Isotopes Environ. Health Stud., Vol. 39, No. 2, June 2003, pp. 141–158

NUTRIENT CYCLING RESPONSES TO FIREFREQUENCY IN THE KRUGER NATIONAL PARK

(SOUTH AFRICA) AS INDICATED BY STABLEISOTOPE ANALYSIS

J. N. ARANIBARa,*, S. A. MACKOa, I. C. ANDERSONb, A. L. F. POTGIETERc,

R. SOWRYc and H. H. SHUGARTa

aDepartment of Environmental Sciences, University of Virginia, USA; bVirginia Institute of MarineSciences, College of William and Mary, USA; cKruger National Park, Skukuza, South Africa

(Received 2 July 2002; In final form 17 November 2002)

Fires, which are an intrinsic feature of southern African ecosystems, produce biogenic and pyrogenic losses ofnitrogen (N) from plants and soils. Because of the long history of fires in these savannas, it was hypothesizedthat N2 fixation by legumes balances the N losses caused by fires. In this study, the N2 fixation activity of woodylegumes was estimated by analyzing foliar d15N and proportional basal area of N2 fixing species alongexperimental fire gradients in the Kruger National Park (South Africa). In addition, soil carbon (C) and N pools,foliar phosphorus (P) and gross N mineralization and nitrification rates were measured, to indicate the effects offires on nutrient stocks and the possible N cycling processes modified by fires. Although observations ofincreased soil C=N and mineralization rates in frequently burned plots support previous reports of N lossescaused by fires, soil %N did not decrease with increasing fire frequency (except in 1 plot), suggesting that Nlosses are replenished in burned areas. However, relative abundance and N2 fixation of woody legumes decreasedwith fire frequency in two of the three fire gradients analyzed, suggesting that woody legume N2 fixation is notthe mechanism that balances N losses. The relatively constant %N along fire gradients suggests that theseecosystems have other mechanisms to balance the N lost by fires, which could include inputs by atmosphericdeposition and N2 fixation by forbs, grasses and soil cyanobacteria.

Soil isotopic signatures have been previously used to infer patterns of fire history. However, the lack of arelationship between soil d15N and fire frequency found in this study indicates that the effects of fires onecosystem d15N are unpredictable. Similar soil d15N along fire gradients may reflect the contrasting effects ofincreased N gaseous emissions (which increases d15N) and N2 fixation other than that associated with woodylegumes (which lowers d15N) on isotopic signatures.

Keywords: Carbon 13; Legumes; Mineralization; Natural variations; Nitrogen 15; Nitrogen fixation; Savannas; Soilcarbon; Soil nitrogen

INTRODUCTION

Regular fires are one of the characteristic features of tropical savannas. For the past tens of

thousands of years, humans have set fires for hunting, improving the quality of grazing,

* Corresponding author. Department of Environmental Sciences, University of Virginia, Charlottesville, VA22903, USA; E-mail: jna3h@virginia.edu

ISSN 1025-6016 print; ISSN 1477-2639 online # 2003 Taylor & Francis LtdDOI: 10.1080=1025601031000096736

preparing the land for cultivation, and controlling the spread of woody plants [1]. Isotopic

and sedimentological studies suggest that anthropogenic fires have been present in sub-

Saharan Africa for the past million years [2]. The tendency for woody plant density to

increase when fires are excluded leads to the view of savannas as subclimax to woodland

or forest. However, the long history of fire in Africa and the numerous adaptations of plants

to survive fires, suggest that fires are modifiers of savannas, rather than the primary determi-

nant of savanna distribution [3]. The direct effects of fire or fire suppression on soils remain

unclear because most soil changes are related to vegetation changes, which may respond to

drivers other than fire [4].

The use of fire as a management tool in savannas has provoked controversy. In the Kruger

National Park (South Africa), early burning policies (before 1960s) ranged from prescribed

burning to fire suppression, without experimental knowledge of the effects of burning on the

ecosystem. A long-term fire research experiment was laid out in the park in the mid 1950s, to

investigate the effects of season and frequency of burning on the vegetation condition of the

main units of the park [5]. This experiment provides us with the opportunity to analyze the

effects of fires on nutrient cycling, without the confounding influences of climate, soil

texture, and anthropogenic disturbances such as cattle grazing, cultivation, fertilization and

logging.

Losses of N by volatilization occur during fires and can lead to substantial net nutrient loss.

About 50% of the N in biomass fuel can be released as N2, which is permanently lost from the

system [6]. Some of the volatilized N may be replaced by nitrogen inputs in wet or dry deposi-

tion or by biogenic nitrogen fixation. Ammonium or its oxidation products, nitrite and nitrate,

and phosphate, which accumulate in ash as a result of decomposition of organic material dur-

ing burning, may be reimmobilized by plant and microbial processes [7]. However, the

enhanced mineral N availability after fires also favors microbial activity (nitrification and deni-

trification) that mediates gaseous N losses from the ecosystem. Biogenic N losses are espe-

cially enhanced after wetting of burned surfaces, reaching values of 77 ng of N m�2 s�1 in

southern African savannas [9–12]. Biomass burning can enhance microbial N emission

rates by 5–10 fold greater than rates observed as a result of rainfall alone [13]. However,

because of the long history of fires on savannas, we hypothesize that savanna ecosystems

have developed mechanisms to enhance the recovery of N lost during and after fires. In addi-

tion to N inputs by atmospheric deposition, N2 fixation by free-living or symbiotic bacteria is

the most likely source of nitrogen to burned savannas. Cyanobacteria and other free-living

microorganisms common in semi-arid areas are known to fix N2 [14, 15]. Vascular plants

from different groups, such as grasses and legumes, may form associations with N2 fixing bac-

teria. Trees from the leguminous family (Fabaceae), where N2 fixation is often present, are

common and some times dominant in these savannas, although they are not always found

to fix N2. The incidence of legumes is often higher in frequently burned plots [1]. It has

also been suggested that legumes increase nodulation (a requirement for N2 fixation) in dis-

turbed environments, and non-symbiotic N2 fixation increases following fires [8]. The

enhanced available phosphorus found in ash after burning is required for the N2 fixing process

and may facilitate N2 fixation [16, 17], but the enhanced mineral N availability in ash may sup-

press it [4]. These two potential responses of legume N2 fixation to fires may have opposite

effects on the N pools of savannas: N replenishment by N2 fixation or N depletion by subse-

quent N losses in the absence of a N replenishment mechanism. Foliar d15N has been used to

assess N2 fixation, as it is generally more enriched in plants whose only N source is soil N,

rather than atmospheric N2 [18, 19]. In this study, we investigate whether soil N decreases

with fire frequency as a consequence of net N losses enhanced by fires. We also analyze

whether N2 fixation by woody legumes increases in frequently burned savannas and wood-

lands, as a mechanism to recover part of the N lost during and after fires.

142 J. N. ARANIBAR et al.

Stable isotope signatures of soil and plant nitrogen indicate patterns of N cycling, integrat-

ing processes occurring in the ecosystem over large temporal scales [20]. Biogenic gaseous

losses of N, such as those enhanced after burning and subsequent wetting, tend to enrich the

system in 15N because of the high isotopic fractionation during the microbial processes

involved (nitrification and denitrification) [21, 22]. Inputs of 14N by biogenic N2 fixation

tend to dilute the15N of the system, decreasing 15N natural abundance (d15N).

The role of fires on ecosystem d15N is not well understood, but fires have been reported to

explain both, depletion and enrichment of leaves in 15N [23, 24]. Foliar d15N within a parti-

cular rainfall regime has even been suggested to indicate historical fire frequencies [25]. In

Kapalga savannas (Australia), however, burned and unburned plots had similar d15N values

[26]. This unclear role of fires on d15N may result from the confounding effects of enhanced

gaseous losses (which enrich the system in 15N) and biologic N2 fixation (which lowers eco-

system d15N) on ecosystem isotopic signatures. In this study, we analyzed soil d15N along

fire gradients located within a similar climate, vegetation and soil, to isolate the effect of

fires on N isotopic signatures. In addition, we investigated N2 fixation by woody legumes

(with foliar d15N and vegetation surveys), foliar P, soil C and N, and soil microbial activity

(gross mineralization and nitrification rates) to infer processes causing the observed isotopic

patterns.

We expected either similar soil N along fire gradients (if N losses caused by fires are

replenished by atmospheric deposition or N2 fixation), or lower soil N in frequently burned

areas (if N losses are higher than N inputs). Woody legume N2 fixation, indicated by foliar

d15N and relative abundance of N2 fixing species, could also show opposite trends with

increasing fire frequency: enhancement of N2 fixation due to increased P availability and

lower total N stocks (if lower soil N is observed in burned areas), or inhibition of N2 fixation

due to enhanced N availability in ash after fires. Gross mineralization and nitrification rates

were hypothesized to increase with fire frequency because of the higher availability of sub-

strates (simple organic N compounds and ammonium) promoted by fires. Finally, soil d15N

could either increase, decrease or remain constant with fire frequency, depending on the

isotopic signatures of N inputs, isotopic fractionation during N losses and inputs, and

whether or not N losses are balanced by N inputs.

METHODS

Study Sites

Between 1926 and 1954, indiscriminate burning was applied to the Kruger National Park to

provide green grazing for wildlife [27]. The Experimental Burning Plot Trial was established

in 1954 in the four most important vegetation types of the Kruger National Park [5], to deter-

mine the effect of season and frequency of burning on the vegetation condition of the park.

The trial was laid out as a randomized block design with four replicates for each vegetation

type. Each of the four replicates had 12 treatments, including different seasons and frequen-

cies of burning, with a total of 292 plots, measuring 360 m�180 m (approximately 7 ha) each

[5, 28]. Plots are separated from each other by wide firebreaks [29].

In this study, experimental plots of only two vegetation types were analyzed: the

Combretum sp.=Terminalia sericea woodland initiated in 1954, and the Sclerocarya

birrea=Acacia nigrescens savanna, initiated in 1958. Only two replicates of the woodland

(Skukuza and Nwaswitshaka) and one replicate of the savanna (Satara) were studied due

to time constrains. The two vegetation types are located within the same rainfall regime

(about 550 mm mean annual precipitation), but have different soil substrates: sandy granitic

ISOTOPES AND FIRES IN SOUTH AFRICA 143

soils in the woodland (Skukuza and Nwaswitshaka), and clay basaltic soils in the savanna

(Satara) [30]. The topography of the plots located in the woodland (Skukuza and

Nwaswitshaka) change from convex to concave, affecting soil texture and moisture.

Samples were taken from the top, middle and bottom of each plot, to reduce the effects of

slope on soil and plant variables along the fire gradients. Because the topography was similar

for all the woodland plots, the differences in the variables measured should reflect the fire

treatments instead of topography. Another important difference between the two vegetation

types is the co-dominance of potentially N2 fixing trees in the savanna vegetation (Acacia

nigrescens), and the dominance of non-N2 fixing trees in the woodland (Combretum sp.

and Terminalia sericea).

The plots selected included controls (where fire has been suppressed since the establish-

ment of the trials), triennial (burned every three years in winter) and annual (burned every

year in winter) treatments. Winter fires are very hot but fast when compared to summer

fires, and occur when the vegetation is dormant. The vegetation is indeed less vulnerable

to fires in winter than in summer, and the influence of fires would be lower [29]. A sexennial

burn (burned every 6 years in spring), initiated in 1984, was only available in the Satara repli-

cate, but it was also sampled to analyze a low fire frequency treatment. Due to time availabil-

ity and accessibility to the sites, the number of samples differed for the different plots. Table I

indicates the number of samples collected, and experiments performed at each of the plots.

During the wet season of year 2000, when the samples were collected, southern Africa

experienced anomalously heavy, at times over two standard deviations above normal rainfall

(M. Rouault, pers.com.) [31]. The plots in Skukuza and Nwaswitshaka had been flooded a

month prior to this study; it is likely that flooding affected soil and plant nutrient cycling

by modifying soil oxygen availability and microbial activity: denitrification is enhanced

under anaerobic conditions, while nitrification is suppressed [32].

Surface soils were collected in all plots to a depth of 5 cm, with equal numbers of samples

under and between tree=shrub canopies for each plot. In Skukuza and Nwaswitshaka, soils

were collected from the top, middle and bottom of the plots, to average out the variability

caused by differences in slope within each plot. The soils were air dried in the field, oven

dried at 60 �C in the laboratory to constant weight, sieved (2 mm) and pulverized. After treat-

ment with HCl (30%) to remove carbonates, the soils were analyzed for %C, %N, d13C and

d15N with an Optima isotope ratio mass spectrometer coupled to an elemental analyzer, with

an overall precision of 0.3‰. The isotopic data are reported relative to a standard, and

expressed in d notation as

dsample(‰) ¼Rsample

Rstandard � 1

� ��1000

where dsample represents either d13C or d15N, and R is the molar ratio of the heavier to the

lighter isotope for the standard or sample. The standards, defined to be 0.0‰, are atmo-

spheric N2 for N and PeeDee Belemnite for C [33].

Mature leaves were collected from woody plants and forbs, legumes and dominant non-

legume species from all the plots. All the leaves sampled from each individual plant (5 to 10

leaves=plant depending on the leaf size) were combined into one sample, air dried in the field,

oven dried in the laboratory at 60 �C to constant weight, and ground in a Wiley mill using a

40 mesh (425 mm) sieve. The ground leaves were analyzed for %C, %N, d13C and d15N with

the same isotope ratio mass spectrometer used for the soil samples. Additionally, foliar P

concentrations were determined from the ground leaves with an Alpkem ‘‘flow solution’’

144 J. N. ARANIBAR et al.

TA

BL

EI

Exper

imen

tal

Des

ign,

Incl

udin

gN

um

ber

of

Sam

ple

sC

oll

ecte

dan

dE

xper

imen

tsP

erfo

rmed

atE

ach

One

of

the

Plo

ts.

Sit

ea

nd

trea

tmen

tN

um

ber

of

soil

sam

ple

sN

um

ber

of

pla

nt

sam

ple

s

Gro

ssm

iner

ali

zati

on

rate

s,n

um

ber

of

incu

ba

tion

s

Gro

ssn

itri

fica

tion

rate

s,n

um

ber

of

incu

ba

tion

sN

um

ber

of

veg

etati

on

tra

nse

cts

Sk

uk

uza

contr

ol

92

83

32

Skukuza

trie

nnia

lburn

11

17

32

Sk

uk

uza

annu

alb

urn

82

03

32

Nw

asw

itsh

aka

con

tro

l6

13

Nw

asw

itsh

aka

trie

nnia

lburn

61

02

Nw

asw

itsh

aka

annu

alb

urn

71

32

Sat

ara

contr

ol

61

12

Sat

ara

sixen

nia

lb

urn

61

6S

atar

atr

iennia

lburn

78

2S

atar

aan

nu

alb

urn

61

02

Aca

cia

savan

na

(19

99

)1

4C

om

bre

tum

savan

na

(19

99

)9

ISOTOPES AND FIRES IN SOUTH AFRICA 145

autoanalyzer, after Kjeldahl digestion and ascorbic acid reduction of a phosphomolybdate

complex [34].

Nitrogen fixation activity by legumes was indicated by foliar d15N and taxonomic classi-

fication. Legume species with lower d15N than non-legumes of the same site were considered

to show indications of N2 fixation. It is not possible to quantify N2 fixation with this method

because non-N2 fixing plants may absorb N from different sources (soil organic N, ammo-

nium and nitrate) [35], and even absorb more than one N form at different times during

the same season [36]. In addition, foliar d15N may be increased if plants reduce nitrate in

their stems [37, 38]. Although the selection of a reference plant to quantify N2 fixation

based on field d15N measurements is invalid, foliar d15N of legumes can indicate whether

part or all the leaf N was recently derived from the atmosphere. This and estimates of relative

abundance of suspected N2 fixing plants (basal area, biomass, or other vegetation surveys)

provide an insight into the importance of N2 fixation across different sites [39].

The botanical composition and structure of the woody layer (trees and shrubs) were sur-

veyed using a belt transect method [40]. Two transects (2 m� 300 m) were established across

each of the two diagonals of the plots. Basal diameter of stems of all rooted woody plants

within the transects were measured at ground level and converted to basal area. In the

case of multi-stemmed plants the basal diameter was recorded as the distance between the

two outermost stems [41]. The relative importance of N2 fixing plants was estimated from

the proportional basal area of species with indications of N2 fixation (based on d15N) relative

to the total basal area of all woody species in each plot.

Gross N mineralization and nitrification rates were estimated using the 15N-NHþ4

(ammonium) and 15N-NO�3 (nitrate) pool dilution techniques respectively [42, 43]. Both

mineralization and nitrification rates were determined under and between woody plants at

the Skukuza control, triennially and annually burned plots; however, because of limited

resources nitrification could not be measured in the triennially burned plot. The top 5 cm

of soil were sampled (n¼ 6) under and between canopies, composited and coarse roots

were removed. The composited sample was weighed into 15 whirlpak bags (20 g of soil

each). Four ml of 15N-NHþ4 (0.4 mM and 30 15N atom%) were added to 6 bags for determi-

nation of gross N mineralization, and four ml of 15N-NO�3 (0.5 mM and 30 15N atom%) were

added to 6 bags for measurement of nitrification. The amount of solution added was meant to

wet the soils to field capacity, but the antecedent moisture conditions caused the labeled solu-

tion to wet the soils above field capacity. The bags were incubated under about 5 cm of soil

with 3 replicates of each process incubated for 0 and 48 hours. The remaining 3 bags were

kept for gravimetric determination of soil moisture by oven drying at 60 �C to constant

weight. At the end of the incubations 60 ml of KCl (2M) were added to each of the bags,

which were then shaken, allowed to settle and filter sterilized (with 0.45 mm, Gelman

supor filters) into sterile whirlpak bags. The KCl extracts were frozen until analysis of nutri-

ents and atom %15N. Ammonium was determined colorimetrically by an automated indophe-

nol method, and NO�3 þ NO�

2 by copperized cadmium reduction in combination with

diazotization. All were analyzed using an Alpkem ‘‘Flow Solution’’ autoanalyzer, equipped

with a model 510 spectrophotometer. The soil extracts (50 ml containing approximately

30 mg N) were placed in sterile, disposable plastic cups and MgO was added. Ammonium

was trapped by diffusion onto KHSO4-soaked filters for a six-day period as described by

Brooks et al. [44] and the filters sent to the University of California, Davis, for analysis of

atom %15N using a CHN analyzer coupled to an isotope ratio mass spectrometer.

Samples for analysis of atom %15N in NO�3 (nitrification study) were prepared by a two-

step procedure. MgO was added to KCl extracts (50 ml containing approximately 30 mg N)

in sterile, disposable plastic cups. The cups were left open for 48 hours to allow any NHþ4

present to diffuse out of the sample. Any NO�3 present in the sample was then converted

146 J. N. ARANIBAR et al.

to NHþ4 by addition of Devarda’s alloy. The NHþ

4 thus produced was trapped and analyzed by

mass spectrometry as described above. Gross nitrogen mineralization and nitrification rates

were determined from measured changes in atom %15N and concentrations of NHþ4 and

NO�3 using the model of Wessel and Tietema [45]. The gross rates are reported as mg of

N (either ammonium or nitrate) produced m�2 day�1 from the surface to 5 cm depth.

Differences of %C, %N, d13C and d15N among plots, functional types and sites were ana-

lyzed with student’s t-tests. Relations between relative abundance of N2 fixers and soil d15N,

and between foliar P and d15N were analyzed by linear regressions, using the coefficient of

determination (r2). Results were considered significant at a p< 0.05. Standard errors of the N

mineralization and nitrification rates were estimated by propagation of errors [46], calculated

with Mathematica 4.2 (Wolfram Research, Inc.).

RESULTS AND DISCUSSION

Soil Carbon and Nitrogen

Soil organic carbon and total N in Satara were much higher than in the Combretum wood-

lands (Skukuza and Nwaswitshaka) ( p< 0.0001, Figs. 1 and 2), perhaps because of the

higher clay content (25–50%) of Satara soils and higher leaching of the other sites [30].

Soil organic carbon (SOC) was not significantly different among plots with different fire

treatments, although there was a trend of decreasing C with fire frequency in the Skukuza

replicate (Fig. 1). Soil total nitrogen was significantly higher (p< 0.05) in the control than

in the annual and triennial treatments of the Skukuza replicate, but it was not significantly

different among the fire treatments at the Nwashwitshaka and Satara sites (Fig. 2). Indeed,

the soil C and N data do not indicate a considerable effect of fire frequency on the concen-

tration of these soil nutrients. Soil C=N, however, increased in the more frequently burned

plots (significant for Skukuza and Satara, Fig. 3), indicating higher losses or lower inputs

of N relative to carbon as a function of burn frequency. Higher SOC, total N, microbial

FIGURE 1 Soil organic carbon (%C) along fire frequency gradients from three replicates of the burningexperiments of the KNP: Skukuza (triangles), Nwaswitshaka (circles) and Satara (squares). The ‘x’ axis indicates firefrequency: C¼ control; 6, 3 and 1 correspond to sixennial, triennial and annual fire treatments, respectively. Barsdenote standard errors of the mean. Different letters indicate significant differences ( p< 0.0001).

ISOTOPES AND FIRES IN SOUTH AFRICA 147

biomass, and surface litter have been reported for control plots compared to burned plots of

another replicate of the Sclerocarya birrea=Acacia nigresces savanna, the Nwanedzi plots

[47]. Other previous studies showed lower C contents in burned than unburned areas, but

also on some occasions the opposite pattern due to charcoal accumulation after fires [1].

The effects of fire on soil carbon and nitrogen contents, thus, are often not clearcut because

losses may be offset by inputs of carbon from charcoal accumulation and of nitrogen from

atmospheric deposition or N2 fixation.

FIGURE 2 Total soil N (%N) along fire frequency gradients in Skukuza (triangles), Nwaswitshaka (circles) andSatara (squares). The ‘x’ axis indicates fire frequency: C¼ control; 6, 3 and 1 correspond to sixennial, triennial andannual fire treatments, respectively. Bars denote standard errors of the mean. Different letters indicate significantdifferences ( p< 0.0001).

FIGURE 3 Soil C=N along fire frequency gradients in Skukuza (triangles), Nwaswitshaka (circles) and Satara(squares). The ‘x’ axis indicates fire frequency: C¼ control; 6, 3 and 1 correspond to sixennial, triennial and annualfire treatments, respectively. Significant differences were calculated for the plots of each site. Different letters inSkukuza and Nwaswitshaka indicate significant differences ( p< 0.05). One similar letter in Satara indicates non-significant differences (the sixennial treatment is significantly different from the annual treatment, but the triennialtreatment was not significantly different from annual and sixennial treatments).

148 J. N. ARANIBAR et al.

Spatial heterogeneity within each plot may also affect total C and N stocks, if random sam-

ples are takes without differentiating between soils located under and between plant canopies.

In this study, soils under woody plant canopies had higher soil C and N content than soils

between canopies, although the differences were significant ( p< 0.05) only in Skukuza

for C and N, and in Satara for N (data not shown). It is possible that previously reported

effects of fires on soil C and N are partially due to the indirect and differential effects of

fires on tree and grass cover, instead of the direct effects on C and N emissions. Because

the sampling in this study included similar numbers of soils under and between plant

canopies for each plot, the effects of relative tree-grass cover on soil nutrients was less likely

to confound the reported effects of fires.

Although almost 50 years of fire suppression did not result in significant increases in soil C

and N stocks (except for N in the control plot of Skukuza), relative to those in frequently

burned areas, the soil C=N ratio did increase as a function of burn frequency. The impacts

of fire frequency on soil elemental composition may be somewhat mitigated by the timing

of burn events. Winter fires, as performed in the annual and triennial treatments and common

in savannas, are not as detrimental to vegetation as are summer fires [29]. These ecosystems,

do indeed, seem to be resilient to the effects of winter fires further supporting the idea of fire

as an intrinsic feature of the savanna landscape. A possible mechanism of N replenishment is

N2 fixation by woody legumes, which was also analyzed in this study.

Soil Microbial Activity

Gross mineralization rates were higher under than between woody canopies in the Skukuza

plots. Comparing the rates for under and within canopy soils in different plots, there was a

trend of increasing mineralization with higher fire frequency (Fig. 4A). Gross nitrification

rates were lower than mineralization rates, and did not show a clear pattern for the two

plots analyzed (Fig. 4B). The high soil moisture of the soils during the floods of 2000

may have created anaerobic conditions, which inhibited nitrification and promoted denitrifi-

cation [32]. Nitrification rates indeed may have been more affected by the anomalous floods

than were the other variables analyzed (soil and plant %C, %N, isotopic signatures and gross

mineralization rates), reflecting local antecedent soil moisture conditions rather than the

effects of fire frequency. Gross mineralization rates instead, provide a better insight into

soil N dynamics along the fire gradients, indicating higher rates, and N availability, with

increasing fire frequency. This enhanced mineralization may provide more substrate for

coupled nitrification–denitrification, favoring gaseous N losses that will eventually result

in net N depletion if not balanced by N2 fixation. In addition, higher mineral N availability

to plants may enhance uptake by potentially N2 fixing legumes, inhibiting N2 fixation.

N2 Fixation

Foliar d15N signatures were significantly different ( p< 0.0001) for forb legumes, tree

legumes, and non-legume plants, with average values of 0.2, 2.5 and 4.5‰ respectively.

The C:N ratios followed a similar pattern (forbs< tree legumes< non-legumes, p¼ 0.002).

Low foliar d15N suggests the occurrence of N2 fixation in legumes [18, 19]. Forb legumes

seem to derive most of their N from N2 fixation. The foliar 15N enrichment in perennial

tree legumes may also reflect the isotopic signatures of soil N sources, absorbed either during

the season of study or in previous years, and which are commonly retranslocated before

senescence and utilized in subsequent growing seasons [48, 49]. However, not all the

legumes sampled had lower d15N than non-legume plants; these few species, which were

not considered N2 fixers, included Peltophorum africanum, Lonchocarpus capassa, and

ISOTOPES AND FIRES IN SOUTH AFRICA 149

Cassia sp. (Tab. II). Cassia sp. does not nodulate [50] and L. capassa of the Kalahari sands

similarly did not show any indication of N2 fixation [39]. The d15N of Acacia tortilis does not

clearly indicate N2 fixation, but it is possible that the signature of atmospheric N2 (0‰) was

masked by the uptake of other soil N sources or nitrate reduction in stems, which would

increase foliar d15N [37, 38]. However, because only one individual of this species was

sampled and did not show strong indication of N2 fixation, Acacia tortilis was not assumed

to fix N2. Although use of 15N signatures to indicate N2 fixation is subject to some error, such

FIGURE 4B Gross nitrification rates (mg of NO�3 -N produced m�2 day�1, up to 5 cm depth) for soils between

(horizontal lines) and under tree canopies (crossed lines) in the Skukuza plots. In the ‘‘x’’ axis, 3 and 1 mean triennialand annual fire treatments, respectively.

FIGURE 4A Gross mineralization rates (mg of NHþ4 -N produced m�2 day�1, up to 5 cm depth) for soils between

(horizontal lines) and under tree canopies (crossed lines) in the Skukuza plots. In the ‘‘x’’ axis, 3 and 1 mean triennialand annual fire treatments, respectively.

150 J. N. ARANIBAR et al.

as the effect of N sources other than N2 on foliar d15N, d15N integrates N processes over long

time scales, provides insights into the relative importance of N2 fixation in the field, and

is especially useful for comparative purposes [39].

The variation of N2 fixation activity with different fire frequencies was estimated by ana-

lyzing the relative abundance of plants with indications of N2 fixation (basal area of N

fixers=total basal area), and foliar d15N of these species across the plots (Tab. III). The

vegetation surveys showed decreasing proportions of N2 fixing species with increasing fire

frequency for the Skukuza and Satara replicates (Tab. III). Moreover, the foliar d15N of N2

fixing trees=shrubs from Skukuza and Satara suggest that N2 fixation was lower in the

annually burned than in the control plots (higher d15N in annual treatments, Tab. III). The

lower N2 fixation observed in annually burned plots compared to control plots may result

TABLE II Foliar d15N and C=N of Plants from the KNP Fire Plots. ‘‘se’’ Denotes Standard Error for n> 3. Plantsare Grouped According to the Potential N2 Fixing Status, Inferred by Lower d15N in Certain Legumes than in OtherPlants.

Species Family=subfamily d15N, ‰ se C=N se

Forbs with indications of N2 fixationChamaecrista absus Caesalpinoideae �0.5 0.2 14.8 2.8Chamaecrista mimosoides Caesalpinoideae 0.3 0.6 14.7 3.8Elephantorrhiza burkei Mimmosoideae 0.8 15.7Indigophera sp. Papilionoideae 0.3 0.3 12.3 0.6Rhinchosia totta 0.5 16.1Stylosanthes fruticosa Papilionoideae 0.1Tephrosia sp. Papilionoideae 0.6 0.3 10.4 0.3Trig. asiatica �0.3 24.4Vigna ungiculata subsp.

stenophyllaCaesalpinoideae �0.8 10.2

Trees with indications of N2 fixationA. exuvialis Mimmosoideae 1.5 0.4 20.1 1.4A. garardii Mimmosoideae 2.5 13.1A. nigrescens Mimmosoideae 2.7 0.4 15.1 0.5A. robusta Mimmosoideae 2.9 0.8 20.3 0.3Albizia harveyi Mimmosoideae 2.1 0.7 20.9 1.2Dalbergia melanoxylon Caesalpinoideae 2.9 1.1 13.2 0.6Dichrostachys cinerea Mimmosoideae 1.7 0.4 18.5 1.1Ormocarpum trichocarpum Caesalpinoideae 0.7 15.6Pterocarpus rotundifolius Papilionoideae 0.9 0.5 14.5 1.4

Plants without indications of N2 fixationA. tortilis Mimmosoideae 3.5 12.0Combretum appiculatum Combretaceae 4.0 0.3 21.6 0.8C. azirec Combretaceae 5.1 21.8C. collinum Combretaceae 3.4 17.6C.imberbeþ zeyeri Combretaceae 4.3 0.4 20.7 0.7Cassia abreviata Caesalpinoideae 4.9 0.1 13.8 1.1Cassia petesiana Caesalpinoideae 5.4 18.5Ehretia rigida Boraginaceae 4.7 18.3Euclea sp. Ebenaceae 3.4 40.7Grewia bicolor Tiliaceae 6.6 24.0Grewia bilubosa Tiliaceae 9.2 17.9Heliothropium sp. Boraginaceae 4.3 13.2Lannea stuhlmani Anacardiaceae 4.1 27.4Lonchocarpus capassa Papilionoideae 4.7 15.0Peltophorum africanum Caesalpinoideae 4.3 0.2 23.3 1.6Sclerocarya birrea Anacardiaceae 5.6 0.6 28.5 2.5Securinega virosa Euphorbiaceae 6.7 12.3Terminalia sericea Combretaceae 4.1 0.5 26.9 0.9Understory legume Fabaceae 5.4 27.8Ziziphus mucronata Rhamnaceae 3.8 0.7 15.2 1.0

ISOTOPES AND FIRES IN SOUTH AFRICA 151

from the higher N availability after fires [7, 51]. The annual treatment of Nwaswitshaka

presents a different response, not only increasing N2 fixer relative basal area, but also

increasing N2 fixation (indicated by lower foliar d15N). Lower soil d15N in Nwaswitshaka

than in the other replicates (Fig. 5) may reflect higher N2 fixation by legumes (as suggested

by Tab. III) in these plots. The reason of the distinct response in Nwaswitshaka is not clear,

TABLE III Indicators of N2 Fixation Activity (Abundance of N2 Fixers and Foliar d15N) and Foliar P Along FireGradients. Relative Abundance of N2 Fixing Trees Decreased with Fire Frequency in Skukuza and Satara. HigherFoliar d15N also Indicates Lower N2 Fixation in Plots of Maximum Fire Frequency (Annual Treatments).Nwaswitshaka Showed a Different Trend, with Higher Relative Abundance of N2 Fixers and Lower Foliar d15N(More N2 Derived N) in the Annual than Triennial Treatments. Foliar P was Positively Correlated with d15N of N2

Fixers (r2¼ 0.38).

Plot and treatment BA N2 fixers=total BA Foliar d15N of N2 fixers, ‰ Foliar P, mg g�1 of leaf

SkukuzaControl 0.27 2.4� 0.5 1.3� 0.1Triennial 0.22 1.8� 0.3 1.2� 0.1Annual 0.13 3.1� 0.9 1.4� 0.1

NwaswitshakaControl 0.6� 0.2 1.2� 0.1Triennial 0.30 2.4� 0.8 1.4� 0.1Annual 0.52 0.5� 0.4 1.4� 0.1

SataraControl 0.48 2.9� 0.3 2.6� 0.4Sixennial 2.2� 0.4 1.5� 0.2Triennial 0.38 2.5� 0.7 2.4� 0.3Annual 0.32 4� 1.3 2.5� 0.5

FIGURE 5 Soil d15N in plots with different fire frequencies (C¼ contro; 6¼ sixennial; 3¼ triennial; 1¼ annualtreatments)in Skukuza (triangles), Nwaswitshaka (circles) and Satara (squares). Bars denote standard errors.Different letters indicate significant differences ( p< 0.05), and ‘‘Da’’ and ‘‘Db’’ in Satara indicate that the annual andtriennial treatments are significantly different from each other, but not significantly different from the sixennial andcontrol treatments.

152 J. N. ARANIBAR et al.

because soil and vegetation types are similar to Skukuza, and both sites were designed to be

replicates of the same vegetation unit.

Because phosphorus may limit N2 fixation, the increased P availability after fires may

enhance N2 fixation [17, 52]. Foliar phosphorus (P) determined for all the plots was signifi-

cantly lower for tree-legumes (mean¼ 1.49 mg P=g plant) than for forb legumes

(mean¼ 1.98 mg P=g plant) and non-legumes (mean¼ 1.95 mg P=g plant) ( p¼ 0.0007,

data not shown). Tree legumes had higher P concentrations in Satara than at the other

sites (significant at p¼ 0.0003, Table), as was observed for soil %C and N. In the

Combretum woodlands (Skukuza and Nwaswitshaka), foliar P was slightly increased with

high fire frequency (in the annual treatment), while in Satara there was no clear trend

(Tab. III). If N2 fixation were limited by P in these ecosystems, a negative relation between

foliar P and d15N would be expected. However, foliar P and d15N of N2 fixing plants were

positively correlated (r2¼ 0.38), suggesting that N2 fixation was not limited or enhanced

by P availability in the sites analyzed. Moreover, the positive relation suggests that higher

P availability may be associated with higher N availability, inhibiting N2 fixation.

Overall, the indicators of N2 fixation analyzed (foliar d15N and proportional basal area of

N2 fixers) do not support the hypothesis of enhanced woody legume N2 fixation in frequently

burned areas, and even suggest an opposite trend: decreased N2 fixation with fire frequency if

Nwaswitshaka is excluded. However, because soil N was not clearly depleted by subsequent

fires (except in Skukuza, Fig. 2), other N inputs may be available and should be analyzed.

Possible N sources include N2 fixation associated with grasses or soil crusts, atmospheric

deposition, and urine deposition by herbivores, which may graze on the frequently burned

plots because of the highly nutritious grasses (A. Potgieter, pers. comm.).

Soil d15N

There was no clear pattern of soil d15N along fire gradients (Fig. 5). The soils from

Nwaswitshaka had lower d15N than those from Satara and Skukuza, which agrees with

the relatively higher N2 fixation in this replicate (indicated by proportion and foliar d15N

of N2 fixing plants on the annually burnt plot). Soil d13C was also different in

Nwaswitshaka, indicating a higher contribution of plants with the C3 metabolism to soil

organic carbon (SOC) (Fig. 6). Soil d13C and d15N in the control plot at Nwaswitshaka

were also lower relative to other sites (Figs. 5 and 6). If the Nwaswitshaka soils are

excluded from data analysis (because of the strong differences between this site and the

others with respect to d15N, d13C, and patterns of N2 fixation), there was a significant

relationship between soil d15N and the proportion of N2 fixing plants at a site (r2¼ 0.36

p¼ 0.04) (Fig. 7). Inputs of N by woody legume N2 fixation seemed to modify the soil iso-

topic signatures, lowering d15N values as expected. The differences between soils located

under and between tree canopies were in some cases greater than differences among

plots (data not shown), indicating that soil d15N may be affected by local environmental

conditions. Although 2000 was an anomalously wet year, isotopic signatures in soils

collected during the dry season of 1999 in other locations near Skukuza (in the Skukuza

research tower site, described by Scholes et al. [53]), were not significantly different

( p< 0.05) from those sampled in 2000 at Satara and Skukuza (Fig. 8). This indicates

that flooding prior to our study in the Combretum woodlands did not cause dramatic

changes in soil d15N.

The soil d15N data indicate that fire frequency did not clearly enrich the system in 15N

(excluding Nwaswitshaka) [54]. Surface burning in a variety of sites has been shown to

increase biogenic emissions of NO and N2O [9–12]. These enhanced N fluxes tend to enrich

the soils in 15N because of the strong fractionations that occur during production of N-trace

ISOTOPES AND FIRES IN SOUTH AFRICA 153

gases by nitrification and denitrification [21, 22, 55–57], unless other processes such as

N2 fixation dilute the 15N of the system. Although N2 fixation seemed to decrease with

fire frequency, which would reinforce the effect of N emissions on 15N enrichment, there

appear to be other mechanisms masking the effect of the fires. N2 fixation by forb legumes

or other soil microorganisms such as cyanobacteria or grass-associated bacteria may be

important in these soils [58]. Our results agree with previous studies that do not show a

consistent pattern of d15N with fires [24–26], probably due to the opposing effects of various

N inputs (other than N2 fixation by woody legumes) and N emissions on isotopic signatures.

FIGURE 7 Correlation between soil d15N and proportion of N2 fixing trees (as indicated by Tab. II), for samplesfrom Skukuza and Satara. Open symbols¼ between tree canopies; closed symbols¼ under tree canopies. Barsdenote standard errors. r2

¼ 0.36; p¼ 0.04.

FIGURE 6 Soil d13C of plots with different fire frequencies (C¼ control; 6¼ sixennial; 3¼ triennial; 1¼ annualtreatments) in Skukuza (triangles), Nwaswitshaka (circles) and Satara (squares). Bars denote standard errors anddifferent letters indicate significant differences ( p< 0.05).

154 J. N. ARANIBAR et al.

Soil d13C

Carbon isotopic signatures reflect the contributions of C3 (generally trees, shrubs and forbs)

and C4 (generally grasses) plants to soil organic matter. Fires are believed to enhance grass

growth and suppress woody vegetation [3]. Even though this pattern may not be clear in

standing vegetation surveys performed during a single time of the year, d13C is thought to

integrate the C3–C4 contributions at longer time scales [59]. Shifts of C3 to C4 vegetation

after conversion of forests to pastures and climate change have been reflected in soil d13C

[60, 61]. Only in the annual and triennial treatments of Satara were the differences between

soils located under and between woody plant canopies significant ( p¼ 0.003), with soils

under canopies indicating more C3 derived carbon than those between canopies (data not

shown). Soils under and between canopies of the other plots were not significantly different.

Soil d13C from the Nwaswitshaka replicate was significant lower than that from the other

sites, indicating higher contributions of C3 plants to SOM than in all the other plots (Figs.

6 and 8). The control had lower d13C than the burned sites, indicating proportionally greater

C3 biomass (woody vegetation and C3 forbs) after several years of fire suppression, as is

expected. The much lower d15N and d13C in the control than in the burned plots of

Nwaswitshaka suggest the possibility of higher abundance of C3, N2 fixing plants at this

site, but unfortunately, vegetation was not surveyed in the control plot. It is surprising that

the Skukuza replicate, which was meant to represent the same vegetation type as the

Nwaswitshaka plots (Combretum woodland) had evidence of higher C4 contributions to

organic matter. Moreover, fires did not seem to affect tree-grass abundance in Skukuza. In

Satara, soils from the triennially and annually burned plots had higher contribution of C4

plants to SOM than soils from the control and sixennial treatments. The lower d13C in control

plots of Satara and Nwaswitshaka suggests that fire suppression over a 42–46 year period

caused a proportionally greater abundance of C3 plants (proportion of woody plants and

forbs over the total biomass, including grasses) to develop, (Fig. 6), which agrees with

reported increases of woody vegetation in the absence of fires [29].

FIGURE 8 C and N isotopic signatures of soils collected during the 2000 and 1999. Stars¼ soils collected nearSkukuza in 1999; squares¼Nwaswitshaka; triangles¼Skukuza; circles¼ Satara. Each symbol represents theaverage and standard errors of soils under and between plant canopies at different plots, in the case of the 2000samples. For the 1999 samples, each symbol represents the average of soils locations under and between plantcanopies of Combretum and Acacia savannas.

ISOTOPES AND FIRES IN SOUTH AFRICA 155

CONCLUSIONS

Our hypothesis that woody legume N2 fixation would increase as a function of fire frequency

was not supported by our data. To the contrary, our results suggest an overall decreased rela-

tive abundance and N2 fixation activity of N2 fixing tree legumes in frequently burned plots

(excluding Nwaswitshaka). Higher mineral N availability in annually burned plots, as indi-

cated by higher gross mineralization rates, may inhibit N2 fixation by tree legumes. Because

soil N content was not significantly lower in annually burned plots (except in Skukuza), N

inputs other than N2 fixation by woody legumes must occur. The mechanisms involved in

recovering N losses caused by frequent winter fires in savannas remain uncertain, but N2

fixation by tree legumes did not seem to contribute to N recovery in frequently burned

areas. Other possible mechanisms of N inputs include atmospheric deposition, N deposition

by grazers, and N2 fixation associated with forbs, grasses and soil free-living microorganisms.

The lack of a relationship between fires and d15N may reflect the opposing effects of N

inputs (other than woody legume N2 fixation) and gaseous N emissions on soil d15N. Our

data indicate that soil d15N is not a good tool to detect fire history, and that the changes

in d15N caused by fires at ecosystem or regional scales are unpredictable.

Acknowledgements

We thank L. Otter, and the personnel from the Skukuza research center for assistance in the

field and with species identification. Thanks to B. Neikirk, D. Lawrence and L. Read for

laboratory guidance. The comments of two reviewers helped us improve the manuscript.

This study was part of the Southern African Regional Science Initiative (SAFARI 2000)

and it was funded by the National Aeronautic and Space Administration (grants 7956;

7266; 7862; 9357) and the department of Environmental Sciences, University of Virginia

(Moore award, and Dupont fellowship awarded to Julieta Aranibar).

References

[1] Frost, P. G. H. and Robertson, F. (1987). The ecological effects of fire in Savannas. In: Walker, B. M. (Ed.),Determinants of Tropical Savannas. IRL, Orlando, pp. 93–140.

[2] Bird, M. I. and Call, J. A. (1998). A million-year record of fire in sub-Saharan Africa. Nature, 394, 767–769.[3] Scholes, R. J. and Walker, B. H. (1993). An African Savanna: Synthesis of the Nylsvley Study. Cambridge

University Press, p. 306.[4] Menaut, J.-C., Abbadie, L. and Vitousek, P. M. (1993). Nutrient and organic matter dynamics in tropical

ecosystems. In: Crutzen, P. J. and Goldammer, J. G. (Eds.), Fire in the Environment: The Ecological,Atmospheric, and Climatic Importance of Vegetation Fires. John Wiley and Sons Ltd., pp. 215–231.

[5] Van der Schiff, H. P. (1958). Inleidende verslag oor veldbrandnavorsing in die Nasionale Krugerwildtuin.Koedoe, 1, 60–92.

[6] Crutzen, P. J. and Andreae, M. O. (1990). Biomass burning in the tropics: impact on atmospheric chemistry andbiogeochemical cycles. Science, 250, 1669–1678.

[7] Raison, R. J. (1979). Modification of the soil environment by vegetation fires, with particular reference tonitrogen transformations: a review. Plant and Soil, 51, 73–108.

[8] Cass, A., Savage, M. J. and Wallis, F. M. (1984). The effect of fire on soil and microclimate. In: Booysen, P. ds V.and Tainton, N. M. (Eds.), Ecological Effects of Fire in South African Ecosystems. Springer-Verlag, Berlin, pp.311–325.

[9] Anderson, I. C., Levine, J. S., Poth, M. A. and Riggan, P. J. (1988). Enhanced biogenic emissions of nitric oxideand nitrous oxide following surface biomass burning. J. Geophys. Res., 93, 3893–3898.

[10] Levine, J. S., Winstead, E. L., Parsons, D. A. B., Scholes, M. C., Scholes, R. J., Cofer, W. R., Cahoon, D. R. andSebacher, D. I. (1996). Biogenic soil emissions of nitric oxide (NO) and nitrous oxide (N2 O) from Savannas inSouth Africa: The impact of wetting and burning. J. Geophys. Res., 101, 23689–23698.

[11] Parsons, D. A. B., Scholes, M. C., Scholes, R. J. and Levine, J. S. (1996). Biogenic NO emissions from Savannasoils as a function of fire regime, soil type, soil nitrogen and water status. J. Geophys. Res., 101, 23683–23688.

[12] Scholes, M. C., Martin, R., Scholes, R. J., Parsons, D. and Winstead, E. (1997). NO and N2O emissions fromSavanna soils following the first simulated rains of the season. Nutrient Cycling in Agroecosystems, 48, 115–122.

156 J. N. ARANIBAR et al.

[13] Swap, R. J., Garstang, M., Macko, S. A., Tyson, P. D. and Per Kallberg (1996). Comparison of biomass burningemissions and biogenic emissions to the tropical South Atlantic. In: Levine, J. S. (Ed.), Biomass Burning andGlobal Change. MIT Press, Cambridge, MA, pp. 396–402.

[14] Sprent, J. I. (1985). Nitrogen fixation in arid environments. In: Wickens, G. E., Goodin, J. R. and Field, D. V.(Eds.), Plants for Arid Lands. Kew International Conference on Economic Plants for Arid Lands. RoyalBotanical Gardens, Kew, England, pp. 215–299.

[15] Skarpe, C. and Henriksson, E. (1987). Nitrogen fixation by cyanobacterial crusts and by associative-symbioticbacteria in Western Kalahari, Botswana. Arid Soil Res. Rehab., 1, 55–59.

[16] Stromgaard, P. (1984). The immediate effect of burning and ash-fertilization. Plant and Soil, 80, 307–320.[17] Israel, D. W. (1987). Investigation of the role of phosphorus in symbiotic dinitrogen fixation. Plant Physiol., 84,

835–840.[18] Shearer, G., Kohl, D. H., Virginia, R. A., Bryan, B. A., Skeeters, J. L., Nilsen, E. T., Sharifi, M. R. and

Rundel, P. W. (1983). Estimates of N2-fixation from variation in the natural abundance of 15 N in SonoranDesert ecosystems. Oecologia, 56, 365–373.

[19] Schulze, E. D., Gebauer, G., Ziegler, H. and Lange, O. L. (1991). Estimates of nitrogen fixation by trees on anaridity gradient in Namibia. Oecologia, 88, 451–455.

[20] Robinson, D. (2001). d15N as integrator of the nitrogen cycle. Trends Ecol. Evol., 16, 153–162.[21] Blackmer, A. M. and Bremmer, J. M. (1977). Nitrogen isotope discrimination in denitrification of nitrate in soils.

Soil Biol. Biochem., 9, 73–77.[22] Mariotti, A., Germon, J. C., Hubert, P., Kaiser, P., Letolle, R., Tardieux, A. and Tardieux, P. (1981). Experimental

determination of nitrogen kinetic isotope fractionation: some principles, illustrations for the denitrification andnitrification process. Plant and Soil, 62, 413–430.

[23] Herman, D. J. and Rundel, P. W. (1989). Nitrogen isotope fractionation in burned and unburned chaparral soils.Soil Sci. Soc. Am. J., 53, 1229–1236.

[24] Grogan, P., Bruns, T. D. and Chapin, F. S., III (2000). Fire effects on ecosystem nitrogen cycling in a CaliforniaBishop pine forest. Oecologia, 122, 537–544.

[25] Cook, G. D. (2001). Effects of frequent fires and grazing on stable nitrogen isotope ratios of vegetation inNorthern Australia. Austral. J. Ecol., 26, 630–636.

[26] Mordelet, P., Cook, G., Abbadie, L., Grably, M. and Mariotti, A. (1996). Natural 15N abundance of vegetationand soil in the Kapalga Savanna, Australia. Austral. J. Ecol., 21, 336–340.

[27] Gertenbach, W. P. D. and Potgieter, A. L. F. (1979). Veldbrandnavorsing in die struikmopanieveld van dieNasionale Krugerwildtuin. Koedoe, 22, 1–28.

[28] Trollope, W. S. W., Potgieter, A. L. F., Biggs, H. C. and Trollope, L. A. (1998) Report on the experimentalburning plots (EBP) trial in the major vegetation types of the Kruger National Park. South African NationalParks, p. 42.

[29] Van Wyk, P. (1971). Veld burning in the Kruger National Park. An interim report of some aspects of research.Proceedings of the Tall Timbers Fire Ecology Conference, 11, 9–31.

[30] Gertenbach, W. P. D. (1983). Landscapes of the Kruger National Park. Koedoe, 26, 9–121.[31] National Oceanic and Atmospheric Administration (2002). Climate Research and Applications. http://

lwf.ncdc.noaa.gov/oa/climate/climateresearch.html[32] Brady, N. C. and Weil, R. R. (1999). The Nature and Properties of Soils. Prentice-Hall Inc., New Jersey, p. 881.[33] Hoefs, J. (1997). Stable Isotope Geochemistry, 4th ed. Springer-Verlag, Berlin, p. 201.[34] Parsons, T. R. (1984). A Manual of Chemical and Biological Methods for Seawater Analysis. Pergamon Press,

New York, p. 173.[35] Schmidt, S., Stewart, G. R., Turnbull, M. H., Erskine, P. D. and Ashwath, N. (1998). Nitrogen relations of natural

and disturbed plant communities in tropical Australia. Oecologia, 117, 95–104.[36] McKane, R. B., Johnson, L. C., Shaver, G. R., Nadelhoffer, K. J., Rastetter, E. B., Fry, B., Giblin, A. E., Kielland,

K., Kwiatkowski, B. L., Laundre, J. A. and Murray, G. (2002). Resource-based niches provide a basis for planyspecies diversity and dominance in Arctic Tundra. Nature, 415, 68–71.

[37] Evans, R. D., Bloom, A. J., Sukrapanna, S. S. and Ehleringer, J. R. (1996). Nitrogen isotope composition oftomato (Lycopersicon asculentum Mill. Cv. T-5) grown under ammonium or nitrate nutrition. Plant CellEnviron., 19, 1317–1323.

[38] Evans, R. D. (2001). Physiological mechanisms influencing plant nitrogen isotope composition. Trends PlantSci., 6, 121–126.

[39] Aranibar, J. N., Otter, L., Macko, S. A., Feral, C. J. W., Epstein, H. E., Dowty, P. R., Eckardt, F., Shugart, H. H.and Swap, R. J. Nitrogen cycling in the soil – plant system along a precipitation gradient in the Kalahari sands.Accepted in Global Change Biol.

[40] Mueller-Dombois, D. and Ellenberg, H. (1974). Aims and Methods of Vegetation Ecology. Wiley, New York.[41] Enslin, B. W., Potgieter, A. L. F., Biggs, H. C. and Biggs, R. (2000). Long term effects of fire frequency and

season on the Woody vegetation dynamics of the Sclerocarya birrea=Acacia nigrescens savanna of the KrugerNational Park. Koedoe, 43, 27–37.

[42] Davidson, E. A., Hart, S. C., Shanks, C. A. and Firestone, M. K. (1991). Measuring gross nitrogenmineralization, immobilization, and nitrification by 15N isotopic pool dilution in intact soil cores. J. Soil Sci.,42, 335–349.

[43] Anderson, I. C. and Poth, M. (1998). Controls of trace gases from Brazilian cerrado soils. J. Environ. Qual., 27,1117–1124.

ISOTOPES AND FIRES IN SOUTH AFRICA 157

[44] Brooks, P. D., Stark, J. M., McInteer, B. B. and Preston, T. (1989). Diffusion method to prepare soil extracts forautomated Nitrogen-15 analysis. Soil Sci. Soc. Am. J., 53, 1707–1711.

[45] Wessel, W. W. and Tietema, A. (1992). Calculating gross N transformation rates of N-15 pool dilutionexperiments with acid forest litter – analytical and numerical approaches. Soil Biol. Biochem., 24, 931–942.

[46] Hagiwara, K., Hikasa, K., Nakamura, K., et al. (2002). Review of particle physics. Phys. Rev. D, 66, 10001.[47] Jones, C. L., Smithers, N. L., Scholes, M. C. and Scholes, R. J. (1990). The effect of fire frequency on the organic

component of a basaltic soil in the Kruger National Park. S. Afr. Tydskr. Plant Ground, 7, 236–238.[48] Ernst, W. (1975). Variation in the mineral content of leaves of trees in Miombo Woodland in South Central

Africa. J. Ecol., 63, 801–807.[49] Schmidt, S. and Stewart, G. R. (1998). Transport, storage and mobilization of nitrogen by trees and shrubs in the

wet=dry tropics of Northern Australia. Tree Physiol., 18, 403–410.[50] Sprent, J. I. (1995). Legume trees and shrubs in the tropics: N2 fixation in perspective. Soil Biol. Biochem., 27,

401–407.[51] Covington, W. W. and Sackett, S. S. (1986). Effect of periodic burning on soil nitrogen concentrations in

Ponderosa pine. Soil Sci. Soc. Am. J., 50, 452–457.[52] Yahiya, M., Samiullah and Fatma, A. (1995). Influence of phosphorus on nitrogen fixation in chickpea cultivars.

J. Plant Nutr., 18, 719–727.[53] Scholes, R. J., Wilson, B., Davidson, N., et al. (2001). The environment and vegetation of the flux measurement

site near Skukuza, Kruger National Park. Koedoe, 44, 73–84.[54] Handley, L. L., Austin, A. T., Robinson, D., Scrimgeour, C. M., Raven, J. A., Heaton, T. H. E., Schmidt, S. and

Stewart, G. R. (1999). The 15N natural abundance of ecosystem samples reflects measures of water availability.Aust. J. Plant Physiol., 26, 185–199.

[55] Wellman, R. P., Cook, F. D. and Krouse, H. R. (1968). Nitrogen-15: Microbiological alteration of abundance.Science, 161, 269–270.

[56] Yoshida, N. (1988). 15N depleted N2O as a product of nitrification. Nature, 335, 528–529.[57] Hogberg, P. and Johannisson, C. (1993). 15N abundance of forests is correlated with losses of nitrogen. Plant and

Soil, 157, 147–150.[58] Sprent, J. I. and Sprent, P. (1990). Nitrogen fixing organisms. Pure and Applied Aspects. University Press,

Cambridge, p. 256.[59] Sage, R. F., Wedin, D. A. and Li, M. (1999). The biogeography of C4 photosynthesis: patterns and controlling

factors. In: Sage, R. F. and Monson, R. K. (Eds.), C4 Plant Biology. Academic Press, San Diego, pp. 313–371.[60] Veldkamp, E. (1994). Organic carbon turnover in three tropical soils under pasture after deforestation. Soil Sci.

Soc. Am. J., 58, 175–180.[61] Victoria, R. L., Fernandes, F., Martinelly, L. A., Piccolc, M. de C., De Camargo, P. B. and Trumbore, S. (1995).

Past vegetation changes in the Brazilian Pantanal arboreal-grasse Savanna ecotone by using carbon isotopes inthe soil organic matter. Global Change Biol., 1, 165–171.

158 J. N. ARANIBAR et al.