Initial litter properties and decay rate: a microcosm experiment on ...

Download PDF
1 downloads 25 Views 634KB Size Report
Comment
GILLON, D., JOFFRE, R., and IBRAHIMA, A. 1994. Initial litter properties and decay rate: a microcosm experiment on Mediter- ranean species. Can. J. Bot.
Initial litter properties and decay rate: a microcosm experiment on Mediterranean species DOMINIQUE GILLON,'RICHARD JOFFRE,AND ADAMOU IBRAHIMA Centre d'Ecologie Fonctionnelle et ~volutive,Centre national de la recherche scientifique, BP 5051, 34033 Montpellier Cidex, France Received October 12, 1993 GILLON,D., JOFFRE, R., and IBRAHIMA, A. 1994. Initial litter properties and decay rate: a microcosm experiment on Mediterranean species. Can. J. Bot. 72: 946-954. Twelve leaf litters belonging to 10 Mediterranean species of coniferous and broad-leaved trees and shrubs and grass species were incubated in microcosms in the laboratory at 22OC and constant humidity for 14 mon'ths. Samples were collected at 0.5, 1, 2, 4, 6, 10, and 14 months, the remaining dry weight being measured at each sampling time. At the end of 14 months, the litters had lost between 52 and 74% of their original mass. The comparison of regressions fitted to the various functions showed that for the species studied, the litter mass loss in relation to incubation time best fitted a double-exponential decay function. The mass loss therefore resulted from the simultaneous decomposition of two main compartments, a labile compartment that decreased rapidly (half-life of 20-60 days under the experimental conditions) and a resistant compartment that depending on the species, either did not decrease significantly or decreased 10 to 20 times slower than the labile compartment (half-life of 320-630 days). The litters studied could be categorized according to the relative importance of these two compartments. This was related to the initial content of water-soluble substances and of carbon in the litters. It was also strongly correlated with the spectral information of the initial litters obtained by near-infrared reflectance spectroscopy. In contrast, the rate at which the labile and resistant compartments decreased was related to the permeability of the leaves for the former and to their thickness and mass per surface area for the latter. Near-infrared reflectance spectroscopy provides new perspectives for characterizing the capacity of litters to decompose. Key words: litter, decomposition, near infrared reflectance spectroscopy. GILLON,D., JOFFRE,R., et IBRAHIMA, A. 1994. Initial litter properties and decay rate: a microcosm experiment on Mediterranean species. Can. J. Bot. 72 : 946-954. Douze litikres de feuilles appartenant i 10 espkces mCditerranCennes, conifere, arbres et arbustes feuillus, herbacCes, ont CtC incubCes en microcosmes au laboratoire a 22°C et B humidit6 constante pendant 14 mois. Les Cchantillons ont CtC prClevCs ? 0,5, i 1, 2, 4, 6, 10 et 14 mois, et i chaque prClkvement le poids sec restant a CtC mesurC. Au bout de 14 mois, les litikres avaient perdu entre 52 et 74% de leur masse initiale. La comparaison des ajustements B differentes fonctions a montrt que, pour l'ensemble des espkces Ctudites, la perte de masse des litieres en fonction du temps d'incubation s'ajustait le mieux a la fonction exponentielle double. La perte en masse a donc CtC la rCsultante de la dCcomposition simultanCe de deux compartiments principaux, l'un labile qui a diminuC rapidement (demie-vie de 20 i 60 jours dans les conditions de l'expkrimentation), l'autre resistant qui, selon les espkces, n'a pas diminuC de f a ~ o nsignificative ou 10 i 20 fois plus lentement que le compartiment labile (demie-vie de 320 i 630 jours). Les litikres CtudiCes ont pu &re classCes selon l'importance relative de ces deux compartiments. Celle-ci est liCe i la teneur initiale des litikres en substances hydrosolubles et en carbone. Elle est Cgalement fortement corrClCe h l'information spectrale des litibres initiales, obtenue par spectromCtrie de riflexion dans le proche infrarouge. Par contre, la vitesse 2 laquelle ont diminuC les compartiments labile et rtsistant est liCe ii la permCabilitC des feuilles pour le premier, et il leur Cpaisseur et leur masse surfacique pour le second. La spectromCtrie proche infrarouge offre de nouvelles perspectives pour la caractkrisation de l'aptitude des litikres ? dCcomposer. i Mots clis : litikre, dCcomposition, spectromCtrie proche infrarouge. 0

,

I

Introduction Litter decomposition is regulated by numerous factors, of which the most important are probably (i) the environment under which decay takes place and (ii) the physicochemical properties of the substrate. Among these initial characteristics, the nitrogen content of the litter and the C:N ratio have often been considered to be the key factors for decomposition (Berg and Staaf 1980; Berg and Ekbohm 1983; McClaugherty and Berg 1987; Taylor et al. 1989). Other studies have shown that the phosphorus content and the C:P ratio may be the determinant factors (Schlesinger and Hasey 1981; Staaf and Berg 1982; Berg and Ekbohm 1991). Research has also demonstrated the major role of water-soluble substances contained in the litter at the start (Berg and Tarnrn 1991; Berg and Ekbohm 1991). Finally, many authors have demonstrated the influence of the initial lignin content in litter decomposition processes (Fogel and Cromack 1977; Meentemeyer 1978; Berendse et al. 1987; Aber et al. 1990; Berg and Tamm 1991), of the contents 'Author to whom all correspondence should be addressed. Printed in Canada 1 lmprim6 au Canada

of lignin plus cellulose (Aber et al. 1990), of the lignin to N ratio (Aber and Melillo 1980; Melillo et al. 1982; Gallardo and Merino 1993), of the ligno-cellulose index (lignin to (lignin + cellulose); LCI) (Melillo et al. 1989), and of the holocellulose to lignocellulose quotient (HLQ) (McClaugherty and Berg 1987). Nevertheless, most of these studies were carried out in northern European or American forests and there are few that concern litters from Mediterranean regions. These include the studies of Lossaint (1973), Rapp and Leonardi (1988), and Van Wesemael (1993) on the litters of Quercus ilex L. in southern Europe, of Hart et al. (1992) on ponderosa pine litter in California, of O'Connel (1987) on Eucalyptus litter in Australia, of Schlesinger and Hasey (1981) and Schlesinger (1985) on litter from Californian chaparral shrubs, of Mitchell et al. (1986) on shrub litter of the South African fynbos, and of Maggs and Pearson (1977) on litters from the Australian scrub. Only a single recent work by Gallardo and Merino (1993) has attempted to determine the influence of the initial properties on the decomposition of the litter of nine different

947

ET AL.

species in Spain. However, under the Mediterranean climate, plants do show adaptations to the strong hydrological and edaphic constraints and their leaf litters possess characteristics (cuticles, wax, thickness, and toughness) that influence their rate of decomposition. For example, Gallardo and Merino (1993) showedthat toughness was a key factor determining the rate of decomposition of litter in the south of Spain. In this study, we attempted to determine the specific roles played in the decomposition of litter by its initial chemical composition and the properties related to anatomy (leaf thickness, degree of impermeability) that could reflect the varying capacities to absorb water and release water-soluble substances. We also used near-infrared reflectance spectroscopy (NIRS) to attempt to relate spectral information on the initial litters with the rate of decomposition. To do this, we chose 12 different leaf litters representative of the dominant species in the calcareous Mediterranean regions of France, presenting a wide range of forms and chemical compositions. To exclude the influence of meteorological factors, the litters were incubated in microcosms in the laboratory, under controlled conditions of temperature and humidity; to obtain litters reaching high percentage mass losses, incubation took place at a high temperature and humidity over a period of 14 months.

Materials and methods Litter studied Twelve different litters, belonging to 10 species, were studied. These included the following: one conifer, Pirzus halepensis Miller, of which two sorts were used, one originating from the spontaneous annual needle fall in summer (PHs), and the other from green needles that dried on the tree following the passage of a low intensity fire in winter (PHw); two grass species, Brachypodiumphoenicoides Roem. and S. (BP) and Brachypodium retusum (Pers.) Beauv. (BR); and seven broad-leaved species including two shrub species, Cistus albidus L. (XA) and C. morzspeliensis L. (XM), and five tree species, the first three deciduous, Castanea sativa Miller (CS), Fagus sylvatica L. (FS), and Quercuspubescens L. (QP), and the others evergreen, Quercus coccifera L. (QC) and Q. ilex, of which two types of litter were used, one composed exclusively of leaves (QI) and the other a mixture of leaves and flowers that is the normal type of litter found with this species during the first month of leaf fall (QIf). The litters were collected at the season of maximum leaf fall, that is in the spring for Q. ilex, Q. coccifera, and the two shrubs, summer for the pine, and in the autumn for Q. pubescens, Castanea, and Fagus. The dried leaves of the two grasses were cut in the summer. Litters were air dried and stored in the laboratory. Microcosms The microcosms were of the type described by Taylor and Parkinson (1988). They consisted of plastic cylinders 15 cm in diameter and 15 cm tall, fitted with a lid and a bottom pierced by a l-cm diameter hole allowing excess water to be drained off. One kilogram of a previously prepared soil mixture consisting of mineral soil and surface organic horizon (3: 1) from a nearby forest plot was placed on a grid situated 2 cm above the bottom of each microcosm. The litter, previously soaked for 24 h in 0.1 L of water, was placed on the soil in the microcosms and enclosed in a thin litter bag of l-mm mesh to recover all the material at each sampling time. So as not to deplete the soluble nutrients in the system, the soaking water was poured into the microcosm, and the quantity of water needed to be added to bring the water content of the microcosm soils up to 80% field capacity was calculated by weighing. The quantity of water needed to replace that evaporating and thus maintain a constant soil moisture during incubation was also calculated each week by weighing the microcosms and added. The microcosms were maintained in the dark at 22OC throughout the experiment.

Litter incubation Thirty-five samples of 7 f 0.0001 g were weighed and placed in microcosms for each of seven of the litters (PHs, XM, CS, FS, QC, QI, and QP). Five additional samples were weighed after drying in an oven at 55OC for 48 h to determine the exact water content of the original litter. Five samples of each species were sampled at seven dates: 2 weeks and 1, 2 , 4 , 6, 10, and 14 months, and the dry weights were determined after drying in a forced draught oven for 48 h at 55°C. For the two grass litters (BP and BR) the incubation only lasted 10 months, so there were only six sampling dates. For three of the litters (PHw, XA, and QIf) there was no sampling at 2 weeks and 1 month, so that there were only five sampling dates. Measurement of the initial chaf.acteristics The leaf specific mass was only measured for the broad-leaved species. Five samples of 20 leaves were dried in an oven for 48 h at 55"C, then weighed. The corresponding leaf areas were measured with a planimeter (Delta-T Area meter MK2) after having dampened them to make them flat. The leaf specific mass (LSM) was then calculated from their dry mass and area. Thickness was measured on the same leaves using a linear displacement transducer. The initial capacities of the litters to absorb water and release water-soluble substances were measured by soaking five samples of 5 f 0.0001 g of each litter for 24 h in 1 L of demineralized water. Additional samples were dried (in an oven for 48 h at 55 OC) to calculate their initial water content. At the end of 24 h the samples of each litter were weighed wet, after having removed surface water by pressing them between two sheets of filter paper. Then they were weighed dry (48 h in an oven at 55°C). The increase in the water content of leaves that had spent 24 h in water gave an estimate of the initial capacity to absorb water (absorbency). The loss in dry mass of the same samples after 24 h in water gave an estimate of their capacity to release water-soluble substances (solubility). The initial litters were analysed chemically after passing through a cyclone mill with a l-mm mesh. Ash content was measured after combustion in a muffle furnace at 550°C for 3 h. Carbon and nitrogen content were determined with a Perkin Elmer elemental analyser (PE 2400 CHN). Fibres were determined using Van Soest procedures (1963, 1965) adjusted for Fibertec (Van Soest and Robertson 1985). The following fractions were determined: hemicellulose, cellulose, and lignin. Soluble sugars were analysed by anion exchange chromatography with pulsed amperometric detection (Rocklin and Pohl 1983). The initial phosphorus concentration was determined with an inductively coupled plasma spectrophotometer (ICP) after acid digestion.

Spectrometry measureynts All samples were scanned with a near-infrared reflectance spectrophotometer (NIRSystems 6500). Two replicate measurements of monochromatic light were made at 2-nm intervals over a range from 400 to 2500 nm, to produce an average spectrum with 1050 data points. Reflectance (R) was converted to absorbance (A) using the following equation: A = log (1tR). Data analysis was conducted using the IS1 software system (Shenk and Westerhaus 1991). Data analysis The ash content of all the litter samples was predicted by NIRS (Joffre et al. 1992), which enabled the ash-free litter mass remaining to be determined for each. The ash-free ovendried litter mass remaining will hereafter be denoted as LMR. The LMR values for each species were fitted to several mathematical models assuming that the litters were composed of three, two, or one compartments with different rates of decomposition:

[l] LMR

+ Ce-d' + E

+ LMR = Aecb' + C

[2] LMR [3]

= AeCb'

[4] LMR

=~

e - ~ Ce-d' '

= Aecb'

where A

+ C = 100

CAN. J. BOT. VOL. 72. 1994

m

o!

w

**t--

3

t to! T

wwC1

'?"?

o o o \ D c l * m t m t0-C1-

where b and d are rate constants over time t for components A , C, and E, and LMR is expressed as a percentage of the original mass and the time in years. The various fitted regressions obtained with each species were compared with the fit to eq. 1, which always obtained the highest coefficient of determination, using an F-test. The F-ratio was calculated from the sum of squares of the two fits and the number of degrees of freedom in each. The multiple comparison between the fitted constants was carried out using the TI-method (Sokal and Rohlf 1981). To assess the influence of the initial characteristics of the litters on mass loss during incubation, the correlation coefficients between the initial characteristics and the fitted constants were calculated.

Results Initial litter characteristics The carbon content of the litters varied little between species (45 -52 %), with the exception of PHs, which contained 57 % carbon (Table 1). In contrast, the nitrogen content varied twofold (0.75 - 1.47 %) and that of phosphorus sixfold (2201450 mg . kg-'). In the group of species in this study, the usual ratios that are used (C:N and C:P) were therefore mainly dependent on the N and P contents (r = -0.93 and -0.81, p < 0.01) and provided no further information; they were not therefore used as indicators of initial litter quality. The litters also differed in their cellulose content (18 -33 %) and lignin content (9-28%). The LC1 varied from 0.21 and 0.24 (for the two grasses, BP and BR) to 0.62 (XM). This index was strongly correlated with the lignin content (r = 0.91, p < 0.01) and so was not retained. It was compared with the HLQ used by McClaugherty and Berg (1987), which was also not retained because it provided no additional information (r between LC1 and HLQ = -0.98). Soluble sugars were very unequally distributed, being 2-8 times more abundant in the litter of grass species (BP and BR) and in QI than in the others. The ash content varied from 4 to 9 %, the litters of the two Cistus species (XA and XM) being particularly rich in inorganic substances. The litters could be distributed along standard quality gradients, in relation to their lignin to N ratio or lignin to P ratio. However, the first depended mainly on the lignin concentration (r = 0.82, p < 0.01), whereas the latter depended on the P abundance.(r = -0.73, p < 0.01). These ratios were not therefore used to describe the initial quality of the litters. The water absorption capacities of the initial litters and their capacity to release water-soluble substances varied fourfold (Table 1). The two pine litters and that of QI were typified by their low absorption capacities, whereas in contrast, CS litter was capable of absorbing 250% of its own mass of water in 24 h. The quantities of water-soluble substances released in 24 h by leaching usually varied between 3 and 8 % of litter mass, with the exception of Aleppo pine needle litter that was dried when still green (PHw) and that released 11% of its mass in 24 h. The leaf specific mass of the broad-leaved species varied between 1 and 3 and the thickness from 1 to 4. Comparison of the various decomposition models To compare the mass losses and the parameters of decay in different species, we needed to find the mathematical function that fitted the data best. Regression fitted to the most complex model (eq. I), that is with two components ( A and C) each decreasing at a different rate (b and d) and a third refractory component (E), had the highest coefficients of determination (Table 2). However, because of the limited number of sampling dates, the number of degrees of freedom was low and the

GILLON ET AL.

TIME

(year)

FIG. 1. Measured values of litter mass remaining (LMR expressed as percentage of ashfree original mass) with incubation time, and predicted double-exponential decay regression lines (LMR = Ae-br + Ce-dr where A + C = 100) of the 12 types of litter studied. 2. Coefficients of determination (?) of the regressions fitted to the different models TABLE

Species

Abbr.

1

2

3

Pinus halepensis (summer) Pinus halepensis (winter) Brachypodiurn phoenicoides Brachypodiurn retusunl Cistus albidus Cistus monspeliensis Castanea sativa Fagus sylvatica Quercus coccifera Quercus ilex (leaves) Quercus ilex (leaves + flowers) Quercus pubescens

PHs PHw BP BR XA XM CS FS QC QI QIf QP

0.988 0.993 0.983 0.978 0.974 0.939 0.980 0.986 0.977 0.982 0.978 0.979

0.986 0.992 0.983 0.975 0.973 0.939 0.978 0.984 0.977 0.982 0.978 0.976

0.987 0.992 0.983 0.978 0.955** 0.927* 0.955*** 0.981* 0.95@ 0.981 0.978 0.975

4

NOTE:The four models were as follows: 1, double-exponential decay function with asymptote (LMR = ~ e -+ ~ '~e-'I' + E); 2, double-exponential decay function (LMR = ~ e - + ~ '~ e - " ' 1 where A + C = 100; 3, asymptotic function (LMR = ~ e - + ~ C); ' 4, single-exponential function (LMR = ~ e - ~ ' Significant ). differences with the fit 1: *, p < 0.05; **, p < 0.01; ***, p < 0.001.

estimates for the regression constants were poor (in 50% of cases the standard error of the parameters was greater than the estimate). The regressions fitted to the double-exponential decay model (eq. 2) had slightly lower coefficients of determination than those fitted to the previous model. However, none of the differences were significant and the parameters of this model were estimated with standard errors that were generally much lower (Table 3). The regressions fitted to an asymptotic model (eq. 3) did not differ significantly from those fitted to eq. 1 for seven species but were significantly less good for the other five species (XA, XM, CS, FS, and QC) (Table 2). Finally, the regressions fitted to the single-exponential decay model

(eq. 4) generally had coefficients of determination that were much lower than the first three models and were always significantly different from those fitted to the most complex function (eq. 1). The regressions fitted to the double-exponential decay model (eq. 2) were therefore adopted (Fig. 1) because they provided a better estimate of the regression parameters and did not differ significantly from those obtained with the more complex model (eq. 1).

Comparison of the regression parameters .. Component A represents the labile component of the litters since it was associated with a high rate constant (6). It varied

950

CAN. J. BOT.

1100

1300

1500

1700

1900

2100

2300

2500

Wavelength ( nm ) FIG.3. Correlation of the labile component values (A) with second derivative spectra for the 12 initial litter types.

o ~ ' " ' " " " " ~ species

FIG.2. Ninety-five percent comparison intervals by the TI-method for labile components A.

TABLE4 . Correlations between the regression parameters and initial litter characteristics A

TABLE 3. Parameters of the exponential regressions LMR = Aecbl Ce-dl, where C = 100 - A, LMR expressed as percentage of original mass, and t in years (estimated SE in parentheses)

+

Species PHs PHw BP BR XA XM CS FS

Qc QI QIf Qp

A

b

d

21.69 (3.04) 65.96 (5.44) 69.00 (6.23) 50.09 (4.49) 49.15 (3.84) 45.14 (4.87) 38.93 (2.58) 18.08 (2.58) 24.86 (1.95) 58.44 (7.80) 67.49 (4.68) 34.05 (7.78)

NOTE:Set Table 2 for names of species. *, not significantly different from 0.

between 18 and 69% of the initial litter mass (Table 3) and showed significant differences between species (Fig. 2). Rate constant b varied from 4 to 13/year, i.e., the labile component A had a half-life of 20 -60 days under the experimental conditions. Rate constant b was significantly lower in PHw, QI, and QIf than in CS and QC, the other species having values between those of these two groups. Component C, complement of component A, represents a more resistant component since it was associated with a low rate constant (d); it varied between 31 and 82% of the initial mass. Rate constant d did not differ significantly from zero for five species (PHw, BP, BR, QI, and QIf) (Table 3). Their component C could therefore be interpreted as a refractory component under the experimental conditions. In these five species, component A always accounted for more than 50% of the initial litter (Table 3). For the other litters, rate constant d varied from 0.4 to 0.8/year, i.e., values 10-20 times lower than rate constant b (Table 3). Under the experimental condi-

b

d

Solubility Absorbency LSM Thickness C N P Ash Sugar Hemicellulose Cellulose Lignin NOTE:n = 12 for A and b, and n = 7 ford values significantly different from 0 (except LSM and thickness where n = 7 for A and b, and n = 6 for d). *, p < 0.05;", p < 0.01.

tions, the half-life of the resistant component C therefore varied between 320 and 630 days. Rate constant d was significantly lower in PHs and QC than in CS and FS, the other species having intermediate values. Correlations between regression parameters and initial characteristics Parameter A was positively correlated with the initial litter solubility and with its concentration of soluble sugars but was negatively correlated with the initial carbon concentration (Table 4). Rate constant b was positively correlated with the absorbency of the initial litters. Rate constant d, when it was significant, was negatively correlated with the thickness and the mass per surface area of the initial litters. Regression parameters and near-infrared spectra of the initial litters The correlogram between the values of component A and the near-infrared spectra of the litters (1100-2400 nm) showed significant correlations in some regions of the spectrum (Fig. 3). The regions of the spectrum for which highly significant correlations (r > 0.71, p < 0.01) with component A occurred are all related to the concentration of starch and sugars. The

95 1

ET AL.

correlogram between rate constants b and d and the nearinfrared spectra of the litters also showed significant but much less strong correlations.

Discussion Initial characteristics The litter studied came from a varied range of plants, i.e., grasses, shrubs, conifers, and broad-leaved trees. The initial nitrogen and lignin contents within the litters studied ranged from 0.75 to 1.47% and from 8.8 to 30.6%, respectively. The following studies used various types of litters to investigate the influence of the initial characteristics on decomposition: Edmonds (1987), who used woody litters with a nitrogen content varying between 0.04 and 0.47% and lignin content between 15.9 and 40.9 % ; Melillo et al. (1982), using deciduous leaves with nitrogen content between 0.6 and 1.2% and lignin content of 10.1 -24.1 % ; Taylor et al. (1989), using litters of conifers, broad-leaved trees, shrubs, and grasses with nitrogen content between 0.52 and 1.31% and lignin between 3.4 and 20.5%; Aber et al. (1990), using leaf litters and roots with nitrogen and lignin contents varying between 0.43 and 1.67% and 12.1 and 33.8%, respectively; and Gallardo and Merino (1993), using litters from broad-leaved trees and shrubs with nitrogen and lignin contents varying from 0.33 to 0.93% and from 3.8 to 11.5% , respectively. Compared with that of other decomposition studies, the range of variation in initial N and lignin contents in the litters in this study was at least of the same order of magnitude. Mass losses After 2 months incubation at 22OC in the microcosms, the litters studied had lost between 20 and 43% of their initial mass and between 30 and 63 % after 4 months. In very similar microcosms but maintained at 26OC, Taylor et al. (1989) obtained losses varying from 16 to 48% after 2 months and from 22 to 57% after 4 months, which is fairly similar. After 14 months incubation (or 10 months in the case of the two grass litters), the litters had lost between 52 and 74% of their mass. In the field, P. halepensis and Q. pubescens litters, enclosed in litter bags and placed on the pine and oak forest soils in the Marseille region of southern France, had lost on average 48 and 55 % of their mass after 26 months (unpublished data), which is equivalent to about 4 to 6 months incubation in our microcosms. There are other field data on decomposition involving the same or similar species to those in this study. According to Gallardo and Merino (1993), Quercus coccifera litter had lost 39% of its original mass after 2 years in the field in the south of Spain (this is equivalent to about 6 months incubation in our microcosms), those of other Quercus species had lost between 31 and 68% (equivalent to about 4-6 months incubation of our Quercus), and that of a Cistus sp. had lost 27-51 % in each of two study sites (i.e., 2-4 months incubation). The speed of decomposition in the microcosms was therefore 4-6 times faster than in natural conditions. It can therefore be estimated that 14 months incubation in the microcosm was equivalent to 5 -6 years decomposition in the field. Taylor et al. (1989) also noted that in their microcosms maintained at 26OC, 2 months of mass loss corresponded to 1 year for the same species in the field and that 4 months incubation was equivalent to 1.5 -2 years in the field. This experimentation therefore provides a test for the relation between long-term decomposition and litter quality characteristics.

Mass loss dynamics The model most commonly used to describe litter mass loss with time is the single-exponential decay model (Olson 1963; Witkamp and Olson 1963). In this study, where the litters lost a total of between 50 and 75% of their initial mass, the data fitted poorly to this function (Table 2). The asymptotic model, used for example by Howard and Howard (1974) and Berg and Ekbohm (1991), brings into play a totally refractory compartment. Our data only fitted this model well for seven species. However, the asymptotic model can be considered to be just a special case of the doubleexponential decay model, in yhich the rate constant d is zero. For all the litters in this study, the double-exponential decay model, used for example by Lousier and Parkinson (1976), O'Conne1 (1987), and Parsons et al. (1990), accounts well for the mass loss during incubation in the microcosms (Fig. 1). Under the experimental conditions, the litters behaved as if they were composed of two compartments, one labile in character that decreased quickly, and the other more resistant that decreased 10-20 times more slowly or showed no significant decrease. The mass loss dynamics observed in this experiment can be classified into two different patterns. The first occurred in litters in which the labile compartment dominates and that have a very refractory compartment that did not decrease significantly in the experiment. In this case, the refractory character of the second compartment could either be due to the experimental conditions (particularly the absence of a detritivore fauna) or to the biochemical nature of this compartment. The second pattern occurred in litters composed of-a small labile compartment and a dominant resistant compartment that decreased slowly but significantly (half-life between 320 and 630 days). The relative contribution of the labile compartment determined the mass loss mainly at the start of decomposition; the correlations between the litter mass remaining in the different litters and the values of component A were highly significant ( p < 0.01) at all sampling dates up until 10 months incubation in the microcosms, with a maximum at 6 months, but were no longer significant at 14 months. In the long term, litters with a dominant resistant compartment can in fact theoretically attain cumulative mass losses greater than those with a dominant labile compartment. For example, at the end of 14 months incubation in the microcosms, the litter of Castanea sativa, with a dominant resistant compartment, in total had lost more mass (74%) than that of Q. ilex (64%), with a dominant labile compartment. d

z

Comparison between species The litters in this study can be classified according to the relative sizes of their labile and resistant compartments. Those of l? halepensis (collected in winter), Q. ilex (QI and QIf), and B. phoenicoides (Fig. 4) were composed of a largely dominant labile compartment (60-70% of their initial mass). The halflife of this compartment in the microcosms varied between 40 and 60 days, so these litters had lost most of their mass in a few months and almost all that remained at the end of 10 or 14 months was the refractory component that only decreased slowly or not at all under the experimental conditions. The litters of B. retusum, Cistus monspeliensis, and Cistus albidus (Fig. 4) were of an intermediate type; their labile and resistant compartments were of approximately the same size and their resistant compartments declined slowly or insignificantly. All

CAN. J. BOT.

952

VOL. 72,

1994

01

I

-0.006

I

I

-0.Q07

I

I

-0.009

-0.008

N l R S absorbance a t 1476 nrn

0

I I t I 0.2 0.4 0.6 0.8

O 1.0

TIME

I

I

I

I

I

I

1

0.2 0.4 0.6 0.8 1.0

(year)

FIG.4. Difference between changes in total mass and resistant component C with time showing the labile component A loss (hatched) for four species.

the other litters were typified by a dominant resistant compartment. The litters of Q. pubescens and Castanea sativa (Fig. 4) had a resistant compartment of moderate proportions (6065%), which declined relatively fast, especially in Castanea sativa. Finally, the litters of I? halepensis (collected in summer), E sylvatica, and Q. coccifera (Fig. 4) were composed almost entirely of a resistant component (75 -80% of their initial mass) that disappeared slowly. These were the litters of the three species that lost the least total mass during this experiment. Litters of the same or related s ~ e c i e ssometimes behaved somewhat similarly in the microcosms; this was the case for the two grass species (BP and BR), the two litters of Q. ilex (QI and QIf), and the two Cistus litters (XA and XM). In contrast, the two pine litters behaved very differently; the needles dried when green on the trees (PHw) included a much larger labile component than the needles falling naturally in summer (PHs) and the mass loss at the end of 14 months incubation was significantly different. Similarly, the three oak litters had contrasting behaviours during this study: Q. ilex had a dominant labile compartment, whereas Q. pubescens and Q. coccifera had a strong resistant compartment. Factors determining litter mass loss The relative importance of the labile and resistant compartment, which to a large extent determined the mass loss dynamics under the conditions of this experiment, was related to the initial contents of water-soluble substances and carbon contents in the litters (Table 4): the richer the litters were in water-soluble substances, particularly soluble sugars, the greater the labile component, and the richer the litters were initially in carbon, the greater the weight of the resistant component. However, no significant correlation was found between the relative proportions of these compartments and the initial contents of nitrogen, phosphorus, or lignin; these chemical char-

01

0

"'1

I

50

I

I

I

I

100 150 200 250 Water absorbency (Ole)

I 300

(c )

LSM ( g m-2) FIG.5. Linear regressions between exponential regression constants and initial litter char&teristics. ( a ) A and second derivative of NIRS absorbance at wavelength 1476 nm. (b) b and water absorbency. (c) d and leaf specific mass (LSM).

acteristics have nevertheless been shown in many studies (see Introduction) to determine the rate of decomposition. The correlations found between the absorbance in certain infrared wavelengths and component A (Fig. 3) show that a part of the spectral information is directly related to the labile (or the resistant) component. For example, the absorbance of the initial litters at a wavelength of 1476 nm alone explains 93 % of the variance in the percentage of the labile component in the 12 litters studied (Fig. 5a). This relation is the strongest obtained in this study between an initial characteristic of the litters and a fitted constant. This supports the hypothesis that the two compartments defined in the double-exponential decay model are related to the biochemical characteristics of the initial litters. Rate b , at which the labile compartment disappeared, was greater the more the initial litters were absorbent or permeable (Fig. 5b). In other words, the greater the water exchanges

GlLLON ET

between the leaf tissues and the outside environment, the faster the labile substances decomposed. Therefore the speed at which the labile compartment disappeared seemed to be partly determined by properties related to the structure of the leaves. Rate d, at which the resistant component declined under the experimental conditions, was greater the thinner of the leaves and the lower their mass per unit area (Fig. 5c). As with the labile compartment, the decrease in the resistant compartment of the litters therefore seemed to depend on the initial physical characteristics of the litters.

Conclusions The litters studied showed a great variety in behaviours in their decomposition dynamics in microcosms at constant temperature and humidity and in the absence of animal detritivore activity. By fitting their mass loss with time to a doubleexponential decay model, it was possible to interpret these behaviours as a function of the relative proportions of two compartments: one rapidly decomposable and the degradation of which is favoured by permeability of the leaf tissues, and the second more resistant, the degradation of which is faster in thin leaves and of low specific mass. In these Mediterranean species, the physical properties of the leaves therefore seem to play a major role during leaf decomposition. This is in agreement with one of the conclusions of the study by Gallardo and Merino (1993) on the decomposition of litters of Mediterranean species. These authors stressed the key role of the toughness of leaves in controlling the rate of decomposition. In contrast, their elementary chemical composition, such as their initial nitrogen or fibre concentration, does not seem to play any particular role in the decomposition dynamics of the litters studied. Nevertheless, the strong relations recorded between the absorbance of litters at certain wavelengths in the infrared spectrum and the values of the labile component confirm that the partition between the labile and resistant components has a biochemical basis. It seems as if the Mediterranean litters studied had different aptitudes for rapid decomposition related to their biochemical composition and that the speed at which the processes took place depended on physical properties related particularly to their anatomy (thickness, specific mass, and degree of permeability). The use of NIRS offers new perspectives. It is probable that NIRS will be able to predict the aptitude of litters to decompose under given environmental conditions on the basis of spectral information, just as it can predict with great precision the in vivo or in vitro digestibility of forages (Holechek et al. 1982; Duncan et al. 1987; Meuret et al. 1993); digestibility does in fact express a property of plant material that has very similar biochemical processes to those that take place during litter decay.

Acknowledgements W e want to thank Bjorn Berg and an anonymous reviewer for their constructive comments. This research was supported by the Programme interdisciplinaire de recherches sur l'environnement (PIREN) of the Centre national de la recherche scientifique. Aber, J.D., and Melillo, J.M. 1980. Litter decomposition: measuring relative contribution of organic matter and nitrogen to forest soils. Can. J. Bot. 58: 416-421. Aber, J.D., Melillo, J.M., and McClaugherty, C. 1990. Predicting long-term patterns of mass loss, nitrogen dynamics, and soil

AL.

953

organic matter formation from initial fine litter chemistry in temperate forest ecosystems. Can. J. Bot. 68: 2201-2208. Berendse, F., Berg, B., and Bosatta, E. 1987. The effect of lignin and nitrogen on the decomposition of litter in nutrient-poor ecosystems: a theoretical approach. Can. J. Bot. 65: 1116- 1120. Berg, B., and Ekbohm, G. 1983. Nitrogen immobilization in decomposing needle litter at variable carbon:nitrogen ratios. Ecology, 64: 63-67. Berg, B., and Ekbohm, G. 1991. Litter mass-loss rates and decomposition patterns in some needle and leaf litter types. Long-term decomposition in a Scots pine forest. VII. Can. J. Bot. 69: 14491456. Berg, B., and Staaf, H. 1980.~Decompositionrate and chemical change of Scots pine needle litter. 11. Influences of chemical composition. In Structure and function of northern coniferous forests an ecosystem study. Edited by T. Persson. Ecol. Bull. 32: 375 -390. Berg, B., and Tamm, C.O. 1991. Decomposition and nutrient dynamics of litter in long-term optimum nutrition experiments. Scand. J. For. Res. 6: 305-321. Duncan, P.A., Clanton, D.C., and Clark, D.H. 1987. Near infrared reflectance spectroscopy for quality analysis of sandhills meadow forage. J. Anim. Sci. 64: 1743- 1750. Edmonds, R.L. 1987. Decomposition rates and nutrient dynamics in small-diameter woody litter in four forest ecosystems in Washington, U.S.A. Can. J. For. Res. 17: 499-509. Fogel, R., and Cromack, K., Jr. 1977. Effect of habitat and substrate quality on Douglas fir litter decomposition in western Oregon. Can. J. Bot. 55: 1632-1640. Gallardo, A., and Merino, J. 1993. Leaf decomposition in two mediterranean ecosystems of southwest Spain: influence of substrate quality. Ecology, 74: 152- 161. Hart, S.C., Firestone, M.K., and Paul, E.A. 1992. Decomposition and nutrient dynamics of ponderosa pine needles in a Mediterraneantype climate. Can. J. For. Res. 22: 306-314. Holechek, J.L., Shenk, J.S., Vavra, M., and Arthun, D. 1982. Prediction of forage quality using near infrared reflectance spectroscopy on esophageal fistula samples from cattle on mountain range. J. Anim. Sci. 55: 971 -975. Howard, P. J. A., and Howard, D .M. 1974. Microbial decomposition of tree and shrub leaf litter. 1. Weight loss and chemical composition of decomposing litter. Oikos, 25: 341 -352. Joffre, R., Gillon, D., Dardenne, P., Agneessens, R., and Biston, R. 1992. The use of near-infrared spectroscopy in litter decomposition studies. Ann. Sci. For. 49: 481-488. Lossaint, P. 1973. Soilvegetation relationships in Mediterranean ecosystems of southern France. In Mediterranean type ecosystems. Edited by F. Di Castri and H.A. Mooney. Springer, New York. pp. 199-210. Lousier, J.D., and Parkinson, D. 1976. Litter decomposition in a temperate deciduous forest. Can. J. Bot. 54: 419-436. Maggs, J., and Pearson, C.J. 1977. Litter fall and litter layer decay in coastal scrub at Sydney, Australia. Oecologia, 31: 239-250. McClaugherty, C.A., and Berg, B. 1987. Holocellulose, lignin and nitrogen levels as rate-regulating factors in late stages of forest litter decomposition. Pedobiologia, 30: 101- 112. Meentemeyer, V. 1978. Macroclimate and lignin control of litter decomposition rates. Ecology, 59: 465 -472. Melillo, J.M., Aber, J.D., Linkins, A.E., Ricca, A., Fry, B., and Nadelhoffer, K.J. 1989. Carbon and nitrogen dynamics along the decay continuum: plant litter to soil organic matter. In Ecology of arable land. Edited by M. Clarhom and L. Bergstrom. Kluwer Academic Publishers, Norwell, Mass. Melillo, J.M., Aber, J.D., and Muratore, J.F. 1982. Nitrogen and lignin control of hardwood leaf decomposition dynamics. Ecology, 63: 621 -626. Meuret, M., Dardenne, P., Biston, R., and Poty, 0 . 1993. The use of NIR in predicting nutritive value of Mediterranean tree and shrub foliage. J. Near Ir. Spectrosc. 1: 45-54.

954

CAN. J. BOT. VOL. 72, 1994

Mitchell, D.T., Coley, P.G.F., Webb, S., and Allsopp, N. 1986. Litterfall and decomposition processes in the coastal fynbos vegetation, south-western Cape, South Africa. J. Ecol. 74: 977-993. O'Connel, A.M. 1987. Litter dynamics in karri (Eucalyptus diversicolor) forests of South-western Australia. J. Ecol. 75: 781 -796. Olson, J.S. 1963. Energy storage and the balance of producers and decomposers in ecological systems. Ecology, 44: 322 -33 1. Parsons, W.J.F., Taylor, B.R., and Parkinson, D. 1990. Decomposition of aspen (Populus tremuloides) leaf litter modified by leaching. Can. J. For. Res. 20: 943 -951. Rapp, M., and Leonardi, S. 1988. Evolution de la litikre au sol au cours d'une annCe dans un taillis de ch&ne vert (Quercus ilex). Pedobiologia, 32: 177- 185. Rocklin, R.D., and Pohl, C.A. 1983. Determination of carbohydrates by anion exchange chromatography with pulsed amperometric detection. J. Liq. Chromatogr. 6: 1577 - 1590. Schlesinger, W.H. 1985. Decomposition of chaparral shrub foliage. Ecology, 66: 1353- 1359. Schlesinger, W.H., and Hasey , M.M. 1981. Decomposition of chaparral shrub foliage: losses of organic and inorganic constituents from deciduous and evergreen leaves. Ecology, 62: 762774. Shenk, J.S., and Westerhaus, M.O. 1991. IS1 NIRS-2.Software for near-infrared instruments. Infrasoft International, Silver Spring, Md.

Sokal, J.R., and Rohlf, R.R. 1981. Biometry. 2nd ed. W.H. Freeman and Co., San Francisco. Staaf, H., and Berg, B. 1982. Accumulation and release of plant nutrients in decomposing Scots pine needle litter. Long-term decomposition in a Scots pine forest. 11. Can. J. Bot. 60: 15611568. Taylor, B., and Parkinson, D. 1988. A new microcosm approach to litter decomposition studies. Can. J. Bot. 66: 1933- 1939. Taylor, B., Parkinson, D., and Parsons, W.F.J. 1989. Nitrogen and lignin content as predictors of litter decay rates: a microcosm test. Ecology, 70: 97-104. Van Soest, P.J. 1963. Use of detergent in the analysis of fibrous feeds. 11. A rapid method-for the determination of fiber and lignin. J. Assoc. Off. Anal. Chem. 46: 829-835. Van Soest, P.J. 1965. Use of detergent in the analysis of fibrous feeds. 111. Study of effects of heating and drying on yield of fiber and lignin in forages. J. Assoc. Off. Anal. Chem. 48: 785-790. Van Soest, P.J., and Robertson, J.B. 1985. Analysis of forages and fibrous foods: a laboratory manual for animal science. Cornell University Press, Ithaca, N.Y. Van Wesemael, B. 1993. Litter decomposition and nutrient distribution in humus profiles in some Mediterranean forests in southern Tuscany. For. Ecol. Manage. 57: 99-114. Witkamp, M., and Olson, J. 1963. Breakdown of confined and nonconfined oak litter. Oikos, 14: 138- 147.