 and 10 years as a result of the different N pools considered by the three methods. Growth analysis results depended on the distinction between live and dead tissues, whereas retrospective analysis gave valuable information on N release from decaying segments but did not measure MRT within the living segments. Valid estimates of MRT were obtained by a 15N-tracer approach.
3 MRT and aNP of H. splendens were similar to values typically found in woody evergreen vascular plants. Efficient nutrient recycling and a relatively long segment life span were responsible for the long residence time of nitrogen. Feathermosses show efficient nutrient acquisition, nutrient recycling and acropetal transport; nutrient losses will therefore be small.
4 Dominant bryophytes may retard the nutrient turnover at the forest floor through their production of acidic nutrient-poor organic matter and their negative effect on soil temperature, and they may therefore function as autogenic ecosystem engineers.
Low temperatures and a short growing season limit plant performance at high altitudes and latitudes. Low temperatures reduce mineralization rates and this – together with a strong microbial competition for free mineral nutrients (Michelsen et al. 1995; Jonasson et al. 1996) – leads to the growth of plants of cold regions generally being nutrient limited. Nevertheless, gradients of nutrient availability exist even within the tundra (e.g. Shaver & Chapin 1991; Giblin et al. 1991; Weih 1998; Weih & Karlsson 1999) and the various combinations of traits exhibited by higher plants in response to such contrasts can be characterized as different strategies of nutrient use (Eckstein & Karlsson 1997). The extremes are represented by fast growing species (herbs and graminoids) with a short retention time of growth limiting nutrients and by slow growing species (mainly evergreens) with a long residence time of these nutrients. The relative performance of plants can be expressed by the index of nutrient use efficiency (NUE) (Berendse & Aerts 1987), which is defined as the annual amount of dry matter produced per unit of nutrient lost. Berendse & Aerts (1987) partitioned NUE into two components: (i) the mean annual nutrient productivity (aNP), which measures the annual dry matter production per unit nutrient in the plant, and (ii) the mean residence time (MRT). Nitrogen productivity depends on photosynthetic nitrogen use efficiency and nitrogen allocation within the plant (Garnier et al. 1995) and represents a measure of the species’ productivity. MRT, which is estimated as the ratio between the annual average nutrient pool and the annual nutrient losses, is an index of nutrient conservation and depends on tissue life span and nutrient resorption (Escudero et al. 1992; Aerts 1995; Eckstein & Karlsson 1997; Garnier & Aronson 1998).
The few available data on the NUE of bryophytes suggest that these plants are either very efficient (Chapin & Shaver 1989) or are intermediate (Oechel & Van Cleve 1986) as compared to various vascular plant life-forms. To my knowledge, there has been no attempt to characterize the nutrient use efficiency sensuBerendse & Aerts (1987) of any individual bryophyte species. There are three reasons why bryophytes do not lend themselves easily to the growth analysis needed for ecophysiological studies. Firstly, the annual dry matter productivity, a central variable in growth analysis, is difficult to estimate in the vast majority of species because they lack innate annual markers of growth. Second, there is no clear-cut distinction between live and dead tissues in bryophytes, nor can these states be defined unambiguously. Third, due to the origin of the gametophytic shoots from microscopic protonema, followed by intense ramification, clonal growth and an unknown degree of clonal integration, it is often difficult to define functional growth units. Therefore, the colony rather than the ‘individual’ is regarded as the functional unit in many bryophyte species (Bates 1998).
Despite these problems, information on the nutrient use strategy of bryophytes is essential in order to understand the role of bryophytes in terrestrial ecosystems. An increasing body of evidence suggests that traits of plants adapted to nutrient-poor or nutrient-rich habitats may have feedback effects on decomposition and ecosystem nutrient turnover (e.g. Hobbie 1992; Berendse 1998) and thus have a stabilizing effect on ecosystem development (Aerts 1999). I studied growth and nutrient use of Hylocomium splendens, an important bryophyte of boreal and subarctic woodlands. My aim was to characterize this species in terms of its nitrogen (N) use strategy and to compare its nitrogen productivity and its ability to conserve N with that of various types of vascular plants. I addressed the following questions:
1 What is the mean residence time of growth limiting nutrients in H. splendens? Mosses characteristic of nutrient-poor habitats should be conservative in terms of their nutrient use. Therefore, I hypothesize that H. splendens will show a long mean residence time of growth limiting nutrients (high nutrient retention) similar to evergreen vascular plants. Since MRT and aNP make conflicting demands on plant performance, a high MRT in H. splendens will be paid for by a low annual productivity per unit nutrient.
Furthermore, given the problems of growth analysis in bryophytes, I used three different methods for estimating the losses and the retention time of N. A traditional growth analysis based on sequential harvests over 2 years was compared with a 15 N-tracer approach where the fate of an initial amount of label was followed through time, and with a retrospective analysis of growth and N and phosphorus (P) content of interconnected annual segments.
Materials and methods
The pleurocarpous, perennial H. splendens, is a characteristic element of the understorey vegetation at high latitudes (Oechel & Van Cleve 1986). Unlike many mosses, H. splendens shows clear innate markers which correspond to its annual growth (e.g. Callaghan et al. 1978; Økland 1995, 1997), and thus distinct annual segments of known age can be identified. A number of interconnected annual segments will be termed a ‘segment chain’ and individual segments designated either according to the year they started to grow (e.g. S 98 for shoots originating in 1998) or according to their age (S 0 for current-year shoots), depending on the analysis.
Each segment of H. splendens is the product of one meristem and therefore represents a ramet. Earlier studies suggest strongly that there is clonal integration among adjacent interconnected ramets and that considerable acropetal transport of nitrogen occurs under field conditions (Økland 1995; Eckstein & Karlsson 1999).
The present study was carried out in an open, bryophyte dominated birch forest in subarctic Sweden in the vicinity of the Abisko Scientific Research Station (68°21′N, 18°49′E, 360 m a.s.l.). The regional climate is subarctic: the average annual temperature is − 0.8 °C (1961–90) and there is an annual total of c. 300 mm precipitation (Andersson et al. 1996). The vegetation can be classified as an Empetro-Betuletum hylocomietosum Nordh. 1943 (Dierßen 1996) with the feathermosses H. splendens and Pleurozium schreberi dominating in the bryophyte layer. Within this forest type, I selected 20 apparently undisturbed populations of H. splendens, all of which had a very low abundance of vascular plants or other mosses. For the 15N study, 30 10 × 10 cm plots with a cover of H. splendens of > 80% and a low abundance of vascular plants were established within an area of about 15 × 15 m adjacent to the 20 study populations.
This analysis was carried out in 1996 and 1997 and H. splendens was harvested on six occasions during each year (at about three-week intervals from early June to late October). On each of the 12 occasions, two to four segment chains were taken from each of the 20 populations, selecting intact, mainly unbranched, sympodial segment chains to reduce the variation in nutrient content due to different modes of ramification (Økland et al. 1997). The chains were divided into their component annual segments (ramets) in the laboratory. The oldest ramets were 5 years old (S 5) and had started growth during 1992. All material was dried at 70 °C for 24 h and weighed to the nearest 0.1 mg. The average dry weight of the segments of each age class was calculated for each population. Variation in the data thus expresses spatial variation among populations.
The labelling and harvesting procedure for the 15N-tracer study is described in detail in Eckstein & Karlsson (1999). In short, the mosses were activated by wetting with distilled water before H. splendens in half of the selected plots was labelled with an ammonium-sulphate solution on two consecutive days (6 and 7 June 1997) using a simple spraying bottle. Each labelled plot received 0.15 mg 15N (15 mg 15N m−2) in total, which accounted for about 1% of the average total N pool per plot while the remaining (control) plots received only distilled water. Ammonium was chosen as the N source because in the study area it is present in larger amounts in precipitation and wet deposition than is nitrate (Kindbom et al. 1994). Furthermore, ammonium is more important as a source of mineral soil nitrogen than is nitrate in dry to mesic heath birch forests (Weih 1998).
Three harvests of all the segment chains within a study plot were carried out: (i) 12 June 1997 (ncontrol= 12, nlabel= 12), (ii) 8 September 1997 (ncontrol= 8, nlabel= 9) and (iii) 22 September 1998 (ncontrol= 8, nlabel= 8). From each sampled plot, 30 segment chains were selected and divided into segments that were current-year growth, 1-year-old growth, other green (usually only 2 to 3-year-old) growth and brown growth. The remaining H. splendens segment chains from each plot were also kept but not separated. All material was dried at 70 °C for 24 h and weighed to the nearest 0.1 mg. During the 1998 harvest all annual segments were separated and only brown segments originating from 1993 and older were pooled. The mass of each segment type per plot was then estimated using the ratio of the total dry weight of H. splendens per plot and the dry weight of the selected segment chains (see Eckstein & Karlsson 1999).
On 26 September 1998, three to five segment chains were collected from each of 16 of the 20 populations used in the growth study. Segment chains were selected as for the earlier analysis but were also chosen to include as many ramets as possible. Each chain was separated into annual segments, dried and weighed. The oldest ramets originated from 1989 and were thus 9 years old. After weighing, ramets of the same age from each population were pooled and homogenized. The dry weights of all segments of the same age class were averaged for each population.
Samples from the growth study and from the retrospective approach were analysed after digestion with concentrated sulphuric acid for total Kjeldahl nitrogen using a flow injection analyser (Fia-star 5012 system, Foss Tecator AB, Höganäs, Sweden). Samples from the retrospective study were also analysed for phosphorus with a flow injection analyser using the ammonium-molybdate method.
Samples from the 15N study were analysed for total N and 15N by the Waikato Stable Isotope Unit (Hamilton, New Zealand) using an ANCA-MS (Europe Scientific Ltd, Crewe, UK).
Calculations and statistics
Differences among the three methods applied in this study with respect to the basic study unit and definitions of nutrient pools and losses used to calculate MRT are summarized in Table 1. According to the concept of Berendse & Aerts (1987), nitrogen flux is measured as N losses and the MRT of nitrogen is defined as the ratio between the average N pool and annual N losses.
Table 1. Summary table of the basic study unit, the definition of nutrient pool and loss related to the different methods applied in this study. The annual nutrient losses are an estimate for nutrient transfer out of the bryophyte. MRT is defined as the ratio between average nutrient pool and annual nutrient loss
Basic study unit
Sum of green segments (S 0, S 1, S 2, S 3 [S3 only until August])
Segments turning brown during the season (S 3 in autumn)
Annual segment (ramet)
Average green segment during growth phase (S 0, S 1, S 2)
Segments during degeneration phase (linear regression)
Initial amount of label
Amount of label not recovered
Dry matter productivity was calculated based on the minimum and maximum mean dry weight of each segment age class (S 0, S 1, etc.) per year, if these were significantly different (P < 0.05) using Tukey’s test after one-way analysis of variance with harvest as a fixed factor. In the current-year segments, productivity was assumed to equal the maximum dry matter reached during the year concerned.
N and P pools were calculated based on the mean dry matter and mean N and P concentration summed over all green (= living) segment types (Table 1).
The fate of the added 15N label was followed by calculation of the excess 15N content of different segments, i.e. the total amount of 15N reduced for the natural abundance of this isotope (0.366% of the total N). By definition, the pool of 15N was the amount of tracer recovered during September 1997, while the difference in 15N between September 1997 and September 1998 was designated as N losses (Table 1).
The retrospective analysis is an approach that differs from both growth analysis and the 15N technique in one important respect: each annual segment is assumed to represent the result of past periods of growth and decay accumulated during its lifetime (cf. Callaghan et al. 1997). All calculations have therefore to be expressed per ramet. Dry matter productivity refers to the maximum average dry weight attained divided by the segment’s age. The N and P pool was calculated as the average N and P content of green segments, i.e. S 98 through S 96 (Table 1). Using the plot of nutrient pool vs. segment age, nutrient losses are expressed as the slope of a linear regression through the degeneration phase (cf. Callaghan et al. 1997), i.e. S 94 through S 89 (Table 1).
Dry matter productivity
Growth analysis showed that both current-year and 1-year-old segments (S 0 and S 1, respectively) showed a significant increase in dry weight over the growing season (Fig. 1). Total dry matter productivity of these segment types was 9.4 and 7.2 mg during 1996 and 1997, respectively (Table 2). Dry matter production of current year’s segments in 1997 (2.7 mg) was only 52% of that of 1996 (5.2 mg). Segments older than one year did not show significant variation in dry weight with time (Tukey’s test, P > 0.05).
Table 2. Dry matter productivity, average nitrogen (N) pool, N losses, annual N productivity (aNP) and mean residence time of N (MRT) of Hylocomium splendens in a subarctic woodland during two study years (1996 and 1997) determined using different methods: a conventional growth analysis (GA) and a 15N-tracer technique (15N). nd = not determined
Retrospective analysis revealed that H. splendens segments achieved a maximum dry weight at an age of 4 years (Fig. 2). The average annual productivity per segment during the growth phase estimated using the retrospective analysis was 2.7 mg (Table 3).
Table 3. Average dry matter production, nitrogen (N) and phosphorus (P) losses, average N and P pool size, annual nutrient productivity (aNP) and mean residence time (MRT) of N and P of Hylocomium splendens from a subarctic woodland determined by retrospective analysis
Refers to all segments during the growth period (S 0 through S 4).
Refers to green segments during the growth period (S 0 through S 2).
Refers to segments during the degeneration period (S 4 and older).
The average N pool was calculated as the average N content of live segments, i.e. S 0 through S 3, during the season. The assumption that these segments were still alive was based on their green colour and on the observation that net photosynthesis becomes negative at an age of about 3 years (Callaghan et al. 1978; Oechel & Van Cleve 1986). Since S 3 segments turn brown during autumn this segment type was counted as alive from June to August. The N pool was slightly smaller in 1997 as compared with 1996 (Table 2).
The retrospective analysis revealed an average N and P pool size of 4.6 and 0.49 µmol per ramet, respectively (Table 3).
Estimation of nitrogen losses using the growth analysis method were based on the above distinction between live and dead segments and yielded a loss per segment chain per year of about 4 µmol (Table 2).
Using the 15N approach, losses averaged 1.1 µmol N per 100 cm2 plot per year between September 1997 and September 1998 (Table 2).
The retrospective analysis used variation in the N and P pool size of differently aged segments in the degeneration phase sensuCallaghan et al. (1997) as an indicator of N and P losses. Both linear regressions were significant: nitrogen pool size = 6.99–0.446 × segment age, R2 = 0.819, P = 0.0096, n = 6; phosphorus pools size = 0.6094–0.0454 × segment age, R2 = 0.952, P = 0.0015, n = 6. Thus, ramets lost an average of 0.446 µmol N and 0.045 µmol P per segment per year during the first 6 years of the degeneration phase (Fig. 3, Table 3).
Nitrogen translocation in h. splendens
15N analyses suggested that considerable acropetal transport of added label occurred during the 1997 growing season (Fig. 4). Current-year segments (S 97) increased their relative share from 2 to 20% of the recovered 15N from June to September 1997. However, N recycling among ramets continued at about the same rate from September 1997 through September 1998 (Fig. 4). Tissues not existing when the labelling was carried out (S 98), had accumulated 1.7 µmol 15N per plot (29% of the total 15N recovered) in September 1998 (Fig. 4). Current segments for that year and 1-year-old segments together contained 61% of the label recovered during autumn 1998, as compared with 71% in autumn 1997 (Fig. 4). The amount of 15N in brown segments increased from 0.3 µmol per plot (4%) in September 1997 to 0.9 µmol (15%) in September 1998.
Nitrogen productivity and mean residence time
Values for mean annual nitrogen productivity (aNP) were similar when determined by growth analysis and by the retrospective approach (means of 615 and 587 g mol N−1 year−1, respectively) (Tables 2 & 3).
Estimates of mean residence time of N in H. splendens ranged between c. 3 and 10 years, depending on the method used (Tables 2 & 3). The lowest MRT was estimated using the growth analysis and the highest using the retrospective approach; the 15N-tracer technique gave an intermediate value.
Recycling and transport of nitrogen among ramets of h. splendens
The distribution of 15N among ramets of H. splendens presents evidence of considerable acropetal transport of nitrogen in this species (Fig. 4). Transport of carbon, phosphorus and metal cations has been found in H. splendens and other ectohydric species (e.g. Rydin & Clymo 1989; Wells & Brown 1996; Brūmelis & Brown 1997; Bates & Bakken 1998), i.e. species apparently lacking differentiated conducting tissue. The recycling of N from senescing segments in favour of the growing points appears to be a highly ordered process. All of the 15N captured by the two already expanded ramets that were present near the tip of the segment chain during labelling (i.e. S 96 and S 95 in June 1997) was recovered 3 months later in the now expanded current ramets (S 97), the 1-year-old ramets and the 2-year-old ramets (Eckstein & Karlsson 1999). Similarly, the label in the two topmost ramets in September 1997 (i.e. S 97 and S 96) accounted for all 15N recovered in current (S 98), and 1- and 2-year-old ramets 1 year later (Fig. 4). Similar patterns of translocation of N during these two observation periods suggest that transport may occur mainly internally, since it appears to be largely independent of weather conditions, which varied considerably between the study years. The year 1997 was relatively warm and dry with c. 77 mm precipitation during the growing season from June to August. In contrast, 1998 was much wetter than a normal year (148 mm compared to 121 mm, based on the means from 1961 to 90). It therefore seems unlikely that patterns of clonal integration in this species (Eckstein & Karlsson 1999) together with strong size relationships found between mature ramets (Økland 1995) are based solely on external transport, which would be vulnerable to leakage of nutrients during heavy rain showers (R.H. Økland, personal communication). Furthermore, Bates & Bakken (1998) present new experimental data which indicate internal (symplast) transport of phosphorus in Pseudoscleropodium purum, another pleurocarpous moss. The recent discovery of polarized conducting cells in ectohydric mosses (Ligrone & Duckett 1994, 1996), possibly capable of transporting relatively large volumes of solutes, helps to explain internal transport mechanisms in ectohydric mosses. In peatmosses (Sphagnum), Rydin & Clymo (1989) have demonstrated the existence of acropetal transport and its ultrastructural basis.
Comparison of different methods to estimate nutrient retention
There is considerable variation among estimates of MRT obtained by the different methods (Tables 2 & 3). These differences are probably related to different nutrient pools and nutrient fluxes estimated by the various methods. The lowest MRT was found using a growth analysis approach, where the estimation of MRT is based on assumptions about the distinction between living and dead segments. This method probably overestimates annual N losses and underestimates MRT, since it appears doubtful that all N in brown segments would be unavailable to the clone. Changing the definition of live and dead segments will significantly alter the estimated N pool, the annual N losses and, consequently, MRT.
The retrospective approach for estimation of nutrient losses by H. splendens yielded a nutrient retention of about 11 and 10 years for N and P, respectively (Table 3). Up to an age of 4 years, segments show a net increase in dry weight (Fig. 2), which is not associated with an increase in N pool size (Fig. 3). Retrospective analysis does not take into account possible nutrient losses in these most active segments and thus overestimates MRT. Retrospective analysis therefore represents an in situ decomposition assay rather than measuring the MRT of nutrients within the living parts of the segment chain, but as such it may provide interesting information at the ecosystem level.
When following the fate of added 15N, the amount of label, which cannot be recovered in the system, is considered as loss. In H. splendens, more than 95% of the recovered 15N was found in green segments 5 days after labelling (Fig. 4). As a consequence, senescent ramets showed 15N concentrations only marginally above the natural abundance (Eckstein & Karlsson 1999). The MRT of nitrogen estimated from losses of 15N through time was about 6 years (Table 2). However, labelling in the field may result in some variation in the total amount of 15N recovered per plot, which may influence the calculation of MRT. The accurate estimation of the average pool of 15N appears thus to be of crucial importance. Weber & Van Cleve (1981, 1984) encountered similar variation in the amount of label recovered during sequential harvests, but in the absence of information on the size of 15N pools over time, MRT cannot be estimated.
Estimated annual N losses might also change if the pulse of 15N were monitored over the whole lifetime of the labelled segment cohorts. However, a study in arctic Alaska showed that even 28 months after labelling feathermosses contained more than 90% of the label recovered and still showed 15N concentrations significantly larger than the natural abundance of the isotope (Weber & Van Cleve 1981). Furthermore, since the label appears to be thoroughly distributed among segments at the end of the present experiment, an annual loss rate of 1.1 µmol 15N plot−1 year−1 appears to be a robust estimate. However, assuming a 36% increase after September 1998, would only reduce MRT estimated over a 4-year interval to 5 years. Therefore, I conclude that the 15N-tracer technique appears to be a valid method for determination of MRT, although a long-term study on 15N dynamics within mats of H. splendens would be worthwhile.
Traits related to nutrient conservation
In flowering plants, loss of nutrients is mainly determined by the life span of parts that are shed and the efficiency with which nutrients are resorbed from these parts before abscission (Escudero et al. 1992; Aerts 1995; Garnier & Aronson 1998; Eckstein et al. 1999). The same holds true for feathermosses like H. splendens where, judging from changes in colouration, ramet life span is about 3 years (a conservative estimate since brown segments are not necessarily dead). This is within the range of leaf life spans found in woody evergreen shrubs (Karlsson 1992). Nutrient recycling is high, with as much as 78% of the N pool resorbed from senescing segments (S 95 + S 94, Fig. 4).
An important difference between vascular plants and bryophytes is that in the latter dead segments are not shed but remain an integral part of the segment chain. They allow living, assimilating ramets to be maintained in a favourable position with respect to incoming radiation and also act in the capillary transport of water to new growing points, thus maintaining a favourable water balance. Mean size of mature H. splendens segments showed positive density dependence in boreal spruce forests (Økland & Økland 1996). Dead moss segments may thus serve similar functions to the stems of woody plants. However, since mosses are growing close to the ground, decomposition of dead material occurs in situ.
Large forest floor mosses exert strong control on ecosystem nutrient flux by capturing mineral nutrients from canopy throughfall, incorporating them into organic N compounds, recycling them over several years and, after this time-lag, slowly releasing nutrients through decomposition. By building a surface layer of nutrient-poor, acidic organic substance and as a result of the negative effects of a thick moss cover on soil temperature (Van Cleve et al. 1983), bryophytes will maintain and reinforce the infertility of the habitat. In this respect, large feathermosses of the forest floor may act as ‘autogenic ecosystem engineers’sensuJones et al. (1994). Likewise, the peatmoss Sphagnum fuscum has been identified as small-scale ecosystem engineer in subarctic peatlands (Svensson 1995). By capturing mineral nutrients early in their circulation through the ecosystem, and subsequent internal recycling, this species retains growth limiting nutrients and counteracts the spread of vascular plants.
Since mosses show a negative response to perturbations that simulate climate change (Potter et al. 1995), accelerated nutrient turnover, increased efflux of mineral nutrients and changes in species composition may be anticipated as a consequence of global warming, especially in boreal and subarctic regions.
I am much indebted to my supervisor, P. S. Karlsson, for his support and encouragement during this study. P. S. Karlsson, H. Rydin, R. Aerts and two anonymous referees offered valuable comments which improved the quality of the manuscript. L. Haddon greatly improved the readability of the final version. K. Anton, J. Gaspar, M. Kardefelt and L. Ericsson helped in the field and during laboratory work. Financial support was obtained from the Swedish Natural Science Research Council and from the Abisko Scientific Research Station.
Received 9 September 1999revision accepted February 2000