You are on page 1of 14

Forest Ecology and Management 171 (2002) 261–274

Phenological differences in tree water use and the timing of


tropical forest inventories: conclusions from patterns
of dry season diameter change
T.R. Bakera,*, K. Affum-Baffoea,b, D.F.R.P. Burslema, M.D. Swainea
a
Department of Plant and Soil Science, St. Machar Drive, University of Aberdeen, Aberdeen AB24 3UU, UK
b
Forest Management Support Centre, P.O. Box 1415, Kumasi, Ghana
Received 20 November 2000; received in revised form 5 October 2001; accepted 10 October 2001

Abstract

Interspecific variation in water-induced fluctuations in stem girth demonstrates the mechanisms promoting coexistence in
seasonally dry tropical forest. In addition, these fluctuations are a potential, but unevaluated, source of bias in measurements of
annual tree growth rates. To examine diurnal and seasonal patterns of stem diameter change, tree girth was measured over 2 years
(1997–1999), using dendrometer bands, for three species (Celtis mildbraedii, C. zenkeri and Strombosia glaucescens) in semi-
deciduous forest in Ghana. Soil matric potential was measured concurrently at 15 cm depth. In addition, measurements of all
trees >20 cm dbh on three, 1 ha plots were made at the beginning and middle of the 1998/1999 dry season. During the severe
1997/1998 dry season, soil matric potential declined below 1.5 MPa and two species showed significant stem shrinkage. For
the evergreen species, C. mildbraedii, there was a significant positive effect of tree diameter on stem shrinkage, and shrinkage
was greater in the second, compared to the first, half of the dry season. For the deciduous species, C. zenkeri, shrinkage was
reduced during the second half of the dry season, following leaf fall. During 1998/1999, soil matric potential, did not decline
below 1.5 MPa, and rates of girth change remained positive for all species. There were no significant effects of size or
phenology on the rate of girth change in the plot-based study. Deviations in annual increment calculated over successive monthly
intervals indicate that a 10-fold difference in soil water availability between measurement occasions can lead to a 4% bias in
estimates of annual growth. Measurements of forest plots should be made when inter-annual variation in soil water availability is
low. In this forest, measurements should, therefore, be made during the wet season, contrary to published recommendations.
# 2002 Elsevier Science B.V. All rights reserved.

Keywords: Dendrometer bands; Ghana; Permanent plots; Tree growth

1. Introduction division, and shorter term fluctuations caused by


changes in tree water status (Kozlowski and Winget,
Changes in tree diameter are a combination of two 1964; Sheil, 1995). These short term fluctuations
factors: the increase in dry matter from cambial cell occur over both diurnal and seasonal timescales.
For example, seasonal fluctuations are significant in
* seasonal tropical climates, where soil water availabil-
Corresponding author. Present address: School of Geography,
University of Leeds, Woodhouse Lane, Leeds LS2 9JT, UK.
ity is periodically low, and tree diameter may decrease
Tel.: þ44-113-233-3361; fax: þ44-113-233-3308. in the dry season (Boaler, 1963; Daubenmire, 1973;
E-mail address: t.baker@geog.leeds.ac.uk (T.R. Baker). Lieberman, 1982; Reich and Borchert, 1984; Swaine

0378-1127/02/$ – see front matter # 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 8 - 1 1 2 7 ( 0 1 ) 0 0 7 8 7 - 3
262 T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274

et al., 1990; Reich, 1995; Baker et al., unpublished). of stored water. However, it is also possible that other
Changes in tree diameter in response to tree water factors that influence tree water relations, such as tree
status are closely related to changes in soil water size and topographic position, may differ between the
availability, and are also influenced by tree water populations that are being compared.
use characteristics. Fluctuations in diameter have been Tree size has a significant effect on water relations,
related to changes in stem water potential (Hinckley although the direction of that effect remains conten-
and Bruckerhoff, 1975; Hinckley et al., 1978; Reich tious (Goldstein et al., 1998; Eamus, 1999; Meinzer
and Borchert, 1982; Garnier and Berger, 1986), and to et al., 1999). Some studies indicate that greater tree
the use of stored water (Reich and Borchert, 1984; size is associated with improved dry season tree
Holbrook, 1995; Reich, 1995). Studies of these water- water status. For example, large trees maintained
related changes in tree diameter in seasonal tropical maximum rates of transpiration for longer during
forests are important for two reasons. Firstly, they the day than small trees in a seasonal forest in
allow hypotheses concerning interspecific differences Panama (Goldstein et al., 1998), and taller saplings
in water use to be tested, illustrating mechanisms maintained higher pre-dawn water potentials during
that may promote species coexistence. Secondly, they the dry season in an Australian savanna (Eamus,
allow us to test whether inter-annual variation in soil 1999). However, in a seasonally dry forest in
water availability between years introduces a signi- Panama, small trees have been shown to use deeper
ficant bias in measurements of annual diameter sources of soil water during the dry season than large
increment. Measurements of tropical forest plots are trees (Meinzer et al., 1999), suggesting that in forest
increasingly being used to monitor long-term changes ecosystems, large trees experience more frequent and
in forest biomass at regional scales (Phillips et al., severe periods of low soil water availability. To
1998). Critiques of this work have focussed on arte- attribute differences in dry season diameter change
facts introduced by the methodology itself (Clark, in to differences in phenology, we need to adequately
press). It is, therefore, important to quantify all the account for variation in tree size. Similarly, we need
potential errors and biases in forest inventory proce- to account for differences in topographic position,
dures (Phillips et al., in press). as topographic differences in soil water availability
One of the major axes of variation allowing species have been shown to cause differences in plant water
coexistence in seasonal tropical forests relates to water potential in a seasonally dry forest in Panama (Becker
use. The different strategies are exemplified by decid- et al., 1988).
uous and evergreen species (Sobrado, 1986, 1993; The possibility that water-induced fluctuations in
Fanjul and Barajas, 1987; Borchert, 1994; Eamus, stem diameter are a significant bias in longer term
1999). In a dry forest in Venezuela, evergreen species measures of growth was suggested following a study
had smaller seasonal changes in leaf water potential of long-term growth rates in moist forest in Uganda
and wood water content than deciduous species (Sheil, 1997). Most stems showed a reduction in
(Sobrado, 1986, 1993), and in a semi-deciduous forest growth rates when measured over wetter seasonal
in Costa Rica dry season stem water potentials periods, possibly as a result of field measurements
declined to 2.5 MPa in evergreen species, compared at these times being made only on bright, sunny days
to <5 MPa for deciduous species (Borchert et al., in when diurnal stem shrinkage is high (Sheil, 1997).
press). At the same site in Costa Rica, patterns of dry However, no previous studies have measured changes
season diameter change correlate with variation in in tree girth at a range of timescales, in order to
phenology (Daubenmire, 1973; Reich and Borchert, directly evaluate whether short-term fluctuations
1984; Reich, 1995). Deciduous species decrease in can significantly bias longer term measurements of
girth substantially as soil water availability declines, growth. Repeated measurements of forest plots, car-
whereas evergreen species maintain positive girth ried out over whole year intervals, assume that tree
increment throughout the dry season. The physiolo- water status is identical on the first and all subsequent
gical studies suggest that the greater contraction of measurement occasions. However, this assumption is
stem diameter in deciduous species is caused by both unlikely to be correct. If tree water status is higher
more negative plant water potentials, and greater use on the second compared to the first measurement
T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274 263

occasion, then diameter increment will be overesti- papaverifera (Sterculiaceae). Common understorey
mated, as increased hydration will add to the annual species are Cleidion gabonicum (Euphorbiaceae),
increment. If tree water status is lower, then annual Microdesmis puberula (Pandaceae) and Baphia nitida
increment will be underestimated. (Papilionaceae).
By studying diurnal and seasonal changes in tree The three principal study species were C. mildbraedii
girth and soil water availability in semi-deciduous Engl. (Ulmaceae), C. zenkeri Engl. and Strombosia
forest in Ghana 1997–1999, this paper tests two glaucescens J. Leonard (Olacaceae). C. mildbraedii is
hypotheses: an evergreen, non-pioneer species that is very common
in semi-deciduous forests in Ghana, found throughout
1. Dry season stem shrinkage, and hence water the Guineo-Congolian forest zone and in East African
stress, is greater in deciduous compared to ever- coastal forests (Hawthorne, 1995). Its abundance
green species, having accounted for differences in declines markedly as the landscape becomes wetter,
tree size and topographic position. at both regional and local scales (Fig. 1(a), Swaine and
2. A significant bias is introduced into measurements Hall, 1986; Hawthorne, 1995). It is an important fuel
of annual increment by inter-annual variation in for domestic use and making pestles (N. Gyakari,
soil water availability. pers. comm.). C. zenkeri is a deciduous, non-pioneer
species found across the Guineo-Congolian forest
zone (Hawthorne, 1995). In Ghana, its distribution is
2. Methods biased towards dry, fertile sites (Fig. 1(b), Swaine,
1996). S. glaucescens is a very common evergreen,
2.1. Study site and species non-pioneer species in Ghana and found throughout the
Guineo-Congolian region. Its distribution in Ghana is
The study site was Tinte Bepo Forest Reserve biased toward wet, infertile sites (Fig. 1(c), Swaine,
(78040 N, 28060 W) which is classified as moist semi- 1996). It is currently important for making telegraph
deciduous forest (Hall and Swaine, 1981). Mean poles (N. Gyakari, pers. comm.).
annual rainfall, 1971–1993, recorded at the nearest
Ghana Meteorological Service station at Bechem, 2.2. Topographic definitions
15 km northeast of the site, is 1288 mm. There is a
strong dry season from December to March, when soil Within areas of uniform soil parent material, the
matric potential, at 20–60 cm depth, may decline water regime is an important determinant of colour in
below 1.5 MPa, and a less severe drier period from tropical soils due to its influence on the content and
July to September (Veenendaal et al., 1996). The site is form of the predominant iron oxide present. Haemi-
at approximately 300 m a.s.l., and in the study area tite, which gives the soil a bright red colour, is the form
overlies fine grained, non-micaceous hornblende gran- typical of well-drained areas whilst the formation of
ite (Adu, 1993). The topography is undulating with an goethite is favoured under wetter conditions (Schwert-
elevational difference of 30–50 m between summits mann and Taylor, 1989). As the particular focus of this
and valleys, over horizontal distances of 300–500 m. study was to examine the influence of topographic
Maximum slopes are approximately 208. Reddish variation on water supply, soil colour was used to
brown sandy clay soils are found on the summits, define landscape position (cf. Clark et al., 1998). Soil
yellowish brown sandy clay soils on the slopes and colour was assessed using a Munsell Soil Colour
yellowish brown sandy clay loams and sandy loams in chart at 15 cm depth, 2 m away on the western side
the valleys (Adu, 1993). of each selected tree (see below). There were clear
There are approximately equal proportions of and consistent changes in soil colour downslope,
individuals of evergreen and deciduous species in and different landscape positions were defined on
semi-deciduous forest (Hall and Swaine, 1981). Com- the basis of the hue (Oyama and Takehara, 1991).
mon canopy species, which may exceed 60 m in Summit soils were red, hue 5 YR, slope soils yellow/
height, are Triplochiton scleroxylon (Sterculiaceae), red, hue 7.5 YR and valley soils pale brown/grey,
Celtis mildbraedii (Ulmaceae) and Nesogordonia hue 10 YR.
264 T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274

2.3. Dendrometer bands 12 ha, over several watersheds. The large study area
indicates that the study trees were typically well
Adult trees greater than 20 cm dbh whose crowns dispersed (>20 m) and were considered independent
received at least some direct side light, i.e. Dawkins data points for the purposes of statistical analysis.
crown classification of 4 or 5 (Dawkins, 1958), were Seasonal changes in tree girth were monitored using
selected for study, over an area of approximately dendrometer bands constructed from 20 mm width,

Fig. 1. Distributions of: (a) C. mildbraedii, (b) C. zenkeri and (c) S. glaucescens in Ghana. The location of Tinte Bepo Forest Reserve, 250 mm
isohyets and extent of the forest zone (shaded area) are also shown.
T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274 265

Fig. 1. (Continued ).

150 mm thickness, hard-tempered aluminium and held start of the study indicated that a short period was
in place with a stainless steel spring. Shoe eyelets were required for the bands to settle (Keeland and Sharitz,
used to strengthen the holes where the spring held the 1993). The results are not influenced by long-term
band. Measurements to 0.01 mm were made using settling of the bands, as a control band placed on one
Vernier calipers of the distance between scribe marks individual of C. mildbraedii that had a badly damaged
on the overlapping front and rear ends of the band. crown, showed no change in girth over the 2 years
Bands were placed at 1.3 m height or 50 cm above the of the study. Therefore, to obtain reliable readings,
top of any fluting or buttress. Trunk climbers were those taken less than 30 days following the attachment
cut prior to fixing the band. From 23 July 1997 to 6 of a band were discarded (cf. Pélissier and Pascal,
August 1997, bands were applied to 28 C. mildbraedii, 2000). On seven occasions, out of a total of 837
five C. zenkeri and three S. glaucescens trees; one measurements, slipped or damaged bands had to be
S. glaucescens was added to the sample on 22 August replaced. For calculations of the rate of girth change,
1997, two C. mildbraedii on 1 October 1997 and two data from these individuals were discarded for at
C. zenkeri and one S. glaucescens on 17 October 1997. least 30 days, and annual increments were only calcu-
The total sample sizes were, therefore, 30, 7 and 5 for lated for trees and periods with continuous, 12 month,
the three species, respectively, which reflects the records.
species relative abundance at the landscape scale in
this forest type. 2.4. Soil matric potential
Measurements of tree girth were made at approxi-
mately monthly intervals until May 1999. In addition, Measurements of soil matric potential were made
on four occasions (18 September 1997, 11 October concurrently with measurements of tree girth. The
1997, 22 October 1997, and 21 November 1997) filter paper method of Deka et al. (1995) was used
measurements were made of diurnal fluctuations for as it gives reliable results down to low matric poten-
C. mildbraedii, by comparing tree girth between tials of approximately 2 MPa, sampling is rapid and
morning (08:00–11:00 h) and late afternoon (15:30– it is inexpensive. A standard filter paper is equilibrated
17:30 h) measurements. Weekly measurements at the over 6 days with a soil sample, the water content of the
266 T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274

filter paper determined and then converted to matric 1. Dry season diameter change of C. mildbraedii was
potential using calibration equations given by Deka analysed with topographic position as a factor, and
et al. (1995). From 11 October 1997, samples at 15 cm tree diameter and annual increment (calculated 18
depth were taken 2 m from the southern side of 14 C. September 1997 to 18 September 1998) as
mildbraedii across the range of landscape positions. covariates.
On the second and subsequent measurement occa- 2. Dry season diameter change of C. mildbraedii and
sions, samples were taken 50 cm from either side of C. zenkeri were compared in summit positions
the original sample point, and then three sets of three with species as a factor, and diameter and annual
sample points, each 50 cm behind the previous set. diameter increment included as covariates.
This systematic sampling pattern was then repeated, 3. Dry season diameter change of C. mildbraedii and
moving clockwise, and assumes there is no systematic S. glaucescens were compared in slope positions
variation in soil matric potential for a 3.5 m radius with species as a factor, and diameter and annual
around each tree. From 5 November 1997 until 8 April diameter increment included as covariates.
1999, samples were taken at 15 cm depth around all
42 trees, at the same time as measurements of the Secondly, the plot-based study allowed the influ-
dendrometer bands. ence of the same factors on dry season diameter
change to be assessed at the stand level. Trees were
2.5. Forest plot study grouped into size and phenology classes. The median
diameter was used as a threshold in a two-way classi-
To test whether size or phenology-related patterns fication of tree size, with individuals less than or
of dry season stem shrinkage could be detected at the equal to 30.5 cm diameter classified as small. Species
stand level, measurements were also made of three were classified as deciduous or evergreen, following
1 ha permanent sample plots (PSPs) in the same forest Hawthorne (1995); species with variable vegetative
reserve, during the 1998/1999 dry season. Measure- phenology, such as Cylicodiscus gabonensis (Mimo-
ments were made at the start (November) and end saceae), were omitted from the analysis. The resulting
(January) of the dry season. The plots showed no dataset comprised 465 trees in four categories. As the
evidence of recent fire damage, and are part of the residuals from parametric analysis were not normally
on-going programme of permanent sample plot enu- distributed, differences between groups were tested
meration run by the Ghana Forest Service (Bird and for using Kruskal–Wallis tests (Sokal and Rohlf,
Sackey, 1991). A full account of the field procedures 1995).
are given in Wong (1997) and Affum-Baffoe (1999).
On each measurement occasion in this study, the 2.7. Data analysis: bias in estimates of tree growth
girth of each tree >20 cm dbh was recorded, at a
height of 1.3 m, or 50 cm above the top of any buttress The effect of inter-annual variation in soil water
(diameter at reference height, drh). availability on measurements of annual increment was
examined for individuals of C. mildbraedii growing
2.6. Data analysis: tree water relations in summit positions. Annual increments were calcu-
lated from successive measurement occasions, with
The relationships between patterns of dry season values for exact years obtained by linear interpolation
stem diameter change, phenology, tree size and topo- between measurement dates in the second year of
graphy were examined in two ways. Firstly, the effect study. To remove the influence of temporal change
of these factors on dry season diameter change was in actual growth over the course of the study, the data
tested during the severe 1997/1998 dry season using were detrended using linear regression. The residuals
the dendrometer band data. Three repeated measures of this regression were then compared with the change
ANOVA were performed, comparing the first and in mean soil matric potential between years. Similar
second half of the dry season, 21 November 1997 to the calculation of successive annual increments,
to 14 January 1998, and 14 January 1998 to 18 March exact values for soil matric potential on equivalent
1998, respectively. dates during the second year of study were calculated
T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274 267

by linear interpolation of mean logged soil matric in all sites (Fig. 2). Less negative minimum soil matric
potential data. potentials during November 1998 to March 1999,
indicate that the dry season was less severe during
the second year of the study (Fig. 2).
3. Results
3.2. Magnitude of fluctuations
3.1. Soil matric potential regime
No significant diurnal girth fluctuations were found
The soil matric potential at 15 cm depth was higher for C. mildbraedii, either on the four occasions when
in valley than slope and summit positions, apart from they were measured, or when the data were combined
during the late dry season of 1997/1998, when all sites (Table 1). However, significant dry season stem
became very dry, and periodically during the wet shrinkage occurred for two of the three species, and
season when water availability was similarly high for C. mildbraedii, the magnitude of the mean dry

Fig. 2. Rates of change of girth for C. mildbraedii in summit and valley positions (slope position not shown), C. zenkeri and S. glaucescens,
and the soil matric potential regime in summit, slope and valley positions, at 15 cm depth, in semi-deciduous forest in Ghana, from October
1997 to April 1999. For the rate of girth change, significant differences between species, tested using univariate F-tests, are indicated,

p < 0:05, p < 0:01, p < 0:005.
268 T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274

Table 1
Diurnal and dry season stem girth change in C. mildbraedii, C. zenkeri and S. glaucescens in semi-deciduous forest in Ghana, 1997/1998a

Diurnal Dry season

Mean  S:E: n Significance Mean  S:E: % of % of annual n Significance


(mm) (mm) diameter increment

C. mildbraedii 0:02  0:02 59 ns 0:68  0:13 0.18 15.2 30 <0.005


C. zenkeri 1:28  0:34 0.51 17.6 7 <0.005
S. glaucescens 0:11  0:14 0.03 17.9 5 ns
a
Mean diurnal fluctuations are calculated from morning (08:00–11:00 h) and late afternoon (15:30–17:30 h) measurements made on 18
September 1997, 11 October 1997, 22 October 1997, and 21 November 1997. The dry season is defined as 21 November 1997 to 18 March
1998. Z-tests were used to test if changes in girth were significantly different from zero.

season change was more than 30 times greater than Table 2


mean daily fluctuation (Table 1). Table of significance of repeated measures ANOVA of the effects
of size, annual increment, size, topographic position and time (first
and second half of the dry season), on dry season diameter change
3.3. Dendrometer study: tree water relations of C. mildbraedii in semi-deciduous forest in Ghana, 1997/1998 (p-
values in bold are significant at p < 0:05)
For C. mildbraedii, tree diameter had a significant F p
main effect on the degree of dry season diameter
change, indicating that larger trees shrank more than Size 10.8 0.00
Annual increment 3.0 0.10
smaller trees during the severe 1997/1998 dry season
Position 3.2 0.06
(Fig. 3, Table 2). Annual diameter increment did Time 2.9 0.10
not have a significant main effect, but there was a Time  size 0.7 0.40
significant time/increment interaction (Table 2). Time  increment 7.5 0.01
Although trees better supplied with water in valley Time  position 2.0 0.15

Fig. 3. Dry season change in diameter, 1997/1998, as a function of tree diameter for C. mildbraedii in a semi-deciduous forest in Ghana.
Linear regression: y ¼ 0:33  0:03x, F ¼ 11:9, p < 0:005.
T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274 269

Fig. 4. Mean and standard error of change in tree girth for C. mildbraedii and C. zenkeri during the early and late dry season 1997/1998,
November–January and January–March, respectively, in semi-deciduous forest in Ghana. For C. zenkeri, leaf fall occurred in all trees between
January and March measurements.

positions tended to shrink less (Fig. 2), the effect of deciduous species, C. zenkeri, following leaf fall,
topographic position on dry season diameter change which occurred between January and March 1998
was marginally non-significant (Table 2). measurements in all trees of this species (Fig. 4).
Evergreen and deciduous species growing in the However, there were no differences between the two
same topographic position showed significant differ- evergreen species, C. mildbraedii and S. glaucescens,
ences in dry season diameter change (Fig. 4, Table 3). when they were compared in slope positions (Table 3).
In the comparison of C. zenkeri and C. mildbraedii in
summit positions, there was a significant time/species 3.4. Plot-based study: tree water relations
interaction (Table 3). The rate of stem shrinkage
increased in the evergreen species, C. mildbraedii, There were no differences in median diameter
as the dry season progressed, but decreased in the change between small and large, deciduous and

Table 3
Table of significance from repeated measures ANOVA of the effects of size, annual increment, size, species and time, on dry season diameter
change of an evergreen and deciduous species (C. mildbraedii and C. zenkeri) and two evergreen species (C. mildbraedii and S. glaucescens)
in equivalent topographic positions in semi-deciduous forest in Ghana, November–January and January–March, 1997/1998 (p-values in bold
are significant at p < 0:05)

C. mildbraedii/C. zenkeri C. mildbraedii/S. glaucescens

F p F p

Size 19.2 0.00 0.29 0.60


Annual increment 0.6 0.44 2.87 0.12
Species 13.4 0.00 0.60 0.46
Time 3.2 0.09 0.17 0.69
Time  size 1.5 0.24 1.49 0.25
Time  increment 14.3 0.00 1.57 0.24
Time  species 52.0 0.00 1.77 0.22
270 T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274

Fig. 5. Successive mean annual increments, S.E., for C. mildbraedii in summit positions in semi-deciduous forest in Ghana, 1997/1999.
Annual increments are calculated from successive measurement occasions during the first year of study. Linear regression: y ¼ 11:8 þ 0:026x,
F ¼ 263, p < 0:005.

evergreen species, in the plot-based study, during the changes on the measurement of annual increment was
1998/1999 dry season (H ¼ 2:73, d:f: ¼ 3, p ¼ 0:434, evaluated further. Annual increments calculated from
overall mean diameter change ¼ 0:03  0:01 cm, successive measurement occasions show a steady
median change ¼ 0:00 cm). The effects of size and significant increase over the study (Fig. 5). However,
phenology may only be important during particularly there is also substantial variation around this relation-
dry years. In the dendrometer study, higher rates of ship, particularly for annual increments initiated dur-
girth change during the 1998/1999 dry season com- ing the course of the dry season (Fig. 5). There was a
pared to 1997/1998, reflected the less severe dry significant positive correlation between the difference,
season and resulted in similar patterns in all three and mean, of soil matric potential measured in suc-
species, in contrast to the previous year (Fig. 2). cessive years (Fig. 6). This result demonstrates that
soil matric potential was more variable at generally
3.5. Dendrometer study: hypothesis 2 dry times of year. In addition, the difference in soil
matric potential between years is correlated with the
As there was no evidence for significant diurnal residual error in the calculation of annual increment
fluctuations in tree girth, only the impact of seasonal (Fig. 7). The significant regression indicates that a 10-

Fig. 6. Relationship between the mean, and variation, in soil matric potential between 1997/1998 and 1998/1999. Linear regression:
logðyÞ ¼ 0:45 logðxÞ  0:013, F ¼ 9:6, p < 0:05.
T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274 271

Fig. 7. Residual error in linear increase of annual increment through time, plotted against the difference in soil matric potential between years.
Linear regression: y ¼ 0:57x  0:31, F ¼ 5:1, p < 0:005.

fold increase in soil matric potential between mea- defined as the change in water content per unit change
surement occasions leads to an overestimate of annual in water potential, per unit volume (Holbrook, 1995),
increment by 0.5 mm, or approximately 4%. In this does, however, vary between trees, due to differences
study, the most extreme errors in the calculation of in tissue structure or the relative proportions of dif-
annual increment were found in measurements at the ferent tissues. For instance, variation in bark thickness
height of the dry season, when annual growth is is important (Borchert, 1994). As phloem is an elastic
overestimated by more than 9% (Fig. 5). tissue, absolute declines in stem diameter, for a given
reduction in plant water potential, will be greater in
trees with thicker bark. In addition, differences in
4. Discussion growth rate may be important, as fast growing trees
produce larger cells that may show a greater reduction
4.1. Tree water relations in volume for the same change in water potential. The
result of these effects is that a given reduction in water
As dry season diameter change only represents an content (stem diameter) in two trees, that are different
indirect measure of tree water status, inferring species species or sizes, may not reflect similar changes in
and size effects on tree water relations from these data plant water potential. In this study, the effect of
must be done with caution. Seasonal decreases in stem differences in growth rate can be discounted, as sig-
diameter are caused by water loss from elastic storage nificant effects of tree size and species on dry season
elements such as the phloem, cambium and sapwood diameter change were found, independently of varia-
parenchyma (Holbrook, 1995). Close correlations tion in annual increment (Table 2). However, differ-
have been obtained between xylem water potential ences in bark thickness are important for interpreting
and stem diameter over diurnal timescales (Worrall, the relationships between dry season diameter shrink-
1966; Hinckley and Bruckerhoff, 1975; Hinckley et al., age and tree water relations.
1978; Reich and Borchert, 1982; Garnier and Berger, The greater shrinkage of larger trees of C. mild-
1986), but extending this relationship over longer, braedii during the 1997/1998 dry season is consistent
seasonal timescales is more complex. Specifically, with the suggestion that large trees experience greater
inferring species and size effects on seasonal changes water deficits than small trees during the dry season.
in plant water potential from patterns of stem diameter The greater transpirational demand of a larger canopy
change assumes that tissue capacitance does not vary apparently exceeds the ability of a larger root system
across species and size classes. Tissue capacitance, to supply water. Research on the sources of dry season
272 T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274

water and patterns of tree mortality has also suggested and wood water content found in deciduous, compared
that large trees are more susceptible to seasonal to evergreen, species in a dry forest in Venezuela
drought. For example, by relating the isotopic signa- (Sobrado, 1986, 1993). Leaf abscission was an effec-
ture of xylem sap to groundwater values, large trees tive strategy for conserving water in C. zenkeri, as the
were found to tap shallower sources of water than rate of shrinkage dropped following leaf fall in the
small trees during the dry season in a seasonally dry second half of the 1997/1998 dry season (Fig. 4).
forest in Panama (Meinzer et al., 1999). In addition, However, two factors may confound the comparison
drought related mortality was 50% higher in trees of deciduous and evergreen species. Firstly, differ-
>60 cm dbh, compared to individuals <30 cm dbh, ences in bark thickness between these two species
following the 1983 drought in Borneo (Wirawan, were not measured, and the magnitude of diameter
unpubl. rep., 1983, in Goldammer and Seibert, 1990). change may not directly reflect differences in plant
However, the suggestion that large trees experience water potential, as discussed above. Secondly, the
greater seasonal water deficits than small trees should species may differ in rooting depth allowing differ-
be treated with caution. Although the increased stem ential access to dry season water. Rooting depth was
shrinkage of larger trees must result from greater not studied here, but the forest occurs on deep soils
water loss from elastic storage tissues, this does not (Adu, 1993) and there was no evidence of the forma-
necessarily reflect the development of more negative tion of a hard pan in soil pits dug to 1 m depth.
plant water potentials. Bark thickness increases with However, studies of the stable hydrogen isotope com-
tree size in C. zenkeri in Ghana (Orgle, 1994); if the position of xylem sap and groundwater in seasonally
same relationship occurs in its congener, the effect of dry forest in Panama have shown that evergreen
size on dry season diameter change in C. mildbraedii species have access to water at greater depth than
may be, at least in part, due to changes in capacitance, deciduous species (Jackson et al., 1995). It is possible
in addition to plant water potential. that either differences in bark thickness or rooting
Topographic differences in soil water availability depth mean that the patterns of dry season diameter
have been shown previously to cause differences in change do not reflect equivalent differences in water
plant water potential in a seasonally dry forest in stress, or are not influenced by leaf phenology. How-
Panama (Becker et al., 1988). In this semi-deciduous ever, the significant species/time interaction related to
forest in Ghana, topographic position did not have a leaf fall in C. zenkeri (Table 2), provides strong
significant independent effect on diameter change, support for the hypothesis that dry season water
although trees in sites with higher soil water avail- deficits are greater in the deciduous compared to
ability did tend to decrease in diameter less during the the evergreen species, and that it is caused by differ-
severe 1997/1998 dry season (Fig. 2). However, soil ences in leaf phenology.
water availability was an important determinant of
inter-annual differences in dry season diameter 4.2. Errors in forest plot measurements
change. No differences were found between species
and size classes during the second year of the study in No evidence was found here for significant diurnal
either the dendrometer or plot-based study, and the fluctuations in stem girth, and hence no evidence that
mean rate of girth change did not decline below zero they can generate significant errors in measurements
for any species (Fig. 2). of annual increment (Sheil, 1997). It is possible that
The greater shrinkage of C. zenkeri compared to C. dendrometer bands are unable to detect very small
mildbraedii during the first half of the 1997/1998 dry changes over this short timescale. In contrast, Sheil
season, suggests that water loss was greater, and more (1997) cites a study where multiple measurements of
negative water potentials were developed, in the tree girth using a tape measure found significant
deciduous compared to the evergreen species. These diurnal changes in stem diameter. Alternatively, the
patterns are consistent both with patterns of dry season shade tolerant species C. mildbraedii, may show little
diameter change in a dry forest in Costa Rica (Reich change in diameter over these timescales, possibly as a
and Borchert, 1984; Borchert et al., in press), and the result of a low proportion of elastic storage tissues in
greater seasonal fluctuations in leaf water potential the trunk.
T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274 273

Significant water-related changes in stem diameter forest type should be made during the driest time of
do result in errors in estimates of growth. The steady year, and of the driest forest plots at the wettest time of
increase in growth measured over successive, over- year, to minimise the effect of inter-annual variation in
lapping, time periods is caused by the decreasing dry season soil water availability on estimates of growth.
contribution of the 1997/1998 dry season to the cal-
culation of annual increment. Rates of wet season
expansion of these trees remained constant between Acknowledgements
years, whereas dry season diameter change was sig-
nificantly higher in the second year of study (Baker This work was funded by a University of Aberdeen
et al., unpublished). The hypothesis that increased soil Faculty Studentship (TB) and the Department for
water availability on successive measurement occa- International Development (TB/KAB). Yaw Atua-
sions leads to an overestimate of growth is supported, hene, William Asante and Raymond Votere provided
as the deviations in annual increment about this trend invaluable assistance with the fieldwork, and Doug
are negatively correlated with the difference in soil Sheil and Rolf Borchert gave very helpful comments
matric potential between years (Fig. 7). Although the on a previous version of the manuscript.
potential error is small, it will be important where
intercensus intervals are short, in forests with a
strongly seasonal climate, and where inventories take References
several months to complete.
Adu, S.V., 1993. Soil map of the Upper Tano Basin, 1:250,000. Soil
Published recommendations on methods of tropical
Research Institute, Kumasi, Ghana.
forest inventory suggest that the optimum time to Affum-Baffoe, K., 1999. Forest dynamics along an environmental
measure forest plots is during the dry season, when gradient in Ghana. M.Phil. Dissertation. University of Aberd-
inter-annual variation in water-induced changes in tree een, Aberdeen, UK
diameter are assumed to be minimal (Sheil, 1995). Becker, P., Rabenold, P.E., Idol, J.R., Smith, A.P., 1988. Water
potential gradients for gaps and slopes in a Panamanian tropical
However, this study indicates that variation in soil
moist forest’s dry season. J. Trop. Ecol. 4, 173–184.
water availability is greatest during the dry season Bird, N.M., Sackey, E.P., 1991. Forestry Department PSP
(Fig. 6). Although this study only compares 2 years, Programme. Ghana Forestry Department, Kumasi, Ghana.
one of which is strongly influenced by an El Niño Boaler, S.B., 1963. The annual cycle of stem girth increment in
event, it is likely that the patterns described here are trees of Pterocarpus angolensis D.C., at Kabungu, Tanganyika.
Comm. For. Rev. 42, 232–236.
representative of West African forests, and will be
Borchert, R., 1994. Soil and stem water storage determine
reflected over longer timescales. Climatic fluctuations phenology and distribution of tropical dry forest trees. Ecology
during El Niño events are likely to be the main 75, 1437–1449.
determinant of inter-annual variation in climate, and Borchert, R., Rivera, G., Hagnauer, W., in press. Modification of
in this region, El Niño events typically lead to more vegetative phenology in a tropical semideciduous forest by
abnormal drought and rain. Biotropica.
severe dry seasons (Swaine et al., 1997). Less bias
Clark, D.A., in press. Are tropical forests an important global
would be introduced into estimates of growth if enu- carbon sink? Revisiting the evidence from long-term inventory
merations were carried out during the wet season, to plots. Ecol. Appl.
avoid the large inter-annual variation in dry season Clark, D.B., Clark, D.A., Read, J.M., 1998. Edaphic variation and
patterns of stem diameter change, caused by variation the mesoscale distribution of tree species in a neotropical rain
forest. J. Ecol. 86, 101–112.
in dry season soil water availability. This study pro-
Daubenmire, R., 1973. Phenology and other characteristics of
vides direct evidence to support a similar recommen- tropical semi-deciduous forest in north-western Costa Rica. J.
dation made for plots in wet evergreen forests with a Ecol. 60, 147–171.
monsoon climate in the Western Ghats of India (Pélis- Dawkins, H.C., 1958. The Management of Tropical High Forest
sier and Pascal, 2000). Finally, in large inventory with special reference to Uganda. Imperial Forestry Institute
Paper 34.
programmes that maintain plots in a variety of forest
Deka, R.N., Wairi, M., Mtakwa, P.W., Mullins, C.E., Veenendaal,
types, with a full-time schedule of remeasurements, E.M., Townend, J., 1995. Use and accuracy of the filter paper
care should be taken in the timing of plot enumeration. technique for measurement of soil matric potential. Eur. J. Soil
Measurements of plots in the climatically wettest Sci. 46, 232–238.
274 T.R. Baker et al. / Forest Ecology and Management 171 (2002) 261–274

Eamus, D., 1999. Ecophysiological traits of deciduous and Phillips, O.L., Malhi, Y., Higuchi, N., Laurance, W.F., Núñez, P.V.,
evergreen woody species in the seasonally dry tropics. Trends Vásquez, R.M., Laurance, S.G., Ferreira, L.V., Stern, M.,
Ecol. Evol. 14, 11–16. Brown, S., Grace, J., 1998. Changes in the carbon balance of
Fanjul, L., Barajas, V.L., 1987. Diurnal and seasonal variation in tropical forests: evidence from long-term plots. Science 282,
the water relations of some deciduous and evergreen trees of a 439–442.
deciduous dry forest of the western coast of Mexico. J. Appl. Phillips, O.L., Malhi, Y., Vinceti, B., Baker, T.R., Lewis, S.L.,
Ecol. 24, 289–303. Higuchi, N., Laurance, W.F., Núñez Vargas, P., Vásquez
Garnier, E., Berger, A., 1986. Effect of water stress on stem Martinez, R., Laurance, S., Ferreira, L.V., Stern, M., Brown,
diameter changes of peach trees growing in the field. J. Appl. S., Grace, J., in press. Biomass dynamics of tropical forests:
Ecol. 23, 193–209. evaluating potential biases. Ecol. Appl.
Goldammer, J.G., Seibert, B., 1990. The impact of droughts and Reich, P.B., 1995. Phenology of tropical forests: patterns, causes
forest fires on tropical lowland rain forest of East Kalimantan. and consequences. Can. J. Bot. 73, 164–174.
In: Goldammer, J.G. (Ed.), Fire in the Tropical Biota. Springer, Reich, P.B., Borchert, R., 1982. Phenology and ecophysiology of
Berlin. the tropical tree, Tabebuia neochrysantha (Bignoniaceae).
Goldstein, G., Andrade, J.L., Meinzer, F.C., Holbrook, N.M., Ecology 63, 294–299.
Cavelier, J., Jackson, P., Celis, A., 1998. Stem water storage Reich, P.B., Borchert, R., 1984. Water stress and tree phenology in
and diurnal patterns of water use in tropical forest canopy trees. a tropical dry forest in the lowlands of Costa Rica. J. Ecol. 72,
Plant Cell Environ. 21, 397–406. 61–74.
Hall, J.B., Swaine, M.D., 1981. Distribution and Ecology of Schwertmann, U., Taylor, R.M., 1989. Iron oxides. In: Dixon, J.B.,
Vascular Plants in a Tropical Rain Forest. Dr. W. Junk Weed, S.B. (Eds.), Minerals in Soil Environments. Soil Science
Publishers, The Hague, the Netherlands. Society of America, Madison, WI.
Hawthorne, W.D., 1995. Ecological profiles of Ghanaian forest Sheil, D., 1995. A critique of permanent plot methods and analysis
trees. Trop. For. Pap. 29, 1–345. with examples from Budongo Forest, Uganda. For. Ecol. Mgmt.
Hinckley, T.H., Bruckerhoff, D.N., 1975. The effects of drought on 77, 11–34.
water relations and stem shrinkage of Quercus alba. Can. J. Sheil, D., 1997. Long-term growth and rainfall in a Ugandan moist
Bot. 53, 62–72. forest: seasonal rhythms and flexing stems. Comm. For. Rev.
Hinckley, T.M., Lassoie, J.P., Running, S.W., 1978. Temporal and 76, 121–127.
spatial variations in the water status of forest trees. For. Sci. Sobrado, M.A., 1986. Aspects of tissue–water relations and
Monogr. 20, 1–72. seasonal changes of leaf water potential components of
Holbrook, N.M., 1995. Stem water storage. In: Gartner, B.L. (Ed.), evergreen and deciduous species coexisting in tropical dry
Plant Stems: Physiology and Functional Morphology. Aca- forest. Oecologia 68, 413–416.
demic Press, San Diego, CA. Sobrado, M.A., 1993. Trade-off between water transport efficiency
Jackson, P.C., Cavelier, J., Goldstein, G., Meinzer, F.C., Holbrook, and leaf life-span in a tropical dry forest. Oecologia 96, 19–23.
N.M., 1995. Partitioning of water resources among plants of a Sokal, R.R., Rohlf, F.J., 1995. Biometry. Freeman, New York.
lowland tropical forest. Oecologia 101, 197–203. Swaine, M.D., 1996. Rainfall and soil fertility as factors limiting
Keeland, B.D., Sharitz, R.R., 1993. Accuracy of tree growth forest species distributions in Ghana. J. Ecol. 84, 419–428.
measurements using dendrometer bands. Can. J. For. Res. 23, Swaine, M.D., Hall, J.B., 1986. Forest structure and dynamics. In:
2454–2457. Lawson, G.W. (Ed.), Plant Ecology in West Africa. Wiley,
Kozlowski, T.T., Winget, C.H., 1964. Diurnal and seasonal Chichester, UK.
variation in radii of tree stems. Ecology 45, 149–154. Swaine, M.D., Lieberman, D., Hall, J.B., 1990. Structure and
Lieberman, D., 1982. Seasonality and phenology in a dry tropical dynamics of a tropical dry forest in Ghana. Vegetatio 88,
forest in Ghana. J. Ecol. 70, 791–806. 31–51.
Meinzer, F.C., Andrade, J.L., Goldstein, G., Holbrook, N.M., Swaine, M.D., Agyeman, V.K., Kyereh, B., Orgle, T.K., Thompson,
Cavelier, J., Wright, S.J., 1999. Partitioning of soil water J., Veenendaal, E.M., 1997. Ecology of Forest Trees in Ghana.
among canopy trees in a seasonally dry tropical forest. ODA, London.
Oecologia 121, 293–301. Veenendaal, E.M., Swaine, M.D., Blay, D., Yelifari, N.B., Mullins,
Orgle, T.K., 1994. Ecology of burnt forests in Ghana. Ph.D. C.E., 1996. Seasonal and long-term water soil water regime in
Dissertation. University of Aberdeen, Aberdeen, UK. West African tropical moist forest. J. Veg. Sci. 7, 473–482.
Oyama, M., Takehara, H., 1991. Revised Standard Soil Colour Wong, J.L.G., 1997. The State of Ghana’s Forests: Growth of the
Charts. Eijelkamp Agrisearch Equipment, Tokyo, Japan. Reserved forest. FRR Limited, Bristol, UK.
Pélissier, R., Pascal, J.-P., 2000. Two-year tree growth patterns Worrall, J., 1966. A method of correcting dendrometer measures of
investigated from monthly girth records using dendrometer tree diameter for variations induced by moisture stress change.
bands in a wet evergreen forest in India. J. Trop. Ecol. 16, 429–446. For. Sci. 12, 427–429.

You might also like