Leaf attributes in the seasonally dry tropics: a comparison of four habitats in northern Australia


  • L. D. Prior,

    Corresponding author
    1. School of Science, Northern Territory University, Darwin, NT 0909,
    2. Key Centre for Tropical Wildlife Management, Northern Territory University, Darwin, Northern Territory 0909, Australia
      †Author to whom correspondence should be addressed. E-mail: lynda.prior@ntu.edu.au
    Search for more papers by this author
  • D. Eamus,

    1. Institute for Water & Environmental Resource Management, University of Technology, Sydney, NSW 2007, and
    Search for more papers by this author
  • D. M. J. S. Bowman

    1. Key Centre for Tropical Wildlife Management, Northern Territory University, Darwin, Northern Territory 0909, Australia
    Search for more papers by this author

†Author to whom correspondence should be addressed. E-mail: lynda.prior@ntu.edu.au


  • 1Patterns of leaf attributes were compared at regional and global scales in relation to the seasonal availability of water.
  • 2Light-saturated assimilation rate (Amass), leaf mass per area (LMA), leaf density, thickness, life span, saturated water content, chlorophyll, nitrogen and phosphorus content were determined during the wet season for 21 tree species in four contrasting habitats in northern Australia. Rainfall in this area is extremely seasonal.
  • 3Amass and foliar chlorophyll, N and P contents were positively correlated with each other, and were all negatively correlated with LMA, leaf thickness, density and life span.
  • 4Deciduous species had smaller LMA and leaf life span, and larger foliar N and P contents than did evergreen species.
  • 5The eight Myrtaceous species had smaller Amass, foliar chlorophyll, N and P contents, and larger LMA, leaf thickness and leaf life span than did the non-Myrtaceous species.
  • 6Leaves from the closed canopy dry monsoon forest had significantly larger Amass, chlorophyll and P contents than did leaves from the three open canopy habitats (eucalypt open forest, mixed woodland and Melaleuca swamp). This reflected the relatively low proportions of evergreen and Myrtaceous species in the dry monsoon forest. There were also significant intraspecific differences among habitats.
  • 7Leaf thickness, density and LMA were lower than predicted from globally derived relationships with temperature and precipitation. Tropical seasonally dry biomes are under-represented in such global analyses.


Measurements on leaves from diverse ecosystems have consistent interrelationships among a range of leaf attributes. For example, mass-based assimilation (Amass) and leaf nitrogen are positively correlated, and both decrease with increasing leaf mass per area (LMA) and leaf life span (Reich, Walters & Ellsworth 1992, 1997; Reich et al. 1999). Variations in leaf attributes, and especially in LMA, are strongly correlated with several whole-plant and ecosystem properties (Lambers & Poorter 1992; Reich et al. 1992). For example, relative growth rates of seedlings (Huante, Rincón & Acosta 1995; Cornelissen, Castro-Díez & Hunt 1996; Wright & Westoby 1999), maximum instantaneous assimilation rates and annual above-ground production of a forest per unit canopy mass (Reich et al. 1997) all scale negatively with LMA. Leaf mass per area, leaf thickness and leaf density (Witkowski & Lamont 1991; Castro-Díez et al. 1997; Bussotti et al. 2000), along with relative growth rates (Huante et al. 1995; Wright & Westoby 1999), are affected by environmental factors such as water and nutrient availability.

Niinemets (2001) highlighted geographical patchiness of available leaf attribute data sets. A notable gap is for seasonally dry tropical environments such as dry tropical forests and savannas, which occupy large areas of Africa, South America, Australia and parts of southern Asia (Murphy & Lugo 1986; Eamus & Prior 2001). In these environments, the major constraints on plant and ecosystem productivity are low soil water availability during the dry season, flooding during the wet season, generally low soil fertility, and fire. Low temperatures are rarely, if ever, experienced. The present study investigated attributes of leaves from four habitats near Darwin, in the extensive savanna region of northern Australia. This area is interesting because rainfall is extremely seasonal, and because dry-season deciduous, semi-deciduous and evergreen tree species are all well represented in the region's flora (Williams et al. 1997) and are likely to employ contrasting physiological strategies. Leaves of deciduous species are present only at relatively favourable times, while evergreen leaves must survive conditions of low soil water availability and high evaporative demand. Deciduous species are also likely to have shorter leaf life spans than evergreen species, and this affects other leaf attributes (Chabot & Hicks 1982; Eamus & Prichard 1998). The native flora is equally dominated by Pan-tropic and Australasian genera, with substantial contributions from Old World Tropic, Indo-malaysian and Cosmopolitan genera (Bowman, Wilson & Dunlop 1988). A variety of habitats are embedded within the savanna matrix of open forest and woodland. Wet monsoon rainforests are associated with sites of perennial moisture availability, such as along streamlines or in seepage areas that are protected from fire, while dry monsoon rainforests occur in drier areas that are topographically protected from fire (Bowman 1992), and Melaleuca swamps are found in seasonally inundated or poorly drained areas that are periodically burnt (Bowman & Minchin 1987; Russell-Smith 1991; Brock 1993). These habitats experience different levels and patterns of soil moisture availability, but share very similar rainfall, temperature and light conditions, thereby excluding these as causes of variation in leaf attributes.

This study considers patterns in leaf attributes in both the global and regional context by addressing the following specific questions: (1) Do leaf attributes differ between species in four contrasting habitats within the Darwin region? (2) Do leaf attributes of individual species differ among habitats? (3) Do differences between the evergreen and dry-season deciduous tree species in this region mirror those between the evergreen and winter-deciduous species of the Northern Hemisphere? (4) Are there systematic differences in leaf characteristics between the Australasian, Pan-tropic and Cosmopolitan genera in this region? and (5) How do leaf attributes of a range of tree species in an Australian savanna biome compare with those from other biomes?

Materials and methods

Study sites

The study sites were located near Darwin, northern Australia, in four distinct habitats common in the northern part of the Northern Territory. The dry monsoon forest site was at Leanyer (12·4 °S, 130·9 °E) while the other sites were located at Berry Springs Wildlife Park and the adjacent Berry Springs Nature Park (12·7° S, 131·0° E). Temperatures in this region are high all year round, with monthly means for Darwin ranging from 24·9 °C in July to 29·2 °C in November. Rainfall is extremely seasonal: 97% of Darwin's average annual 1710 mm rain falls between October and April; June, July and August are virtually rainless (Commonwealth Bureau of Meteorology, Australia 2001).

The dry monsoon forest at Leanyer has been largely actively regenerated since 1990, following cyclone damage, weed invasion and wildfire (Panton 1993). Remnant trees were used as the seed source for regeneration, so that all study species occur naturally in the area. Study trees were between 2 and 6 m tall and included remnant individuals from the original forest and planted trees that were 4–5 years old.

The vegetation, soils and landforms of the Berry Springs locations have been described by Sivertsen, McLeod & Henderson (1980) and by Bowman & Minchin (1987). Each study site was about 0·5–1 ha in area. The open forest was on deep (>1 m) red massive earth, developed on highly weathered parent material. The woodland was located on shallow earthy sand with extensive surface gravel and low in nutrients (Sivertsen et al. 1980). The Melaleuca swamp was in a seasonally inundated section of a poorly drained depression. Neither open forest nor woodland had been burnt for at least 10 years, in contrast to similar areas outside the Territory Wildlife Park, where fire is frequent (Russell-Smith, Ryan & Durieu 1997).

Soil measurements

In March 2002, four to six samples of soil were taken from 0·2 m depth from each habitat for qualitative texture determination (by feel; Brady & Weil 1999) and nutrient analysis. Approximately 600–800 g soil was dried at 60 °C, ground, and passed through a 2 mm sieve to determine percentage gravel. Soil was then passed through a 1 mm sieve, and total Kjeldahl N and P were determined using flow-injection analysis (Lachat Instruments, Milwaukee, WI, USA).

Study species

At each site, all tree species with six or more representatives at least 2 m tall and with accessible leaves (lower than about 2 m from the ground) were included in the study, giving a total of 21 species from the four habitats described above. Three species (Melaleuca viridiflora, Terminalia ferdinandiana and Planchonia careya) were found in two habitats (Table 1). A full set of measurements was made at both habitats for these species. Nomenclature follows that of the Northern Territory Herbarium (2003).

Table 1.  Study species listed by habitat, with details of family and phenological guild. Nomenclature according to Northern Territory Herbarium (2003)
SpeciesFamilyPhenological guild
Dry monsoon forest
Ficus racemosaMoraceaeSemi-deciduous
Ficus scobinaMoraceaeSemi-deciduous
Mallotus nesophilusEuphorbiaceaeSemi-deciduous
Melaleuca leucadendraMyrtaceaeEvergreen
Terminalia microcarpa (formerly T. sericocarpa)CombretaceaeDeciduous
Wrightia pubescensApocynaceaeDeciduous
Open forest
Brachychiton megaphyllusSterculiaceaeDeciduous
Buchanania obovataAnacardiaceaeSemi-deciduous
Erythrophleum chlorostachysCaesalpiniaceaeSemi-deciduous
Eucalyptus tetrodontaMyrtaceaeEvergreen
Planchonia careyaLecythidaceaeDeciduous
Syzygium suborbiculareMyrtaceaeSemi-deciduous
Terminalia ferdinandianaCombretaceaeDeciduous
Callitris intratropicaCupressaceaeEvergreen
Cochlospermum fraseriBixaceaeDeciduous
Corymbia foelscheana (formerly Eucalyptus foelscheana)MyrtaceaeSemi-deciduous
Melaleuca viridifloraMyrtaceaeEvergreen
Planchonia careyaLecythidaceaeDeciduous
Syzygium eucalyptoides ssp. bleeseriMyrtaceaeSemi-deciduous
Terminalia ferdinandianaCombretaceaeDeciduous
Xanthostemon paradoxusMyrtaceaeSemi-deciduous
Acacia auriculiformisMimosaceaeEvergreen
Lophostemon lactifluusMyrtaceaeSemi-deciduous
Melaleuca viridifloraMyrtaceaeEvergreen

The study species included evergreen, semi-deciduous and deciduous trees from 12 families (Table 1). Eight species are Myrtaceous and thus of Gondwanan origin, as is the coniferous family Cupressaceae; all other families are considered either Pan-tropic or Cosmopolitan (Thiele & Adams 1999). Phenological classifica-tion was based on observation during this study, as there are slight inconsistencies between Brock (1993) and Williams et al. (1997). Trees were considered evergreen if they maintained at least 60% canopy throughout the year. Deciduous species were defined as those in which every individual lost all its leaves for a period of at least 1 month (Williams et al. 1997). Other species were considered semi-deciduous. Two species (Acacia auriculiformis and Erythrophleum chlorostachys) are considered nitrogen-fixers.

Leaf measurements

Measurements were made on well lit leaves between 0·5 and 2·5 m high. Leaflets of E. chlorostachys, phyllodes of A. auriculiformis and small terminal branchlets of Callitris intratropica (which are completely green and covered with tiny, scale-like leaves), were treated as leaves. Sampled leaves were fully expanded and thickened, and for most species had only recently matured. Three leaves from each of six trees of each species (18 leaves per species) were measured in the wet season.

Assimilation rate per unit area (Aarea) was measured under light-saturating conditions (photon flux density >1200 µmol m−2 s−1) with a portable photosynthesis system (Li-Cor Li 6200, Lincoln, NE, USA) between 9.30 am and 12.30 pm (solar noon). Measured leaves were then harvested and sealed in plastic bags containing damp paper towel before being fully rehydrated overnight in an air-conditioned laboratory. Petioles were then removed and the leaves blotted dry and quickly weighed to determine saturated mass.

Leaf thickness was measured in four places between major veins with a micrometer (Moore & Wright, Sheffield, UK). Two readings were then taken on each leaf with a chlorophyll meter (SPAD-502; Minolta, Osaka, Japan), which was calibrated for each species individually using the extraction method of Coombs et al. (1985). Area of each leaf was then estimated using a Delta-T Scan image analysis system (Delta-T Devices, Cambridge, UK). For the C. intratropica branchlets, leaf area was estimated from total branchlet length and average branchlet diameter. Dry mass was determined after leaves were dried for a minimum of 48 h at 70 °C. Leaf density was calculated as LMA divided by leaf thickness. Assimilation per unit mass (Amass) was calculated as Aarea divided by LMA.

Total Kjeldahl nitrogen (Nmass) and phosphorus (Pmass) were determined using flow-injection analysis (Lachat Instruments, WI, USA). One leaf per tree (six leaves per species) was ground to a powder, and a subsample taken for the micro-Kjeldahl procedure.

Median leaf life span was determined for each tree by tagging new branchlets during major flushes. Three branchlets from each of six trees per species were selected, and individual leaves present at that time (60–199 per species, average 110) were tagged with thin, plastic-coated electrical wire. Remaining tagged leaves were counted monthly.

Data analyses

Data were transformed (square root or log10) as required so that standard deviations were independent of means, and to comply with Cochran's test. These transformed data were used in all statistical tests described below. Tukey's HSD tests were used to determine significance of differences between pairs of means. Differences are termed ‘significant’ where P < 0·05 and ‘marginally significant’ where 0·05 < P < 0·10. Analyses were performed using statistica versions 5·5 and 6 (StatSoft Inc., OK, USA).

Relationships among leaf attributes were investigated by calculating correlation matrices based on species means, and by principal component analysis. Interactions with habitat were investigated by testing for homogeneity of slopes. Where slopes were not different, ancova was used to determine the significance of differences between habitats in these relationships.

anova was used to determine the significance of differences in soil properties among habitats, intraspecific differences between habitats, and differences between species and individual trees within species. Multi-factor anovas of the complete leaf data set were problematic because of the incomplete design – for example, no deciduous species were found in the swamp, and only three species were found in more than one habitat. To overcome these difficulties, several sets of anovas were performed. Differences between habitats, leaf phenological guilds and Myrtaceous vs non Myrtaceous species (family) were investigated for all broad-leaved species. (Data for the coniferous C. intratropica were incomplete, and assimilation rates were very low, so this species was excluded.) The incomplete nested design offered two alternatives for analysis of differences between habitats: habitat variance could be tested against that of individual trees within species (within habitats) or, more conservatively, against variance of species within habitats. (This is essentially the difference between using trees as the experimental unit, or using species.) Similarly, differences between leaf phenological guilds and between families could be tested using individual trees, or species, as the experimental unit. We chose to use the conservative, species-based analyses, as characteristics of vegetation within habitats, phenological guilds or families are inherently determined by attributes of constituent species.


Soil properties of habitats

Soil texture varied from loam in the dry monsoon forest to sandy loam in the open forest and swamp, and loamy sand in the woodland. Kjeldahl N and P content in the <1 mm soil fraction were highest in the dry monsoon forest and lowest in the swamp and woodland (Table 2). Soils of the dry monsoon forest and woodland had a high gravel content, while those of the swamp and open forest had virtually no gravel (Table 2). Therefore, per mass of substrate (soil plus gravel), Kjeldahl N and P content would be similar in the dry monsoon forest and the open forest, and lowest in the woodland.

Table 2.  Soil properties (means ± SE) at 0·2 m depth for the different habitats. Total Kjeldahl nitrogen (TKN) and phosphorus (TKP) were measured on the soil fraction that passed through a 1 mm sieve
HabitatSoil textureGravel (%)TKN (mg g−1)TKP (mg g−1)
Dry monsoon forestLoam  36 ± 61·25 ± 0·290·273 ± 0·059
Open forestSandy loam0·18 ± 0·050·69 ± 0·140·105 ± 0·006
WoodlandLoamy sand  47 ± 90·24 ± 0·030·045 ± 0·004
SwampSandy loam 2·5 ± 1·90·22 ± 0·030·030 ± 0·000

Relationships among leaf attributes

Leaf attributes were strongly intercorrelated. Mass-based assimilation, foliar chlorophyll, N and P content were all positively correlated with each other, and were all negatively correlated with LMA, leaf thickness, density and life span (Table 3; Fig. 1). There were strong negative correlations between leaf density and saturated water content. Principal components analysis identified combinations of variables that best summarized the data. The first principal component (PC1) represented a contrast between variables associated with high assimilation rates (Amass, Nmass, Pmass and Chlmass) and high LMA, and accounted for 60·2% of the total variation (Fig. 2a). The second principal component (PC2), accounting for 13·7% of the total variation, represented a contrast between high leaf density (and also area per leaf) and high saturated water content (and also leaf thickness). The third principal component (PC3; 10·1% variation) was largely a measure of area per leaf, which was not significantly correlated with any other attribute.

Table 3.  Correlation coefficients between pairs of leaf attributes, using species means (n = 23). Data for Callitris intratropica were incomplete and therefore excluded from the analysis. Log10 transformations were used for Amass, H2O/DW, LMA, leaf thickness, Chlmass, Nmass, Pmass and leaf life span. Aarea and area per leaf were not significantly correlated with any other attribute, so are not listed. Results were similar when measurements for individual leaves were used, although more correlations were then significant. Significant (P < 0·05) correlations shown in bold
 H2OLMAThicknessDensityChlmassNmassPmassLife span
Amass0·48−0·87−0·77−0·55 0·89 0·85 0·88−0·62
H2O/DW −0·51−0·21−0·79 0·48 0·44 0·52−0·32
LMA   0·90 0·61−0·90−0·89−0·85 0·67
Thickness    0·21−0·80−0·79−0·70 0·63
Density    −0·56−0·57−0·64 0·36
Chlmass      0·89 0·89−0·47
Nmass       0·94−0·50
Pmass       −0·53
Figure 1.

Significant correlations were found between 26 pairs of attributes (Table 3), including those shown here. Means for dry monsoon forest (circles), open forest (squares), woodland (triangles) and swamp (diamonds) species are plotted on log scales, except for leaf density. Regression lines are shown separately where ancova indicated significant differences between habitats (dry monsoon forest, solid line; open forest, long dashes; woodland, medium dashes; swamp, dashes alternating with dots). Slopes were significantly different only for (f) LMA vs. Nmass.

Figure 2.

(a) The first principal component (PC1) accounted for 60% of variation in the data set, and represented a contrast between (i) Nmass, Amass, Pmass and Chlmass and (ii) LMA. Principal component 2 (PC2) represented a contrast between (i) leaf density and (ii) leaf saturated water content and leaf thickness. (b) A scattergram of PC1 plotted against PC2 showed that Myrtaceous trees (circles) segregated from non-Myrtaceous (squares) (the few overlapping points represent the non-Myrtaceous Buchanania obovata trees). Separation was largely in the direction of leaf thickness, but Myrtaceous trees also had higher LMA and lower Amass, Nmass and Pmass.

Differences among species

Leaf attributes for all species are listed in Table 4. Differences among species, among individual trees within species, and among species within habitats, were highly significant for all measured leaf attributes (P < 0·0001).

Table 4.  Species means for mass-based (Amass) and area-based (Aarea) assimilation, saturated water content (H2O/DW), area per leaf, leaf mass per area (LMA), leaf thickness, density, area-based (Chlarea) and mass-based (Chlmass) chlorophyll content, mass-based nitrogen (Nmass) and phosphorus (Pmass) content, and leaf life span. Values are means of three leaves from each of six trees per species, except for Nmass and Pmass, which are means of one leaf from each of six trees; and life span, which is based on median values for each of six trees (total of 60–190 leaves per species). Estimates of Aarea and LMA for Callitris intratropica are based on total branchlet length and average branchlet thickness. Standard errors were pooled for all species and are applicable to species means. n.d. = not determined
SpeciesAarea (µmol m−2 s−1)Amass (nmol g−1 s−1)H2O/DW (g g−1)Area (cm2)LMA (g m−2)Thickness (mm)Density (kg m−3)Chlarea (mmol m−2)Chlmass (mmol kg−1)Nmass (mg g−1)Pmass (mg g−1)Life span (month)
Dry monsoon forest
F. racemosa16·42502·07 72 660·183610·507·6824·11·74 6·0
F. scobina12·31522·44 39 830·302750·445·3530·11·86 4·8
M. nesophilus13·21751·35 42 760·243110·385·1422·41·47 5·0
M. leucadendra18·71292·31 241500·453360·543·6810·50·82 8·2
T. microcarpa13·31671·52 41 830·194240·485·9122·81·68 7·0
W. pubescens 9·61822·57 27 520·202680·417·9632·41·64 6·8
Open forest
B. megaphyllus10·61081·49304 980·313230·495·0220·21·27 7·0
B. obovata10·6 641·621571650·374500·553·3411·50·7211·0
E. chlorostachys10·91101·16 221000·214770·444·4222·80·81 9·5
E. tetrodonta11·1 621·43 851820·384760·442·47 7·10·33 8·2
P. careya10·61272·08 68 840·233600·404·7618·21·11 7·0
S. suborbiculare11·1 821·67 631380·373780·513·7712·00·63 9·5
T. ferdinandiana 9·4 941·761211010·263970·383·7515·10·90 5·2
C. intratropica 7·8 361·54n.d.220n.d.n.d.n.d.n.d.11·10·69n.d.
C. fraseri 9·41571·93 82 600·163870·305·0717·60·80 3·3
C. foelscheana12·9 590·99 912210·415410·542·47 7·80·37 9·8
M. viridiflora14·1 611·59 242330·564170·602·59 7·70·3710·2
P. careya10·21001·75 641010·254110·444·3516·00·85 8·5
S. eucalyptoides 8·3 662·18 451260·423010·403·2211·60·6411·3
T. ferdinandiana12·91081·601661210·284390·403·2915·20·81 4·8
X. paradoxus11·7 851·521021370·363870·402·8911·50·4910·5
A. auriculiformis14·41361·68 401050·254150·625·9724·81·2010·0
L. lactifluus 8·4 681·37 541210·294210·484·11 9·60·49 8·7
M. viridiflora14·1 641·37 322250·484660·673·09 6·10·28 9·8
Standard error 0·64  6·70·059  9·0  4·60·008  9·90·0140·282 0·770·074 0·73

Differences among habitats

Differences among habitats, based on species means, were significant for Amass, Chlmass and Pmass, with differences in LMA, area per leaf, leaf density and Nmass marginally significant (Table 5). Except for area per leaf, the dry monsoon forest was consistently different from the three open-canopy habitats; it had the largest Amass, Aarea, saturated leaf water content, Chlmass, Nmass and Pmass, and least LMA and leaf density (Fig. 3). Differences between habitats, based on tree means, were highly significant (P < 0·0001) for all attributes (not shown).

Table 5.  Summary of results of anovas to determine the significance of differences between habitats, leaf phenological guilds and Myrtaceous vs. non-Myrtaceous species (family). Intraspecific habitat differences are also indicated for the three species (P. careya, T. ferdinandiana and M. viridiflora) found in more than one habitat. Values given are the probability (P) that indicated leaf attributes do not differ. Callitris intratropica was excluded from these analyses. *, P < 0·05; **, P < 0·01; ***, P < 0·001; M, 0·05 < P < 0·10; NS, P > 0·10
All habitats (all species)Woodland vs open forest (P. careya, T. ferdinandiana)Woodland vs swamp (M. viridiflora)PhenologyFamily
Area leaf−1MMMMNS
Life spanNSNSNS***
Figure 3.

Values for Amass, LMA, leaf density, Nmass, Pmass and Chlmass in the four habitats (swamp, open forest, woodland and dry monsoon forest: Sw, OF, W and DMF, respectively). Boxes indicate median, 25th and 75th percentile values, with error bars showing 10th and 90th percentile values and solid circles indicating outliers. Differences between habitats were significant, or marginally so, for these attributes because species from the dry monsoon forest were different from the others.

Relationships between pairs of attributes often differed among habitats. Slopes differed among habitats only in the case of LMA vs. Nmass (Fig. 1f), but intercepts were different for several pairs of attributes, especially those involving Amass, Chlmass and Pmass (Table 6). This was because Amass, Chlmass and Pmass were generally higher in dry monsoon forest species than those from other habitats for any given value of the other attributes (Fig. 1a–c,h).

Table 6.  Significant habitat effects indicated by ancova, with habitat as a covariate. Slopes were heterogeneous only for LMA vs Nmass, indicated by X. *, P < 0·05; **, P < 0·01; ***, P < 0·001
Independent variableDependent variable
AmassH2OLMAThickDensityChlmassNmassPmassLife span
Amass   *     
H2O/DW*      * 
LMA**    ** * 
Thick***    ** ** 
Chlmass* *  *   
Nmass* X  * * 
Pmass     *   
Life span**    * * 

Intra-specific differences among habitats

There were some significant differences between habitats for all three species common to more than one habitat (Table 5). Assimilation, on either area or mass basis, did not differ significantly between habitats for any of the three species. Leaves were thicker, with a higher LMA and lower Chlmass, in the woodland than the open forest or swamp for all species (Table 4). Leaves of P. careya and T. ferdinandiana were more dense, and had lower saturated water content and Pmass, in the woodland than in the open forest. By contrast, M. viridiflora leaves were less dense, and had larger saturated water content and Nmass and Pmass, in the woodland than in the swamp.

Differences among phenological guilds

Evergreen species had larger Aarea, LMA, leaf thickness, Chlarea and life span, and smaller area per leaf, Nmass and Pmass, than deciduous species (Table 5; Fig. 4). Values for semi-deciduous species were intermediate, and differences in Amass, Chlmass and saturated water content were not significant.

Figure 4.

There were significant differences between evergreen (E), semi-deciduous (SD) and deciduous (D) phenological guilds in Aarea (but not for Amass), LMA, leaf thickness, Nmass, Pmass and leaf life span. Boxes indicate median, 25th and 75th percentile values, with error bars showing 10th and 90th percentile values and solid circles indicating outliers.

Myrtaceousvsnon-myrtaceous species

Non-Myrtaceous species had larger Amass, Chlmass, Nmass and Pmass, and smaller LMA, leaf thickness and life span, than Myrtaceous species (Table 5; Fig. 5). However, Aarea was similar for the two groups, as were area per leaf, saturated water content and density.

Figure 5.

There were significant differences between non-Myrtaceous (non-Myrt) and Myrtaceous (Myrt) species in Amass, LMA, leaf thickness, life span, Nmass and Pmass. Boxes indicate median, 25th and 75th percentile values, with error bars showing 10th and 90th percentile values and solid circles indicating outliers.


Global context

Our study provides the most extensive comparat-ive data set for seasonally dry tropical areas, an area that has been largely neglected by ecophysiological researchers (Niinemets 2001). The results highlight the importance of including studies from a wide range of biomes, and habitats within biomes, in order to derive globally applicable relationships between ecosystem function and leaf attributes. While our mean values for Amass, LMA, leaf density, leaf thickness and water content are similar to those reported in the global meta-analysis of Niinemets (1999), we found that Amass and Aarea were generally larger than found by Reich et al. (1999), who studied shrubs and herbs as well as trees, and sampled from biomes ranging from alpine tundra to tropical rainforest. Average leaf N, on either an area or a mass basis, was similar to that reported by Niinemets (1999) and Reich et al. (1999). This was unexpected, given the generally very infertile soils in northern Australia. Niinemets (1999) also derived a linear relationship between leaf thickness and annual average temperature, and attributed this to higher photosynthetic rates and shorter payback times at near-optimal temperatures. Leaf thickness measured in our study is only half what would be expected from this linear relationship; it is also lower in the few other studies where mean annual temperature exceeded 20 °C. It therefore appears that the leaf thickness–temperature relationship levels out, and perhaps even declines, at temperatures above 20 °C. Our leaf density and LMA values are also smaller than would be expected from regressions against precipitation in the three driest months (Niinemets 2001). The Darwin region is more seasonal in terms of rainfall, with both a drier 3-month period (total mean rainfall is 8 mm for June to August inclusive), and a higher maximum monthly rainfall (426 mm in January), than any included in the Niinemets (2001) study. An alternative measure of water availability that incorporates potential evapotranspiration and soil water storage would be more appropriate when such seasonal sites are included (Eamus 2003).

Reich et al. (1999) showed that there were similar slopes, but often different intercepts, in the equations relating pairs of leaf attributes from the six different biomes they studied. Our findings for the four habitats within one biome (tropical savanna) were generally consistent with this, except that the slopes varied with habitat for LMA vs Nmass. For any given value of any attribute except Pmass, Amass was larger in the dry monsoon forest than in the other habitats (Fig. 1). By contrast, differences found by Reich et al. (1999) between biomes were not so consistent; for a given LMA, Amass was highest in desert shrub species, but for a given life span, Amass was highest in humid temperate and tropical forests.

Relationships found in this study among LMA, leaf life span, Nmass, Amass and Aarea are similar to those found by Reich and co-workers (Reich et al. 1992; Reich et al. 1999). As in their studies, correlations with life span, Nmass and LMA were stronger for Amass than Aarea. Niinemets (1999) found that Amass was negatively correlated with LMA because of the negative correlation between Amass and leaf density, while there was little relationship between Amass and leaf thickness. In our study, however, Amass was more strongly correlated with leaf thickness than density (Table 3). The components of LMA, leaf thickness and density were not significantly correlated with each other, as has been found in other studies (Witkowski & Lamont 1991; Niinemets 1999). The strong negative correlation between leaf density and water content can be explained by the observation of Roderick et al. (1999) that, if a leaf has a high liquid content, then it cannot have a large volumetric fraction of dry matter and will thus have a low density.

Leaf P has seldom been included in studies such as ours, which is surprising given its suggested role in characterizing sclerophyllous vegetation (Loveless 1961). Our study found a stronger correlation between Pmass and Amass than any reported by Wright, Reich & Westoby (2001) in southern Australia, and contrasts with that of Tuohy, Prior & Stewart (1991), who found no correlation between these attributes in Zimbabwean tree species. We also found a closer correlation between Nmass and Pmass than found by Wright et al. (2001).

We found that evergreen species had larger LMA, leaf thickness and leaf life span, and smaller Nmass and Pmass than deciduous species, but there was no significant difference in Amass. (Differences in Amass were significant if trees, rather than species, were considered the experimental unit.) Smaller LMA and leaf thickness in deciduous compared with evergreen tree species have been reported in a range of woody species (Castro-Díez, Puyravaud & Cornilessen 2000), but these have generally been accompanied by larger Amass (Chabot & Hicks 1982; Sobrado 1991; Prado & Moraes 1997; Eamus & Prichard 1998). In Eurasian tree species, Nmass and Pmass are smaller in evergreen than in deciduous trees (Chabot & Hicks 1982). The longer life span of evergreen compared with deciduous leaves offsets their higher construction costs (Chabot & Hicks 1982; Sobrado 1991; Eamus & Prichard 1998). Given difficulties in precisely defining phenological categories, it is more useful to classify trees according to leaf life span than leaf phenology (Reich et al. 1992). In this study there was considerable overlap in leaf life span between the phenological guilds, especially between evergreen and semi-deciduous species (Table 4; Fig. 4). Median leaf life span varied between 3 and 12 months, and was shorter on average than in a Malaysian rainforest (Osada et al. 2001), a Mexican cloud forest (Williams-Linera 2000), Venezuelan rainforest (Reich, Walters & Ellsworth 1991), or in five of the six biomes studied by Reich et al. (1999). Our findings are consistent with those of Myers et al. (1998), who reported that leaf life span of the Eucalyptus/Corymbia group appears to be shorter in northern than in southern Australia.

There were significant differences among species in all attributes studied, and these differences were larger when expressed on a mass than on an area basis. For example, area per leaf varied 13-fold, LMA 4·5-fold, and Amass varied sevenfold, while Aarea varied only twofold. These interspecific variations were smaller than found by Reich et al. (1999). Callitris intratropica was included in this study because there have been few physiological measurements made on tropical or Southern Hemisphere conifers. In common with Northern Hemisphere conifers (Ackerly & Reich 1999), it had lower Aarea, Amass and Nmass, and higher LMA than most or all of its broad-leaved counterparts, evergreen or deciduous. In the present study, species differences were strongly associated with leaf phenology and whether they were Myrtaceous or not, and are related to habitat types.

Regional patterns

Variation in leaf attributes found between species within the three non-rainforest habitats was larger than the variation between these habitats. This is surprising given their different physical characteristics. The swamp is inundated for several months a year, and the soil has a high organic content with high moisture availability throughout the dry season [unpublished data, and as described by Kelley (2002) for a similar site]. The open forest has deep, well drained, relatively fertile soils, free of rock. By contrast, woodland soils are stony and infertile, and have poor subsurface drainage. As expected, based on the higher soil N and P at that site, species in the dry monsoon forest had higher mean Amass and Nmass, and lower LMA and leaf density than species in the other habitats. These characteristics, typical of fast-growing species in relatively favourable environments, were reflected in higher tree growth rates, basal area, stem density and canopy cover in this habitat (L.D.P., unpublished results).

Higher levels of soil N and P in Australian monsoon forest compared with adjacent savanna have been reported previously, and are attributed, at least in part, to less frequent fire in the monsoon forest (Bowman 1992). High nutrient content in leaves of dry monsoon forest relative to savanna species has also been recorded (Fensham & Bowman 1995; Schmidt et al. 1998). However, Fensham & Bowman (1995) found that leaves of one sclerophyllous monsoon forest species had similar nutrient concentrations to leaves of savanna species, suggesting that differences between the habitats were related to intrinsic leaf properties as well as soil fertility. This is consistent with our study, in which M. leucadendra had leaf N and P concentrations less than half those of the other species in the dry monsoon forest.

Differences between Myrtaceous and non-Myrtaceous species were generally larger than differences between habitats or leaf phenological guilds, and were clearly separated by principal components analysis (Fig. 2b; Table 5). Myrtaceous species were typically sclerophyllous, with larger LMA, leaf thickness and life span, and smaller Amass, Chlmass, Nmass and Pmass than non-Myrtaceous species. Only one non-Myrtaceous species, Buchanania obovata, was comparable to Myrtaceous species.

Myrtaceous species do not comprise the majority of species in north Australian savannas, but they often contribute the largest component of woody biomass in frequently burnt habitats (O’Grady et al. 2000). Myrtaceous leaves are characteristically sclerophyllous, and sclerophyllous leaves are usually associated with slow relative growth rates (Lambers & Poorter 1992; Wright & Westoby 2000). Therefore other factor(s) must confer a competitive advantage to Myrtaceous species in these tropical savanna environments. One probable factor is fire tolerance, as extensive, frequent fires are a feature of Australian savannas (Russell-Smith et al. 1997). Williams et al. (1999) found that deciduous non-eucalypt species comprised the group most susceptible to fire in Australian savanna. Based on their data, Myrtaceous and non-Myrtaceous species had similar plant survival rates after a single, high-intensity fire (86 vs 82%), but stem survival was higher in the Myrtaceous species (43 vs 15%). Our findings help explain the success of non-Myrtaceous species in areas relatively protected from fire, such as those where dry monsoon forests are found (Bowman 1992).

In terms of their leaf attributes, the four habitats dichotomized into the closed canopy (dry monsoon forest) and the open canopy habitats (woodland, open forest and swamp). These differences in leaf traits could be largely explained by differences between the mesophytic, often deciduous, species from non-Myrtaceous genera that prevailed in the dry monsoon forest, and the sclerophyllous, Myrtaceous species common in the other three habitats. Thus habitat differences were linked to the phylogeny and leaf phenology of the species found there. In a broad Australian context, there is an often sharp ecological and floristic dichotomy between quintessentially Australian vegetation (dominated by Eucalyptus and Acacia) and the rainforests that occur in the most humid climates and in fire-protected refugia (Bowman 2000). This study has shown that in northern Australia the dichotomy between rainforest and non-rainforest vegetation apparently extends to, and can be inferred from, leaf attributes.


We wish to thank staff of the Parks and Wildlife Commission of the NT for allowing access to the three study sites, and especially to Bill Panton for his enthusiasm and information about the Leanyer rainforest. Dr Keith McGuinness advised on statistical procedures. Thanks also go to Françoise Foti for meticulous N and P analyses. This work was funded by Australian Research Council Large Grant A00001382.