C.K. Ong et al. Agriculture, Ecosystems and Environment 80 2000 121–141 123
runoff or deep drainage resulting from increased shad- ing, reduced soil temperature, the windbreak effect of
the trees, or increased abstraction of water at depth or during the dry season would increase the proportion of
rainfall used for transpiration. A possible disadvantage is that interception losses resulting from the evapora-
tion of rainfall from the tree canopy may range from 10 to 30 in agroforestry systems Ong and Black,
1994, although interception losses are lower when the tree canopy is sparse 5–10; Wallace et al., 1995.
The hypothesis that agroforestry may increase pro- ductivity by capturing a larger proportion of the annual
rainfall Ong et al., 1992 was supported by the Hyder- abad studies, which demonstrated that improvements in
annual rainfall utilisation from 40 to 80 were possi- ble in perennial pigeonpea Cajanus cajangroundnut
agroforestry systems, primarily because the use of off- season rainfall was increased Marshall, 1995. These
observations demonstrate the potential of agroforestry for temporal complementarity in areas where signifi-
cant rainfall occurs outside the normal cropping season. However, the short-term nature of these experiments,
often involving only one or 2 years of measurements, made it impossible to assess the long term implications.
The presence of trees may also modify microclimatic conditions in ways that improve the water use effi-
ciency of understorey crops, although regular pruning limits the extent of such effects in alley cropping sys-
tems Wallace, 1996. Several factors may be involved. Firstly, shading by the trees may increase the fraction
of available water used for transpiration by decreasing soil evaporation, particularly when the crop canopy is
sparse and rain is received as frequent, low intensity events. Under these circumstances, any reduction in
the quantity of radiation reaching the soil decreases evaporation as this process is primarily energy-limited.
Decreased windspeed at ground level may also limit evaporation. Secondly, agroforestry may confer mi-
croclimatic benefits by decreasing the air temperature, windspeed and saturation deficit experienced by un-
derstorey crops, thereby reducing evaporative demand Monteith et al., 1991. In C3 crops, in which pho-
tosynthesis becomes light saturated at relatively low irradiances, the reduced flux of photosynthetically ac-
tive radiation PAR resulting from partial shading may have little effect on assimilation Stirling et al., 1990,
although this is less likely to apply to C4 species with their much higher light-saturated photosynthetic rates.
The potential benefits of shade are therefore likely to depend on tree spacing and age, canopy structure, in-
cident radiation, shading intensity and the photosyn- thetic pathway of the understorey crop. Thirdly, shad-
ing may alter the surface temperature of understorey crops in ways that benefit their phenology and produc-
tivity Monteith et al., 1991; Vandenbelt and Williams, 1992. In areas of high incident radiation and ambi-
ent temperature, tissue temperatures frequently exceed optimal levels in unshaded crops, particularly during
drought periods; under such conditions, partial shade may exert an ameliorative influence by bringing tem-
peratures within the optimum range.
There are, therefore, several mechanisms whereby agroforestry may use available water more effectively
than mono-crops and improve microclimatic condi- tions for understorey crops. The key question is whether
the potential benefits outweigh the detrimental influ- ence of competition for water and nutrients between
trees and crops. The present study, extending over a pe- riod of almost 6 years, is one of the most comprehensive
and highly instrumented agroforestry experiments ever attempted. Its objectives were to quantify the changing
influence of trees on the microclimatic and hydrolog- ical conditions experienced by understorey crops un-
der semi-arid conditions as the system matured, and to establish the consequences for water use, light inter-
ception and tree and crop performance. Fig. 1 shows the main categories of measurements, the periods when
they were conducted, and seasonal rainfall. This paper describes the experimental design, instrumentation and
measurement protocols, evaluates effects on the atmo- spheric and soil environment, and examines their influ-
ence on tree and crop growth. More detailed consider- ation of the soil water balance, partitioning of light and
water between the trees and crops, system productiv- ity and the modelling aspects of the work is presented
elsewhere.
2. Materials and methods
2.1. Site description The Complementarity In Resource Use on Sloping
land CIRUS trial was located at ICRAF’s Machakos Research Station, Kenya 1
◦
33
′
S, 37
◦
8
′
E, altitude 1560 m on a south-westerly facing slope 18–22
124 C.K. Ong et al. Agriculture, Ecosystems and Environment 80 2000 121–141
C.K. Ong et al. Agriculture, Ecosystems and Environment 80 2000 121–141 125
Table 1 Seasonal and annual rainfall mm at Machakos: historical means and recorded values for CIRUS between planting and project end October
1991 to June 1997 Average
a
1991 1992
1993 1994
1995 1996
1997 Long growing season 1 March to 31 July
359 229
261 112
199 302
235 245
b
Dry season 1 August to 30 September 9
16 2
4 8
11 8
na 1991–92
1992–93 1993–94
1994–95 1995–96
1996–97 Short growing season 1 October to 2829 February
414 404
773 381
628 337
168 Annual total
782 653
675 799
810 666
507 na
a
Data for 9-year period 1963–1971 from Machakos Maruba Dam station.
b
Data available only until 7 July 1997.
above the Maruba river. Climate and soil were typical of the Kenyan uplands Scott et al., 1971, where rainfall
is bimodal with rainy seasons extending from March to June long rains and October to December short
rains; little rain is received between July and Septem- ber. Mean rainfall during the short and long growing
seasons between 1963–1971 was, respectively, 414 and 359 mm, with an annual average of 782 mm Table 1.
Seasonal rainfall during the experimental period ranged from 112 to 773 mm, but the variation in annual rainfall
was smaller 651–811 mm. Monthly potential evapo- ration varies between 95 and 165 mm, giving an annual
total of 1450 mm Huxley et al., 1989. The climate is relatively cool, with annual mean, minimum and max-
imum temperatures of 20.1, 13.8 and 23.2
◦
C; monthly mean relative humidity ranged from 56 during the
day to 96 at night. 2.2. Experimental design and management
CIRUS examined the extent of complementarity be- tween G. robusta A. Cunn.; grevillea and associated
crops in the use of above and below-ground resources. The 0.6 ha site had no previous cropping history be-
fore being cleared of scrubby vegetation dominated by Acacia species in July 1991. The trial comprised
a balanced incomplete block design containing five treatments replicated four times Fig. 2a; the remain-
ing five plots were mono-crop controls used for mea- surements not reported here. The five treatments
were: •
Cg — mono-crop maize Z. mays L, Katumani com- posite or cowpea V. unguiculata L. Walp cv.‘K80’
planted at intra- and inter-row spacings of 0.3 m×1 m and 0.15 m×0.5 m
• Td — dispersed mono-culture trees planted at
3 m×4 m spacing, 35 trees per plot 833 trees ha
− 1
• CTd — dispersed trees with crops; trees planted at
3 m×4 m spacing, 35 trees per plot 833 trees ha
− 1
• CTc — contour-planted tree rows with crops; tree
spacing 2 m×9 m, 30 trees per plot 640 trees ha
− 1
• CTa — across-contour tree rows with crops; tree
population as for CTc. Three-month-old grevillea saplings were planted in
October 1991. A rotation of maize and cowpea was initially adopted during the short and long growing
seasons, but maize was grown continuously from the 1994 long rains onwards because infection by root-rot
Fusarium udum caused substantial yield losses in cowpea. The mono-crop plots Fig. 2a had a maximum
of two sides adjacent to agroforestry plots to minimise the risk of interference. The plots were separated by
grass walkways, and a strip of vetiver grass Vetiver zizanoides
L. was contour-planted across the centre of each plot to control erosion. These were cut to a height
of 10–15 cm at 7–14 day intervals to minimise compe- tition with the trees and crops and ensure they did not
constitute a significant component of the water balance. No fertilisers or organic residues were applied.
The trees were managed to produce single stems and maintain leaf area index between 0.25 and 0.50 by
cutting back the side branches at the first pruning, 21 months after planting 6 June 1993, and removing the
lower branches at subsequent annual prunings Lott, 1998. Termites damaged some trees, causing death
when the bark was completely removed. The trunks were painted with wood preservative twice annually
to a height of 0.3 m above the ground to avoid further attack. Trees which died were replaced, but these and
adjacent cropping areas were excluded from further
126 C.K. Ong et al. Agriculture, Ecosystems and Environment 80 2000 121–141
Fig. 2. a Experimental design of CIRUS; the rhomboidal shape of some plots resulted from the contour planting of the tree rows; b soil depth measured at 4 m×4 m grid intervals across the site; depths between grid intersections were interpolated to produce the depth contours.
observations. Lateral extension of grevillea roots was examined before each rainy season to ensure they had
not extended into adjacent plots; trenches were dug to a depth of 0.75 m to sever such roots.
2.3. Soil characteristics The soil was a well-drained, shallow to moderately
deep 0–2.5 m sandy clay loam characterised as a Khandic Rhodustalf overlying petroplinthite murram,
and was stony with gravel bands. Five distinct hori- zons were apparent; the uppermost stone-free layer was
sub-divided into till 0–0.4 m, sub-surface 0.4–0.8 m and clay 0.8–1.0 m layers, underlain by a claygravel
layer and saprolite eroded gneiss bedrock at depths of 1.0–1.2 and 1.2–1.6 m, respectively. Bulk density and
particle density increased from 1.19 and 2.49 g cm
− 3
, respectively, in the surface horizon to 1.67 and
C.K. Ong et al. Agriculture, Ecosystems and Environment 80 2000 121–141 127
2.62 g cm
− 3
in the eroded bedrock; all horizons had moisture retention pF characteristics typical of sandy
or sandy clay soils. Further details are given by Jackson et al. 1997. Chemical analyses showed the soil was
not nutrient-limiting for crop growth before CIRUS was established, and had an organic carbon content
of 1 Kiepe, 1995. Soil pH, organic carbon con- tent and available ammonium, nitrate, potassium, phos-
phate, calcium and magnesium levels were determined in May 1996 to establish treatment effects on nutri-
ent status. The only detectable effects were that NH
4 +
and NO
3 −
levels were higher and phosphate content was lower in mono-crop plots than in the agroforestry
treatments p0.01; Lott, 1998. None of the nutrient concentrations or other soil variables examined were
limiting for crop growth.
A survey conducted at 4 m×4 m grid intervals in February 1993 using a 5 cm diameter auger showed that
soil depth above the bedrock varied greatly Howard, 1997. Additional points were added during the instal-
lation of neutron probe access tubes June to October 1993 and root distribution studies carried out between
October 1995 and August 1997. Fig. 2b shows that a band of relatively shallow soil 0.2–0.6 m extended
between the upper north-west and the lower south-east corner of the site. Soil depth was included as a variate
in the ANCOVA analysis of crop yields described by Howard 1997.
Detailed studies of climatic and microclimatic con- ditions, resource capture and productivity were con-
centrated on the Cg, CTd and Td treatments using the approaches outlined below.
2.4. Meteorology Dry and aspirated wet-bulb air temperatures, wind
speed and incident solar radiation were measured 2 m above the tree canopy in the centre of the site, and at
adjacent positions under the tree canopy, 0.3 and 2.5 m from the base of trees. The height of the above-canopy
instruments was adjusted before each rainy season to allow for tree growth. Rainfall was measured using an
automatic tipping-bucket raingauge located 20 m uphill from the experimental site.
2.5. Radiation interception Incident solar radiation, net radiation and photosyn-
thetically active radiation PAR were determined. In- terception by the trees and crops was measured using
unscreened tube solarimeters Delta-T Devices, Cam- bridge, UK located beneath both canopies and refer-
enced against incident values recorded above the tree canopy or at ground-level outside the experimental area.
Groups of three solarimeters were used to integrate in- terception across adjacent crop rows or concentric rings
around trees in the Cg, Td and CTd treatments during the four growing seasons between the 1994 long L94
and 199596 short S9596 seasons. Detailed measure- ments of light interception by the trees were made us-
ing six Kipp solarimeters mounted on a 3 m arm which moved over the ground area enclosed between four ad-
jacent trees in the Td and CTd treatments. This sys- tem provided 48 measurements within the 12 m
2
sam- pling area hereafter defined as ‘cell’, thereby allowing
spatial variation in the radiation environment beneath the trees to be quantified. Net radiation was measured
0.75 m above an area of bare soil and at the same height beneath the tree canopy 0.3 and 2.5 m from trees in the
Td and CTd treatments.
2.6. Hydrological measurements 2.6.1. Rainfall interception and runoff
Throughfall beneath the tree and crop canopies in the Cg, Td and CTd treatments was measured using man-
ually recorded raingauges 127 mm diameter refer- enced against the tipping bucket gauge and an identical
manual gauge sited on bare soil. Stemflow gauges were attached to trees in the Td and CTd treatments. These
comprised a flexible plastic collar attached to the trunk 0.75 m above the ground using silicone sealant RS
Components, Corby, Northants, UK which drained into plastic 35 l Jerrycans. 2.5 m×20 m runoff plots
leading to 1 m
3
collection tanks were installed in three replicate Cg, Td and CTd plots. Sub-samples of the
watersoil slurry mixture were evaporated to dryness to determine the quantities of water lost as runoff and
soil eroded.
2.6.2. Soil evaporation Soil evaporation E
s
was measured using lysimeters 160 mm diameter×200 mm deep; Jackson and Wal-
lace, 1999 located on bare soil or beneath the Td or Cg canopies; lysimeters were located 0.3 and 2.5 m
from trees, or on and off crop rows. E
s
was also mea- sured beneath the combined tree and crop canopies
128 C.K. Ong et al. Agriculture, Ecosystems and Environment 80 2000 121–141
in the CTd treatment. The lysimeters were weighed twice daily 08:00–09:00 and 17:00–18:00 hours
following rainfall during cropping seasons. Potential soil evaporation E
so
was calculated using the Penman–Monteith formula Monteith, 1965, and
hourly means were recorded for wet and dry-bulb air temperatures, wind speed and net radiation. Hourly E
so
values were summed for comparison with daily mea- surements of E
s
. 2.6.3. Soil water storage
A TDR system SoilMoisture, Trase I was used; this comprised ‘triple-wire’ sensors multiplexed to a central
signal processing and logging unit. The waveguides were inserted horizontally in groups of four, at depths
of 0.05, 0.15, 0.25 and 0.35 m in the Cg plots and at distances of 0.3, 1.0, 1.2, 1.5, 2.0 and 2.5 m from trees in
the Td and CTd treatments. A site-specific calibration was established, and the sensors were logged hourly
from November 1993 onwards.
Aluminium access tubes 44.5 mm diameter were installed to 0.6–1.8 m, depending on soil depth. Eight
tubes were installed in three replicates of the Td and CTd treatments, and four in three replicate Cg plots.
Tube locations in the Td and CTd plots were identical to the TDR sensor positions, and were chosen to allow
the influence of proximity to trees to be examined. Soil water content was measured weekly at 0.2 m depth in-
tervals using an IH neutron probe Didcot Instruments, Abingdon, UK. Calibrations were established for each
depth interval and tube by Centre d’Etudes Nucleaires de Cadarache France using thermal neutron proper-
ties. Over 300 calibrations relating soil water content θ
v
to count rate and bulk density were performed. Moisture deficits were calculated by subtracting θ
v
val- ues for specific horizons from the field capacity at a
water potential of −0.05 MPa determined by moisture release analysis.
2.7. Tree growth Allometric approaches were used to determine leaf,
branch, trunk and total above-ground biomass and leaf area. During the first 28 months, these were based on
measurements of tree height H, basal diameter BD, diameter at breast height DBH and total leaf number.
Trunk volume was calculated as Howard, 1997: V =
BD × 0.33π BD
2
2
H 1
Trunk biomass was calculated as the product of trunk volume and the density of grevillea wood 0.47 g cm
− 3
determined by destructive analysis. Leaf area was es- timated as the product of leaf number per tree, mean
leaf dry weight 1.52 g per leaf, and specific leaf area SLA; 50.2 cm
2
g
− 1
; mean leaf weight and SLA were determined for 1000 leaves sampled from three to six
trees in each plot. This approach could not be used during the later
stages of the trial because the allometric relationships based on tree height and basal diameter were altered
by pruning Lott, 1998. The pipe model theory War- ing et al., 1982; Whitehead et al., 1984 was therefore
modified to allow growth to be determined from mea- surements of trunk diameter immediately below the
lowest branch. Allometric relationships based on lin- ear regression analyses of log-transformed data were
developed to provide reliable estimates of leaf area, to- tal above-ground biomass and the biomass of leaves,
branches and trunks; these enabled tree growth to be determined non-destructively throughout the experi-
mental period Lott et al., 2000a.
2.8. Crop growth Growth analysis during the first two seasons was re-
stricted to the fresh and dry weights of pods and haulms at final harvest in cowpea S9192 and the stover, cobs
and grain of maize L92. Maize failed completely during the 1993 long rains L93 because of the very
low 25 of seasonal average and poorly distributed rainfall Table 1. Regular destructive growth analy-
ses of cowpea were carried out during the S9293 and S9394 growing seasons, and row-wise measure-
ments of pod and haulm yields were made at maturity.
As the much lower population of maize than cowpea 3.3 versus 13.3 plants m
− 2
precluded regular destruc- tive analysis, allometric procedures were developed to
establish treatment effects on crop performance in the Cg and CTd treatments and examine the influence of
distance from trees on maize growth. Allometric rela- tionships providing reliable estimates of leaf area and
above-ground biomass were developed for each sea- son between the 1994 long and 199596 short growing
C.K. Ong et al. Agriculture, Ecosystems and Environment 80 2000 121–141 129
seasons using linear regression analyses based on paired weekly destructive and non-destructive measurements
of plants reflecting the full range of growing conditions and plant sizes present. The relationships were based
on measurements of basal stem diameter, height to the youngest expanded leaf and the top of the canopy, and
leaf number. Leaf, stem, cob and grain fresh and dry weights were determined at maturity, and land equiva-
lent ratios LER and performance ratios PR; Willey, 1985; Rao et al., 1990, 1991 were calculated to es-
tablish treatment effects on system performance. Full details are given by Lott et al. 2000c.
The timing of germination, floral initiation, anthesis, silking and physiological maturity was determined to
establish treatment effects on the duration of the veg- etative, reproductive and grain-filling periods. Floral
initiation was determined by dissecting the whorl to establish the presence of tassle primordia. Anthesis
and silking were defined as occurring when spikelets emerged from the leaf whorl and silks stigmata ap-
peared from the husk; maturity was defined either by the formation of a black layer at the pedicel of at least
one grain per cob Daynard and Duncan, 1969, or when 90 of the leaf area had senesced during sea-
sons when CTd maize failed to set grain.
2.9. Root studies Roots were sampled twice during each of the seasons
indicated in Fig. 1, when the maize had 6–8 leaves and at anthesis. On each occasion, soil cores were taken to
the bedrock in three replicate Cg, Td and CTd plots and separated into 0.2 m depth increments. These were
sampled along a transect between a maize plant or grevillea tree and the mid-point between rows. The
cores were collected using a 4.8 cm i.d. auger driven into the soil using a hydraulic hammer and extracted
using a jack. Roots were washed from the soil, placed in vinegar and stored at 4
◦
C before being separated into maize and grevillea roots, which were distinguish-
able on the basis of colour and morphology; dead roots were discarded. Root lengths were determined using a
flat-bed scanner ScanJet IIcx, Hewlett Packard, Palo Alto, CA, USA and image-analysis software DT-
Scan, Delta-T Devices, Cambridge, UK. The lengths of proteoid roots were measured after shaving off the
rootlets into water and collecting them by filtration be- fore scanning. Root dry weights were also determined.
Mini-rhizotrons were used to allow changes in root length to be estimated over a shorter time-scale than is
possible using coring approaches. A new design based on innovations described by Gijsman et al. 1991 was
used. Aluminium frames with open sides were installed in square access holes in the soil profile and lined with
inflatable rubber bladders which eliminated air-filled gaps between the mini-rhizotron and the surrounding
soil, minimising root proliferation along the walls of the access frames. To observe root growth, the bladders
were deflated and removed before inserting an endo- scope. A 30 cm
2
circular area of soil was photographed at each sampling depth using a 35 mm SLR camera.
Fifteen mini-rhizotrons were installed to the bedrock in December 1995, six each in randomly selected Td
and CTd plots and three in the Cg treatment, using a sampling pattern similar to that used for the coring
measurements. Photographs taken at 0.1 m depth in- crements at weekly intervals during the 1996 long and
short rains and 1997 long rains were used to examine distribution of roots in maize and grevillea.
2.10. Sap flow Sap flow through the trunks and lateral roots of gre-
villea was measured using heat balance gauges; tech- nical details are given by Khan and Ong 1995, Lott
et al. 1996 and Howard et al. 1997. Trunk gauges were calibrated as described by Khan and Ong 1995
and sap flow was corrected for variation in trunk di- ameter as reported by Lott 1998. These procedures
enabled sap flow to be determined to within 7–10 for trunks and roots up to 10 cm in diameter Khan and
Ong, 1995; Lott et al., 1996; Howard, 1997, a level of accuracy consistent with that reported for other species
Baker and van Bavel, 1987; Ishida et al., 1991; Valan- cogne and Nasr, 1993. Measurements were made dur-
ing all cropping seasons from the 199293 short rains onwards and periodically during the dry seasons Fig. 1.
2.11. Meristem and soil temperatures The thermal environments experienced by maize in
the Cg and CTd treatments were examined during four seasons when the trees were well established
L94-S9596; Fig. 1. Soil and meristem temperatures were determined using copperconstantan thermo-
couples. Meristem temperature was estimated from
130 C.K. Ong et al. Agriculture, Ecosystems and Environment 80 2000 121–141
measurements of soil temperature at a depth of 2 cm immediately adjacent to maize plants until ca. 45 days
after sowing DAS, when stem elongation began; there- after thermocouples were inserted into the meristem.
Soil temperature initially provides a good approxima- tion of meristem temperature in cereals as the meristem
remains below the soil surface until stem elongation begins Norman et al., 1984.
3. Results and discussion