You are on page 1of 13

Forest Ecology and Management 188 (2004) 77–89

Spatial distribution of root length density and soil water of


linear agroforestry systems in sub-humid Kenya:
implications for agroforestry models
Simone Radersmaa,b,*, Chin K. Ongc
a
Department of Soil Quality, Agricultural University, P.O. Box 8005, 6700 EC Wageningen, The Netherlands
b
International Center for Research in Agroforestry (ICRAF), Maseno, Kenya
c
ICRAF, Nairobi, Kenya
Received 14 March 2002; received in revised form 6 May 2003; accepted 14 July 2003

Abstract

In simultaneous agroforestry systems trees can compete with crops for water, especially in semi-arid areas. However, in the
(sub)humid tropics, on P-fixing Oxisols/Ferralsols small decreases in soil water content caused a decrease in P-transport to roots
and therewith a soil-drying induced P-deficiency.
The aim of this study was to assess the spatial distribution of soil water content in crop fields bordering tree lines and its
relation with root length density distribution of the trees throughout the soil profile. To achieve this, soil water content and tree
root length densities throughout the soil profile were measured over a period of 2 years in an experiment with lines of four tree
species in the middle of maize fields in sub-humid western Kenya.
Soil water content was significantly reduced (2–7 vol.%) near two of the three fast-growing tree species, Eucalyptus grandis
and Grevillea robusta, but not near Cedrella serrata and the slower growing Markhamia lutea. These differences were related to
differences in water use. Eucalyptus and Grevillea showed high water use and Cedrella and Markhamia low water use. However,
soil water content distribution was not related to root length density distribution. Root length densities hardly decreased with
distance to Grevillea and clearly decreased with distance to Cedrella.
Most water-uptake models, including those of agroforestry models, assume that root length density distribution throughout the
profile is proportional to water extraction throughout the profile. The absence of a clear relation between root length density and
water extraction near Grevillea tree lines opposed this view. It can be explained by a decrease in water-potential gradient
between root and soil at increasing distance from the tree base. If the change in root length density is similar or smaller than the
change in water-potential gradient between root and soil, the decrease in water-potential gradient between root and soil is of
similar or larger importance for determining tree-water extraction distribution throughout the profile than root length density.
Thus, modeling of spatial agroforestry systems cannot assume a direct relation between tree-water extraction and root length
density, but needs to include decreasing water-potential gradient between root and soil along roots with increasing distance to the
stem base, especially over the horizontal dimension.
# 2003 Elsevier B.V. All rights reserved.

Keywords: Agroforestry; Water extraction; Root distribution; Simulation models

*
Corresponding author. Tel.: þ320-291-352; fax: þ320-230-479.
E-mail address: s.radersma@freeler.nl (S. Radersma).

0378-1127/$ – see front matter # 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.foreco.2003.07.021
78 S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89

1. Introduction is closely related to root length density (Rowse et al.,


1978; Belmans et al., 1979; Lascano and van Bavel,
Water deficiency can be an important limitation to 1984; Barataud et al., 1995). This relationship between
crop-growth, not only in semi-arid environments, but water extraction and root length is also used in agro-
also in environments where phosphorus (P) is the main forestry models like WaNuLCAS (Van Noordwijk
limitation to crop-growth, because transport of P in and Lusiana, 2000), Hypar (Mobbs et al., 1998) and
soil is highly dependent on soil water content. On a P- WIMISA (Mayus et al., 1999). Therefore, we consid-
fixing Oxisol/Ferralsol in western Kenya we found ered root length density distribution as main deter-
that a significant absolute soil water content reduction minant of soil water content distribution in the soil
of 2–3% (from 38 to 39 to 36 vol.%) at relatively high volume near trees.
pF, as measured close by tree lines, caused a decrease In the research described in this paper we examined
in maize-production of 30–40%. Using model simula- the profiles of water extraction with distance from tree
tions, the decrease in crop-growth near the tree lines lines and their relation with tree root distribution in a
could only be explained by decreased P-transport field experiment on a deep P-fixing Ferralsol/Oxisol in
towards roots at the lower soil water contents, with sub-humid western Kenya. Lines of four tree species
its cumulative effects on crop and crop-root growth. It selected for pole-production were grown within maize
could not be explained by direct competition for fields. During 2 years, soil water contents were mon-
nutrients or water (unpublished data). In this way, itored on a monthly basis and roots were sampled
trees affected crop-growth by soil-drying induced twice, over depth and with distance from the tree line,
P-deficiency. in order to determine (i) the effect of the tree lines on
Trees can affect soil water contents in different ways. soil water contents over depth and with distance from
They can increase soil water content by decreasing soil the tree line, (ii) differences in tree root distribution
evaporation (Wallace et al., 1999; Jackson and Wallace, among species and over time, and (iii) the relation
1999), by hydraulic lift (Caldwell and Richards, 1989; between root length density distribution and soil water
Dawson, 1993) and by hampering growth of other extraction.
plants with a higher water use per unit surface area.
They decrease soil water contents by their own water
use, or by interception of rain (Broadhead, 2000; Wal- 2. Material and methods
lace et al., in press). In tropical climates with an annual
rainfall of 1000–1500 mm, interception losses decrease 2.1. Site description
soil water inputs under the trees. However, this decrease
in water input to the soil is only slightly larger than An experiment was conducted on a degraded farm-
evaporation reductions under canopies, which decrease er’s field at Nyabeda, western Kenya (latitude 0.088N,
soil water losses under the trees. This leads to only a few longitude 34.248E, altitude 1300 m). The site was sub-
percent lower total water input under the tree canopies humid with an annual rainfall of about 1500 mm,
(Wallace et al., 1999) compared to water inputs outside distributed over two main cropping seasons; the long
the area shaded by tree canopies. Because of this rainy season from March to June/July (900 mm) and
balance between canopy effects on interception and the short rainy season from September to December/
evaporation in this climate, we considered tree water January (600 mm). The experiment was established
uptake as the main determinant of soil water depletion on level land, which after maize cropping was left
in the soil volume near tree lines. fallow the last year (three blocks)/last 2–3 years (one
Root density distribution of trees and crops is usually block).
seen as a main indicator of competition between trees The soil was an Oxisol/Ferralsol, which is a major
and crops (Schroth, 1995; Smith et al., 1999), assuming soil type in the area (Shepherd et al., 1992). Soil
that at equal supply of growth resources in the soil, the texture was 58% clay, 16% sand and 26% silt in the
resource uptake is related to the amount of root length top 0–15 cm. Clay contents increased to 73% at depth
per unit soil volume. Similarly, most crop and forest (3 m). Bulk density remained <1.3 kg dm3 over the
water-uptake models assume that water extraction rate first 1.6 m depth, due to high sesquioxide content and a
S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89 79

good soil structure. Soil pH (1:2.5 soil/water suspen- distance between maize rows and at increasing dis-
sion) was 5.6 over the entire depth, and Olsen-P was tance to the tree lines. Measurements were done in
1.2 mg kg1 in the top 15 cm, decreasing with depth all four agroforestry treatments and all four replicates
to 0.3 mg kg1. Soil organic carbon decreased from at 0.20 m intervals to 1.60 m depth. Neutron-probe
1.7% in the top-soil to 0.3% at 2–3 m depth. measurements were calibrated in the same field. First,
10 neutron-probe measurements were taken in a tube
2.2. Experimental design and management in a water-barrel (as usual), than each depth in an extra
tube in the field was measured seven times, and finally
The experiment was a 5  2 factorial design, with again 10 water-barrel measurements were taken. This
four replicates. The treatments were a complete fac- resulted together in the usual soil-count:water-count
torial combination of two phosphorus (P) levels: one ratios. On the same day four fixed volume rings per
with P (þP) and one without P (P) and five crop/ measuring depth were inserted horizontally in the soil,
agroforestry treatments: four treatments of a tree line dug-out, leveled to contain exactly 100 cm3, weighed
in the middle of maize plots and a sole maize treatment moist, oven-dried for 4 days at 108 8C, and weighed
without a tree line. The tree species were Cedrella dry. Seven access tubes all around the experiment were
serrata, Grevillea robusta, Eucalyptus grandis and used for calibration, three tubes in the driest periods,
Markhamia lutea. The þP treatments received an two tubes in the wettest periods and two tubes in
addition of 250 kg P ha1 at the start of the experiment periods with intermediate soil water contents.
and 50 kg P ha1 in April 2000. The P treatments Root densities were measured in three replicates
did not receive P-fertilizer. The plot size was 13:5 m of maize þ Grevillea, maize þ Eucalyptus, maize þ
15 m with the tree lines in the middle of the plots Cedrella and sole maize treatments at 15 and 27
parallel to the 13.5 m length side. Trees were trans- months after transplanting the tree-seedlings. Mea-
planted to the field experiment in April 1997 at an age surement dates coincided with maize tasseling in the
of 4–6 months. Tree spacing was 0.5 m in the first year long rainy season of 1998 and 1999. A bucket-auger of
thinned to 1 m thereafter. 0.10 m diameter (Ø) was used to sample pairs of
Nitrogen (70 kg ha1 per year) and potassium points (one within a maize row and one in between
(15 kg ha1 per year) were applied regularly in all two maize rows) for the layers 0.10–0.30, 0.30–0.50,
plots, to prevent these elements from becoming a 0.50–1.00, 1.00–1.50 and 1.50–2.00 m (in 1998 the
limitation to crop and tree growth. 1.00–2.00 m was sampled as one layer). In 1998 the
Trees were pruned on 22 December 1997, 14 May average distance of the pairs of points was 0, 0.94,
and 30 September 1998, 20 April and 27 October 2.44 and 4.69 m, like the distances of soil water
1999, according to farmers’ practice to produce poles content measurements. In 1998 when the trees were
(cutting all branches up to about 3/4 of the tree height), still small and the spread of roots limited, samples
to decrease shading of crops. The plots were weeded were taken at 4.69 m instead of 6.94 m and a sampling
manually by hoe, twice during a cropping season. at 0 m was included. In 1999 the 0 m distance was quit
Growth of tree roots into neighboring plots was because of the too large woody roots at the base of the
avoided by trenching up to 1.5 m depth in February trees and the 6.94 m distance was added because of
1999 and February 2000. the wider spread of tree roots at this age. The layer
0–0.10 m was not sampled, because weeding twice
2.3. Measurements during the cropping season cut most roots. If an auger
hole could not reach the specified depth due to occa-
Soil water contents were measured by neutron sional stones, final depth of the hole was measured and
probe (Didcot Instruments, Wallingford, UK), at sample volume adjusted accordingly before calculat-
monthly intervals between July 1998 and August ing root length densities. Both sides of the tree lines
2000 in all treatments. Access tubes for these mea- were sampled in this way. In 1998 the samples of the
surements were installed at three distances from two sides were taken separately, total root length
the tree line, 0.94, 2.44 and 6.94 m. These distances measured per sample and divided by the total volume
were chosen to be at one-quarter/three-quarter of the of each sample, after which the mean of both sides was
80 S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89

used for further analysis. In 1999, samples from both After adjustment of all root length densities to this
sides were bulked, total root length measured and average S, the means of the three replicates of root
divided by the total volume of the two sides. length density profiles with distance from the tree
Samples of depth 1.00–2.00 m, which were too line were analyzed by non-linear fitting to derive
large volumes (>20 l), were mixed thoroughly in the the variables of the elliptical function as described
field and a sub-sample of half or quarter of the total by (Van Noordwijk and Lusiana, 2000):
weight was taken.
rldði; jÞ
Buckets containing soil þ root samples were soaked
overnight. The next day the mixture was washed ¼ X0  DD expðDDðDepth2j þ ðDS  Disti Þ2 Þ1=2 Þ
through a double sieve with mesh size 0.5 mm for
(1)
the bottom sieve. Further measurement of root length
densities was done according to the procedure of hand where rld(i, j) is the root length density at distance i
sorting, staining, spreading and scanning following the and depth j in cm cm3, X0 the total root length per
method described by Smith et al. (1999). unit area (cm cm2) at a distance of 0 m from the tree
stem, DD the parameter (cm1) for the distribution of
2.4. Statistical analysis and calculations total root length per unit area with depth, and DS the
dimensionless parameter governing the shape of the
Soil water content data of the four most contrasting tree root system (when rld at 5 m distance is the same
depths: 0.20, 0.60, 1.20 and 1.60 m, were analyzed as at 1 m depth than DS ¼ 0:2)
separately. Data for measurement times either within
the wet seasons or within the dry seasons showed
similar trends as far as differences in soil water con- 3. Results and discussion
tents with distance from the tree lines were concerned.
Therefore the average of all measurement times in the 3.1. Tree growth
last three wet seasons (long rains of 1999, short rains
of 1999 and long rains of 2000) and the average of Fig. 1 shows the growth in height and diameter of
all measurements times in the last two dry seasons the trees over time in the different treatments. Growth
(dry season 1999 and dry season 2000) were used for rates of Eucalyptus þP (þP ¼ with P-application)
analysis of soil water content differences with distance were clearly the highest, followed by Eucalyptus
from the tree line (Dd), with Dd ¼ yfar  ynear , and yfar P (P ¼ without P-application), Cedrella (þP and
the average volumetric soil water content of the mea- P) and Grevillea (þP and P). Grevillea trees were
surements at 2.44 and 6.94 m away from the tree line slightly taller than Cedrella trees, but Cedrella had a
and ynear the average volumetric soil water the content larger diameter than Grevillea. Markhamia grew very
at 0.94 m away from the tree line. These Dd values little compared to the other trees. The slow-down in
were analyzed by ANOVA as 5  2 factorial design diameter-growth between the last two measurements
with four replicates. before 900tp (¼900 days after transplanting), is due to
Because the trees were not all of the same size, the a severe pruning event (April 1999). The first mea-
root density data for each plot were adjusted to the surement after 900tp marks another change: the height
average of basal stem cross-sectional surfaces (S) of of diameter measurements changed from 0.05 to
all trees per treatment (average of three replicates). S is 0.30 m above soil surface.
related to total cross-sectional area of finer roots and
root length according to the fractal branching theory 3.2. Soil water content
(Van Noordwijk et al., 1994). Thus root densities in
each plot/replicate were multiplied by the ratio of Average soil water contents over 2 years in sole
Smeasured:Saverage, where Smeasured is the sum of the maize fields are shown in Fig. 2. The small time-lag
surface of the three tree-stem cross-sections closest to in drying and wetting cycles of the different soil
the sampling transect in one plot/replicate, and Saverage depths, shows that hydraulic conductivity of the soil
the average of the three replicates of Smeasured. is quite high, in both wet and dry circumstances, as
S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89 81

Tree height over time Tree diameter over time

11,0 120
10,0 110
9,0 100
8,0 Mark.-P
90

basal diameter (mm)


Cedrel.-P
7,0
Grevil.-P 80
height (m)

6,0 Eucal.-P
70
5,0 Mar.+P
Cedrel.+P 60
4,0
Grevil.+P
50
3,0 Eucal.+P
2,0 40

1,0 30

0,0 20
450 900 1350 450 900 1350
20/7/98 15/10/99 8/1/00 20/7/98 15/10/99 8/1/00
tp (days) and date tp (days) and date

Fig. 1. Tree height and tree-stem diameter over time, with tp as time in days after transplanting to the field: Mark: ¼ Markhamia lutea;
Cedrel: ¼ Cedrella serrata; Grevil: ¼ Grevillea robusta; Eucal: ¼ Eucalyptus grandis; P ¼ without phosphorus application; þP ¼ with
phosphorus application. Standard errors for tree height and basal stem diameter were between 0.4 and 1% of the means plotted in the figures.

may be expected for an Oxisol/Ferralsol (El-Swaify, layer (around 11%), with y at a mean value of 32%. At
1980). 0.60 and 1.60 m depth amplitudes were lower, 6 and
The amplitude and average level of soil water 3.5 vol.%, respectively, with y at higher mean values
contents during the dry and wet season, differed for of 43 and 38%, respectively. In between these layers,
the different layers. The amplitude in volumetric soil at 1.20 m depth, the amplitude was larger again
water content (y) was the largest for the upper soil (10 vol.%) with y at a mean value of 38%.

50

45
vol.% soil water

40 depth 0.20 m
depth 0.60 m
35 depth 1.20 m
depth 1.60 m
30

25
490 690 890 1090
1/9/98 19/3/99 5/10/99 22/4/00
tp (days) and date

Fig. 2. Mean volumetric soil water content and standard errors in sole maize fields over time, with tp as time after transplanting.
82 S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89

Because water contents at 2.44 and 6.94 m distance of soil water content at 2.44 and 6.94 m minus the soil
away from the tree lines did not yet differ significantly, water content at 0.94 m distance from the tree lines,
their means were used for further analyses to reduce was used to calculate Dd values, shown in Table 1a
spatial variability in soil water content. This average and b. Thus Dd ¼ y244694  y94 and Dd > 0 if
Table 1
Changes in soil water contents (Dd) in maize fields approaching the tree line (average over last three (a) wet seasons and (b) dry seasons),
calculated by Dd ¼ yfar  ynear, with yfar as the volumetric soil water content far from the tree lines (mean of measurements at 2.44 and 6.94 m
distance from the tree lines) and ynear as the volumetric soil water content at 0.94 m from the tree linesa

Species P-supply Dd (vol.%)

Depth 0.20 m Depth 0.60 m Depth 1.20 m Depth 1.60 m

(a) Average over last three wet seasons: March–July 1999, September–November 1999 and March–July 2000
Cedrella þP 0.13 0.34 1.39 0.58
Cedrella P 0.00 1.28 0.43 0.92
Eucalyptus þP 2.51 2.72 6.92 3.86
Eucalyptus P 1.68 1.22 2.45 1.07
Grevillea þP 0.57 1.22 3.17 1.28
Grevillea P 0.23 0.50 2.07 2.40
Markhamia þP 0.72 0.51 0.61 0.71
Markhamia P 0.57 0.93 0.22 0.56
Sole maize þP 0.60 0.34 0.56 0.9
Sole maize P 0.94 0.48 0.91 0.17
S.E.D. 1.423 1.108 1.309 1.207
Cedrella 0.07 0.47 0.48 0.17
Eucalyptus 2.09 1.97 4.69 2.46
Grevillea 0.17 0.86 2.62 1.84
Markhamia 0.64 0.72 0.19 0.63
Sole maize 0.17 0.07 0.74 0.54
S.E.D. 1.006 0.784 0.925 0.853
þP 0.85 0.75 2.53 1.46
P 0.22 0.88 0.96 0.66
S.E.D. 0.636 0.496 0.585 0.54
(b) Average over last three dry seasons: December 1998–February 1999 and December 1999–February 2000
Cedrella þP 0.58 0.14 1.12 0.60
Cedrella P 0.09 0.91 0.11 1.13
Eucalyptus þP 0.24 0.88 3.51 3.11
Eucalyptus P 1.31 0.61 0.66 0.99
Grevillea þP 0.97 0.34 2.16 1.01
Grevillea P 1.54 0.58 2.44 2.74
Markhamia þP 1.80 0.34 1.22 0.94
Markhamia P 0.35 0.53 0.18 0.38
Sole maize þP 1.17 0.57 0.27 0.22
Sole maize P 0.56 0.73 1.95 0.28
S.E.D. 1.535 1.255 1.677 1.316
Cedrella 0.24 0.39 0.50 0.27
Eucalyptus 0.54 0.13 2.08 2.05
Grevillea 1.25 0.46 2.30 1.87
Markhamia 0.72 0.10 0.52 0.66
Sole maize 0.31 0.08 1.11 0.25
S.E.D. 1.085 0.888 1.186 0.931
þP 0.33 0.03 1.17 1.18
P 0.74 0.43 1.03 0.65
S.E.D. 0.686 0.561 0.750 0.589
a
A significant positive Dd indicates a decrease in soil water content near the tree line.
S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89 83

y244694 > y94 or positive Dd values indicate decrea- (0.20 m depth and 0.60 m depth) were generally not
sed soil water content near the tree lines. Table 1a significantly different from zero ðDd < 2 S:E:D:Þ, and
shows the Dd values for the means of the three wet thus soil water contents did not change significantly
seasons, (which showed similar trends and were there- with distance from the tree lines. The Dd values in the
fore averaged). Positive Dd values were the clearest lower layers (1.20 and 1.60 m depth for Eucalyptus
(2.5–6.9%) in Eucalyptus þP, and significantly dif- þP and 1.60 m for Grevillea P) were still significant
ferent from 0 (Dd > 2 S:E:D: at a ¼ 0:05) in all layers in the dry season.
except the first. Positive Dd values (1.2–2.5%) seemed The decrease in Dd in the dry season, indicating
to occur in all layers of Eucalyptus P as well, but decreased differences in soil water content close by
these were not significant ða ¼ 0:05Þ. Positive Dd and far from the trees can be explained as follows.
values (1.2–3.2%) occur in the three lower layers of Over moist and wet soil water content ranges (like in
Grevillea þP, but are only significant at 1.20 m depth. our wet season), root resistance usually exceeded soil
In Grevillea P positive Dd values occurred in the resistance (Belmans et al., 1979; Reid and Hutchinson,
two deepest layers, but this Dd was only significant at 1986) and suction by the root was larger than suction
1.60 m depth. by the soil. The situation in which soil resistance
The significance of Dd values could be partly exceeds root resistance occurs only in drier soil or
explained by tree-growth. Eucalyptus þP was the at high transpiration rates in coarse soils (Hillel et al.,
largest and showed the largest Dd. Markhamia was 1976; Weatherley, 1979; Hainsworth and Aylmore,
very small and did not show any significant Dd. 1989). Soil suction becomes higher than root suction
Another part of the differences in Dd among species when the soil dries out, causing plant water uptake to
could be explained by daily water use of a species at a cease. This happened likely earlier in drier soil near
particular size. Cedrella and Grevillea of about the the tree lines than in the wetter soil far from the tree
same size, showed different daily water use (unpub- lines and caused the decrease in Dd values in the dry
lished data). Decreased water use by maize, due to its seasons as compared to the wet season.
slow growth caused by competition near especially
Eucalyptus, would have counteracted the development 3.3. Tree roots
of large Dd values in the top layers.
If significant positive Dd values of 2–7 vol.% are The elliptical-function parameters differed per spe-
compared with the normal annual amplitude (Amp) in cies but were not affected by P treatment. Therefore,
a sole maize field (Fig. 2), by calculating Dd:Amp. the root length density (rld) data were multiplied by a
The Dd:Amp at depth 0.20 m is 2.51%:11% ¼ 23%, factor Smeasured:Saverage for all three þP and P repli-
at depth 0.60 m is 2.71%:6% ¼ 45%, at depth cates (where S is the cross-sectional surface of the
1.20 m is 6.92%:10% ¼ 70% and at depth 1.60 m is stem base and the calculation is similar to the one
3.86%:3:5% > 100%. This shows that the soil near described in Section 2) before taking the average of
Eucalyptus þP dried out to 25–100% of the normal þP and P and determining the ellipse function
annual soil water content minimum in sole maize parameters of the species as given in Table 2, or
fields. Near the other tree lines, drying was less severe, plotting root distribution per species in Fig. 3. Because
but compared with seasonal fluctuations it was still of the exponential decrease in root length density over
considerable. depth, the depth taken to represent the layers in Fig. 3
A reason for the strong tree-caused drying of deeper is at 1/4 of the layer thickness below the upper
layers compared to seasonal drying at depth may have boundary of the layer, e.g. the average rld of layer
been that the roots caused a ‘shortcut’ contact between 0.1–0.3 m depth is plotted at 0.15 m depth.
the deep layers and the atmosphere. This shortcut was The elliptical parameter, X0, is the total root length
absent without roots (e.g. maize field in dry season), per unit soil surface (cm cm2) at distance 0 m from the
and caused that the low soil conductivity determined tree line. Thus high maximum and average rld values,
the drying rate. especially near the stem base are reflected in higher X0
The Dd values in the dry season are shown in values. The X0 values were partly dependent on tree-
Table 1b. The Dd values for the first two layers size (average cross-sectional surface of stem base, Sav),
84 S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89

Table 2
Ellipse function parameters of Eq. (1), rldði; jÞ ¼ X0  DD expðDDðDepth2j þ ðDS  Disti Þ2 Þ1=2 Þ, all calculated with depth and distance
in (cm)a

1998 1999 1998 1999 1998 1999

Sav X0 S.E. Sav X0 S.E. DD S.E. DD S.E. DS S.E. DS S.E.


(cm2) (cm cm2) (cm2) (cm cm2) (cm1) (cm1) (–) (–)

Cedrella 13.5 23.0 6.76 49.8 15.3 5.51 0.025 0.0059 0.027 0.0040 0.21 0.055 0.14 0.031
Eucalyptus 11.7 19.0 5.92 55.9 16.1 3.93 0.033 0.0057 0.020 0.0030 0.14 0.039 0.17 0.032
Grevillea 5.2 10.1 1.78 30.4 4.1 0.85 0.034 0.0034 0.022 0.0036 0.04 0.033 1E11 *

With rld(i, j) as root length density at distance i and depth j from the stem base, X0—total root length per unit area (cm cm2) at a
a

distance of 0 cm from the tree stem, DD—parameter (cm1) governing the decrease with depth of root length density (for X ¼ 0 at a depth of
0.699/DD the root length density has half of its value at the soil surface), DS—parameter governing the shape of the tree root system (when rld
at 5 m distance is the same as at 1 m depth, then DS ¼ 0:2) and Sav—mean basal stem surface of trees next to root-sampling transect.

and because Eucalyptus and Cedrella were much larger decrease rapidly with increasing distance from the
than Grevillea, their X0 value was larger as well. How- tree line and rld lines in Fig. 3 should be far apart.
ever, when the two sampling times (1998 vs 1999) were Thus, the very low DS of Grevillea, which was not
compared, there was no relation between Sav and X0. In significantly different from 0, and the overlap of the
1999 the trees and the Sav were larger, but the X0 was rld lines in Fig. 3, showed that Grevillea root length
lower than in 1998. Fig. 3 also shows that absolute density did hardly decrease with increasing distance to
root densities were lower in 1999 compared to 1998. the tree. DS values of Eucalyptus and Cedrella were
The decrease in absolute root length densities and in significantly higher than 0, hence in the figure the rld
X0 values of the elliptical function, was likely due to lines at increasing distance to the trees became closer
tree-management and root-system growth dynamics to the Y-axes, and root length density decreased with
as also found by Smith et al. (1999) and Odhiambo increasing distance to the trees. DS values did not
et al. (1999). Before the long rains of 1999, tree cano- change significantly between 1998 and 1999.
pies were severely pruned, which likely caused root- Root distribution profiles with distance from a tree
mortality and thus reduced the X0:Sav ratio. line differed between tree species, and changed over
The DD variable describes the shape of the vertical time. Over a period of 1 year the root systems generally
root profile. A high DD value in Table 2 is represented extended horizontally without similar vertical exten-
by a less steep slope of the rld lines in Fig. 3. The sion. High root density at the first root sampling in 1998
DD values were not significantly different for the tree might indicate that if a tree grows fast over a short
species. However, the DD value of 1999 root-data was period of time in a wet period, its increase in root length
significantly smaller than that of 1998 for Eucalyptus increases root length density, thus intensifying extrac-
and Grevillea but did not change for Cedrella. This tion of resource from a small soil volume. On the other
implies that for Eucalyptus and Grevillea relatively hand after a dry season and a severe pruning event, fine
more roots had grown at depth than in the surface roots may have died and new roots grew mainly to
layers. Smith et al. (1999) mentioned a relatively extend the root system in order to cover a larger soil
high root length density of Grevillea at depth as well. volume. Fitter et al. (1991) mentioned that a rooting
However, the exponential decrease in rld of Grevillea architecture with long links between branching points
over depth they mentioned, is much less than what we and branching predominantly on the main axis,
found, likely due to the dryer environment of their although expensive in use of carbohydrates, would
experiment. mainly occur in nutrient poor soils. This indicates that
The DS variable of the elliptical function describes changes in root distribution over time may not only be
the shape of horizontal root distribution in relation to caused by management, but also by a root-system
the vertical distribution. If DS is 0, root densities do distribution strategy adapted to depleted soils. Whether
not decrease over distance, and the rld lines in Fig. 3 caused by management or by resource availability, not
should overlap. If DS is large, the root densities only total root length and extent changed over time, but
S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89 85

Eucalyptus (1998) Grevillea (1998) Cedrella (1998)


-3 -3 -3
rld (cm cm ) rld (cm cm ) rld (cm cm )

0,001 0,010 0,100 1,000 0,001 0,010 0,100 1,000 0,001 0,010 0,100 1,000
0 0 0

20 20 20

0m

0m

0m
40 40 40
0.94 m

0.94 m

0.94 m
depth depth depth
(cm) 60 (cm) 60 (cm) 60

80 80 80
2.44 m

2.44 m

2.44 m
100 100 100

120 120 120


4.69 m

4.69 m

4.69 m
(a) 140 140 140

Eucalyptus (1999) Grevillea (1999) Cedrella (1999)


-3 -3 -3
rld (cm cm ) rld (cm cm ) rld (cm cm )

0,001 0,010 0,100 1,000 0,001 0,010 0,100 1,000 0,001 0,010 0,100 1,000
0 0 0

20 20 20

0.94 m
0.94 m

0.94 m

40 40 40

60 60 60

depth depth depth

2.44 m
2.44 m

2.44 m

(cm) 80 (cm ) 80 (cm ) 80

100 100 100


4.69 m

4.69 m

4.69 m
120 120 120

140 140 140

160 160 160


6.94 m

6.94 m

6.94 m

(b) 180 180 180

Fig. 3. Tree root length density (rld) distribution and standard errors in (a) 1998 and (b) 1999.

also tree root distribution. This change of root distribu- 3.4. Root length densities and soil water content
tion profiles over time may pose a problem to agrofor-
estry models, like WaNuLCAS, because the elliptical Comparing the root profiles of Table 2 and Fig. 3
parameters derived from measurements at one point with tree-water extraction as indicated by Dd in the
in time may not be valid during the whole growth of wet season (Table 1a), tree-water extraction and root
the tree. density distribution show no clear relation. We see that
86 S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89

the root density of Grevillea hardly changed with root length density, and our observation near Grevillea
distance from the tree line, but soil water contents tree lines that root length density distribution and
were significantly lower close by the Grevillea tree distribution of water extraction are not closely related,
line ðDd > 0Þ. On the other hand, the root density of can be explained by Fig. 4a. In Fig. 4a water flow in a
Cedrella decreased with distance to the tree line but branching root, is represented by a branched electrical
soil water contents were not significantly different circuit. The principles used here are those of water
when measured close by or far from the tree line. extraction by plants analogous to Ohm’s electrical
This absence of a relation between root density dis- circuit, with a potential gradient as driving force
tribution and soil water contents is in contrast with the and resistances limiting the flow rate: q ¼ Dc=r
relations used in most computer models (Rowse et al., (Hillel et al., 1976; Scott Russell, 1977), with q as
1978; Belmans et al., 1979; Lascano and van Bavel, flux/flow, Dc as potential gradient and r as resistance.
1984; Barataud et al., 1995), the general descriptions Fig. 4a represents a main root with branches, in which
of how water extraction is modeled (Campbell, 1991; each branch/resistance between main flow path and
Simmonds and Kuruppuarachchi, 1995), and with the earth represents a similar total root length (case 1).
way it is employed in agroforestry models such as Fig. 4a can also represent a tiny piece of root in which
WaNuLCAS (Van Noordwijk and Lusiana, 2000) each branch/resistance between main flow path and
Hypar (Mobbs et al., 1998) and WIMISA (Mayus earth represents a channel for radial water inflow
et al., 1999). In all these models water uptake is into the root (case 2). The þ side of the electric circuit
closely related to relative root length density. in Fig. 4a represents the stem base. The earth-sign
We consider soil water content in the wet season to (potential c ¼ 0) represents sites where water uptake
be mainly determined by tree and crop water uptake from the soil into the root takes place. The resistances
rather than by rainfall interception of the trees. This (r) connecting earth to main flow path are all chosen
assumption is supported by a water balance study of to be the same size ðr ¼ 1Þ. This same resistance size
G. robusta–maize agroforestry systems in semi-arid shows that either the total root length after each
Kenya (750 mm of annual rainfall), which showed branching point is equal (case 1) like the rld of
that the small reduction in rainfall input due to canopy Grevillea with distance from the tree line, or radial
interception was partly offset by a reduction in soil resistance along the root is the same (case 2) whether
evaporation (Wallace et al., 1999). It is therefore close by or further from the stem base. The resistances
unlikely that interception explains the large reduction along the main flow path are chosen to decrease
in soil water content observed in our present study and towards the þ, like can be expected from the axial
we ascribe water extraction from soil by the tree as the resistance of a main root towards the stem base. After
main cause of decreased soil water contents near the this choice of resistances and earth, the flux (q) and
tree lines. the potential (c) along the different parts of the
The absence of Dd in fields with a Cedrella tree line, branched electrical circuit were calculated by
although the tree roots were concentrated near the q ¼ Dc=r. This shows that with increasing distance
stem base, may be explained by Cedrella’s relatively away from the þ or stem base, the potential (c)
low water use (unpublished data) and low water decreases resulting in a drop in influx (q) through
extraction from soil. This low water use is likely the resistances connecting earth to main flow path
easily balanced by precipitation in the wet season, ðr ¼ 1Þ. Thus, although the root length density (case
as shown by soil water content values, which are not 1) or radial resistance (case 2) is the same close by the
significantly different from those in sole maize fields. stem base (þ) as further away from the stem base, the
However, the clear Dd in fields with a Grevillea tree water inflow (q) decreases with increasing distance to
line, showed that water extraction near the stem base the stem base, due to a decrease in potential gradient
was higher than far from the tree, although root length between root and soil.
densities did not decrease with distance to the stem If these small sections were multiplied to form a
base. whole root, the decline in potential gradient between
The discrepancy between the general assumption root and soil (or þ and earth) and water uptake with
(and models), that water uptake is closely related with distance from the stem base would show a logarithmic
S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89 87

Fig. 4. (a) Model of an electric circuit representing a branching root, in which the root-branches have similar hypothetical uptake-resistances
(chosen r ¼ 1) and the main root has smaller hypothetical axial resistances (chosen r ¼ 1=2; 1=3; 1=4). The earth-sign ðc ¼ 0Þ represents the
water-uptake site/soil–root interface. The flux (q) and potential (c) at different sites along the branched circuit/root were calculated by
q ¼ Dc=r. (b) Hypothetical course of decrease in hydraulic potential along a root with increasing distance to the stem base.

course like in Fig. 4b. This course can be described by Grevillea tree base, differences in other flux determin-
a function exy, in which x is the distance and y would ing factors like water-potential gradient between root
be a function of radial resistance, root-branching and and soil gain importance.
axial resistance. Taylor et al. (1992) and Doussan et al. (1999) show
The origin of the assumption that water uptake is that indeed there is a severe drop in water-potential
closely related to root length density can be found in gradient between root and soil along roots (cotton,
earlier soil water extraction models. These models peach) away from the stem base. Decreasing water
were made for crop fields or forest stands with a one- uptake per unit root length as a consequence of
dimensional profile, in which root density, water- increasing axial resistance and decreasing water-
potential gradient between root and soil and water potential gradient between root and soil has also been
extraction only varied vertically, with depth. Gener- mentioned by Hainsworth and Aylmore (1989). From
ally, root length densities decreased exponentially water profile figures of Green and Clothier (1999) it is
over depth, hence water-uptake resistance of the root clear that the highest water extraction occurred around
system increased exponentially over depth, and was the tree bases, although root densities did not show
the main determinant of soil water extraction. As a horizontal variation with distance to the tree base in
result, in models with only a vertical dimension, water their orchard.
extraction was modeled as proportional to (relative) Decreases in axial resistance and potential gradient
root length density. However, if the root length density between root and soil occur in the vertical direction
does not decrease, like horizontally away from the as well. However, in the vertical direction the decrease
88 S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89

in water-potential gradient between root and soil is Campbell, G.S., 1991. Simulation of water uptake by plant
roots. In: Hanks, J., Ritchie, J.T. (Eds.), Modelling Plant
smaller than the exponential decrease in root length
and Soil Systems. ASA/CSSA/SSSA, Madison, WI, pp. 273–
density causing the exponential increase in radial resis- 285.
tance. Therefore over the vertical direction, changes Dawson, T.E., 1993. Hydraulic lift and water use by plants:
in water potential between root and soil could be implications for water balance performance and plant–plant
neglected. Nevertheless, Belmans et al. (1979) mention interactions. Oecologia 95, 565–574.
that the fit of modeled and experimental data would Doussan, C., Vercambre, G., Pages, L., 1999. Water uptake by two
contrasting root systems (maize, peach tree): results from a
probably be better if axial resistance and therewith model of hydraulic architecture. Agronomie 19, 255–263.
change in water-potential gradient between root and El-Swaify, S.A., 1980. Physical and mechanical properties of
soil was included. Oxisols. In: Theng, B.K.G. (Ed.), Soils with Variable Charge.
Summarizing, if the root length density decreases New Zealand Society of Soil Science, Palmerston North,
pp. 303–323.
far more than the decrease in water-potential gradient
Fitter, A.H., Stickland, T.R., Harvey, M.L., Wilson, G.W., 1991.
between root and soil, water extraction can be mod- Architectural analysis of plant root systems. I. Architectural
eled as inversely proportional to (relative) root length correlates of exploitation efficiency. New Phytol. 118, 375–
densities. In the soil volume this occurs usually in the 382.
vertical direction. However, if root length density does Green, S., Clothier, B., 1999. The root zone dynamics of water
not change far more than water-potential gradient uptake by a mature apple tree. Plant Soil 206, 61–77.
Hainsworth, J.M., Aylmore, L.A.G., 1989. Non-uniform soil water
between root and soil, the decrease in water-potential extraction by plant roots. Plant Soil 113, 121–124.
gradient between root and soil and/or axial resistance Hillel, D., Talpaz, H., Van Keulen, H., 1976. A macroscopic scale
needs to be included in attempts to model water model of water uptake by a non-uniform root system and of
uptake, to avoid large errors in water-uptake distribu- water and salt movement in the soil profile. Soil Sci. 121 (4),
tion. This occurred near our Grevillea tree lines in the 242–255.
Jackson, N.A., Wallace, J.S., 1999. Soil evaporation measurements
horizontal direction. in an agroforestry system in Kenya. Agric. For. Manage. 94,
203–215.
Lascano, R.J., van Bavel, C.H.M., 1984. Root water uptake and soil
Acknowledgements water distribution: test of an availability concept. Soil Sci. Soc.
Am. J. 48, 233–237.
Mayus, M., Van Keulen, H., Stroosnijder, L., 1999. A model of
Thanks are due to Sammy Kyalo and Japheth tree–crop competition for windbreak systems in the Sahel:
Kyengo and all the casuals for field and root work, description and evaluation. Agrofor. Syst. 43, 183–201.
to Meine van Noordwijk, Jan Goudriaan and Herman Mobbs, D.C., Cannell, M.G.R., Crout, N.M.J., Lawson, G.J.,
van Keulen for their analysis suggestions, and to the Friend, A.D., Arah, J., 1998. Complementarity of light
KEFRI/ICRAF EU-funded project and the Dutch and water-use in tropical agroforests. I. Theoretical model
outline, performance and sensitivity. For. Ecol. Manage. 102,
government covering the labor-costs of this research. 259–274.
Odhiambo, H.O., Ong, C.K., Wilson, J., Deans, J.D., Broadhead, J.,
Black, C., 1999. Tree–crop interactions for below-ground
References resources in drylands: root structure and function. Ann. Arid
Zone 38 (3), 221–237.
Reid, J.B., Hutchinson, B., 1986. Soil and plant resistances to water
Barataud, F., Moyne, C., Breda, N., Granier, A., 1995. Soil
uptake by Vicia faba L. Plant Soil 92, 431–441.
water dynamics in an oak stand II. A model of the soil–root
Rowse, H.R., Stone, D.A., Gerwitz, A., 1978. Simulation of the
network compared with experimental data. Plant Soil 172,
water distribution in soil. II. The model for cropped soil and its
29–43.
comparison with experiment. Plant Soil 49, 533–550.
Belmans, C., Feyen, J., Hillel, D., 1979. An attempt at experimental
Schroth, G., 1995. Tree root characteristics as criteria for species
validation of macroscopic-scale models of soil moisture
selection and systems design in agroforestry. Agrofor. Syst. 30,
extraction by roots. Soil Sci. 127, 174–185.
125–143.
Broadhead, J., 2000. Ecophysiology of Indigenous Trees in
Scott Russell, R., 1977. Plant Root Systems: Their Function and
Agroforestry Systems in the Semi-arid Tropics. University of
Interaction with the Soil. McGraw-Hill, London, 298 pp.
Nottingham, Nottingham, 259 pp.
Shepherd, K.D., Swinkels, R., Muturi, W.M., Ohlsson, E., Ndufa,
Caldwell, M.M., Richards, J.H., 1989. Hydraulic lift: water efflux
J.K., 1992. Evaluation of hedgerow intercropping on farms in
from upper roots improves effectiveness of water uptake by
western Kenya. I. Diagnosis of technology potential and
deep roots. Oecologia 79, 1–5.
S. Radersma, C.K. Ong / Forest Ecology and Management 188 (2004) 77–89 89

farmers interest. AFRENA Report No. 50. ICRAF, Nairobi, Van Noordwijk, M., Spek, L.Y., De Willigen, P., 1994. Proximal
Kenya, pp. 1–29. root diameters as predictors of total root system size for fractal
Simmonds, L.P., Kuruppuarachchi, D.S.P., 1995. Dynamics of plant branching models. I. Theory. Plant Soil 164, 107–117.
water uptake. In: Kabat, P., Marshall, B., van den Broek, B.J., Wallace, J.S., Jackson, N.A., Ong, C.K., 1999. Modelling soil
Vos, J., Van Keulen, H. (Eds.), Modelling and Parameterization evaporation in an agroforestry system in Kenya. Agric. For.
of the Soil–Plant–Atmosphere System, A Comparison of Potato Meteorol. 94, 189–202.
Growth Models. Wageningen Pers, Wageningen, pp. 77–101. Wallace, J.S., Young, A., Ong, C.K., in press. The potential of
Smith, D.M., Jackson, N.A., Roberts, J.M., Ong, C.K., 1999. Root agroforestry for sustainable land and water management.
distributions in Grevillea robusta–maize agroforestry system in In: Proceedings of the UNESCO/IUFRO Workshop on Forests–
semi-arid Kenya. Plant Soil 211, 191–205. Water–People in the Humid Tropics: Past, Present and Future
Taylor, H.M., Upchurch, D.R., McMichael, B.L., 1992. Root Hydrological Research for Integrated Land and Water Manage-
hydraulic resistance: implications in modelling nutrient and ment, Kuala Lumpur, Malaysia, July 30–August 4, 2000. J.
water uptake. J. Plant Nutr. 15 (6–7), 727–736. Hydrol.
Van Noordwijk, M., Lusiana, B., 2000. Wanulcas Version 2.0, Weatherley, P.E., 1979. The hydraulic resistance of the soil–root
Background on a Model of Water, Nutrient and Light Capture interface—a cause of water stress in plants. In: Harley, J.L.,
in Agroforestry Systems. International Centre for Research in Scott Russell, R. (Eds.), The Soil Root Interface. Academic
Agroforestry, Bogor, Indonesia, 186 pp. Press, London, pp. 275–286.

You might also like