Modelling daily root interactions for water in a tropical shrub and grass alley cropping system

Modelling daily root interactions for water in a tropicalshrub and grass alley cropping system

J. F. SILLON1, H. OZIER-LAFONTAINE2, * and N. BRISSON3

1

INRA, Unité d’Agronomie de Laon-Péronne, Rue Fernand Christ, 02007 Laon Cedex France;2 INRA, Unité Apropédoclimatique de la Zone Caraïbe and 3 INRA, Agroparc Unité deBioclimatologie d’Avignon, Domaine Duclos, BP 515, 97165 Pointe à Pitre Cedex, GuadeloupeFWI (*Author for correspondence: E-mail: [email protected])

Key words: belowground competition, Digitaria decumbens, Gliricidia sepium, root distribu-tion, two-dimensional soil-root water transport model

Abstract. A two-dimensional physically-based model for the daily simulation of root com-petition for water in an alley cropping system associating Gliricidia sepium with Digitariadecumbens is developed. This paper deals with the impact of root distribution on soil waterpartitioning. By adapting existing principles of root water uptake modelling for pure crops, themodel accounts simultaneously for the sink terms of each species in a defined soil domain.Soil-root water transport functions are solved at the level of discrete volumes of soil; each ofthem are characterized by the inherent soil physical properties, root length density, soil-rootdistances, and the calculated sink terms of each species. The above ground boundary condi-tions, such as transpiration and soil evaporation, were managed by simple equations found fromthe literature or provided by experimental measurements. Running the model with two con-trasting observed root maps, an evaluation was carried out over a 10-day period following arainfall event. With both root maps, the simulated soil water potential profiles at the row, at0.75 m and 1.50 m from the row did not differ significantly, and were in good agreement withthe measurements. However, although water was not limiting during this period, the simulatedcumulative water absorption profiles of G. sepium and D. decumbens contrasted markedly, andmatched their observed root distribution. This model, although still under further development,forms the basis for development of an above and below ground coupled model to simulate plantinteractions for water in intercrops or agroforestry.

Introduction

Belowground interactions are now considered as a decisive issue that needsto be investigated to improve our understanding and manipulation of agro-forestry systems (Huxley, 1996). Numerous failures attributed to below groundcompetition have been observed when transposing plant associations to variousecological environments, particularly in situations of low-moisture and highlyacidic soils (Ong, 1994; Govindarajan et al., 1996). In the case of simulta-neous systems, complementarity in root distribution as regards soil water andnutrient availability may be considered as a key to success in plant associa-tions. (Ong, 1995; Van Noordwijk, 1996). However, the respective contribu-tion of species in interaction for soil resources is not easy to measure. Methodsbased on the use of non-radioactive tracers (Martin et al., 1982; Tofinga andSnaydon, 1992), or natural isotopes of water (Le Roux et al., 1995), although

Agroforestry Systems 49: 131–152, 2000. 2000 Kluwer Academic Publishers. Printed in the Netherlands.

allowing identification of preferential zones of water uptake in plant associ-ations, are destructive and labor intensive. Hence, to account for the numerouscombinations possible between plants, environments and techniques, model-ling may be useful in testing hypothesis on the functioning of such systems,and models should support experimentation as much as possible, by testingand predicting the most suitable plant associations.

In plant associations, the sharing of water depends on the partitioning ofevaporative demand between the crop components (Ozier-Lafontaine et al.,1997), soil-water availability and mobility, root systems distribution andfunctionality, and biophysical regulation of water flow (Ozier-Lafontaine etal., 1998). As root system’s ability to uptake soil-water depends both upongeometrical distribution of roots (De Willigen and Van Noordwijk, 1987;Tardieu et al., 1992), and root elongation (Caldwell and Richards, 1986),simulation models of root competition in plant associations need to accountfor these factors. The necessity of at least a two-dimensional subdivision ofthe root environment for the description of vertical and horizontal water-flowto test the degree of complementarity of the component crops, has beenpointed out by Van Noordwijk et al. (1996), Ozier-Lafontaine et al. (1998)and Lafolie et al. (1999). Hence, realistic simulations of below groundinteractions for water to predict the most useful plant associations in dif-ferent environments depend mainly on the soil-root water transport modelused.

Available approaches designed for studying root competition for water are,however, still partial: they can be roughly classified into two types:

1) Approaches devoted to simulate root competition at the scale of a plantassociation. As part of this, defined models commonly run at a daily timestep to solve soil-root water transport equations, assuming that root systemsare completely mixed (Adiku et al., 1995), or conversely fully separated(Kiniry et al., 1992; Kiniry and Williams, 1995). The sink term is gener-ally estimated using a one-dimensional function and the concept of ‘meandistance between roots’ (Gardner, 1960) is commonly used to describe rootdistribution into soil layers.

2) More mechanistic approaches that account for heterogeneous distributionof roots, allowing a two-dimensional solution of soil-root water transporton short time-step (hourly or less) – see the 2D-model of Ozier-Lafontaineet al. (1998) that simulates water competition in intercrops. Although wellsuited to analyze the role of root geometrical arrangements in competi-tion for water, this type of model does not account of root growth, mainlybecause of the time-consuming calculations required by the numericalcomputational of individual sinks. Among the collection of models dealingwith water partitioning in intercropped or agroforestry systems (see thesynthesis of Caldwell, 1995), this model may be considered as a tool moreappropriate to study the exploitation of environmental heterogeneity atmicrosite scales, that is a domain which has not been thoroughly exploredbefore (Caldwell, 1994).

132

Consequently, our aim in this paper is to develop a physically-based two-dimensional soil-root water transport model for plant associations, interme-diary between the two reported types of models:

1) that accounts, as simply as possible, for soil and root spatial heterogeneityby designing a soil profile with regular cells of required dimensions;

2) that solves root competition mechanisms by calculating the soil water-flowand the sink term for each species at a daily time step, in order to facili-tate the integration of the processes at the scale of the whole crop cycle.

So, at the first stage of a more elaborate model that may simulate dynam-ically above and below ground interactions for water at the scale of completecrop cycles, the presented study focuses on root-water competition in an alleycropping system associating a nitrogen fixing tree (Gliricidia sepium) with aforage (Digitaria decumbens). The study was conducted during a shortdesiccation period in tropical conditions. As the root systems of the two peren-nial species were not likely to change markedly during this period, first testsconcerned the sensitivity of the model to belowground interactions for waterwith respect to contrasted observed root maps.

Theoretical background

Soil-root water transport

Soil-root water transport functions were solved at the level of discrete volumesof soil or cells. As shown in Figure 1, following Hutson and Wagenet (1995),the soil profile is divided into n horizontal layers of thickness

∆z, furthersubdivided into m regions of width ∆x. The subscripts i and j refer to thecoordinates of the cells, and the subscripts G and D refer to G. sepium andD. decumbens, respectively. The cell dimension can be adapted to the desiredlevel of resolution. Vertical and horizontal water flows are assumed to operateat a daily interval. A finite difference module (Feddes et al., 1975) is usedfor the numerical solution of vertical and horizontal water flows. At the endof each day, the water content of each cell is updated.

Each cell of soil is specified by its hydraulic properties (Ψ(θ) and K(θ)functions) and the root length densities (L) of each species. Vertical andhorizontal water flows are solved successively using Richards’ equation viz.:

133

[ ( )]Vertical flows

Horizontal flows

Ci, j × ∂ψ∂t

∂ψ∂z

=∂∂z

K(ψi, j) × – 1 – SGi, j – SD

i, j (1)

[ ( )]Ci, j × ∂ψ∂t

∂ψ∂x

=∂∂x

K(ψi, j) × (2)

where Ψ is the soil water pressure head (cm), t is time (day), z is the depth(cm) considered to be positive downwards, x is the horizontal distance fromthe row center, C is the soil water capacity (cm–1), K(Ψ) is the hydraulicconductivity (cm day–1). SG and SD are the root water uptake terms (day–1) orsink term for G. sepium and D. decumbens, respectively. These sink termsare included in the vertical flow equation, assuming that the total uptake ofwater is the sum of D. sepium and D. decumbens root water uptake in eachcell. The sink term relies on the Ohm’s law analogy (Van den Honert, 1948;Cowan, 1965), and the simplified representation of the root system, on Gardner(1960). In each cell and for each species, the root is considered to be a uniformcylinder of radius rroot (cm). According to Tardieu and Manichon (1986a) orDe Willigen and Van Noordwijk (1987), the tesselation low boundary, alsonamed the soil-root distance (SRD), is the equivalent of the radius of thecylinder of soil through which the water moves, analogous to the half distancebetween roots (Gardner, 1960). Unlike Gardner (1960), the SRD conceptemphasizes that this distance is not constant, but varies with root distributiongeometry (Tardieu et al., 1992). Assuming negligible osmotic potential insteady conditions, the daily term S (cm3 water cm–3 soil day–1) for each speciesis:

134

Figure 1. Schematic description of the soil profile and the water fluxes considered in the model.

( )(3)Si, j

G, D = 2π × Ksysi, j × Li, jG, D ×

∆ψi, jG, D

SRDi, jG, D

rrootln

where Ksys is an hydraulic conductivity (cm day–1) characterizing soil watermovement between soil and roots, and

DC(cm) is the soil-plant driving force.In existing models, it is common usage to identify soil-root conductance tosoil hydraulic conductivity (see Moltz, 1981 for a review). Nevertheless Taylorand Keppler (1975), and Reid and Hutchison (1986) have demonstrated thatthe bulk soil hydraulic conductivity affects root absorption only for low levelsof soil water contents, justifying the use of Ksys which is different from Ksoil

(Habib et al., 1991).The definition of Ksys also depends on the choice of the soil-plant driving

force for water uptake. For short time step models – microscopic models –which solve the water transport equations at the root level (Lafolie et al., 1991;Habib et al., 1991), it is common to use the soil-root gradient of potential asthe driving force. In our cases, given the daily time step and the difficultyassessing root water potential readily, we use the soil-leaf gradient ofpotential as the driving force for water uptake (Slabbers, 1980; Ritchie andOtter, 1985; Brisson, 1998). In order to match the daily time step, we usedthe lowest leaf water potential over the day, namely the midday leaf waterpotential (Eq. 4). This assumption agrees with Van Bravel and Ahmed (1976),Brisson et al. (1993).

where ψ1 is the midday leaf water potential (cm).Concerning the calculation of Ksys, we follow the suggestion of Brisson

(1998) relying on Taylor and Keppler (1975). Ksys is equal to K, the soilhydraulic conductivity, for very low levels of soil water potential until athreshold value. This threshold, hereafter referred to as Klim, stands for a limitin plant absorption, varying between 10–7 and 10–6 cm day–1 according to rootspecies (Taylor and Klepper, 1975; Brisson, 1998).

Boundary conditions

The boundary conditions were specified either by simple equations found fromthe literature or by experimental measurements. The boundary condition atthe bottom of the soil profile is the measured water potential at that depth.The infiltration at the soil surface was estimated, assuming that all rainfallenters the soil, neglecting runoff process. Soil evaporation was estimated usingthe Deardorff function (1977), viz:

135

( )(4)Si, j

G, D = 2π × Ksysi, j × Li, jG, D ×

ψlG, D – ψi, j

SRDi, jG, D

rrootln

Esj = Ep(1 – FjD)

Esj =θsurfj

0.75 × θfc

× Ep(1 – FjD)

if θsurfj > 0.75 θfc

if θsurfj < 0.75 θfc

(5)

where Ep is the reference climate evaporation, (1 – FjD) is the shading factor

(see Eq. 8) due to the presence of the canopy above the soil, θsurf is the watercontent of the first 0.5 cm layer and θfc is the water content at field capacity.

Transpiration of the two species

The actual transpiration of each component is derived from the climateevapotranspiration Ep, a radiation partitioning factor F, a water stress factorα (Figure 2), and the soil evaporation Es. In our specific case, G. sepiumwas taller than D. decumbens and the foliage of the two species never mixed.Thus, the calculations of transpiration are made globally for the G. sepiumdominant species, while they are distributed by cells for the subordinate D.decumbens species.

where Esj and FjG are averages over the interrow. The fraction of incident

radiation intercepted by G. sepium (FjG) is considered as independent of the

D. decumbens one (FjD). Fj

G can be either calculated from a coupled model ofradiative transfer accounting for the geometry of G. sepium canopy, or directlymeasured. Fj

D is evaluated using Beer’s law:

FjD = (1 – Fj

G) × [1 – exp(–kD × LAIjD)] (8)

where kD is the daily extinction coefficient for global radiation, and LAIjD the

leaf area index of D. decumbens.As shown in Figure 2, stress patterns of each component are expressed as

a function of leaf water potential (Rose et al., 1976; Slabbers, 1980; Brissonet al., 1993). We assume that stress does not occur for values of leaf waterpotential lesser than a critical value (Ψl

cr). From this value a linear decrease

136

For G. sepium

For D. decumbens

TRG = α(ψlG) × Fj

G × (Ep – Esj) (6)

TRjD = α(ψl

D) × FjD × (Ep – Esj) (7)

Figure 2. Stress factor α as a function of leaf water potential.

is assumed between Ψlcr and Ψl

sc, this last value corresponding to the completestomatal closure.

Model dynamics

In order to solve Eq. 4 and Eq. 6, the leaf water potential was adjusted bymaking both daily fluxes of transpiration and absorption converge, viz.:

where Vij is the elementary cell volume, and A is the interrow distance. As asimplified assumption, given the vertical pattern of the D. decumbens rootsystem, we do not allow it to uptake water out of the vertical limits of eachregion (see Figure 1). The least square mathematical method is used itera-tively to solve Eq. 9, giving the sink term of each cell (Si, j

G, D). By solving thewater balance, the soil water potential profile is then updated. Figure 3 depictsthe algorithm of the model.

Materials and methods

Agronomic design

The experiment was carried out at the IRNA Research Center of DomaineDuclos in Guadeloupe (16°17′ N, 61°16′ W, elevation 125 m) on ferralliticsoils (Oxisols, FAO-UNESCO). These soils are well drained, high in CEC(19 cMc kg–1), with a pH(H2O) of 5.9 to 6.7. The soil organic matter contentin the topsoil is 9.9 (g kg–1). Although the experiment took place during thedry season (March–April 1997), the abundance of rainfall limited the durationof desiccation periods. Hence, the presented results are limited to a 10-dayrain-free period (Figure 4). Observations were made in a three years old alleycropping plot (0.30 ha), associating Gliricidia sepium with a forage crop:Digitaria decumbens between March and July 1997. In this stand, G. sepiumwas planted in 3 m apart NW – SW rows, and was spaced at 0.75 m intervalin the row. G. sepium was regularly pruned every two months.

Measurements

The instrumented soil volume was 0.9 m deep, 1.5 m wide (row–interrowdistance) and 0.75 m long (in-between shrubs distance on the row). Soil watercontent was measured with TDR probes (Trase Systems, Soil MoistureEquipment Crop., Santa Barbara, USA), and soil water potential with tensio-meters (SDEC, Paris, France). Measurements were made in three profilepositions located at 0, 0.75 and 1.5 m from the row: each profile had tworeplicates. Soil water content was measured at 5 depths (0.04, 0.1, 0.25, 0.4,

137

1A

∑i , j

∑j

(Si, jG, D × Vi, j) → TRj

D + TRG (9)

0.6, 0.9 m) while for soil water potential, the starting depth was 0.25 m. Datawere automatically recorded hourly (Campbell CR21X data logger), and dailyaverages were calculated.

Vertical root maps were made during the 10-day period at four locations,independently of soil moisture measurements. Observations were performedby applying a grid of 2 cm × 2 cm mesh over a trench wall of 1.2 m deep ×1.5 m wide. Each grid mesh was assigned binary value: presence/absence ofroots.

138

Figure 3. Algorithm for the calculation of the sink terms and the soil water potential profiles.

Leaf area index (LAI) of both species were measured weekly with aplanimeter (LI-COR LAI 3000, Nebraska, USA), from five leaf samplescollected on 0.44 m2 area.

Additional data were used from two previous experiments carried in similarconditions: the relative balance of G. sepium from Tournebize and Sinoquet(1995), and Tournebize et al. (1996), and saturated hydraulic soil conductivityfrom Ozier-Lafontaine et al. (1998).

Framework of model application

The modelling framework considers the soil to consist of 15 layers and 25regions (Figure 1: n = 15, m = 25) with ∆x = ∆z = 0.6 m, resulting in 375cells. For each cell, the Ψ(θ) and K(θ)functions, soil-root distances and rootdensities, plant parameters governing water stress and radiation partitioningbetween the species are specified.

The application of the model also accounts for, soil-root distances and rootdensities, and plant parameters governing water stress and radiation parti-tioning between the species.

139

Figure 4. Rainfall and potential evaporation (PET) variations during the evaluation period.

Soil hydraulic properties

The soil water retention curve was obtained by fitting Van Genuchten’s (1980)relationship:

with θr and θs denoting residual and saturated volumetric water contentrespectively. The parameters of Eq. 11 were estimated by minimising the sumof squared differences between measured and simulated water contents.Results (Figure 5a and Table 1) show that it was necessary to distinguish twolayers (0–0.48 m and 0.48–0.90 m). Furthermore, the unsaturated hydraulicconductivity relationship (Figure 5b) was obtained from:

where Ks is the saturated hydraulic conductivity, obtained for the layers 0–0.48m and 0.48–0.90 m, from measurements made in laboratory with permeameterusing undisturbed samples taken at 0.25 and 0.60 m depths (Ozier-Lafontaineet al., 1998).

Soil-root distances and root length densities

As recommended by Tardieu and Manichon (1986b), soil-root distances (SRD)were calculated directly from the root map –presence/absence of roots in a2 × 2 cm grid size. A procedure was developed allowing the automatic reck-oning of the distance between each center of a cell – of the 2 × 2 grid – andthe center of the nearest cell where roots have been observed. These resultswere reported for the 6 × 6 cm grid size used for water flow calculation. Ineach 6 × 6 cm cell, the SRD was considered as the median of SRD calculatedfrom the nine cells surrounding any given 2 × 2 cm grid cell. The root length

140

Θ = ( )11 + (αψ)n

m

, m = 1 –1n

(10)

to measured data:

where α and n are fitting parameters. Θ is the normalized volumetric watercontent, defined as

Θ =θ – θr

θs – θr(11)

Table 1. Estimated hydrodynamic properties of the soil profile.

Soil layer (m) Ks (m day–1) θs (m3 m–3) θr (m3 m–3) α (m–1) n

0–0.48 0.15 0.38 0 0.472 1.0480.48–0.90 3.60 0.5 0 0.528 1.116

K = KsΘ12 [1 – (1 – Θ

1m )m]2 (12)

141

Figure 5. Relationships between a) soil water potential and water content, and between b) watercontent and unsaturated hydraulic conductivity.

density (L) was deduced from SRD using the formula which combinesGardner’s (1960) and Newman’s (1969) equations to give:

Above ground parametrization

The measured LAI was considered as constant during the 10 day-period ofconcern. For G. sepium, an average value of 1.5 was measured. For D. decum-bens, LAI was 1.5 on the row (0 – 0.3 m from the row) and 2.5 in the interrow(0.3 m to 1.5 m from the row). The fractional incident radiation interceptedby G. sepium was estimated using an empirical ‘Fj

G – distance from the row’relationship. This relationship was fitted with a logistic function (Figure 6),using data from Tournebize (1995) corresponding to the same LAI. The valuesof the five plant parameters were taken from the literature (Table 2).

Results

Root maps and soil-root distance profiles

Figures 7a and b show typical root distribution and soil-root distance mapsfor both species for two profiles selected from the four, to illustrate thevariability of root distribution in field conditions.

142

Li, jvG, D =

1π(SRDi, j

G, D)2 (13)

Figure 6. Amount of light transmitted throughout G. sepium canopy as a function of the distancefrom the row.

The general trends are a more superficial rooting for D. decumbens,occurring mainly in the 0–0.45 m first layer, and a deeper colonization forG. sepium. However these maps differ first by the quantity of G. sepium roots,and second, by a preferential deep colonization zone by D. decumbens locatedat around 0.70 m from the row. Translated in soil-root distances, the patternsbecome clearer. For D. decumbens, soil-root distances above 0.60 m arearound 0.02 m (the minimum soil-root distance allowed by the grid), exceptfor the deeper layers. For G. sepium the soil-root distance profiles are muchmore variable, and differ with depth and position with respect to the rowdistance.

Water potential profiles

Figure 8a, b, c shows the variation of the water potential with depth at threedistances from the row, during the period of desiccation for the two root maps.The initial profile was measured on day 91, just after the rainfall event (seeFigure 4), and used as an input for those simulations. Each potential valuecorresponds to the average of two measurements, which is graphicallycompared to the simulated values calculated with the two observed rootdistributions. The three profiles are quite similar. We note that the soil des-iccation is in agreement with the simulations, especially for the simulatedprofile calculated with the second root map. It can be pointed out that thevariability of the measurements decreases in depth.

G. sepium and D. decumbens sink terms

The water uptake at each depth was cumulated over the 10-day period for eachspecies at the same three locations. In all simulations, the derived plant waterpotential never reached the critical water potential (–13 bars), that indicatesthat neither species suffered from water stress. Figure 9a, b, c, d, e, f shows

143

Table 2. Values of plant parameters, taken from literature, for running the model.

Parameters Values Source

Threshold Ksys (cm day–1) Klim = 5 × 106 Taylor and Klepper (1978),Brisson (1998)

Root diameter (cm) rroot = 0.02 Brisson (1998)

Critical value of leaf water potential Ψlcr = –13 Slabbers (1982),

(bar) for G. sepium and D. decumbens Tournebize (1996)

Leaf water potential for stomatal closure Ψlsc = –20 Tournebize (1996)

(bar) for G. sepium and D. decumbens

Daily extinction coefficient for kP = 0.7 Varlet-Grancher (1989)D. decumbens

144a) Map 1

145

Figure 7. Presentation of two root maps and their corresponding soil-root distances profiles taken on the row, at 0.75 m of the row, and on the inter-row.

b) Map 2

146

Figure 8. Comparison between measured and simulated soil water potential profiles a) on therow, b) at 0.75 m from the row and c) on the inter-row.

those results for both root distributions (map 1 and map 2). These figures pointout that the root uptake patterns are much more variable for G. sepium thanfor D. decumbens. This trend is accentuated as we get further from the row.Furthermore, D. decumbens uptake is reduced in the row position, but similarin the other locations.

Discussion

As it was not possible to measure precisely the contribution of each speciesto soil water uptake, model evaluation was done by comparison with theevolution of the water status of different portions of the considered soildomain. The general hypotheses of the model are in agreement with other soil-root water transport approaches that have been already tested in previousworks with pure crops (Brisson et al., 1993). It was not necessary to add newhypotheses or functions, even to account of root systems intermingling.

Although the crops were not water stressed, the simulation exhibits strongdifferences in root water-uptake patterns (Figure 9). The behavior of G.sepium, and to a certain extent, the behavior of D. decumbens in a part of theinter-row, were essentially related to root distribution as it is depicted in Figure8. In the row position, the lesser water consumption of D. decumbens wasmainly due to the reduction of the climate water demand by the presence ofthe tree crown. When analyzing these processes in non-limiting water condi-tions with a variable such as soil water potential which tends to equilibrate,these disparities seem to be reduced. However, even under such conditions,water uptake patterns can vary considerably within the row and with depth(Figure 9). One of the interests of this study lies in identifying the sources ofheterogeneity that may produce responses of tree roots. As the root systemof G. sepium is subjected to spatially vary more than the root system of D.decumbens, the use of root architectural models (see the G. sepium 3D fractalroot model of Ozier-Lafontaine et al., 1999) to simulate various rootingpatterns for tree stands might be of great interest. The horizontal modifica-tions of the hydrodynamic properties of superficial layers due to the trafficof field machinery were not taken into account, and may contribute to thisheterogeneity (Tardieu, 1988). As the dimensions of the cells can be adaptedto the level of heterogeneity of the soil, it is possible to integrate easily thesesources of heterogeneity. It is worth noting that, even though the use ofphysical laws to model soil-root water transport is well suited to simulatebelow-ground interactions, the difficultly of parameterization and the sensi-tivity of soil models to these functions must be considered.

The particular strengths of our model may become apparent when seenagainst other concepts used in modelling water competition in mixed plantsystems. Table 3 summarizes some of the published mixed crop models.Models such as ALMANAC (Kiniry et al., 1992) and CropSys (Caldwell andHansen, 1993) consider root distribution and water uptake in a very simplistic

147

148

Figure 9. Comparison of G. sepium and D. decumbens cumulated water absorption profiles after ten days during the evaluation period:– with map 1: a) on the row, b) at 0.75 m from the row and c) on the inter-row.– with map 2: d) on the row, e) at 0.75 m from the row and f) on the inter-row.

149

Table 3. Summary of water competition models for intercropping or agroforestry systems.

Year Author Model No. of Time step Soil Root system Model designspecies description descriptioncompeting

1992 Kiniry et al. ALMANAC 2 day 1 layer Dyanmic Crops and weeds1D 1D Crop growth model

1993 Caldwell and CropSys 2 day n layers Dyanmic IntercropsHansen 1D 1D Physiologically and

physically process-based

1994 Grant Ecosys 2 hour n layers Dyanmic Multiple ecosystems1D 1D Physiologically and

physically process-based

1998 Ozier-Lafontaine Watercomp 2 0.5 hour Tesselation (n cells) Static Intercropset al. 2D 2D root dots Physically process-based

1999 Van Noordwijk WaNulCAS 2 hour/day 16 compartments Dynamic Agroforestryand Lusiana 1D with hydraulic 1D Physiologically and

lift and sink physically process-based

1999 Sillon et al. – 2 day n compartments Static Agroforestry(this issue) 2D Soil-root distances Physically process-based

manner. Evaluation of such models on biophysical basis is difficult. Manycurrent approaches to water competition modelling such as WaNulCAS aremore process-related, and include wider water transport phenomena such aswater cycling by the tree components of agroforestry systems (Van Noordwijkand Luisana, 1999). It must be pointed out that the choice of a daily timestep in the current model to simulate water uptake patterns is not suitable tothe integration of phenomena within a daily cycle such as hydraulic lift orsink (Van Noordwijk and Luisana, 1999). On the other hand, most of themodels continue to consider only 1D water flow, although the importance ofa 2D flow in intercropping systems has been recognized (Caldwell, 1995;Ozier-Lafontaine et al., 1998; Lafolie et al., 1999). Our current model seeksto combine a 2D root distribution with a 2D water flow model. Its ability topredict the simultaneous water uptake of each plant in each cell is a tremen-dous improvement in simulating the root distribution effect on water balance.Hence, this approach should lead to possible connections with functionalmodels such as WaNulCAS, which is of wider applicability in agroforestrysystems.

Acknowledgements

We are grateful to T. Bajazet, J. André, A. Mulciba, S.A. Sophie, F. Solvarand J. Dezac for helping to obtain the data.

References

Adiku SGK, Carberry PS, Rose CW, McCown R and Bradock R (1995) A maize (Zea mays) –cowpea (Vigna unguiculata) intercrop model. In: Sinoquet H and Cruz P (eds) Ecophysiologyof Tropical Intercropping, pp 396–406. INRA Editions, Paris, France

Brisson N (1998) An analytical solution for estimation of the critical available soil water fractionfor the water balance of growing crops. J Hydrol, in press

Brisson N, Olioso A and Clastre P (1993) Daily transpiration of field soybeans as related tohydraulic conductance, root distribution, soil potential and midday leaf potential. Plant andSoil 154: 227–237

Caldwell MM and Richards JH (1986) Competing root systems: morphology and models ofabsorption. In: Givnish T (ed.) On the the economy of plant form and function, pp 251–273.Cambridge University Press, UK

Caldwell MM (1994) Exploiting nutrients in fertile soil microsites. In: Caldwell MM and PearcyRW (eds) Exploitation of environmental heterogeneity by plants. Ecophysiological processesabove- and belowground, pp 325–347. Academic Press, Inc., San Diego, USA

Caldwell RM (1995) Simulation models for intercropping systems. In: Sinoquet H and Cruz P(eds) Ecophysiology of Tropical Intercropping, pp 353–368. INRA Editions, Paris, France

Caldwell RM and Hansen JW (1993) Simulation of multiple cropping systems with CropSys.In: Penning de Vries F, Teng P, Metselaar K (eds) Systems Approaches for AgriculturalDevelopment, pp 397–412. Kluwer Academic Publishers, Dordrecht, The Netherlands

Cowan IR (1965) Tranpsort of water in the soil-plant-atompshere system. J Appl Ecol 2: 221–239Deardorff JW (1977) A parametrization of ground surface moisture content for use in atomos-

phere prediction models. J Appl Meteorol 16: 1182–1185

150

De Willigen P and Van Noordwijk M (1987) Uptake potential of non-regularly distributed roots.J Plant Nut 10: 1273–1280

Feddes RA, Neuman SP and Bresler E (1975) Finite element analysis of two-dimensional flowin soils considering water uptake by roots: II. Field applications. Soil Sci Am Proc 39:231–237

Gardner WR (1960) Dynamics aspects of soil water availability to plants. Soil Sci 89: 63–73Grant RF (1992) Simulation of competition among plant populations under different manage-

ments and climates. Agron Abstr, p 16. American Society of Agronomy, Madison, Wisconsin,USA

Govindarajan M, Rao MR, Mathuva MN and Ramachandra Nair PK (1996) Soil-water and rootdynamics under hedgerow intercropping in semiarid Kenya. Agron J 88: 513–520

Habib R, Pagès L, Jordan MO, Simonneau T and Sébillotte M (1991) Approche à l’échelle dusystème racinaire de l’absorption hydro-minérale. Conséquences en matière de modélisation.Agronomie 11: 623–643

Hutson JL and Wagenet RJ (1995) A multiregion model describing water flow and solute trans-port in heterogeneous soils. Soil Sci Soc Amer J 59: 743–751

Huxley P (1996) Biological factors affecting form and function in woody-non-woody plantmixtures. In: Ong CK and Huxley P (eds) Tree-Crop Interactions – A Physiological Approach,pp 235–299. CAB International, Wallingford, UK

Kiniry JR (1995) Simulating intercropping with the ALMANACH model. In: Sinoquet H andCruz P (eds) Ecophysiology of Tropical Intercropping, pp 387–396. INRA Editions, Paris,France

Kiniry JR, Williams JR, Gassman PW and Debaeke P (1992) A general, process-oriented modelfor two competing plant species. Transactions of the ASAE 35(3): 801–810

Lafolie F, Bruckler L and Tardieu F (1991) Modeling root water potential and soil-root watertransport: I. Model presentation. Soil Sci Soc Am J 55: 1203–1212

Lafolie F, Bruckler L, Ozier-Lafontaine H, Tournebize R and Mollier A (1999) Modellingsoil-root water transport and competition for single and mixed crops. Plant and Soil 210:127–143

Le Roux X, Bariac T and Mariotti A (1995) Spatial partitioning of the soil water resourcebetween grass and shrub components in a West African humid savana. Oecologia 104:147–155

Martin MPLD and Snaydon RW (1982) Root and shoot interactions between barley and fieldbeans when intercropped. J Appl Ecol 19: 263–272

Molz FJ (1981) Models of water transport in the soil-plant system: A review. Water ResourRes 17(5): 1245–1260

Newman EI (1969) Resistance to water flow in soil and plant. I. Soil resistance in relationamounts of roots: theoritical estimates. J Appl Ecol 6: 261–272

Ong CK (1994) Alley cropping – ecological pie in the sky? Agroforestry today 6(3): 8–10Ong CK (1995) The ‘dark side’ of intercropping: manipulation of soil resources In: Sinoquet

H and Cruz P (eds) Ecophysiology of Tropical Intercropping, pp 45–65. INRA Editions,Paris, France

Ozier-Lafontaine H, Lafolie F, Bruckler L, Tournebize R and Mollier A (1998) Modelling com-petition for water in intercrops: Theory and comparison with field experiments. Plant andSoil, 204: 183–201

Ozier-Lafontaine H, Lecompte F and Sillon JF (1999) Fractal analysis of the root architectureof Gliricidia sepium for the spatial prediction of root branching, size and mass. Modeldevelopment and evaluation in agroforestry. Plant and Soil 209: 167–180

Ozier-Lafontaine H, Vercambre G and Tournebize R (1997) Radiation and transpirationpartitioning in a maize-sorghum intercrop: test and evaluation of two models. Field cropRes 49: 127–145

Reid JB and Hutchinson B (1986) Soil and plant resistance to water uptake by Vicia faba. Plantand Soil 92: 431–441

Ritchie JT and Otter S (1985) Description and performance of CERES-wheat, a user-oriented

151

wheat yeild model. In: Willis WO (ed) ARS Wheat Yield Project, USDA-ARS, ARS-38,USDA, pp 159–175. Washington, DC, USA

Rose CW, Byrne GF and Hansen GK (1976) Water transport from soil through plant toatmosphere: a lumped-parameter model. Agric Meteorol 16: 171–184

Slabbers PJ (1980) Practical prediction of actual evapotranspiration. Irrig Sci 1: 185–196Tardieu F and Manichon H (1986a) Caractérisation en tant que capteur d’eau de l’enracine-

ment du maïs en parcelle cultivée. I. Discussion des critères d’étude. Agronomie 6: 345–354Tardieu F and Manichon H (1986b) Caractérisation en tant que capteur d’eau de l’enracine-

ment du maïs en parcelle cultivée. Il. Une méthode d’étude de la répartition verticale ethorizontale des racines. Agronomie 6: 415–425

Tardieu F (1988) Analysis of spatial variability of maize root density. I. Effect of wheel com-paction on the spatial arrangement of roots. Plant and Soil 107: 259–266

Tardieu F, Bruckler L and Lafolie F (1992) Root clumping may affect the root water potentialand the resistance to soil-root water transport. Plant and Soil 140: 291–301

Taylor HM and Klepper B (1975) Water uptake by cotton root systems: an examination ofassumptions in the single root model. Soil Sci 120: 57–67

Tofinga MP and Snaydon RW (1992) The root activity of cereals and peas when grown in purestands and mixtures. Plant and Soil 142: 281–285

Tournebize R and Sinoquet H (1995) Light interception and partitioning in a shrub/grass mixture.Agric For Meteorol 72: 277–294

Tournebize R, Sinoquet H and Bussière F (1996) Modelling evapotranspiration partitioning ina shrub/grass alley crop. Agric For Meteorol 51: 255–272

Van Bavel CHM and Ahmed J (1976) Dyanmic simulation of water depletion in the root zone.Ecol Modelling 2: 189–212

Van Den Honert TH (1948) Water transport as a catenary process. Faraday Soc Discuss 3:146–153

Van Genuchten MT (1980) A closed-form equation for predicting the hydraulic conductivity ofunsaturated soils. Soil Sci Soc Am J 44: 892–896

Van Noordwijk M, Lawson G, Soumare A, Groot JJR and Hairiah K (1996) Root distributionof trees and crops: Competition and/or complementarity. In: Ong CK and Huxley P (eds)Tree-Crop Interactions – A Physiological Approach, pp 319–365. CAB International,Wallingford, UK

Van Noordwijk M and Lusiana B (1999) WaNulCAS, a model of water, nutrient and light capturein agroforestry systems. Agroforestry Systems 43: 217–242

152