Species differences in timing of leaf fall and foliage chemistry modify nutrient resorption efficiency in deciduous temperate forest stands

Transcription

1 Tree Physiology 25, Heron Publishing Victoria, Canada Species differences in timing of leaf fall and foliage chemistry modify nutrient resorption efficiency in deciduous temperate forest stands ÜLO NIINEMETS 1 3 and ÜLO TAMM 4 1 Department of Plant Physiology, University of Tartu, Riia 23, Tartu 51011, Estonia 2 Centro di Ecologia Alpina, I Viote del Monte Bondone (TN), Italy 3 Corresponding author (ylon@ut.ee) 4 Institute of Forest Research, Estonian Agricultural University, Kreutzwaldi 5, Tartu 51014, Estonia Received October 15, 2004; accepted December 14, 2004, published online June 1, 2005 Summary Extensive variation in fractional resorption of mineral elements from plant leaves is still not fully understood. In multi-species forest stands, species leaf fall phenology and leaf constitution may significantly modify the timing of nutrient return to the soil and overall plant nutrient loss. We studied leaf fall and nutrient loss kinetics, and leaf composition in three natural, temperate, deciduous broadleaf forest stands to determine the role of timing of leaf abscission and nutrient immobilization in cell walls on nutrient resorption efficiency of senescing leaves. Nitrogen (N), phosphorus and potassium contents decreased continuously in attached leaves after peak physiological activity during mid-season. Changes in nutrient contents of attached leaves were paralleled by decreases in nutrient contents in freshly fallen leaf litter. In different species and for different nutrients, resorption of nutrients from senescing leaves proceeded with different kinetics. The maximum nutrient resorption efficiency (the fraction of specific nutrient resorbed from the leaves at the end of leaf fall) did not depend on the mid-seasonal nutrient concentration. Species with earlier leaf fall resorbed leaf nutrients at a faster rate, partly compensating for the earlier leaf fall. Nevertheless, the litter-mass weighted mean nutrient contents in leaf litter were still larger in species with earlier leaf fall, demonstrating an inherent trade-off between early leaf fall and efficient nutrient resorption. This trade-off was most important for N. Losses of the non-mobile nutrients calcium and magnesium were unaffected by the timing of leaf fall. There was large variation in the maximum N resorption efficiency among species. Correlations among leaf chemical variables suggested that the maximum N resorption efficiency decreased with the increasing fraction of cell walls in the leaves, possibly due to a greater fraction of N occluded in cell wall matrix. We conclude that species leaf fall phenology and leaf chemistry modify the timing and quantities of plant nutrient losses, and that more diverse forest stands supporting a spectrum of species with different phenologies and leaf types produce litter with more variable chemical characteristics than monotypic stands. Keywords: calcium, cell wall content, litter quality, nitrogen, phosphorus, potassium, species diversity, temporal variation. Introduction The major determinant of nutrient cycling in forest communities is nutrient return in the form of leaf litter (Prescott 2002). There is a large variation in the fraction of mineral nutrients resorbed from the leaves before leaf fall as well as in the minimum nutrient contents of senescing leaves that is not entirely understood (Pugnaire and Chapin 1993, Killingbeck 1996, Eckstein et al. 1999, Aerts and Chapin 2000, Pugnaire 2001). Differences in translocation have been associated with leaf life span (evergreen versus deciduous) (Schlesinger and Hasey 1981, del Arco et al. 1991, Escudero et al. 1992b, Killingbeck 1996, Wright and Westoby 2003, but see Sheriff et al. 1995), overall leaf nutrient content (Chapin and Kedrowski 1983, del Arco et al. 1991) and site nutrient availability (Chapin and Moilanen 1991, Enoki and Kawaguchi 1999, Wright and Westoby 2003, but see del Arco et al. 1991, Escudero et al. 1992a, Knops and Koenig 1997). Apart from the broad patterns, there is large variation in the fraction of nutrients removed depending on plant functional type and environment (del Arco et al. 1991, Killingbeck 1992, Eckstein et al. 1999, Aerts and Chapin 2000, Wright and Westoby 2003), implying that large-scale general relationships may not necessarily be applicable at a smaller scale and within specific plant functional types. Complete nutrient resorption is time consuming and potential nutrient resorption is rarely, if ever, achieved in natural environments before leaf fall occurs (Killingbeck 1996). As a result of modifications of leaf phenology and fall kinetics, there may be strong year-to-year variations in minimum leaf nutrient contents in leaf litter and in the fraction of nutrients resorbed from senescent leaves at any single experimental site (Killingbeck 1988, 1992, Killingbeck et al. 1990, Lusk et al. 2003). Furthermore, tree species significantly differ in leaf senescence and fall dynamics (Dixon 1976, Koike 1990, 1995, Eckstein et al. 1999, Wilson et al. 2000), but the interaction of leaf nutrient resorption with leaf fall has been considered in only a few cases (Killingbeck et al. 1990, Chapin and Moilanen 1991, del Arco et al. 1991, Escudero et al. 1992b). Because species phenology differs in multi-species forest stands,

2 1002 NIINEMETS AND TAMM species-specific nutrient resorption and leaf fall kinetics may play an important role in modifying total nutrient return to the soil as well as in determining the overall variability in litter chemistry, thereby strongly affecting soil turnover and the performance of the forest community. There is a series of mechanisms by which species diversity affects community performance (Tilman et al. 1997, Grime 1998, Reich et al. 2001), but this potentially important control of species diversity on nutrient cycling has not been investigated extensively. The concentrations of mobile elements such as nitrogen (N), phosphorus (P) and potassium (K) are often strongly correlated during mid-season and in senescent leaves (Nordén 1994, Reich et al. 1995, Cornelissen et al. 1997, Wright et al. 2001, but see Niinemets and Kull 2003). However, the resorption efficiencies may vary significantly for these nutrients for reasons not yet fully understood (Schlesinger and Hasey 1981, Killingbeck et al. 1990, Collier and Thibodeau 1995, Aerts and Chapin 2000, Palma et al. 2000). Nutrient-specific rates of mobilization may interact significantly with the timing of leaf fall, thereby modifying the realized resorption efficiency of different nutrients. To our knowledge, such an interaction has not yet been examined. Apart from time-dependent controls on leaf nutrient resorption, potential nutrient resorption efficiencies differ widely among species (Chapin and Kedrowski 1983, Killingbeck 1996, Aerts and Chapin 2000). Nutrients associated with different leaf compartments are mobilized in varying degrees. Such variation may result from species differences in leaf structure and chemistry. In particular, there is a large fraction of leaf N that is bound to cell walls (Cassab 1998), which is only partly retranslocatable. Thus, the potential N resorption efficiency may depend on the distribution of N between bound and soluble pools, but surprisingly few studies have examined the dependence of potential N resorption efficiency on foliar chemistry (Chapin and Kedrowski 1983, Chapin et al. 1986, Chapin and Moilanen 1991, Pugnaire and Chapin 1993). The rate of nutrient resorption from senescing leaves may also vary with the avaliability of nutrients for resorption. This implies that, in addition to leaf fall patterns, leaf chemistry can further amend the time-dependent controls on nutrient losses. We hypothesized that the timing and speed of leaf fall affects the nutrient content of leaf litter, and that cell wall content affects nutrient resorption efficiency. Leaf fall dynamics and changes in nutrient contents were studied in three early- to mid-successional mixed broad-leaved deciduous forest stands. The results of this analysis highlight large time-dependent and leaf-chemistry-driven variations in nutrient resorption efficiency in important deciduous forest species. Materials and methods The sites The study was conducted at three Populus tremula L. dominated natural deciduous mixed broad-leaved forest stands in Estonia. The first stand was located at Järvselja (58 22 N, E, m a.s.l.), and the second and third stands were located at Kärkna (58 28 N, E, m a.s.l.), approximately 200 m apart (Kärkna-1 and Kärkna-2). At Järvselja, the dominant P. tremula trees were years old and m tall, whereas at Kärkna-1 and Kärkna-2, P. tremula trees were years old and m tall at the time of foliage sampling. The three stands varied widely in soil acidity, nutrient availability, total leaf area index (Table 1) and codominant species composition. The soil was an acid loamy gleyed pseudopodsol at Järvselja and less acidic loamy sand gleyed pseudopodsol in the Kärkna stands. At Järvselja, the upper canopy (17 27 m) was co-dominated by the early-successional shade-intolerant species Betula pendula Roth., and the sub-canopy and shrub layer were formed by the late-successional shade-tolerant species Tilia cordata Mill. and Corylus avellana L. At Kärkna-1, the upper canopy was co-dominated by Alnus glutinosa (L.) Gaert., A. incana (L.) Moench and B. pendula, and the lower canopy was dominated by Fraxinus excelsior L. At Kärkna-2, the upper canopy was co-dominated by A. incana and B. pendula, and the lower canopy and shrub layer by C. avellana, F. excelsior and Lonicera xylosteum L. Leaf and litter sampling The study was conducted in the 1995 and 1997 growing seasons at Järvselja and at the Kärkna, respectively. Nutrient contents in mid-season and time-dependent changes in litter chemistry were studied in all stands, whereas time-dependent variability in nutrient contents of the attached leaves was studied in the Järvselja stand only, and the cumulative leaf area and dry mass loss in the Kärkna stands. In Järvselja, attached foliage from different canopy positions was collected at 1-week intervals starting from bud burst Table 1. Characteristics of three Populus tremula dominated forest stands (mean ± SD of five replicates). Soil ph was measured in 1 M KCl solution, total nitrogen (N) and carbon (C) percentages by a C/N analyzer (CHN-O-Rapid, Foss Heraeus GmbH, Hanau, Germany), and acetate-soluble phosphorus (P) content (ph = 3.75, ammonium acetate 0.1 M, acetic acid 0.3 M, lactic acid 0.1 M) by the molybdenum blue method (AL-method, Swedish standard SS , Egner et al. 1960). Abbreviation: LAI = leaf area index. Site Dominant tree Total LAI LAI of P. tremula Soil A horizon age (year) (m 2 m 2 ) (m 2 m 2 ) ph KCl Total N (%) Total C (%) Soluble P (mg kg 1 ) Järvselja ± ± ± ± ± ± 0.21 Kärkna ± ± ± ± ± ± 0.19 Kärkna ± ± ± ± ± ± 0.17 TREE PHYSIOLOGY VOLUME 25, 2005

3 TEMPORAL AND SPECIES EFFECTS ON NUTRIENTS IN LEAF LITTER 1003 until leaf fall as described in Niinemets et al. (2004). Permanent scaffoldings (25 m high) located at the study site provided access to foliage. In the Kärkna stands, attached leaves were sampled during the period of peak leaf physiological activity in July from the mid-canopy, gathered either from felled P. tremula trees (for methods, cf. Mandre et al. 1998), or by climbing the taller trees or by using a ladder for shrubs and trees of shorter stature. Leaf litter was sampled by litter traps of 0.13 m 2 area at weekly intervals in all stands. In every stand, five litter traps were positioned randomly. Determination of leaf area and leaf dry mass per unit area For each sampling date, 5 10 leaves were randomly selected for every species from every litter trap. The perimeter of each leaf was traced with a computer digitizer (QD-1212, QTronix, Taiwan) and the area determined with a custom-written software package. The leaves were dried at 70 C for at least 48 h and weighed, and leaf dry mass per unit area was calculated (M A ). The value of M A for a specific stand, date and species was calculated as a mean for all litter traps. In the Kärkna stands, the remaining leaves were also dried and weighed, and species leaf area index lost during a specific time period t (normally, t was 7 days), L( t)(m 2 m 2 ), calculated as the total leaf litter dry mass divided by M A, and the area of the litter trap, A L. The values of L( t) were further integrated from the start of leaf fall to obtain the cumulative leaf area index lost by time t from the start of leaf fall, L c (t). Fitting of time-dependent changes in leaf fall To account for species differences in total leaf area index, L c (t) was standardized with respect to total species cumulative leaf area observed after all leaves had fallen, L c,t. We fit the relative cumulative leaf area index (ω L (t)=l c (t)/l c,t ) versus time relations according to Dixon (1976) and del Arco et al. (1991) as: ωl () 1 t = 22. τ1 1 + e ( τ 2 t) (1) where τ 1 (days) is the time between 0.1 and 0.5 fractions of leaf fall and τ 2 is the day of year of maximum leaf fall. This equation assumes that the time-dependence of leaf fall is symmetrical, i.e., that it proceeds at the same rate at equal distances from the mean at the beginning and closing stages of leaf fall (Escudero and del Arco 1987). Previous studies support the assumption of time-symmetric leaf fall (Dixon 1976, Escudero and del Arco 1987, del Arco et al. 1991). Analogously, cumulative changes in relative leaf mass (ω M (t), leaf litter dry mass until the date t standardized with respect to total leaf mass) were also fitted by Equation 1. Although leaf litter was sampled in weekly intervals, fitting the data by Equation 1 allowed us to compare species maximum litter fall times and characterize the species differences in the uniformity of leaf fall (Figure 1A). Chemical analyses Leaf N and carbon contents per unit dry mass were estimated with an elemental analyzer (CHN-O-Rapid, Foss Heraeus GmbH, Hanau, Germany). Calcium, magnesium, P and K contents per unit dry mass were determined by inductively coupled plasma emission spectroscopy (INTEGRA XMP, GBC Scientific Instruments, Melbourne, Australia) after the digestion of fine-ground leaves in 65% HNO 3. Neutral detergent fiber content (NDF) of P. tremula leaves was determined according to Van Soest (Van Soest 1963, Van Soest et al. 1991). The NDF provides an estimate of total cell wall content (Van Soest et al. 1991). Because most leaf N is present in organic form, and a major fraction of organic N is a constituent of proteins that also contain carbon, leaf carbon content of senescing leaves may change as a result of protein degradation. To account for changes in C content because of protein losses, we calculated the protein-free leaf carbon content (C S ) as: C S = C N N where C is leaf carbon, N is leaf N percentage, 6.25 converts leaf N content to protein content, and 53.5% is the carbon per- (2) Figure 1. Sample relationships of cumulative leaf fall (A) and time-dependent changes in fresh leaf litter nitrogen (N) concentration (B) in three species from Kärkna-1 site (see Table 1 for site details). Data in A were fitted by Equation 1, and corresponding regression coefficients and r 2 values are reported in Table 2. Data in B were fitted by linear regressions to determine the mean rate of nutrient resorption. In B, the mid-seasonal values of fully active leaves are also reported. TREE PHYSIOLOGY ONLINE at

4 1004 NIINEMETS AND TAMM Table 2. Nonlinear fitting of leaf fall (Equation 1) in two P. tremula dominated forests. Equation 1 provides a function that is symmetrical with respect to the time for half total leaf fall. The coefficient τ 1 (d) represents the time between 10 and 50% of leaf fall (equal to the time between 50 and 90% of leaf fall), and τ 2 is the day of year of maximum leaf fall. Thus, the leaf fall is distributed over a longer time period for larger τ 1 values, and proceeds later for greater τ 2 values. Figure 1A provides the sample fits for the three species from the Kärkna-1 site. Species Kärkna-1 Kärkna-2 τ 1 (d) τ 2 (d) r 2 τ 1 (d) τ 2 (d) r 2 Alnus glutinosa Alnus incana Betula pendula Corylus avellana Fraxinus excelsior Lonicera xylosteum Populus tremula These species were present only at one of the two sites. centage of proteins (Vertregt and Penning de Vries 1987, Fleck et al. 2003). Nutrient resorption rate in senesced leaves Percentages of mobile elements (N, P, K) in leaf litter (all stands) or in attached senescent leaves (Järvselja stand) versus sampling date relations were fitted by linear regressions (Figure 1B). The slope of these regressions, (% day 1 ), was taken as an estimate of the nutrient resorption rate. Across different sets of data, explained variance (r 2 ) of these relations varied from (mean ± SE = 0.52 ± 0.07 for N, 0.64 ± 0.08 for P and 0.69 ± 0.05 for K). Although the r 2 value was low in some cases, this was not because of strong fluctuations, but because of small time-dependent variability in foliar nutrient contents that was reflected in close to zero values (e.g., B. pendula in Figure 1B). In some cases, foliar nutrient versus time dependencies were curvilinear (e.g., P. tremula in Figure 1B). As a second estimate of, we calculated the total time-dependent change in leaf nutrients as a difference between the foliar nutrient contents at the start of leaf fall and at the end, divided by the period of leaf fall ( S ). However, both and S were strongly correlated. For instance, for N, r 2 = 0.89, P < for all species and stands pooled, suggesting that the use of a linear approximation did not bias our estimates of the overall time-dependent change in foliar nutrient contents. Implicit in this analysis is that the changes in nutrient contents are entirely due to resorption. It may be argued that leaching of leaf nutrients also significantly contributes to the time-dependent patterns in litter nutrients. However, in natural nutrient-limited environments, the fraction of nutrients lost due to leaching constitutes a minor part of total nutrient loss during senescence (Chapin and Kedrowski 1983, Aerts and Chapin 2000), contrary to the case in environments where nutrients are in surplus (Chapin and Moilanen 1991). This experimental evidence is further strengthened by laboratory findings that membrane integrity and subcellular compartmentation are maintained until the latest phases of leaf senescence (Feller and Fischer 1994, Matile 1997). There is further evidence that essentially no N and little P is lost by leaching from tree leaf litter, whereas significant leaching losses may occur for K (Rutigliano et al. 1998). Given these observations, we suggest that possible leaching losses of N and P do not qualitatively modify our conclusions with respect to nutrient resorption. Unweighted and litter-mass weighted mean nutrient contents, and final nutrient content Litter nutrient percentages determined for all sampling dates were averaged to yield unweighted mean nutrient content, X (%). Whenever foliar nutrient contents decrease with a different time constant than that for leaf fall, a biased estimate of mean nutrient content in leaf litter is provided. We calculated a litter mass-weighted mean nutrient content for each species and site as: X t = tn t t W = = 1 t = tn XtMt () () t = t1 Mt () where X(t) is nutrient content of leaf litter taken at the sampling date t, M(t) is total species leaf litter mass at that sampling occasion, and t 1 and t n are the first and last sampling dates, respectively. In addition to these two integral characteristics, we also calculated the final litter nutrient content, X F,as a mean of foliar nutrient contents for the two last sampling dates. Determination of nutrient resorption efficiency There are many definitions of nutrient resorption efficiency in the literature. Because nutrient contents steadily decrease in senescing leaves, the potential nutrient resorption efficiency is obtained for the latest-falling leaves, whereas stand-level nutrient resorption efficiency is obtained when the mean nutrient content of all fallen leaves is employed. We determined the nutrient resorption efficiency (fraction of specific leaf nutrients resorbed) for leaves fallen at time t, R T (t), as: (3) TREE PHYSIOLOGY VOLUME 25, 2005

5 TEMPORAL AND SPECIES EFFECTS ON NUTRIENTS IN LEAF LITTER 1005 X R () t = T M X() t X M where X M (%) is mean leaf nutrient content in mid-season during peak leaf physiological activity (second week of July), and X(t) is nutrient content in leaf litter or in attached leaves at time t. The maximum resorption efficiency was defined as: R T,Max X = X X M M F where X F (%) is final leaf nutrient content of leaf litter at the end of leaf fall. The value of X M for a specific nutrient was determined as a mean for leaves sampled from 20 different canopy positions for P. tremula and T. cordata leaves at the Järvselja site. Because there was only a minor within-canopy gradient in foliar N contents per unit dry mass (Mandre et al. 1998, Niinemets et al. 2004), a mean mid-seasonal value of N(N M ) was determined for mid-canopy leaves of all species in Kärkna stands. For Kärkna stands, litter-mass weighted mean N resorption efficiency was determined as: R T N = M N N M W where N W (%) is litter-mass weighted mean N content (Equation 3). Previous studies have demonstrated different values of nutrient resorption efficiency, depending on whether it was calculated based on N content per unit leaf area, leaf dry mass or per single leaf (Killingbeck 1996, Hevia et al. 1999). Such differences may occur if nutrients and bulk leaf carbon are translocated with a different time constant, biasing the dry massbased resorption efficiency estimates. These discrepancies may also result from a temporal variation in leaf dry mass per unit area (M A ) of shed leaves (Roberts et al. 1999). Because of acclimation to local light environment that mainly involves changes in M A, there is an approximately twofold species-specific within-canopy gradient in leaf N content per unit area (N area 1 ), and in N contents per unit single leaf (N leaf 1 ) (Niinemets et al. 2004). This implies that whenever there is within-canopy variation in leaf fall kinetics, determination of an appropriate mid-seasonal N content for each specific sampling date will complicate the calculation of resorption efficiency based on the values of litter N area 1 or N leaf 1. In the Järvselja stand, the variation in N area 1 and N leaf 1 had been characterized with respect to canopy height and long-term light conditions at different times during the growing season (Niinemets et al. 2004). In this stand, we compared different methods of R T (t) determination using attached leaves at various stages of senescence at different canopy heights (h) in P. tremula and T. cordata. The hypothetical mid-season N leaf 1 values at specific canopy heights and dates of collection of senesced leaves were calculated from the regressions with h (r 2 > 0.80). According to Figure 2, different methods yield (4) (5) (6) comparable estimates of R T (t) that are only slightly biased, justifying the use of mass-based resorption efficiency estimates. A similar result was obtained for K and P resorption efficiencies (data not shown). Statistical analyses Equation 1 was fitted to the data by nonlinear regression analysis (Figure 1A). Linear regression analyses were carried out to test for significant relationships between the other measured variables. Leaf nutrient content and nutrient resorption efficiency versus time relationships between attached and fallen leaves (leaf status) were compared by covariance analyses (ANCOVA). A separate slope model that includes an interaction term, Date X (leaf status), was used first. When the interaction term was insignificant, the model was refitted according to a common slope model that lacks the interaction term to test for the intercept differences (Sokal and Rohlf 1995). All relationships were considered significant at P < All statistical analyses were conducted with Systat 8.0 software (Systat Software, Richmond, CA). Overall, there were eight sampling dates during leaf fall, and for most analyses, 14 independent species-site pairs. For mass-weighted litter nutrient contents (Equation 3), data were unavailable for the Järvselja stand. In this stand, the litter mass of different species was determined for a pooled sample of several sampling dates, whereas Figure 2. Comparison of the fractional leaf nitrogen (N) resorption (resorption efficiency, Equation 4) determined based on whole leaf N contents (R T,W ; N leaf 1 ), and leaf N percentages (R T,% ). The measurements were conducted at the Järvselja forest from September to October. Attached leaves at different heights in the canopy were sampled. A species-specific mean value of leaf N percentage at mid-season was used to determine R T,% (Niinemets et al. 2004). For R T,W, mid-seasonal values of N leaf 1 at different canopy heights were determined from the linear regressions with canopy height (Niinemets et al. 2004). Data were fitted by linear regressions (r 2 = 0.92, P < for P. tremula and r 2 = 0.83, P < for T. cordata). Intercepts are not significantly different from zero for either regression. For comparison, a 1:1 line is shown. TREE PHYSIOLOGY ONLINE at

6 1006 NIINEMETS AND TAMM nutrient contents in a subsample were determinedly separately for each sampling date. Results Timing of leaf fall in different species and stands Leaf fall started in all species in early September, but the intensity of leaf fall at any date differed between species. In particular, maximum leaf fall (parameter τ 2 in Equation 1) occurred first in N-fixing Alnus species, and last in B. pendula (Figure 1A, Table 2). However, in both A. incana and B. pendula, leaf fall was distributed more uniformly in time (parameter τ 1 in Equation 1, the time for 10 to 50% loss of leaf area and mass), whereas the bulk of leaf area was lost during a short time period of about 1 week in the other species (Figure 1A, Table 2). Parameters τ 1 and τ 2 were not correlated for the whole set of data (r 2 = 0.12, P > 0.2). The mean ± SE explained variance for linear regression of M A versus day of year was 0.25 ± 0.06, indicating that leaf dry mass per unit area (M A ) of freshly fallen leaves was essentially constant over time. Given that there is strong within-canopy variation in M A, this non-correlation demonstrates that the contribution of leaves from different canopy positions to the overall leaf litter did not vary between dates. The only exception was a significant (P < 0.01) increase of mean M A from 53.7 ± 3.3 g m 2 at the start of leaf fall to 65.2 ± 2.9 g m 2 at the end of leaf fall in B. pendula at the Kärkna-2 site. Variation in foliar nutrient contents and resorption efficiency in attached and fallen leaves In the Järvselja stand, we compared time-dependent changes in the content of nutrients in attached leaves after peak physiological activity had been attained (July August) and in freshly fallen leaves. For mobile elements N, P and K, the contents in attached leaves decreased in both P. tremula and T. cordata after mid-august (Figure 3). Leaf fall started in the first week of September, and the nutrient contents of attached senescent leaves were generally similar to those in freshly fallen leaves (Figure 3). The contents of Ca and Mg that are considered non-mobile were generally independent of date of sampling for both attached and freshly fallen leaves. In attached P. tremula leaves only, Ca contents increased with time (r 2 = 0.80, P < 0.005) and Mg contents decreased in freshly fallen leaves of T. cordata (r 2 = 0.74, P < 0.01). According to ANCOVA analyses, the slopes and intercepts of the leaf nutrient content versus day of year relationships did not differ between the attached senescent and freshly fallen Figure 3. Time-dependent variations in nitrogen (N), phosphorus (P) and potassium (K) contents and resorption efficiency (Equation 4) of attached ( ) and freshly fallen leaves ( )ofp. tremula and T. cordata in Järvselja stand (Table 1). Data were fitted by linear regressions. Arrows denote the nutrient contents in mid-season. TREE PHYSIOLOGY VOLUME 25, 2005

7 TEMPORAL AND SPECIES EFFECTS ON NUTRIENTS IN LEAF LITTER 1007 leaves for N (P > 0.3; Figure 3A) and P (P > 0.5; Figure 3E) in P. tremula and for P (P > 0.2; Figure 3F) and K (P > 0.3; Figure 3J) in T. cordata. However, the slope of the K content versus day of year relationship was larger in freshly fallen leaves of P. tremula (P < 0.03; Figure 3I), and the slope of the N versus day of year plot was larger in attached leaves of T. cordata (P < 0.005; Figure 3B). Time-dependent decreases in nutrient contents of mobile elements were paralleled by increases in instant nutrient resorption efficiency (R T (t); Equation 4, Figure 3). Nutrient resorption efficiencies of attached senescent leaves and freshly fallen leaves were essentially the same on a common date (Figure 3). Mean nutrient content in relation to the timing of leaf fall and the rate of change of nutrient contents Across the entire set of data, mean N (Figure 4A) and P (Figure 4B) contents of leaf litter (Equation 3) were negatively related to the day of year of maximum leaf fall (parameter τ 2 in Equation 1, Table 2), demonstrating that an earlier leaf fall was associated with larger ecosystem nutrient losses. However, this relationship was not significant for K (Figure 4C) or for other cations (data not shown). Nutrient contents were not related to the coefficient τ 1 of Equation 1, which measures the uniformity of leaf fall in time. Mean nutrient content of leaf litter was negatively correlated with rate of change in nutrient content for P (Figure 4E) and K (Figure 4F), and the same trend was observed for N (Figure 4D). Thus, more rapid nutrient retranslocation only partly compensated for higher initial nutrient contents, and the litter nutrient contents were still larger in species with higher initial contents. Although the data of litter N content of N-fixing Alnus species stand out, the litter N content was also high in F. excelsior (Table 3). In fact, the trends were qualitatively the same when the regression models were refitted using a data set-specific Figure 4. (A, D) Litter-mass weighted mean (Equation 3) leaf litter nitrogen (N), (B, E) phosphorus (P) and (C, F) potassium (K) contents in relation to the day of maximum leaf fall (A C; parameter τ 2 in Equation 1, Table 2), and the rate of change of leaf nutrient content since the start of leaf fall (D F; slope of leaf nutrient content versus day of year relationship, Figures 1A and 3) in the Kärkna stands. Inset in A shows the relationship between litter mass-weighted mean N content and mid-seasonal N content. Each data point corresponds to a different species/stand combination. Data were fitted by linear regressions. Nonsignificant regressions in C and D are indicated by a dashed line. TREE PHYSIOLOGY ONLINE at

8 1008 NIINEMETS AND TAMM Table 3. Unweighted (X) and litter-mass weighted (XW) (Equation 3), mean and final (XF) nutrient contents of leaf litter in two P. tremula dominated forest stands. Abbreviations: N = nitrogen concentration; P = phosphorus concentration; Ca = calcium concentration; K = potassium concentration; and Mg = magnesium concentration. Site Species N (%) P (%) Ca (%) K (%) Mg (%) X ±SE XW XF X ±SE XW XF X ±SE XW XF X ±SE XW XF X ±SE XW XF Kärkna-1 A. glutinosa 2.58 ± ± ± ± ± A. incana 2.69 ± ± ± ± ± B. pendula ± ± ± ± ± F. excelsior 2.26 ± ± ± ± ± P. tremula 1.31 ± ± ± ± ± Kärkna-2 A. incana 2.85 ± ± ± ± ± B. pendula ± ± ± ± ± C. avellana 1.27 ± ± ± ± ± F. excelsior 1.92 ± ± ± ± ± L. xylosteum 1.07 ± ± ± ± ± P. tremula ± ± ± ± ± Figure 5. Correlation between the rate of change in nitrogen (N) and phosphorus (P) (slope of leaf nutrient content versus day of year relationship) of leaf litter in the three studied stands. Each value corresponds to a different species/stand combination. Data were fitted by linear regressions. Sample relations of time-dependent depletion of nutrients in leaf litter are demonstrated in Figures 1B and 3. mean litter N value for the Alnus species. For instance, for the correlation depicted in Figure 4A, r 2 = 0.43, P < Rate of nutrient loss versus the timing of leaf fall The rates of change in N and P were correlated, but the data were scattered (Figure 5), demonstrating significant speciesspecific nutrient resorption rates for different nutrients. The rates of change for N and K (r 2 = 0.05, P > 0.4) and for P and K (r 2 = 0.18, P > 0.1) were not correlated. We observed a relatively faster decrease in leaf N (r 2 = 0.49, P < 0.02) and P (r 2 = 0.85, P < 0.001) contents in species with earlier leaf fall (rate of change in nutrient content in Figure 4 versus the coefficient τ 2 in Equation 1). This relationship was stronger for P than for N. There was no significant correlation between the rate of K change and τ 2 (r 2 = 0.05, P > 0.5). Initial, mean and final nutrient contents Because of nutrient- and species-specific resorption kinetics (Figures 4 and 5) and species-specific leaf fall timing and kinetics (Figure 1A, Table 2), unweighted mean nutrient contents, litter-mass weighted mean nutrient contents (Equation 3) and final nutrient contents differed for mobile elements N, P and K, but were essentially the same for the immobile elements Ca and Mg (Table 3). Because of differing loss kinetics, correlations among elements differed depending on how the litter nutrients are expressed. Litter-mass weighted mean N and P (r 2 = 0.43, P < 0.03), and K and P (r 2 = 0.43, P < 0.03) contents were significantly correlated, as observed for nutrient contents in mid-season (data not shown). Final N and P contents at the end of leaf fall (r 2 = 0.04, P > 0.5; Figure 6A) were not correlated, whereas contents of the non-mobile cations Mg and Ca were correlated (r 2 > 0.49, P < 0.01). There were also strong corre- TREE PHYSIOLOGY VOLUME 25, 2005

9 TEMPORAL AND SPECIES EFFECTS ON NUTRIENTS IN LEAF LITTER 1009 Figure 6. Correlations between N and P (A) and Ca, Mg and K (B) contents of leaves abscised at the end of leaf fall in all three stands. Each value corresponds to a different species/ stand combination. Data were fitted by linear regressions. Figure 7. Maximum nitrogen (N) resorption efficiency (Equation 5) in relation to mid-season N content of leaves (A), and correlation between mean N resorption efficiency (Equation 6) and litter-mass weighted mean N content of leaf litter (Equation 3) (B) in the three studied stands. Inset in B demonstrates the correlation between mean and maximum N resorption efficiencies and the 1:1 line. Figure 8. Relationships between maximum nitrogen (N) resorption efficiency (Equation 5) and protein-free carbon content (C S ; Equation 2) (A) and calcium content (B) in fresh leaf litter in the studied stands. Protein-free carbon content provides an estimate of leaf carbon that is independent of leaf protein losses. Data presentation and fitting as in Figure 4. lations between mobile cations (e.g., K) and the non-mobile cations Mg and Ca (Figure 6B). Nitrogen content versus resorption efficiency The rate of N resorption was greater for species with greater mid-season leaf N contents (r 2 = 0.32, P < 0.05). Maximum N resorption efficiency (Equation 5) was not related to mid-season N content (Figure 7A), demonstrating that the species with the greatest leaf N contents are not necessarily less efficient in nutrient resorption than species with lower N contents. However, mid-season leaf N content and litter-mass weighted mean N content of leaf litter were strongly correlated (r 2 = 0.84, P < 0.001). Mean N resorption efficiency (Equation 6) was negatively associated with the mean N content of leaf litter (Figure 7B), suggesting that the constraints on leaf N resorption due to timing of leaf fall and the rate of N resorption resulted in lower realized N resorption efficiency in species with higher leaf N contents. Maximum and mean N resorption efficiencies were strongly correlated (inset in Figure 7B), but because of time-dependent TREE PHYSIOLOGY ONLINE at

10 1010 NIINEMETS AND TAMM Figure 9. Correlations between neutral detergent fiber (NDF) and leaf calcium contents (A), and NDF and C S (B) in P. tremula leaves, and the correlation between calcium content and C S for the entire data set (all samples from all sampling dates pooled). Data were fitted by linear regressions. decreases in N contents, litter-mass weighted mean resorption efficiency was always lower by, on average, 15 ± 23% than the maximum N resorption efficiency. Foliage constitution and N resorption efficiency Maximum N resorption efficiency was positively correlated with protein-free carbon content (Figure 8A) and negatively correlated with leaf calcium content of fallen leaves (Figure 8B). Because proteins that include the bulk of leaf N also contain carbon (53.5%), leaf carbon contents may vary in senescing leaves as a result of protein degradation. Therefore, we took protein-free carbon content (C S ; Equation 2) as the explaining variable in this correlation. Neither Ca content nor C S (r 2 = 0.00, P > 0.9 for both) was related to the rate of N resorption. To gain insight into the mechanistic relationships between N resorption efficiency and Ca and C S, we analyzed the relationships of foliage Ca content and C S with the total fraction of cell wall material in leaves. Neutral detergent fiber content was taken as a proxy for the content of cell walls. For P. tremula leaves, Ca and NDF were positively correlated (Figure 9A), whereas NDF and C S were negatively correlated (Figure 9B). Furthermore, there was a strong negative correlation between Ca and C S for all data sampled (Figure 9C). Discussion Timing of leaf litter fall and litter quality Previous studies have shown that there is strong year-to-year variation in nutrient contents in, and the fraction of resorbed nutrients from, senescent leaves (Killingbeck 1988, 1992, Killingbeck et al. 1990, Lusk et al. 2003). Our study demonstrates that, in most species, there is a continuous decrease in mobile nutrients in attached leaves after peak leaf physiological activity has been attained (Figures 1B and 3). This decrease is also paralleled by a decrease in nutrient contents in freshly fallen leaf litter and time-dependent increases in nutrient resorption efficiency (Figure 3). Leaf dry mass per unit area of non-senescent leaves at any specific canopy position was stable after leaves had reached physiological maturity (Niinemets et al. 2004), and the moderate decrease in M A of attached leaves paralleled that in freshly fallen leaves (Niinemets et al. 2004). Thus, the time-dependent changes in nutrient contents in attached leaves were not caused by nutrient dilution that may occur if M A is increasing in time, but truly reflected nutrient resorption. Collectively, these data indicate that the timedependent decrease in nutrient content in freshly fallen litter reflects a greater efficiency of nutrient resorption in leaves falling later. The dismantling of cell compartments is an energy-dependent process associated with de novo synthesis of catabolic enzymes (Smart 1994, Feller and Fischer 1994), which explains the time-dependence of nutrient remobilization. Leaf nutrient loss during senescence is accompanied by near-proportional decreases in leaf photosynthetic and respiratory activities (Collier and Thibodeau 1995, Rosenthal and Camm 1997, Oleksyn et al. 2000), further demonstrating that leaf senescence is a highly regulated process. The time-dependent pattern of leaf nutrient loss is further affected by differences in the time when leaves enter senescence. The pattern of autumn leaf coloring varies among forest tree species (Koike 1990), suggesting that the start of senescence and spread of senescence in the crown is partly under genetic control. We observed the time-dependent decrease in nutrient content of freshly fallen leaf litter in most, but not in all, species (Figure 1B). Such a time dependence can arise if the leaves that have been shed at different times also have entered senescence at different times, but have undergone resorption for the same time period, resulting in equal nutrient withdrawal from the leaves. Although this is a reasonable hypothesis that requires further experimental study, at the northern latitudes where our sites were located, all mature forest trees tend to be flush type, and we have found no strong canopy gradients in leaf age and senescence in the canopies of these temperate deciduous trees (Niinemets et al. 2004). There is large variation in the decomposition rate among temperate woody species (Cornelissen 1996), but it is unclear TREE PHYSIOLOGY VOLUME 25, 2005

11 TEMPORAL AND SPECIES EFFECTS ON NUTRIENTS IN LEAF LITTER 1011 to what extent this variation reflects phenological controls on litter quality. Our results demonstrate that leaves that fall earlier contain more nutrients than leaves that fall later. Several studies have provided evidence that the control of tree litter decomposition rates is exerted by the initial litter N to lignin ratio (Melillo et al. 1982, Aerts 1997, Scott and Binkley 1997, but see Schaefer et al. 1985), or by litter P to lignin ratio (Schlesinger and Hasey 1981, Aerts and de Caluwe 1997), with minor effects of initial litter nutrient content alone (Scott and Binkley 1997). As lignin is not reabsorbed from the senescing leaves, larger nutrient contents of freshly fallen leaves are also compatible with their higher nutrient to lignin ratios, and accordingly greater decomposition rates. However, most conclusions on chemical control of litter decomposition rely on short-term experiments ( 1 year). It is important to note that long-term litter decomposition may be inversely related to the initial decomposition rate (Berg et al. 1995, 1996). In fact, the overall nutrient return to the ecosystem may be better predicted from total leaf litter produced by the vegetation and litter nutrient concentration than from the initial decomposition rate (Prescott 2002). Interaction of the timing and the rate of leaf fall with nutrient resorption rate Previous discussion highlights the importance of the timing of leaf nutrient losses for the overall return of nutrients to the ecosystem. The litter mass-weighted mean nutrient content of leaf litter is a function of time-dependent changes in leaf litter nutrient contents and the kinetics of leaf fall. We observed significant variation among species in the timing and rate of leaf fall (Figure 1A, Table 2). Earlier leaf fall was associated with higher N (Figure 4A) and P (Figure 4B) contents of leaf litter, underscoring the importance of accounting for the timing of leaf fall. For a broad range of plant communities, the fraction of N resorbed has been shown to be negatively related to the gradualness of leaf abscission (coefficient τ 1 in Equation 1; del Arco et al. 1991). This relationship was not evident in our study, but the range of τ 1 values of ~5 20 days in our stands was smaller than the range of τ 1 values of ~15 80 days in the study of del Arco et al. (1991). Partial deviation of our work from the general pattern may also result from the different time steps used (1 month in del Arco et al versus 1 week in our study). In addition, for many Mediterranean species studied by del Arco et al. (1991), leaf abscission depends strongly on drought stress, but drought was not a factor in our study at the end of the temperate growing season. Our study demonstrates large variation among species in nutrient decrease kinetics (Figures 1B, 3 and 4). Overall, N resorption was faster for leaves with greater N content, agreeing with previous observations that physiologically more active leaves resorb N at a greater rate (Osaki and Shinano 2001). These relationships with N and P (Figures 4 and 5) imply that, in plants with higher foliar N and P contents, earlier leaf fall was partly compensated for by faster resorption rate. However, litter mass-weighted mean nutrient contents tended to scale with the rate of change in nutrient contents (Figures 4D F), demonstrating that the enhanced rate of resorption in species with greater N contents did not fully offset the initial differences in nutrient contents between species with early and late leaf fall (Figure 4). The lack of compensation of higher nutrient contents by greater resorption rate was also because leaf fall tended to occur earlier in species with greater N and P resorption rates. Thus, as a result of the correlations between the timing of leaf litter fall and initial N and P contents, species with greater N and P contents lost a larger fraction of their nutrients as litter. There is evidence of significant tree-to-tree differences in N resorption efficiency within the same population (Nordell and Karlsson 1995). We found that leaf abscission kinetics and nutrient resorption efficiency were similar for the same species growing in different stands, suggesting that the within-population differences in litter quality are superimposed on species effects. On the whole, the results of this study demonstrate an important species control on litter nutrient contents resulting from species-specific resorption patterns and leaf fall dynamics. Thus, more diverse forests produce litter with more strongly varying chemical characteristics. Given that the nutrient versus carbon requirements differ for soil bacterial and fungal populations (Scheu and Parkinson 1995) as well as for soil microfauna (Kaneko and Salamanca 1999), the forests producing litter with widely varying content of chemicals, may support more diverse soil microfauna and microbial populations. Although there have been few studies, the currently available experimental evidence supports the hypothesis that mixing the litter of different species increases the overall litter decomposition rate (Taylor et al. 1989, Blair et al. 1990, Fyles and Fyles 1993, Kaneko and Salamanca 1999). Nitrogen content and N resorption efficiency Several studies have reported a positive correlation between the fraction of N resorbed and N content of non-senescent leaves (Chapin and Kedrowski 1983, del Arco et al. 1991, Escudero et al. 1992a). Other studies have found no clear relationship between nutrient resorption efficiency and mid-season nutrient contents (e.g., Lajtha and Klein 1988, Chapin and Moilanen 1991, Enoki and Kawaguchi 1999). In our study, there was no significant correlation between mid-season N and maximum N resorption efficiency (Figure 7A), whereas the litter mass-weighted mean N resorption efficiency was negatively related to the initial N content and mean N content of leaf litter. Thus, our analysis suggests that species differences in leaf fall dynamics and kinetics of leaf nutrient resorption provide the primary explanation for the scaling of mean leaf N contents and resorption efficiency with mid-season N contents. We further hypothesize that modification of timing of leaf fall in nutrient-poor sites (Wendler and Millard 1996, Niinemets and Lukjanova 2003) may provide a general explanation of soil nutrient availability effects on litter nutrient contents as observed in some studies. It may be disputed that the insignificant relationship between the mid-season leaf N content and N resorption efficiency is biased by N-fixing Alnus species that have previously been shown to have a low N resorption efficiency (Stachurski TREE PHYSIOLOGY ONLINE at

12 1012 NIINEMETS AND TAMM and Zimka 1975, Chapin and Kedrowski 1983, Clein and Schimel 1995). However, in our study, non-n-fixing species F. excelsior and C. avellana were characterized by low N resorption efficiency, suggesting that the results are robust. Low N resorption potentials in F. excelsior and high resorption potentials in B. pendula and P. tremula agree with previous estimates for these species (Escudero et al. 1992a). Leaf carbon fractions and N resorption efficiency Previous studies have emphasized the importance of leaf chemical composition in modifying nutrient resorption efficiency (Chapin and Kedrowski 1983, Chapin et al. 1986, Chapin and Moilanen 1991, Pugnaire and Chapin 1993). It has been hypothesized that high concentrations of phenolic compounds in leaves reduce leaf nutrient resorption efficiency by precipitating proteins prior to hydrolysis (Schlesinger and Hasey 1981, Nicolai 1988). However, phenolic compounds such as condensed tannins, on average, have a higher carbon content (> 60%) than leaves. Thus, the variation in nutrient resorption efficiency with leaf carbon content (Figure 8A) cannot be explained by variation in leaf tannins. As discussed above, subcellular compartmentation and membrane integrity are maintained until the final stages of senescence, suggesting that the release of tannins from vacuoles during senescence is unlikely. Litter decomposition studies have demonstrated the existence of several litter N pools differing in accessibility to decomposers (Melillo et al. 1982, Müller et al. 1988, Santa Regina and Tarazona 1995). We suggest that an analogous situation with N occlusion exists in the case of N remobilization from attached senescing leaves. This is because plant cell walls contain a significant fraction of total leaf proteins. Some of these proteins such as arabinogalactan proteins are readily soluble, but a major fraction of cell wall proteins is covalently linked to cell wall polysaccharides, and to some lignin chains. Such cross-links with cell wall carbon constituents are formed enzymatically, but may also result from protein peroxidation by reactive oxygen species (Cassab 1998). Furthermore, during leaf development, there is a continuous accumulation of cell wall polysaccharides as the secondary cell wall develops, such that a significant fraction of cell wall proteins are trapped in the cell wall matrix. Thus, the potential N resorption efficiency may vary with the fraction of cell walls in the leaves. Although the existence of a certain cell wall-bound N pool is well documented in biochemical studies and may have a large impact on potential N resorption efficiency, quantitative data on the relevance of this pool are scarce (Pugnaire and Chapin 1993). Pugnaire and Chapin (1993) observed a positive correlation between the fraction of soluble leaf protein content and the nutrient resorption efficiency, supporting the hypothesis that potential resorption efficiency is controlled by the recalcitrant N pool. However, the insoluble N fraction in their study included thylakoid proteins that are also dismantled during senescence (Feller and Fischer 1994). We found that the N resorption efficiency scaled positively with leaf protein-free carbon content (Figure 8A) and negatively with leaf calcium content (Figure 8B). Because the bulk of leaf calcium is associated with cell walls (Figure 9A; Demarty et al. 1984), this correlation supports the hypothesis of the control of N resorption efficiency by total cell wall content. Furthermore, given that the bulk of cell walls consists of polysaccharides such as cellulose and pectin with lower carbon content (~40%) relative to the mean of entire leaves (~45%; Figures 9B and 9C), the fraction of cell wall material increases with decreasing leaf carbon content. This suggestion is supported by the negative correlation between leaf Ca and C S and total leaf carbon (Figure 9A; Niinemets 1995), and NDF and C S (Figure 9B). Consequently, the positive correlation between N resorption efficiency and protein-free carbon content likely reflects the decrease of the fraction of cell walls with increasing carbon content. These data collectively demonstrate a major species control on potential N resorption efficiency. This species control is at least partly driven by species-specific investments in cell walls and cell wall composition. It is significant that even for a single species there is significant year-to-year variation in leaf chemical composition (Taylor and Parkinson 1988) that may partly explain year-to-year variations in N resorption efficiency (Killingbeck 1992). Conclusions There is an extensive body of empirical information of enormous variation in leaf litter nutrient contents, but this large variability is currently only partly understood. Our study indicates that the timing of leaf fall and nutrient resorption kinetics are important drivers of litter nutrient contents to be added to the key determinants of nutrient cycling in temperate forests. In particular, species inherent potentials for nutrient resorption do not depend on the initial nutrient contents, but larger nutrient losses in species with greater nutrient contents are associated with earlier leaf fall. As a result of a range in phenologies and initial nutrient contents, multispecies forest communities appear to produce litter with more widely varying chemical composition than monospecific stands. Apart from the speed of nutrient resorption and loss of foliage, there are important species differences in potential N resorption efficiency that are at least partly associated with species differences in a recalcitrant pool of leaf N that is trapped in cell walls. Such species differences in the total amount of cell wall substances and the fraction of non-removable N further amplify the species control on leaf litter nutrient content. Acknowledgments We thank Lea Hallik for expert assistance. This work was financially supported by the Estonian Science Foundation (Grant 5702) and by the Estonian Ministry of Education and Science (Grant No As03). References Aerts, R Climate, leaf litter chemistry and leaf litter decomposition in terrestrial ecosystems: a triangular relationship. Oikos 79: TREE PHYSIOLOGY VOLUME 25, 2005