STRUCTURE, FUNCTION, AND STABILITY
OF INTERCROPPING SYSTEMS IN TANZANIA
FAYE FRANCES BENEDICT
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
Faye Frances Benedict
This project was made possible by the support of many people and
agencies. The Center for African Studies of the University of Florida
provided National Defense Foreign Language Fellowship support both
before and after the field research phase, which was funded by Fulbright-
Hays and American Association of University Women doctoral research
grants. I am also grateful to the Tanzania National Scientific Research
Council, the Tanzania Ministry of Agriculture, and the University of Dar
es Salaam for permission to conduct the research in Tanzania, and for
the substantial logistic support they provided. Jack Ewel, Ariel Lugo,
Hugh Popenoe, Haig Der-Houssikian, and Hossiah Kayumbo stimulated me to
initiate the project, and provided invaluable guidance and support to
carry it through to completion. R. Hunt Davis, Jr., also provided useful
comments on the final draft.
I am particularly grateful to Asha, Abbas, Ngatimwa, and the other
Tanzanian workers who not only provided superb field assistance, but also
shared their ideas, humor, and culture with me. Special thanks go to my
parents, who have supported my educational efforts for many years, and to
my husband, Harald, for the many kinds of help and support he has given.
TABLE OF CONTENTS
ACKNOWLEDGEMENTS. . . ... ...... .iii
ABSTRACT. . . ... . x
ONE INTRODUCTION. . . ... 1
The Ecological Issues . . 2
Diversity, Stability and Competition ........ 2
Definitional Problems. . 4
Stress and its Relation to Productivity and
Stability .................. .. 5
The Agronomic Issues ................. 7
Productivity in Crop Mixtures .......... 8
Resource Partitioning and Competition Reduction. 9
Pests and Diseases . .. 11
Stability and Risk . ... 12
Methodological Problems. . ... 13
A New Index ................... 16
Hypotheses. . . .. 19
Experimental Approach to the Hypotheses .. 21
The Tanzanian Agricultural Setting.. .. 21
Overview ............... 21
Survey of Local Crop Systems ............ 24
Methods . . .. .24
Results ................ .. 25
TWO METHODS . ...... ...... 32
The Study Site. . . .. .... 32
Location and History. ... . 32
Climate. . .. 33
Soils. . . .. 34
Experimental Design . . .
Crop Systems, Stress Treatments, and
Planting Density . . .
Plot Layout. . . .
Plot Management . . .
Agronomic Management . .
Application of Stress Treatments . .
Measures of System and Species Response .
Growth and Productivity. . .
Total and edible biomass. . .
Biomass at physiological maturity (flowering)
Leaf mass at defoliation. . .
Miscellaneous productivity measures .
LAI and canopy cover . .
Cowpea third-leaf area. . .
Specific leaf mass. . .
Root biomass. . .
Stem length . . .
Survivorship. . .
Resource Use . . .
LAI . . .
Root biomass. . .
Soil moisture . .
Labor . . .
Resistance to Natural Stressors. . .
Weed growth . . .
Lodging . . .
Pests and diseases. . .
Methods of Analysis . . .
Three Types of Analysis. . .
Absolute comparison of system performance
Comparison of intercrops and
corresponding monocultures. . .
Comparison of species performance in
intercrops and monocultures . .
Analyses of Variance . .
Analysis of Stability. . ....
THREE GROWTH AND PRODUCTIVITY
Introduction. . . .
Results: Productivity Differences Among Systems .. ...
Measures of Biomass Accretion. . .
Edible and total aboveground biomass
at harvest. . . .
Total aboveground biomass at flowering. ... 86
LAI and canopy cover. . ... 86
LAI at flowering. . .. 95
Leaf mass at defoliation. . ... 98
Root biomass. . ... 101
System fullstandedness. . ... 104
Measures of Biomass Distribution .. .107
Percent monocot LAI . ... 107
Allocation ratios . .. 111
Root/shoot ratios .. . .. 111
Correlations Among Productivity Measures .116
Year-to-Year Productivity Differences. .. .119
Overall System Performance . ... 119
Overall Response to the Stress Treatments. ... .123
Results: Productivity in Intercrops and Corresponding
Mlonocultures. . . 123
Edible and Total Aboveground Biomass at Harvest. 127
LAI and Canopy Cover .. .... .135
Leaf Mass at Defoliation . ... 135
Root Biomass, Year 1 ............... 135
Aboveground Biomass, LAI, and Root Biomass
at Flowering, Year 2 . ... 140
Fullstandedness. . . ... 140
Yield Equivalent Ratios (YERs) . .. 140
Results: Productivity of Species from System to System. 141
Introduction . .... ... 141
Measures of Biomass Accretion. . ... 144
Edible and total aboveground biomass
at harvest. . . 144
Aboveground biomass at flowering. .. .161
LAI . . 161
LAI at flowering. . ... 165
Leaf mass at defoliation. . ... 165
Root biomass. ............... 168
Survivorship and mortality. . ... 168
Stem length . ... 179
Measures of Biomass Distribution ... .186
Allocation ratios . ... 186
Root/shoot ratios . ... 187
Specific leaf mass. . ... 193
Miscellaneous yield parameters. ... 193
Correlations Among Direct Productivity Measures. .. 204
Correlations Among Indirect and Direct
Productivity Measures. . ... 215
Summary of Year-to-Year Differences in
Productivity . .... .222
Summary of System-to-System Productivity
Differences. . .... .222
Summary of Response to Stress Treatments 233
Discussion. . . .
Comparison with Regional Yields. . .
Food Yield as a Measure of Productivity. .
Absolute Comparisons Among Systems .. ....
Comparisons of Intercrops with
Corresponding Monocultures . .
Response to the Stress Treatments .. ...
Effects of the Stress Treatments on Intercrop
Advantage . . .
Stress and Biomass Distribution. . .
Species YER as a Measure of Competitiveness
and Aggressiveness . .
Height and aggressiveness . .
YIELD STABILITY . . .
Introduction . . .
Results . . .
System Stability . . .
Year-to-year fluctuations . .
Plot-to-plot variability within years .
Responsiveness to the stress treatments .
Stability at different intensities of stress.
Stability with respect to competition ..
Summary . . .
Species Stability. . .
Year-to-year fluctuations . .
Plot-to-plot variability within years .
Responsiveness to the stress treatments .
Summary of species stability in
intercrops and monocultures . .
Stability with respect to competition ..
Discussion: Sources of System Stability .
An Artifact of Coefficient of Variation
Calculations? . .
Masking of Species' Fluctuations .
Buffering of Species' Fluctuations .
Increased Stability of Component Species .
Stress Intensity and Stability .
Interaction Among Stressors. .
PESTS, DISEASES, AND OTHER NATURAL STRESSORS. .
Results . . .
Weed Biomass . .
Maize and Sorghum Lodging. .
Striga . . .
Rats . . ... .. .. 294
Maize Pests and Diseases . ... 294
Sorghum Pests and Diseases . ... 294
Cowpea Pests and Diseases. . ... 298
Pumpkin Pests and Diseases . ... 308
Summary of Pests and Diseases in Intercrops
and Monocultures . ... .312
Effects of Pests on Productivity ... 312
Weeds . . ... .. 317
Maize and sorghum pests . ... 321
Cowpea pests. . ... 325
Pumpkin pests . ... 327
Summary of Diversity-Pest Level Productivity
Relationships. . . ... 328
Pest Stability . . ... 331
Discussion. . . ... .332
Previous Studies . ... 332
Pest Levels. . .... .334
Pest Stability . ... 336
Plot Size, and the Range of Diversity Tested 337
SIX RESOURCE USE. . . ... .338
Introduction. . . ... 338
Light, Water, and Nutrient Use . 338
Results. . . ... .338
Amount and vertical distribution of LAI 338
Amount and vertical distribution of roots 343
Soil moisture . ... 343
Discussion . . ... ... .345
Efficiency of total resource use. .. .345
LAI and light use . ... 348
Root biomass amount and distribution:
use of soil resources ............ 351
Complementary moisture and
nutrient response . ... 352
Efficiency of nutrient use. . ... 353
Efficiency of water use . ... 354
Compensation. . ... 355
Temporal partitioning . ... 359
Human Labor Use . . ... .. .360
Results. . . ... .360
Labor use efficiency. . ... 367
Discussion . . ... ... .371
SEVEN IMPLICATIONS FOR AGROECOSYSTEM DESIGN . 374
Species Composition of Intercrops . .. 374
Environmental Stability and Intercropping .. 378
EIGHT IMPLICATIONS FOR TANZANIAN AGRICULTURE. .. 379
NINE CONCLUSIONS .................... . 381
LIST OF REFERENCES. . . .. 388
BIOGRAPHICAL SKETCH ........................ 399
Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy
STRUCTURE, FUNCTION, AND STABILITY
OF INTERCROPPING SYSTEMS IN TANZANIA
Faye Frances Benedict
Chairperson: John J. Ewel
Major Department: Department of Botany
Effects of crop diversity on system productivity and stability
formed the subject of a two-year study of nine Tanzanian cropping systems
and successional vegetation. Planting density was controlled so that
intercrop-monoculture differences were due solely to diversity, not
Various measures of seasonal biomass accretion were used as
indicators of net productivity. The ordering of systems by productivity,
from high to low, was successional vegetation, maize-sorghum, maize-
pumpkin, sorghum, maize-sorghum-cowpea-pumpkin, sparse maize, recommended
density maize, maize-cowpea, cowpea, and pumpkin. Intercrop performance
was compared with that of a mixture of corresponding monocultures by
Yield Equivalent Ratio (YER), a new index defined as the ratio of yield
in intercrop to expected yield based on planting densities and monoculture
yields. By this measure, intercrops were more productive than correspond-
ing monocultures, but the advantage was lowest under a 50 percent
defoliation treatment. Species YERs were greater than one for maize,
sorghum, and cowpea, and less than one for pumpkin.
The productivity advantage of diverse systems was due primarily to
a greater amount and a more even distribution of leaves and roots, temporal
partitioning of growth, complementary resource response curves, and
compensatory growth. Several of the 23 pests and diseases inventoried
were correlated with crop yield, but few were significantly more or less
abundant in intercrops than monocultures. Efficiency of labor use in
intercrops was equal to or greater than that in monocultures.
The ordering of systems by yield stability (constancy), from high
to low, was sorghum, maize, succession, four-crop intercrop, cowpea, and
pumpkin. Intercrop stability was higher than that of component crops
grown as monocultures. Yield of sorghum, cowpea, and pumpkin, but not
maize, was less stable in intercrops than monocultures; decreased species
stability in diverse systems may enhance system stability through compen-
satory growth. Stability was greatest under low levels of stress.
Parallel lines of research in the agronomic and ecological sciences
have focused on the effects of vegetation diversity on ecosystem pro-
ductivity and stability. The agronomic approach has consisted mostly
of experimental studies of various crop combinations,with relatively
little attention (until recently) to the causes of differences in
yield, and considerable confusion over the indices used to compare
the performance of diverse and simple systems. Comparisons of systems
of unlike density have further confused the issue. The ecological
approach has emphasized theoretical mechanisms underlying the
diversity-stability relationship, mathematical models of competition
and resource partitioning, computer simulation of diverse and simple
systems, and broad comparative ecosystem studies, with relatively few
controlled experimental studies. In this study I integrate these two
approaches by using agronomic systems to conduct a controlled
experiment testing the effects of diversity on various aspects of
ecosystem structure and function. The experimental design allows
conclusions to be made regarding the effects of spatial diversity,
not confounded by varying species composition and planting density.
The Ecological Issues
Diversity, Stability, and Competition
The question of whether increasing species diversity increases
stability, and the relationship of diversity and stability to productivity,
have been recurrent issues in the ecological literature for many years.
Several volumes have been devoted to these topics, drawing on evidence
ranging from global latitudinal comparisons to laboratory microcosm
experiments, mathematical models, and computer simulations (Woodwell and
Smith 1969, May 1973, Pielou 1975, van Dobben and Lowe-McConnell 1975),
van Voris 1976). The issue is still far from being resolved, but
Goodman (1975) and Murdoch (1975) concluded that although species
diversity per se is not a good predictor of stability (especially in
agricultural and model systems), certain correlates of diversity may be
causally related to stability.
Current literature tends to substitute more specific, mechanistic
hypotheses for the generalized diversity-stability question. A large
increase in entomological studies, for example, has begun to elucidate
the roles of resource concentration, pest dilution, feedbacks between
trophic levels, time lags, search behavior, microclimate, and natural
enemy dynamics on pest levels, pest stability, and various methods of
pest control in diverse and simple systems (reviewed by de Loach 1970,
Southwood and Way 1970, Hassell and May 1973, Comins and Blatt 1974,
Huffaker 1974, van Emden and Williams 1974, Barclay and van den Driessche
1975, Murdoch 1975, Murdoch and Oaten 1975, Shattock 1976, Trenbath
1976, Way 1976, van Emden 1977, Levins and Wilson 1980). No clear
conclusions can be drawn on theoretical grounds regarding the effects
of diversity on pests; there are many mechanisms that may stabilize and
reduce pest levels in simple systems as well as complex systems. Per-
haps the best way to summarize the theoretical effect of plant diversity
on pest levels and pest stability is that it is unpredictable due to the
complexity of interactions involved.
There has also been a resurgence of interest in competition
reduction through resource partitioning as a mechanism that may increase
productivity and stability in diverse systems (reviewed by Pianka 1981).
May (1981) and Nunney (1980) modeled growth rates of two interacting
populations using Lotka-Volterra competition equations. Pianka (1981)
noted, however, that while these models are useful, they greatly
oversimplify the competitive mechanisms involved among species. It is
also likely that the competition coefficients in these equations vary
with growing conditions and planting diversity (Vandermeer 1981).
The theory of niche separation was presented by Pianka (1981),
and biological mechanisms of resource partitioning were reviewed by
Schoener (1974). Increased light use efficiency in canopies composed
of sun- and shade-adapted species was discussed by Donald (1961), Black
(1975), and Boardman (1977). Bray (1974) noted that root competition
for nitrogen and water can occur at low root densities due to those
substances' high mobility in soil, and Litav and Wolovitch (1971),
Parrish and Bazzaz (1976) and Berendse (1979) discussed the temporal
and spatial separation of root zones. Partitioning of nutrient element
demand was suggested by the results of Garten (1978), who showed that
concentrations of elements in plant tissue may vary from species to
species at a site.
The question of whether natural successional regrowth con-
stitutes the most effective possible use of resources at a site has
not been completely resolved. It may not be, at least in the
short term; pastures and sugar cane are two examples of human-
managed systems composed of a few highly productive species that
may produce more biomass on a site than succession (Odum 1971, Mott
and Popenoe 1977). Whether, and by what biological mechanisms,
the productivity and stability of successional systems can be improved
upon is still an open question.
The definitions of diversity and stability have been a source of
confusion. Hurlbert (1971) reviewed the various diversity indices in
use and concludes that they are not necessarily correlated, cannot be
shown to be useful measures of a meaningful community property, and
should be replaced by other measures of ecosystem characteristics more
appropriate to the specific issue at hand. Stability is an even more
confused concept; its meaning includes properties of constancy (both
absolute and in relation to the mean), predictability, resistance, and
resilience (Holling 1973, Goodman 1975, Murdoch 1975 and Orians 1975).
The most commonly used measures of species constancy are coefficient
of variation (e.g., Rao and Willey 1980 and other studies cited in
Trenbath 1974), and standard error of population levels or changes
(Watt 1965). Other measures of stability, less commonly used, relate
to degree and rate of response to perturbations (Orians 1975).
Constancy on the community level could be measured as coefficient
of variation or standard error of system attributes such as total
productivity or leaf area, rather than as population counts; Murdoch
et al. (1972) evaluated insect community stability with a similarity
index of two years' sweep net samples. Margalef (1969) proposed a
community stability index that is the sum of the quantity (biomass'biomass
half-life) for component species. Other measures of community stability
with respect to perturbations have been suggested by Orians (1975).
Stress and its Relation to Productivity and
The general definition of stress by Odum (1967) as diversion of
potential energy in response to a stressor is robust, and will be used
in this study in preference to that of Grime (1979), who limited the
term to chronic low-productivity conditions and excluded other possible
energy drains such as competition or disturbance. Lugo (1978) pointed
to the type of stressor, its duration (chronic vs. acute), point of
impingement, and degree of system adaptation as factors that will
determine the rate and degree of system response to stressors. Rapport
(1981) differentiated between "alarm" reactions in non-preadapted systems
and "coping" reactions in resistant, preadapted systems (an analogy with
human response to disease). Stress can be measured as decreased
productivity, increased energy drains, or acceleration of repair within
the system (Odum 1967). Stressors, like limiting factors, are thought
to interact multiplicatively, rather than additively, with respect to
productivity (Odum 1971, Lugo 1978).
If we accept the definitions of stress as reduced productivity,
and stability as low fluctuation in productivity, a link between the
two concepts becomes apparent. Stability can be viewed as a system's
ability to adjust to varying intensities of stressors to maintain
productivity at a constant level. Degree of preadaptation will largely
determine the stability of the system with respect to a stressor (as was
pointed out, in the context of pest populations, by Huffaker 1974,
Goodman 1975, and Murdoch 1975). If stressors do interact multi-
plicatively, then reduced intensity of some stressors (such as competition
or herbivore drains) should reduce the magnitude of productivity response
to other stressors. If diverse systems are more productive than simple
ones due to resource partitioning, reduced drains, or other factors, it
would not, then, be surprising to find that they are also more stable
than simple systems.
Environmental parameters such as level, constancy, and predictability
of limiting factors, describe the setting in which stabilizing biological
adjustment occurs. Such parameters may be critical in determining eco-
system productivity and stability across a range of environments (Zaret
1982). Environmental effects may also explain the noncorrelation of
productivity and stability in (high-productivity, low-stability) marsh
and pond ecosystems, and (low-productivity, high-stability) paramo
vegetation and marine benthos systems. It is within a given environment
that the biological mechanisms leading to system stability can best be
explored, since differences in system behavior due to differences in
environmental parameters are controlled.
The Agronomic Issues
Green Revolution technology assumed abundant supplies of fossil-
fuel energy at affordable prices. Dramatic increases in the cost of
energy inputs to agriculture (mechanization, improved seed, fertilization,
pesticide, irrigation) have raised the issue of how these inputs can be
used most effectively. Subsistence agricultural systems that evolved in
low-fossil-fuel-energy environments might contain features that optimize
the use of resources and/or reduce the risk of crop failure. One such
feature is the high diversity of crops frequently planted side-by-side
in traditional tropical agricultural systems.
The term intercropping has been widely used to designate simultaneous
planting of more than one species in the same field at the same time,
while "multiple cropping" is a more general term that includes relay
and series planting of the same or different species (Dalrymple 1971,
Beets 1975, Norman 1979). Kass (1978) used "polyculture" in reference
to any situation in which two or more crops are simultaneously grown
together; "multiple cropping,"to any situation in which more than one
crop is grown on a given area in one year; and "intercropping," to
simultaneous alternate-row plantings. The planting patterns used in
this study sometimes mixed species within rows, and therefore do not
conform to Kass's use of the term intercrop, but they do fit the
generally accepted use of the word by other researchers.
Intercropping, and other forms of multiple cropping, have been
credited with increased productivity, reduced risk of crop failure,
and reduced incidence of pests and diseases (Janzen 1973, Dasman et al.
1974, Trenbath 1975a, Gibson and Jones 1976, Papendick et al. 1976,
Pimentel 1976, Sanchez 1976). These systems are looked to with great
hope as potentially stable, productive, low-energy-input cropping
systems for the developing tropics.
Productivity in Crop Mixtures
The experimental evidence to back up claims of increased productivity
in crop mixtures is abundant, but the results are seriously confused by
the varying methods used to determine intercrop planting densities and
compare productivities. Kass (1978) gave an excellent review of the
massive literature on this topic. Regarding yield and resource use,
he concluded that by many measures polycultures are clearly superior to
monocultures in terms of yield. Trenbath (1974) reviewed 344 data
papers in which 60 percent of the two-crop mixtures studied yielded
more than the mean yield of corresponding monocultures. Significantly
more of the 572 mixtures reviewed by van den Bergh (1968) yielded more
than monocultures as measured by the Land Equivalent Ratio (see below)
than yielded less than monocultures.
Trenbath (1977) reviewed possible mechanisms responsible for
productivity increases in mixtures. Competition for light, water, and
nutrients may be reduced through temporal partitioning of growth and
resource demand, spatial partitioning of root and leaf systems,
differences in limiting factor response curves (different resource
"requirements"), shading and leaf angle effects on light utilization,
and shading effects on water use. These mechanisms typically result
in higher-than-monoculture yields of one species (the "aggressor"),
the lower-than-monoculture yields of the second species (the "sub-
ordinate"), and increased productivity of the system as a whole.
Increased productivity in mixtures may also be due to allelopathic or
allelophilic interactions, changes in the rhizosphere flora, mechanical
factors (e.g., wind protection and support for understory crops), and
reduced pest and disease impact as a result of reduced resource concen-
tration and microclimate effects (Trenbath 1974). Mixtures also have the
capacity for compensatory growth, or the exploitation by one species of
resources liberated through the demise of an unlike neighbor.
Resource Partitioning and Competition Reduction
The most convincing body of evidence of resource partitioning in
crop mixtures regards spatial partitioning of rooting zones (Kurtz et al. 1952,
Litav and Harper 1967, O'Brien et al. 1967, Whittington and O'Brian 1968,
Ellern et al. 1970, Raper and Barber 1970, Baldwin and Tinker 1972,
Kauraw and Minko 1972, Portas 1973, Nelliat et al. 1974, Allmaras et al.
1975a, 1975b, Trenbath 1975b, Nair 1979). Other aspects of competition
reduction have been less well studied. Complementarity of nutritional
and water demands was suggested by lover (1948, 1959), Kolb (1962), and
Davies and Snaydon (1973). Reduced competition for nitrogen may be
especially important in mixtures containing legumes (de Wit et al. 1966).
Light distribution in canopies, reduced leaf angles in the lower canopy,
and displacement of shade-adapted species with large leaves and lower
optimum leaf temperatures lower in the canopy can all contribute to
improved light use efficiency in mixed crop systems (Monteith 1965,
Parkhurst and Loucks 1972, Trenbath 1974, Terjung and Louie 1973).
Temporal partitioning of leaf and root development and grain-filling
was thought to be responsible for high intercropping yields in two
studies (van den Bergh and de Wit 1960, Kassam and.Stockinger
1973). Labor inputs have been quantified in a number of studies
(reviewed by Kass 1978), but the results are not clear, and there is
no obvious reason to expect labor inputs to be reduced in intercropping
systems. In the studies reviewed by Kass, labor use was usually higher
in intercrops than monocultures, but the reverse was sometimes also
found to be true. The reasons for differences in labor use have not
A number of researchers have proposed models and/or field measures
of competition among crops in mixtures. Vandermeer (1982) obtained
whole-system yield graphs (at all density combinations) from intercrop
competition equations based on the Lotka-Volterra equations. He also
showed (1981) that the same nonlinearity of yield-density functions
that allows species coexistence in nature (Lotka-Volterra equations) also
occurs in intercrops that yield more than corresponding monocultures
(Land Equivalent Ratio > 1). That the mathematical expression for
nonlinearity is of the same form in both cases should perhaps not, however,
be taken as proof that the competitive function and Land Equivalent
Ratio are "mathematically identical" criteria, as Vandermeer states.
Other models of competition are not derived from the Lotka-
Volterra equations, and seem to lack a theoretical common ground; the
relations between de Wit's (1960) crowding coefficient, McGilchrist's
(1965) measure of aggressivity in mixtures, Donald's (1963) competitive
index, and Willey and Rao's (1980) competitive ratio are not clear.
Part of the confusion stems from the varying use of per-plant and per-
area measures, and the lack of differentiation between competition within
a species and competition among species, which creates difficulties in
situations where overall planting density is not held constant. The
de Wit and Lotka-Volterra equations contain competition coefficients
that vary under different growing conditions and overall planting
densities; the other three measures do not contain such coefficients,
and are much more readily applied to field observations. Only the
de Wit and Lotka-Volterra equations can be easily expanded to include
more than two species, however.
As our awareness of the complexity of biological interactions in
intercrops increases, the use of simplified models and indices of
competition to explain yield differences become less and less tenable
on theoretical grounds. It is nevertheless desirable that researchers
agree on a few measures of species and system performance that are
robust under varying planting densities, crop composition, and number
of crops interplanted.
Pests and Diseases
Insect ecology has contributed much to our understanding of
mechanisms operating to determine pest levels and stabilities in agri-
cultural systems (see the numerous reviews listed above). Due to the
many complex interactions involved, differences in pest mobilities, plot
size effects, and many other factors, generalizations cannot be made
regarding the effects of intercropping on pests and diseases. The
field data available support this conclusion: .Kass (1978) and Trenbath
(1976) gave numerous examples of both increased and decreased pests and
diseases in intercrop systems. Pest stability is much less well
researched; virtually no data are available on that topic, and both
stabilizing and destabilizing mechanisms may operate in intercrop
Stability and Risk
Risk reduction is an especially critical issue in the developing
tropics, where obtaining at least a minimum harvest each year may be
more important than obtaining maximum average yield over a period of
several years. The term risk is not synonymous with ecological
instabilityt, however. Risk contains both a productivity and a stability
component, since a system having generally high productivity is less
likely to fall below a given disaster level than a less-productive
system. Productivity rather than stability may be the most important
factor determining degree of risk (Rao and Willey 1980). Stability
rather than risk was the parameter evaluated in the present study, but
inferences regarding risk reduction are drawn in the discussion.
Stability in intercrop systems is much less well-researched than
productivity, resource partitioning, and pest and disease levels. Two
mechanisms underlying the hypothetically greater stability of intercrops
are compensatory growth and greater stability of pest populations.
Donald (1963) found that in 51 of 70 mixtures examined, per-plant
yield of one component increased and the other decreased (compared to
monocultures), but it was not possible to distinguish between uneven
resource partitioning and compensatory growth as possible causes of
the imbalance. Few clear examples exist in the literature, but increased
growth of understory plants in response to natural and artificial
damage to the overstory was reported by Fisher (1977) and Liboon et al.
Trenbath (1974) reviewed several data papers in which seed yield
stability of a mixture of two genotypes was greater than that of the
more stable component. This situation is the exception rather than the
rule; he stated that the stability of mixtures usually lies between
that of the component species or genotypes. Kass (1978)) cited several
studies in which year-to-year variability was less in mixtures than
in monocultures. The usual measure of yield variability in such studies
has been the coefficient of variation (C.V.) over time or space.
Rao and Willey (1980) found that stability of a sorghum-pigeonpea
intercrop in 51 experiments over a period of six years was higher than
either monoculture (C.V.=39 percent in intercrop, 49 and 44 percent in
monocultures), and very slightly higher than that of a weighted mean of
the monocultures (C.V.=39 vs. 42 percent). These differences are not
striking, however, and are confounded by the fact that the intercrop
consisted of full stands of both component crops. Rao and Willey
also determined degree of regression of intercrop and monoculture
yields with a site favorability index, and found that the intercrop
system response was intermediate between that of the less responsive
sorghum and the more responsive pigeonpea. Risk of crop failure was
much lower in intercrop than monoculture, due primarily to differences
The major difficulty with interpretation of the wealth of experi-
mental intercropping data concerns uncontrolled planting densities.
Overall level of biological activity in intercrops may be higher due
to planting more seed of species that grow into larger plants for
seeds of small plants. This increased "functional density" (density
of biological function) may bias yield results in favor of the intercrop.
Fisher (1977) found, for example, no yield advantage in maize-bean
mixtures beyond that due to increased planting densities. Increased
functional density of plants also leads to early cessation of growth
and altered microclimate (Pimentel et al. 1962), reduced root/shoot
ratios (Milthorpe and Moorby 1975) and increased risk of moisture stress
(Dowker 1963, Milthorpe and Moorby 1975). The question arises as to
how effects of inter- and intraspecific competition can be compared
without altering overall planting density. The problem becomes critical
for species of very different plant size, and is a serious factor that
confounds the results of most intercropping experiments.
One of three methods is normally used to determine densities
of the two interplanted species (Haizel 1974). In the additive method,
normal full stands of two crops are superimposed to give a double stand.
In the substitutive method, a given number of individuals of species A
are replaced by the same number of individuals of species B, giving either
a less or more full stand than normal, depending on which species uses
more resources per plant. In the replacement series method (Osiru and
Willey 1972, Willey and Osiru 1972), a given fraction of the optimum
stand of species A is replaced by the same fraction of the optimum
stand of species B. In addition to the above three methods, in many
intercropping studies the component species are planted in unusual
combinations of densities. Any of these experimental designs might
be suitable if one's goal is to find the crop or crop combination having
maximum yield. Only the "replacement series" method controls for over-
all functional density of plants, however, and can show the effects of
spatial diversity (not density) on system structure and function.
A second problem with the literature on intercropping concerns
methods used to compare intercrop and monoculture yields. The simplest
method is to compare the total yield from an intercrop with that from
monocultures of the component crops. Trenbath (1974), in a review of
over 300 studies, found that situations in which the intercrop yielded
either more or less than both components transgressiveve yielding")
were relatively rare. When two crops have been planted in equal pro-
portions in the mixture (by either the substitutive or replacement
method), the intercrop yield is most often compared with the mean yield
of the two monocultures (also known as the mid-monoculture yield or the
sum of the half-hectare yields). When used with the substitutive
design, the meaning of this comparison is not clear, since the density
of the species that was "substituted in" is usually not equal to half
its density in monoculture.
The most common method for comparing intercrop and monoculture
yields is the Land Equivalent Ratio (LER), also called the Relative
Yield Total (RYT) and defined as the sum of the component species'
per-hectare yield in intercrop, divided by per-hectare yield in
optimum-density monoculture. Each species' ratio represents the area
of monoculture required to give the same yield as one hectare of
intercrop; the sum of the ratios represents the area required to equal
the yield of all crops from one hectare of intercrop. Land Equivalent
Ratio has the advantage that it is easy to conceptualize and calculate.
One disadvantage is that it may have little meaning when applied to
systems having different planting densities, especially if the mono-
cultures are more sparsely planted than the intercrops, as is very often
the case, giving a falsely high LER. Land Equivalent Ratio also should
not be used when intercrop/monoculture yield differences are large,
because yield increases in the intercrop are given greater weight than
decreases (an artifact of the ratio calculations). For example, if
species A yields twice as much in intercrop as in monoculture (LERA = 2)
but species B yields four times as much in monoculture as in inter-
crop (LERB = .25), the overall performance of the intercrop is judged
highly favorable (LER = 2.25). Perhaps the most serious criticism of
LER is that it gives equal weight to both high- and low-yielding
components of an intercrop. This weighting has been justified by the
argument that both crops may be necessary to the farmer, but can result
in high LER's in situations where a minor crop failed in the monoculture.
Table 1 gives a hypothetical example of this situation, where LER = 2.3
even though the mid-monoculture yield was greater than that of the
intercrop! This distortion occurs because the species' performances
are not weighted by yield.
A New Index
A more meaningful index of intercrop performance than the area
of land required to produce the exact same yield of all crops as the
intercrop would be the yield difference of intercrops and monocultures
grown on the same land area. Yield could be expressed in terms of any
parameter of interest (biomass, kcals, protein, money, etc.). I propose
to call this new index the Yield Equivalent Ratio (YER) because it is
the ratio of yields from equivalent land areas, in contrast to the Land
Equivalent Ratio, which is the ratio of land areas for equivalent yields.
YER is defined as the ratio of intercrop yield on an area to yield of
Table 1. Distortion of Land Equivalent Ratio (LER) by
failure of a minor crop in monoculture. This hypothetical
example assumes equal planting densities (by the substitute
or replacement methods) in the intercrop.
SPECIES 1 SPECIES 2 TOTAL YIELD
A.INTERCROP YIELD 300 200 500
B. MONOCULTURE YIELD 1000 100 550
A/B .3 2
LER = .3 + 2 = 2.3
monocultures of the component crops grown on the same area, divided among
the monocultures in the same proportion as their intercrop planting
proportions. Planting proportions are expressed as percent of optimum
monoculture stands, and should add to one. With modifications the ratio
could also be applied where planting proportions added to more than one
(overpacked intercrops); the resulting ratio would then reflect packing
effects as well as diversity effects.
In mathematical form,
Y +Y +Y + Y
YER = _Y1,I 2,I 3,I 1 Yi,I
Y1,M P1) + Y2,M P2) + (Y3,M P3) + (Yi,M Pi
where Yi,I = per-area yield of species i in intercrop
Yi,M = per-area yield of species i in optimum-density monoculture
pi = proportion of optimum density monoculture planted of
species i planted in the intercrop
Yield Equivalent Ratios greater than one indicate that greater
yield per area is obtained from intercropping than from the same area
divided among monocultures of the same functional density, and reflect
reduced competition or other effects associated with increased spatial
diversity. When applied to a half-full-stand + half-full-stand mixture
of two crops, YERreduces to the ratio of intercrop yield to the mid-
monoculture yield, a comparison that has been widely used to evaluate
the performance of two-crop intercrops (Trenbath 1974). Unlike LER,
YER can be evaluated statistically through comparisons of the numerator
Performance of each species in intercrop and monoculture can also
be compared with YERs calculated on the basis of intercrop and monoculture
yields of that species only:
YER = 1,I
species 1 Y,
The goal of this study was to compare the productivity and
stability of diverse and simple systems of equivalent density, and
to elucidate some of the mechanisms responsible for differences
found. The following hypotheses and corollaries formed a framework
for the study.
Hypothesis 1. Productivity of diverse cropping systems is
greater than a weighted mean of monocultures of their
Corollary A. Intercrop productivity is intermediate
between that of the most productive and least
Corollary B. The greater the intensity of stress (the less
productive the growing conditions), the greater the
productivity advantage of intercrops compared with a
weighted mean of corresponding monocultures.
Hypothesis 2. Yield stability (constancy) of diverse systems is
greater than the stability of a weighted mean of the yields
of corresponding monocultures.
Corollary A. Intercrop stability is intermediate between
that of its most stable and least stable components.
Corollary B. Stability of all systems is greatest under
the most productive conditions (lowest intensity of
Hypothesis 3. Some species are more productive in diverse systems
than in monoculture; others are more productive in monoculture.
Hypothesis 4. Some species are more stable in diverse systems than
in monoculture; others are more stable in monoculture.
Hypothesis 5. Pest and disease levels are lower and pest
stability higher in diverse systems than in monoculture.
Hypothesis 6. Efficiency of resource use (light, water, nutrients,
human labor) is greater in diverse systems than in monocultures.
Corollary A. Root and leaf systems of different species
occupy different vertical zones, and vertical profiles
of leaves and roots are more evenly filled in diverse
systems than in monocultures.
Corollary B. Compensation occurs among components of diverse
systems; low productivity of one species causes increased
productivity of others.
Corollary C. Intercrop systems composed of species with
unlike resource-response curves and temporal growth
patterns have the greatest productivity advantage
compared with corresponding monocultures.
Hypothesis 7. Natural successional vegetation of the same age
as the crop systems is the most stable and productive of
all systems and has the most evenly-filled root and leaf
Experimental Approach to the Hypotheses
Intercrop systems selected for study were based on those commonly
used in the area. A four-crop system was included as a more diverse
system than the usual two-crop intercrop systems. Planting densities were
determined by the replacement method in order to maintain equal functional
densities. Monocultures of each crop were grown, and successional
vegetation and bare ground plots were included as controls. Four "stress
treatments" were applied: fertilization, pesticide spraying, defoliation,
and watering. Effects of these four types of stressors were compared
with a control treatment that was an imitation of local, low-energy-
input farming practices.
Response parameters included measures of biomass accretion and
distribution, pest and disease incidence, and resource use. Hypotheses
were tested concerning system performance, species performance in various
systems, and comparisons of intercrop performance with that of correspond-
The Tanzanian Agricultural Setting
The economy of Tanzania is largely dependent upon the productivity
of its small-scale farmers, who comprise an estimated 90 percent of the
population. Agriculture produces 40 percent of the country's GNP and
80 percent of its import earnings (Witucki 1978). Fifty percent of
food production is consumed on-farm. Tanzania is a net importer of
food, and a pattern of frequent food shortages has been documented from
1850 to recent years (Brooke 1967a, 1967b). The most frequent cause of
crop failure over this period was drought, but other causes included
excessive rainfall, locusts, birds, and armyworms. The level and
stability of agricultural production on small farms are therefore
questions of prime importance both to the national economy and the
welfare of the Tanzanian people.
Agricultural change has been rapid in Tanzania in the 21 years
since independence. In 1967 President Nyerere introduced a socialist
development policy called ujamaa. Development efforts were concentrated
in newly formed ujamaa villages, consolidated from previously scattered
settlements to facilitate introduction of improved social services and
agricultural technology (McKay 1968). This unusual development approach
and the philosophy of self-reliance are most eloquently expressed in
Nyerere's own writings (1973, 1974).
Under the new development approach, agricultural productivity
rose at a rate exceeding the population growth rate from 1965 to 1975
(Ruthenberg 1973, Witucki 1978). A campaign to increase farmer's
agricultural efforts on both private and communal fields stimulated
small farmer activity. International aid projects provided assistance
for improved seed, crop agronomy, and appropriate agricultural
technology. The most notable project was the National Maize Project,
begun in 1975. Its goal was to increase maize production in favorable
areas through improved seed and fertilizer inputs and improved crop
management (Fortmann 1976). Villages and individuals were particularly
encouraged to increase the area of maize and cotton planted and to
plant maize at higher densities. In many cases partially subsidized
inputs including improved seeds, fertilizer, pesticide, and tractors
Several years of bad weather in the mid-1970's, involvement in
the Uganda war in 1978-1979, increasing oil prices, and internal
management problems all contributed to a reversal of the progress made
from 1965 to 1975. Tanzania's foreign exchange situation is now
extremely serious, and the country is highly dependent on foreign
grants and loans to finance the present five-year development plan
In the present energy and economic environment, the best strategy
for agricultural development may be a two-pronged one. Higher productivity
through use of fossil-fuel inputs is desirable where economically
feasible; other, low-risk systems that make the most efficient use of
available resources are also desirable to minimize small farmer
dependence on unpredictable fossil-fuel-based resources. Development
of traditional, diverse cropping systems for increased productivity
may be one way to achieve this second goal. Identification of specific
mechanisms that may operate in those systems to reduce labor demand,
reduce risk of pest and disease attack, increase efficiency of water,
light, and nutrient use, and reduce the risk of crop failure is the
first step in the process of adapting traditional systems for increased
productivity without increasing risks for small farmers.
Traditional agricultural systems in Tanzania are extremely diverse
and highly adapted to the local climate and soils (Ruthenberg 1964, 1968).
Fortmann (1976) surveyed 96 farmers in Morogoro District (the district
where this study was conducted) and found that 64 percent of the area
planted to maize was intercropped wili one or more other crops. Seventy
percent of total maize area planted had no improved inputs; another
22 percent received only improved seed input, and a very low percent
was fertilized. The villages she surveyed were generally located on
favorable mountain slopes and river floodplains. The most common
species intercropped with maize in those areas were beans, rice, sun-
flower, and bananas.
Survey of Local Crop Systems
Local farmers' crop systems and cultural practices were used to the
greatest extent possible in this study, for three reasons. First,
results of this study will be of more practical value in suggesting
possible changes and improvements in cropping systems if they are taken
from authentic systems widely used by farmers. Second, little experi-
mental ecological research has been done with indigenous systems (and
cultural practices), especially intercropping system composed of three
or more species. Third, if intercropping is advantageous in terms of
stability and yield, one might hypothesize that it is most advantageous
for the unpredictable and low-resource environments in which small-
scale farmers operate, and for cropping systems that have persisted
over time, presumably because farmers found them to be successful.
To determine the most common cropping systems in the study area,
I conducted 30 interviews with farmers in 15 villages in the Kilosa
district of Morogoro Region in October and November, 1978. The District
Agricultural Development Officer of Kilosa District kindly accompanied
me, provided transportation, and introduced me to village leaders.
Selection of farmers was not random, but rather an attempt was made to
select experienced, longtime residents who were willing and able to
provide detailed descriptions of their farming practices. Interviewees
were actually selected by the Village Chairman (or his representative)
or the village agricultural officer, and so may represent the modernized
sector of the farming population. The sample was approximately 30
percent female and 70 percent male.
Interviewees were asked to describe qualitatively each of the
fields they had cultivated the previous year, including species
composition, plant arrangement, cultivation practices, and use of high-
energy inputs (seeds, tractors, fertilizer, pesticide). They were also
asked how long they had used each system and what they felt its advantages
and disadvantages were, particularly with regard to mixing species in
the same field. Finally, they were asked to describe systems they had
used in the past, and to tell what they felt their main agricultural
problems were. Survey results pertaining to crop combinations used
and farmers' perceptions of their advantages and disadvantages are
The survey revealed that an agricultural transition of surprising
depth and breadth is taking place in the surveyed villages, that is
probably typical of the situation in villages throughout Tanzania.
Farmers' ideas about their traditional cropping systems are changing
under the influence of agricultural officers and programs aimed at
introducing Green Revolution technology and increasing the area of
maize and cotton planted. Farmers are experimenting with the new
systems, but almost always keep some fields planted in traditional
Great diversity of cropping systems was present, both among
each farmer's fields and among different farmers' fields. All 30
farmers used more than one cropping system; a maximum of 9 and mean
of 3.9 systems per farmer were reported. Almost all farmers (28 out
of 30) planted at least one kind of monoculture, of which maize was
the most common (22 farmers out of 28). More farmers used monocots as
monocultures than used dicots as monocultures (26 and 18, respectively.
In addition to the major-season monoculture field crops, many farmers
(10) relay-planted bean monocultures after monocultures of Gramineae,
and many (9 of 20 queried) planted a mixed vegetable garden, invariably
composed of several species planted in separate rows, or more commonly
separate sections, of the garden. Most farmers interviewed (19 of 30)
cultivated some fields as monocultures and some as intercrop systems.
Of the remaining 11 farmers, 9 planted exclusively monocultures and
two planted exclusively intercrops.
Most farmers reported that the monoculture systems were relatively
new in the last 2-20 years, although sorghum, cassava, greengram, and
sparsely spaced maize were mentioned as monoculture systems used in the
past that are less widespread now. Many farmers (11 of 30) also
described intercrop combinations they had planted previously but no
longer used at the time of the interview. Two farmers mentioned that
they had changed from scattered to row planting, and two more noted
that they used to plant two or more species in the same hole but now
planted them in separate holes in the same field.
A total of 33 intercrop species combinations were given by the
interviewed farmers; these are broadly categorized and listed in
Table 2. The great majority of the systems were combinations of
monocots and dicots; only six of the 33 intercrop systems were all-
monocot or all-dicot. Three all-grass systems were mentioned, one
being in quite widespread use (maize-sorghum, 8 of 33 farmers). Three
systems containing no monocots were used, all of them composed of one
legume and one "other dicot" (= not Leguminosae or Cucurbitaceae).
Interestingly, two of the "other dicots" were tall species (sunflower
and cassava) which may function much like graminaceous species in these
all-dicot intercrop systems.
Grass-legume systems accounted for 13 and grass-legume-cucurbit
systems for 8 more, of the 33 intercrop combinations. Many of these
contained more than one grass and/or legume. In addition to the grass-
legume and grass-legume-cucurbit systems, six other monocot-dicot
systems were given: two grass-cucurbit, two grass-other dicot, and
two grass-legume-other dicot. The most common spatial arrangement for
monocot-dicot systems was relatively widely spaced rows of grass species
(in alternate rows if more than one grass species was present) with
legumes or other dicots scattered, or in rows, between the rows of
graminaceous species. Cucurbits were described as being planted "here
and there" or at several-meter intervals between the rows of monocot
Interviewees' responses were mixed regarding the advantages and
disadvantages of intercropping; 24 of the 30 farmers interviewed did,
however, have one or more comments to make on the subject. The responses
('dds mn-CTds soD) NO.IO *O
(Enn~rpu-F umnmusas) 3NVS3S
.o (IuaeTn3sa oqJue) VAVSSVD
o (Slunnloi sSnu-rdT?) O0$SV3
a1 _____(auq (MOD
S("dds BTrqlnono) NI-IAfld
Sqeqe______ ) IMI___I
o z (LaLeodAq STqpe-IV) SiflNV~d
0 (urLBnn5 snu e -) VaNOiadMO
ca (sna(in snToasBqa) 0v) DNj 2O1
; ., 4.1 __
u *^ (BBTnr un BU3TA) vaIdMOD X<>X X
lu .11 (sap o
U fln4sTUUBd) IMaIN HSn'dxg5
(ao[o0zq mnq3.os) WfHOtHlOS X +X
4 J SINaaOdS H. dO HaagfIN rI o r- -iM G 0 -In
rH M H lM H l-H
X X X X XXX
were categorized, and for the sake of simplicity the specific intercrop
systems involved were ignored. In almost all cases the farmer was
comparing maize-legume or maize-legume-cucurbit systems with maize
monoculture or monocultures of the species comprising the intercrop
Of nine farmers who mentioned intercrop/monoculture yield differences,
seven said intercrops had higher yield than monocultures and two said
monocultures yield more. Two farmers preferred intercrops because of
the diversity of crops one harvests and the early harvest of legumes
from intercrop systems. Two interviewees said that the chance of total
crop failure was reduced in intercrops, but ten noted that at least one
crop is likely to fail in an intercrop. Of those ten, six were referring
to the high risk of maize failure in two-grass systems, and one gave
shading as the reason. One farmer felt that maize is stronger in
intercrops in general. These comments are not necessarily incompatible:
risk of losing one crop and risk of losing all crops are not the same
thing, and it is quite possible that maize is at high risk in all-monocot
systems but not in other intercrop combinations.
One farmer said intercrop was better than monoculture during
drought, and one noticed reduced erosion on intercropped fields. One
stated that intercrops had more pests than monocultures; two said there
was no difference. Two said monocultures had more weeds than (maize-
pumpkin) intercrops; five said there was no difference.
The numbers of farmers who noted higher or lower labor requirements
for various tasks in intercrops compared to monocultures were as follows:
planting--three (intercrop more), five (less), one (no difference);
weeding--six (intercrop more), three (less), one (no difference);
harvest--one (intercrop more), one (less); total labor--two (intercrop
more), one (less). These results suggest that intercrops may be easier
to plant but harder to weed than monocultures. One interviewee felt
that monocultures are preferable for village communal fields (ujamaa
fields) because they are easier to manage when many people are involved.
Finally, three farmers mentioned crop mixtures that do not work well:
maize-fiwi bean, maize-rice, and any intercrop with cassava. Two more
admonished against planting two species in the same hole; one of these
specified that two grasses or two legumes should not share the same
Planting density is an issue of some interest in Tanzania because
farmers are being encouraged to plant at higher densities than they
traditionally have. Of six farmers who mentioned planting density, five
preferred sparse to dense plantings for the following reasons: survives
better in drought (four respondents); wilts less (one); lodges less in
drought (one); stronger roots (one); easier to plant (1). One farmer
commented that sparse plantings yield the same as dense; one that insect
levels are the same in both; and one that weed growth is the same in
both. One farmer preferred dense plantings because of higher yields in
The Study Site
Location and History
The study was conducted during the 1978 and 1979 growing seasons
(called Year 1 and Year 2) near the town of Morogoro, Tanzania. At
380E and 7S, Morogoro lies approximately 200 km west of Dar es
Salaam and the Indian Ocean, and sits at the base of the Uluguru
Field plots occupied 2 ha on the campus of the Faculty of
Agriculture, Forestry, and Veterinary Science, a branch of the University
of Dar es Salaam situated a few kilometers from Morogoro town. The
plots were located approximately 1 km north of the main campus build-
ings, occupying nearly flat land 550 m in elevation, 60 m above the
present level of the Ngerengere River to the west, and between small
streams flowing from the mountains toward the river. The plots were
bordered on three sides by a strip of young successional vegetation and
by a dirt road on the fourth. Beyond the successional strip was a
variety of field crop and pasture experiments.
The land on which the study site was located was part of the
Tanke Sisal Estate until the founding of the Faculty of Agriculture in
1970, after which it was used for agricultural experiments. Small
patches of Brachystegia- and Combretum-dominant miombo woodland occur
locally at uncultivated sites and suggest that this fire-adapted
vegetation once covered the area. The study site was in fallow for
two years preceding this study and was covered with a dense 2-3 m tall
brush dominated by a diverse mixture of grasses, including Chloris gayana,
Panicum maximum, Digitaria scalarum, Rottboellia exaltata, and Cynodon
dactylon. The successional regrowth also contained occasional sedges
(Cyperus rotunda and C. tuberosus), and an understory of herbaceous
and occasionally woody dicots including Celosia trigyna and C. laxa,
Bidens pilosa, Sonchus examiculatus, Crotelaria spp., Ipomoea mombasana
and 1. tenurostris, Galinosoga parviflora, Tridax procumbens, Amaranthus
spp., and many others. Several small vines were present, but they did
not dominate the grassy overstory.
Amount and distribution of rainfall are the most critical climatic
factors affecting crop production in the Morogoro area. As would be
expected for a site of low latitude and moderate elevation, temperature
and solar radiation do not limit the production of the major food
crops. Slightly lower temperatures occur in June and July (mean
21-22C) than during the rest of the year (mean 23-260C). Lowest
insolation is during the high-cloud-cover rainy period from March
through May (Jackson 19 75b).
Rainfall patterns in Tanzania result from the interaction of two
seasonal monsoons that dictate the movement of the intertropical
convergence zone, and complex local convergence patterns (Jackson 1975i).
Most of the midelevation Central Plateau of Tanzania is in the Tropical
Very Dry Forest life zone of Holdrige (1967), receiving 400-900 mm annual
rainfall in a single season from approximately November to May. Morogoro,
because of its proximity to the Uluguru Mountains, receives about 1000
mm annually, distributed rather unpredictably from November to May,
and is therefore at the edge of the Tropical Dry Forest life zone.
In most years a one- to three-week dry period sufficient to kill young
crops occurs in January or February. This short drought divides the
season into the "short rains" (November to January), generally considered
too unpredictable to be relied upon for the main crop, and the "long
rains," beginning in late February or March, during which most agricul-
tural activity takes place. The unpredictability of both amount and
distribution of rainfall makes annual cropping a risky occupation on
the Central Plateau, although less so in Morogoro due to the local
higher total precipitation.
The rains were considered by local people to be slightly heavier
than average in both study years. Tanzania Meteorological Station No.
96-3776 (less than 1 km from the study site) recorded 1117 mm for 1978
and 735/882 mm for the main growing season from January through July
in 1978 and 1979, respectively. Measures of precipitation taken daily
on the site between Harch 27 and July 28, 1978,were only 3 percent
higher than those from the meteorological station.
Soils on the study site are thought to be the product of uplift,
colluvial and alluvial deposition, and erosion by water. The Morogoro
area is geologically young compared with the sountern part of Tanzania
(Temple 1975, Pratt and Gwynne 1978). The yellow-red soil (color
7.5YR 4/6) was tentatively identified as an alfisol. Alfisols and oxisols
are both present in the Morogoro area (National Academy of Science
1972, Dregne 1976) but subsurface clay accumulation (see below), pre-
dominance of kaolinite clays, and moderate levels of weatherable
minerals at the site (Nicholaides and King 1980) suggest that the soil
is an alfisol.
To determine the overall nutrient status of the soil, 33 composite
samples of surface soil (10, 0-15 cm deep cores from each of 33, 15 x
21 m experimental plots) were collected. They were analyzed by the
Department of Soil Science in Morogoro for pH (1:1 in water), total
nitrogen (micro-Kjeldahl, Black 1965), available phosphorus (Bray and
Kurtz 1945, method no. 1), exchangeable bases, and organic carbon.
While the soils are not excessively acidic, they are low in total
nitrogen, available phosphorus, and organic carbon, and have low cation
exchange capacity and high carbon/nitrogen ratio. Ranges and means are
given in Table 3. The broad ranges of element concentrations and pH
indicate high spatial heterogeneity, but no clear spatial trends across
the study site were evident for any of the parameters.
In addition, the distinguishable soil horizons were sampled to a
depth of 100-130 cm in five soil pits located throughout the study
site (Figure 1). Determinations of pH, organic carbon, total nitrogen,
and particle size distribution were made by the Department of Soil
Science in Morogoro. Particle size distribution was by the Bouyoucous
hydrometer method; other methods were as above. Values of pH, organic
carbon, and total nitrogen in the top 20 cm agreed with those from the
Table 3. Surface soil pH, total N, available P, organic C,
and base concentrations. Range and mean are given for 33
composite samples of 10 0-15 cm deep cores each, except for
total N, where three anomalously low readings were discarded.
pH 4.8 6.0
TOTAL N (percent)
AVAILABLE P (PPM)
ORGANIC C (percent)
Na (me/100 g soil)
K (me/100 g soil)
Ca (me/100 g soil)
Mg (me/100 g soil)
5.0 6.0 7.0 0
%C % TOTAL N %CLAY-SILT-SAND
.5 1.0 1.50 .05.10 .150 25 50 75 100
5060 700 5 10 150 05 10 150 25 50 75 100
pH %C % TOTAL N %CLAY-SILT-SAND
Figure 1. pH, percent C, percent total N, and particle size
distribution in five soil profiles (I-V) taken at the study
composite samples. Organic carbon and total nitrogen were very low
below 20-40 cm in all profiles: less than 0.5 percent organic carbon
(minimum 0.1 percent) and less than 0.05 percent total nitrogen
(minimum 0.01 percent). Profiles I and II, located in the slightly
elevated southeast corner of the site, had increasing clay particles
and decreasing pH below 80 cm, suggesting an argyllic horizon, but
profiles III, IV and V did not show consistent trends in pH and particle
size distribution with depth. The silt fraction was consistently low
(less than 12 percent), and in all profiles-the higher pH values were
associated with a greater sand fraction. Soil texture was categorized
as clay to sandy clay.
Some parts of the site had a tendency to waterlog, but most of
the site experienced only temporary standing water after long and/or
hard rainy periods.
Crop System, Stress Treatments, and Planting Density
Based on the farmer survey, four intercrop combinations were
chosen for study: maize-sorghum, maize-cowpea, maize-pumpkin, and a
four-crop system, maize-sorghum-cowpea-pumpkin. These four represent
a range of types of crop combinations (grass-grass, grass-legume,
grass-cucurbit, and diverse grass-legume-cucurbit). For comparison,
each crop was also grown as a monoculture. Sorghum, cowpea, and
pumpkin densities were those recommended by the Ministry of Agriculture
for local conditions, while maize monoculture was grown both at the
recommended density and half the recommended density (hereon called
"sparse maize"). Two more systems were included as controls for
measurements of yield, Leaf Area Index (LAI), weed growth, soil moisture,
and recovery after defoliation: natural succession (on ground prepared
the same as the rest), and a vegetation-free or "bare ground" system. In
the second year of the study the maize-sorghum, maize-cowpea, maize-
pumpkin, sparse maize, and bare ground systems were omitted to allow
greater replication and inclusion of more "stress treatments."
The planting density of each crop in the intercrop system was
adjusted so that all crop systems except sparse maize had the same
functional density of plants; that is, differences in system structure
and function were due solely to differences in diverstiy and not to
differences in crowding or plant density. Equal functional density
was achieved by planting the species in a mixture in proportions adding
to one, for example one-half maize and one-half sorghum. The pro-
portion allocated to each species, multiplied by its monoculture or
full-stand density, gives that crop's planting density in the intercrop.
In the example given, a hectare of maize-sorghum intercrop would contain
the same number of maize plants as one-half hectare of maize monoculture
(being planted at half the monoculture density), and the same number of
sorghum plants as one-half hectare of sorghum monoculture. This method
of determining planting density so that all systems have the same
functional density is based on the "replacement series" methods of
Willey and Osiru (1972). To compare performance of either species or
whole systems, yields (or other measures) are weighted according to
The proportions for mixtures used in this study were based on the
farmer interviews, and were as follows: maize-sorghum (1/2 + 1/2),
maize-cowpea (1/2 + 1/2), maize-pumpkin (2/3 + 1/3), four-crop (1/4 +
1/4 + 1/4 + 1/4). The actual planting densities these proportions
represent, and the full-stand densities on which they are based, are
given in Table 4. Since individuals of different species do not replace
each other one-to-one in this method, total plant densities of the
Plant arrangements in the intercrop systems are shown in Figure 2,
and the spacing for each species in both intercrop and monoculture
systems is given in Table 4. Intercrop arrangements were patterned
after those described by local farmers, with the exception of maize-
sorghum, which consisted of two rows of sorghum 40 cm apart alternating
with one row of maize. This arrangement prevented crowding of sorghum
within rows and gave a more even spatial distribution than would
alternating single rows. Local farmers interplant maize and sorghum
in a variety of geometric and nongeometric patterns.
Four stress treatments were applied to the systems. In both
Year 1 and 2 half the leaves were systematically removed from the site
from all species in the defoliation treatment. In Year 2 resource
levels were increased by fertilization and watering treatments; losses
to herbivores were reduced in a pesticide-sprayed treatment. The
"control" treatment was managed with typical local methods (periodic
hand weeding and no fossil fuel inputs).
Table 4. Planting proportions, densities, and
spacings in the experi-
90 x 25
90 x 50
60 x 15
75 x 20
120 x 120a
180 x 25
90 x 20
two plants per hole
every fifth hole in every other row planted to pumpkin, not cowpea
90 x 37.5
240 x 180a
* U A
* F* *
o=SORGHUM = COWPEA
Figure 2. Intercrop planting patterns. Inter-row distance
of maize is 90 cm in the maize-cowpea and maize-pumpkin
intercrops, and 180 cm in the maize-sorghum and four-crop
intercrops; other distances are to scale.
In both years a rectangular array of plots was constructed, then
systems and/or treatments were randomly assigned to the array. In Year
1, two separate experiments were conducted (Figure 3). In the main
experiment, the 11 experimental systems were compared in the "control"
treatment only (farmer's conditions). Three replicates per system, a
total of 33, 15 x 21 m plots, were separated by 2 m walkways. A 2 m
buffer strip within each plot surrounded the sampling area, which was
divided into 6, 5 x 5 m subplots with 1 m walkways between for easier
management. In the second experiment, response to defoliation was
investigated in an adjacent set of 60, 8 x 10 m plots (2 treatments x
10 systems x 3 replicates). These plots were also separated by 2 m
walkways and contained 2 m buffer strips surrounding 4 x 6 m sampling
Results from Year 1 suggested several changes in design for Year
2. Due to extremely high plot-to-plot variability, more replicates of
smaller plot size were planted. In addition, the three two-crop systems,
sparse maize, and the bare ground system were eliminated to allow
greater experimentation with.stress treatments. The stress treatments
and main plots were combined into a single experiment in Year 2, shown
in Figure 3. All plots were 8 x 10 m, separated by 1 m walkways and
containing 1 m buffer strips around 6 x 8 m sampling areas. The
watering and pesticide treatments were isolated in blocks separated by
3 m walkways from the other three treatments to avoid contamination.
The watering treatment was located in the southeast corner of the site,
near the water source, and pesticide plots were located along the
- I II
south edge. (Daytime wind direction was unpredictable.) Plots of the
six experimental systems were randomly assigned within each of these
two treatments, and plots for the other three treatments were randomly
assigned tc the remaining area. To save labor and increase replication
of the control plots, watered plots were replicated only four times
(24 plots total), control treatment plots six times. The other three
treatments were replicated five times per system. Two plots were
eliminated due to standing water on the entire plot, and three due
to planting or treatment errors. Plots covered incompletely or briefly
with standing water were not eliminated. No system-by-treatment
combination contained fewer than four replicates.
The chronology of various agronomic and experimental activities
in Years 1 and 2 is shown in Figure 4. Weekly rainfall is given as a
corresponding histogram. Each cultural operation (planting, thinning,
weeding, harvesting) took a maximum of six days to complete (usually
fewer than four). To minimize age differences, operations were conducted
systematically throughout the field, going up one row of plots and
down the next, following the original planting order. The succession
system was slightly older than the crop systems due to natural seeding,
and a correction for this was applied in the results where possible.
In Year 1 the land was plowed and harrowed just before planting.
Judging from the survey of farmers, this was a common practice, and
i- "E 5
< I 5
: E- lo
0 0 0 0
C cOD q-
( 3aaM jad wu) 77VjNIV8
was almost a necessity in view of the time requirements and expense of
hand-clearing fallow land. In Year 2 the land was too wet to plow
during the last two weeks of February, and finally a crew of 40
workers hand-cultivated the site in time for planting in mid-March.
In both years the ground was cleared of weedy residue by hand and
quickly surface-hoed a second time to make a smooth surface for
The middle day of planting (March 18 in Year 1, March 12 in Year
2) was designated day 0; planting took five days in Year 1, 3 in Year
2. Between-row and within-row planting distances were measured with
marked ropes. Using the traditional dibble-stick method, maize
(variety Ilonga Composite) was planted 3 seeds/hole; sorghum (Serena),
5 seeds/hole; cowpea (a nonclimbing SVS3 variety), 3 seeds/hole; and
pumpkin (local variety), 4 seeds/hole. Good germination occurred in
both years, but a few seedlings were transplanted when 5-10 cm high
to fill gaps and achieve the desired planting density. The stands
were thinned at two to three weeks to one plant per hole for all
species except pumpkin (two per hole). All'systems except natural
succession were weeded three times, as necessary, in both years. The
third weeding was used to monitor weed growth and was not necessary
from an agronomic standpoint. Cowpea pods did not mature synchronously
and were harvested several times over the period shown in Figure 4.
Pumpkins were also continuously harvested, but all other crops were
harvested only once. All aboveground biomass was harvested and
separated into edible and nonedible components.
Application of Stress Treatments
On the fertilizer-treated plots, P205 from triple super phosphate
was hoed into shallow furrows along the rows or (for pumpkin only) holes
at planting, at the rate of 40 kg/ha. Nitrogen was applied as a
top-dressing along the rows at day 28 at the rate of 40 kg/ha as
200 kg/ha ammonium sulfate, the most commonly used and available nitrogen
fertilizer. In successional plots, fertilizer was applied along
imaginary rows 90 cm apart. Fertilization rates were those recommended
by the Soil Science Department in Morogoro for local conditions.
Plots designated for the watering treatment were in fact not
watered much, due to abundant rainfall during most of the growing season.
Seedlings were watered once after emergence (day 21) and from day 63
until harvest. Water was applied directly to the plant bases using
a hose. Successional plants were watered by standing outside the
sampling area as much as possible, inserting the hose into the vegetation
layer, and letting it run for several minutes before moving it to the
next point 1-2 m away.
Plots designated for defoliation were cut on day 20-25 (Year 1)
and day 49-50 (Year 2) and the cuttings removed from the plots. The
delay in Year 2 was due to the later first weeding, creating a labor
shortage during the hectic third through seventh weeks. Whenever
possible, every other leaf was cut where it joined the stem. This
system worked well for maize, sorghum, and pumpkin, but was difficult
in cowpeas due to the high degree of branching and in succession because
of the density of stems. Instead, in cowpeas, all leaves on one side
of the plant (partitioned by an imaginary line along the row) were taken.
In the successional plots, all biomass above 15 cm (Year 1) and 20 cm
(Year 2) was removed. This represented approximately 71 percent of
successional leaf area in Year 1 (estimated from the first LAI measure-
ment) and 57 percent of successional LAI in Year 2 (estimated from the
second LAI measurement and linearly interpolated by height). While the
cowpea and succession defoliation methods do not imitate natural leaf-
removing processes as well as the alternate-leaf method, they could be
used reliably by workers with minimal additional damage to the vegetation
by stem breakage.
The "pesticide" treatment was intended to reduce but not to
completely eliminate pest populations in a regime a farmer might
conceivably use. Spraying was stopped more than two weeks before the
harvest to avoid contamination. Sixty percent diazinon was selected
as a broad spectrum organo-phosphate pesticide effective against most
groups of insect pests. The pesticide-treated plots were sprayed five
times with diazinon at approximately ten-day intervals, beginning
day 17 (Figure 4). Application rate was 2.2 liters emulsifiable
concentrate/ha (recommended rate is 0.4-3.5 liters/ha). The second
diazinon spraying was mixed with the rainfast fungicide Blitox (50
percent copper oxychloride) at an application rate of 1 kg/ha. The
mixture caused moderate leaf damage to maize and sorghum, and was
clearly phytotoxic to pumpkin. Finally, as systematic protection
against sap feeders and stem borers during crop maturation, dimecron
(phosphamidon) was sprayed at approximately 300 g active ingredient/ha
on day 58.
Measures of System and Species Response
The response variables may be conceptualized as falling into four
groups: growth and productivity, resource use, resistance to natural
stressors, and response to artificial stress augmentation or amelioration.
These groups are not entirely exclusive; one variable may contribute
information in two or more of these categories.
Sampling regimes for the variables measured are described below.
Sampling took four or fewer days except where noted (the days given are
mid-days of the sampling periods). Systematic, rather than random,
sampling was frequently used to save time. Systematic samples were taken
such that individuals were selected from all locations in all plots;
sample size varied somewhat to meet this condition. The first individual
to be sampled in each row or plot was chosen randomly, and the interval
between individuals did not correspond to any regularity in the planting
patterns. For example, in the four-crop system cowpeas were not sampled
at an interval of four, since this is the interval at which pumpkins
Growth and Productivity
The term productivity is used here in the broad sense of biomass
accumulation per unit time (one growing season), and not in the usual
sense of net primary productivity (NPP). Net primary productivity is
rarely measured, and would theoretically include biomass that dies,
drops off the plant, decomposes, or is consumed by herbivores. Turnover
of plant parts was not measured in this study because of the short
growing period involved and high rate of biomass accumulation of the
systems being studied, and because the primary purpose of the study was
to make comparisons among systems rather than to determine NPP. In
mature ecosystems, changes in standing biomass may be unsatisfactory
indicators of net productivity because of high rates of death, decom-
position, and consumption. For agricultural and early successional
systems, however, where biomass accumulates rapidly, changes in stand-
ing crop of biomass, leaves, roots, etc., should be reasonably reliable
measures of productivity. Accumulation of edible biomass is an important
system attribute in its own right. Throughout the presentation and
discussion of the results of this study, the unit of time over which
growth or biomass accretion was measured is either given or assumed to
be the entire growing season (final harvest data).
Total and edible biomass
Total and edible biomass production, perhaps the best indicators
of species and system performance, were determined at harvest. All
crops were harvested at their appropriate time rather than synchronously.
Cowpea pods were harvested three times, pumpkins were harvested as they
ripened, and maize and sorghum were harvested nearly synchronously
(Figure 4). Natural succession plots were harvested after the cropping
Although some systems (e.g., cowpea) matured slightly before
others, all systems essentially filled the main growing season. There-
fore, all parameters related to yield are implicitly expressed as
production per growing season, rather than per day or per month. Chang-
ing the time unit might affect tests of differences among systems but
should not influence comparisons of a given species' performance
among systems or differences between intercrops and corresponding
All aboveground biomass, including dead material lying on the
ground and/or attached to the plants, was collected, divided into various
components, subsampled if necessary, dried at 800C, and weighed. Edible
and total aboveground yield per area, and edible and total yield per
plant were calculated. Edible yield from the succession plots was
assumed to be zero, and per-plant calculations were not performed for
the succession system. Total biomass at harvest of defoliated plots
did not include the biomass of defoliated leaves unless specified.
Allocation ratio was calculated as.(edible biomass)/(total aboveground biomass).
Biomass at physiological maturity (flowering)
In Year 2, subplots of all systems in the control and fertilized
treatments were harvested approximately at flowering (day 52, cowpea;
day 71, succession; day 73, sorghum; day 77, maize; day 79, pumpkin).
Plants in a 12 m2 agronomicc systems) or 6 m2 (successional system)
area were harvested by hoe, including roots to a depth of approximately
15 cm. In the successional system, monocots and dicots were separated;
the roots cut off at soil level; and the roots and shoots washed, air
dried, subsampled, ovendried at 800C, and weighed. Maize plants were
divided into leaves, stems, tassels, ears, and roots; sorghum into
leaves, stems, panicles, and roots; cowpeas into leaves, stems, pods,
and roots; and pumpkin into leaves, stems, flowers, fruits, roots, and
shoots. Each category was washed, subsampled, dried, and weighed.
Moisture content of the subsamples of each category was averaged for
each system-by-treatment combination. Root/shoot ratios were calculated
for the four agronomic species, successional monocots, and successional
Leaf mass at defoliation
The mass of defoliated leaves represents one-half the standing crop
of leaves at the time of defoliation, and was used to compare leaf
production up to the time of defoliation, and to calculate a measure of
total biomass production that included cut leaves. Defoliated leaves
of each species were collected, subsampled if necessary, dried at 800C,
Miscellaneous productivity measures
In addition to the basic productivity measures, other measures of
yield or biomass distribution were derived from the harvest data, which
included counts of cowpea seeds and pods, maize cobs, sorghum panicles,
and pumpkin fruits. For cowpeas, these measures were percent edible
seeds (by mass, Years and 2), percent edible seeds (by number, Year 1
and 2), number of pods per plant (Year 1), number of seeds per pod
(Year 1), and number of edible seeds per pod (Year 1). For sorghum,
the additional yield parameters were number of heads per plant (all
plots, Year 2) and percent of heads with greater than 75 percent of
the head surface area filled with grain (all plots, Year 2). Tillering
was also monitored in sorghum. Number of maize cobs per plant and
maize grain yield per cob were determined for all plots in Year 1;
number of pumpkin fruits per plant and edible yield per pumpkin fruit
were determined for all plots in Year 2.
LAI and canopy cover
Leaf Area Index (LAI) and canopy cover are measures of leaf production
as well as a species' or system's potential to capture light (discussed
in the section on resource use, below). Leaf Area Index was measured
by the "plumb-bob" method using a fishing rod and reel apparatus (Benedict
1976) to lower a weighted string vertically through the canopy. The
number of leaves hitting the string, averaged over a number of trials
at different points, equals LAI. This method underestimates actual leaf
surface area because leaves oriented vertically are less likely to be
touched by the string than horizontal leaves of the same area. This
error may be rather significant when comparing LAI of different systems
on an absolute basis, but should not be important when comparing a
given species' LAI among systems or LAI of intercrops and corresponding
Leaf Area Index was measured in control-treatment plots on days
28, 46, and 74 in Year 1 and in control, fertilized, and pesticide-
sprayed plots on days 40 and 65 in Year 2. Samples were not taken in
wet or windy weather; each sample took two people a maximum of seven
days to complete. Sixty measurements were made on each plot: six at
each of 10 random points located at least 2 m from each other and 1 m
from the edge of the sampling area. The average LAI from the 60
points was treated as a single replicate in the analysis. Height by
'15 cm intervals and by the species of each leaf the line touched was
recorded in all systems except succession, where only plant class
monocott or dicot) was noted. Since successional vegetation had a
"head start" on the crop systems, the LAI of the successional system
was set back seven days on each sampling date by linear interpolation.
The "system LAI" of an intercrop system is the sum of the LAIs of
its component species. Canopy cover was calculated as the percent of
trials (of the 60 per plot) in which one or more leaves was hit. Canopy
cover isa system-level variable; ground cover by species was not
Cowpea third-leaf area
A measure related to cowpea LAI, the area of the third leaf up
from the cotyledon, was obtained as part of herbivory measurements on
day 22 (Year 1) and day 30 (Year 2). The sampling and area measure-
ment techniques are given in the section on cowpea leaf herbivory,
Specific leaf mass
Specific leaf mass (g/m2 leaf) was determined on day 71, Year 2.
Leaves of three systematically selected plants of each crop species
were harvested from the border strips of each plot, the leaf area
determined, and the samples dried at 800C and weighed to determine mass
per unit leaf area. In the successional system, specific leaf mass of
monocots and dicots was determined for a composite sample of three
systematically-selected, approximately 10-cm-diameter samples of
vegetation from the border strip of each plot. Leaf Area Index of the
crop species was calculated as leaf biomass divided by specific leaf
mass. Biomass of leaves and stems was not separated in the successional
plots,so LAI could not be determined.
In Year 1, root biomass was determined by taking soil cores to a
depth of 40 cm with an 8 cm diameter root auger on day 96. The cores
were taken in 10 cm depth intervals and the soil carefully rinsed off
each segment of each core through approximately 12 strand/cm screens.
The roots were then dried at 80C and weighed to 10 g. Living and
dead roots, and roots of different species, could not be differentiated.
Root biomass data were corrected for roots remaining from the previous
fallow by subtracting (by depth interval) root biomass determined on
the bare ground plots.
Root cores were taken along a line perpendicular to planting rows,
at a systematic interval that did not correspond to the inter-row
distance. It was felt that this gave a more representative sample than
would coring at random points, in view of the very limited sample size.
Ten cores were taken from two plots in the maize-sorghum and maize-cowpea
systems, five cores from one plot in each of the other nine systems.
Roots were also sampled by a complete harvest method at flowering
in Year 2 (see above). Year 2 values were not corrected for residual
roots because the previous vegetation (the Year 1 experiment) was less
than one year old and residual roots were few at the start of the Year
2 growing season.
Root biomass was used as an indicator of root activity in this
study. Root surface area may be a better measure because large roots
are not active in proportion to their biomass; all systems in this
study were of equal and low age, however, and large roots were absent.
Height or length of individuals of each crop species was measured
on days 28, 46, and 89 (Year 1) and days 36 and 59 (Year 2). Control
plots (Year 1) and plots of all treatments (Year 2) were sampled.
Systematic samples were taken in each system, ranging from every 15th
plant in every other row (sorghum monoculture) to every plant (pumpkin
in four-crop system). Sample size was 39-51 in Year 1 and 13-32 in
Year 2 (except pumpkin,n = 1-30). Maize and sorghum plants were
measured from the soil surface to the straightened tip of the longest
leaf; cowpea was measured along the primary stem to the main meristem;
pumpkin was measured along its longest branch to the farthest meristem.
Survivorship was calculated from stand counts taken on days 24,
53, and 74 (Year 1) and days 28, 56, and 96 (Year 2). In Year 1, the
subplots of each main plot were surveyed and the results summed; in
Year 2 the 6 x 8 m sampling areas of all plots were inventoried.
Natural successional vegetation was not inventoried due to the difficulty
of distinguishing individuals. Mortality/30 days was calculated from
survivorship for the intervals between samples and for the growing
season as a whole. Stand counts were also converted to percent of
full stand (recommended or target planting density). At the first
sampling date, percent full stand ("fullstandedness") gives a measure
of how well the desired planting density was achieved (success of
crop establishment). For later sampling dates, fullstandedness of
species in a system can be summed to give a measure of plant survival
on a system level, a property I called "system fullstandedness." If
no mortality occurred, system fullstandedness would be one in all
systems except sparse maize, where it should be 0.5.
Ability of the experimental systems to exploit available resources
was of interest as a possible explanation for differences in productivity.
While it was difficult in most cases to obtain direct measures of resource
utilization, several collateral measures that should be correlated with
use of certain resources were obtained.
Amount and vertical distribution of LAI were used as indicators of
light interception in the experimental systems. Sampling methods are
Amount and distribution of root biomass are indicators of utili-
zation of the soil volume and soil resources (primarily, water and
nutrients). Profiles of root biomass in 10-cm intervals from 0-40 cm
depth were constructed for all system in Year 1, as described above.
Soil moisture was monitored by gravimetric methods in Year 2. A
tube-type soil auger was used to collect soil samples every two to three
days from the 15-30 cm depth zone. The samples were plastic-bagged in
the field, weighed, dried at 800C, reweighed, and the moisture content
calculated as (wet weight-dry weight)/(dry weight). Each day's samples
were taken from random points in a random selection of 25 plots; these
were later grouped into ten-day intervals to allow sufficient
replication by system and treatment to perform analysis of variance
for each interval.
A resource that is not usually monitored but is very important
to small farmers is labor. It was evident from the interviews that
human labor often limits the area planted. In Year 1, planting labor
was measured for the 60 small control and defoliation plots and weeding
labor was measured in the 33 main control plots, for the first and
second weedings. In Year 2, planting labor and first and second weeding
labor were measured in all plots. Crews of five to seven workers were
timed to the nearest five seconds and this was converted to person-
hours/ha. Since planting time included time to align and hold ropes and
to plant at the proper marked intervals, these determinations of planting
labor almost certainly overestimate the labor a small farmer would use.
The same is true of weeding labor, since it included thoroughly cleaning
the ground of weeds and collecting them for weighing. The measures
are nevertheless useful for comparative purposes. Efficiency of labor
use was also calculated, as the ratio of edible and total biomass output
to labor input.
Resistance to Natural Stressors
Stressors provided by nature were used to test hypotheses about
the resistance of the experimental systems to energy drains. All of
the growth and productivity measures in control plots reflect the
species' and systems' reaction to the sum of all naturally occurring
stressors, and are therefore included in this category. In addition,
the impact of a number of specific natural stressors was monitored,
including weed growth; lodging in windstorms; and population levels
of various pests and diseases, usually measured in terms of frequency
of affected host plants.
Competition by weeds is especially important to small farmers,
who often do not weed their fields as often or as thoroughly as was
done in this study. Weeds were hoed, collected, trimmed of roots,
subsampled, dried at 800C, and weighed. Samples were taken from control
plots in the first, second, and third weedings in Year 1 (days 17,
39 and 78) and from control, fertilized, and watered plots in the
second and third weedings in Year 2 (days 44 and 107). The first weed
sample in Year 2 was discarded due to a sampling error.
In both years a windstorm occurred as the crops were approaching
maturity, conveniently providing an opportunity to measure susceptibility
to lodging. The number of maize and sorghum plants inclined at greater
than 450 from vertical was recorded for all control plots on day 76 in
Year 1, and for control and defoliated plots on day 74, Year 2. No
measure could be devised that would meaningfully compare the degree
of lodging in maize and sorghum with that in the other experimental
Pests and diseases
The surveys conducted were intended to be representative, rather
than exhaustive, of the important pests and diseases, and were limited
by time, the need to isolate and identify pests, and constraints on
destructive sampling. In some cases, non-pest-specific plant symptoms
(e.g., leaf discoloration) were used as variables due to the difficulty
of separating and identifying the multiple causes of those symptoms.
Monitoring methods are described below by crop. Most pest and disease
levels are expressed as percent of plants, leaves, pods, or other plant
Rat holes. Shortly after planting, rats excavated some of the
planting holes, eating the seeds. The number of rat holes per plot was
counted in the 60 small control and defoliated plots on approximately
day 4, Year 1, for possible correlation with crop system.
Maize streak virus. Maize streak is a viral disease carried by
the maize leafhopper Cicadulina mbila (Homoptera: Cicadellidae), and
is easily identified by broken white lines parallel to the leaf veins.
Systematic samples of maize plants were scored by degree of infection:
0 (no symptoms), 1 (five or fewer leaves with symptoms), or 2 (more
than five leaves with symptoms). From 91-367 plants were scored in
each control plot on day 31 of Year 1. From 27-91 plants were sampled
on all plots on day 34, and 17-59 plants per plot on day 80, in Year
2. The data were later reduced to percent of plants having any symptoms.
Maize stalk borer. The maize stalk borer Busseola fusca
(Lepidoptera: Noctuidae) completes one generation in the stem of the
growing maize plant and a second generation in the cobs and stems of
maturing plants and trash left on the field after harvest. Populations
of stalk borers. were counted in systematic samples of 100 seedlings per
plot on day 13 in Year 2 by counting the number with "windowing,"
characteristic strips of holes in leaves that result from borers
entering the stem.
Sorghum shoot fly and tillering. Maggots of the sorghum shoot fly
Atherigona indica (Diptera: Muscidae) bore into young sorghum shoots,
causing death of the main shoot ("dead heart") and a compensatory
sprouting of new side shoots (tillering). Systematic samples of 380-676
plants were monitored for dead heart in control plots on day 32, Year 1,
and complete samples of 104-533 plants were taken during the day 22
stand count for all plots in Year 2. Percent of plants tillering was
also measured on day 47, Year 1 (50 systematically selected individuals
per control plot) and on day 57 in Year 2 (complete samples of 105-512
plants in each plot, measured in plots of all treatments).
Sorghum stalk borer. Sorghum stalk borers, primarily Busseola
fusca but possibly some individuals of Chilo and Sesamia species also,
were counted at harvest on day 130 in Year 2. Systematic samples of
25 main stems per plot were scored by number of borer exit holes. These
data were later reduced to percent of stems having any exit holes.
Striga. Striga hermontheca and Striga asiatica (Scrophulariaceae)
are plant parasites that attach to and feed on roots of maize and
sorghum. The Striga individuals were counted in all maize and sorghum
plots on day 77, Year 1, and the data expressed as individuals per
Pumpkin melonfly. The pumpkin melonfly Dacus cucurbitae (Diptera:
Trypetidae) oviposits under the skin of developing pumpkin fruits, where
the larvae cause the fruit to partially or totally discolor and in
extreme cases to rot and fall off the vine. Melonfly damage was
assessed for all fruits in control plots on approximately day 70,
Year 1, and in all plots on day 80, Year 2. Data from plots having
fewer than five fruits were deleted. Fruits were scored by degree of
melonfly attack: 0 (no visible damage), 1 (one or more entry holes
visible), or 2 (fruit discolored or fallen off the vine). These data
were later reduced to percent of fruits having any visible damage.
Pumpkin leaf discoloration. Pumpkin leaves are susceptible to a
number of leaf diseases, of which powdery mildew (Erysiphe cichoracearum),
downy mildew (Pseudoperonospora cubensis) and an unidentified mosaic
virus were present on the experimental pumpkin plants. (Other diseases
may also have been present.) Rather than attempt to differentiate the
symptoms of these diseases, percent of leaves with any kind of leaf
discoloration was measured. Leaf miner trails, discoloration around
holes, etc., were not included, so the discoloration variable should
reflect the status of the plant with respect to leaf diseases in
general. A systematic sample of 44-56 pumpkin plants was surveyed in
each control plot on day 76, Year 1. Each plant was scored according
to the number of leaves that were partially or totally discolored:
0 (none discolored), 1 (five or fewer leaves discolored), or 2 (more
than five leaves discolored). These data were later reduced to percent
of plants with more than five leaves discolored.
Cowpea leaf herbivory. Young cowpea plants were heavily infested
with the beetle Ootheca bennigseni, causing considerable loss of leaf
area through herbivory. Leaf area loss was determined for systematic
samples of 43-71 leaves per plot (day 22, Year 1) and 10-17 leaves per
plot (day 30, Year 2). Only the third leaf up from the cotyledon was
sampled for each plant since the object was to make comparisons among
systems rather than to measure total leaf consumption. No correlation
was made for leaf expansion because the data were to be used to compare
herbivory on fully expanded leaves of like ages. Total leaf area (area
if herbivory had been zero) was determined by tracing the leaves on
opaque paper, approximating the original margins where missing, cutting
out the traces, and measuring their area on a leaf area meter. Area of
tracings were adjusted downward by 8.7 percent to correct for error
during tracing and cutt!iAg,- a determined by a sample of 55 uneaten
leaves and their paper counterparts. The "residual area," or leaf
area remaining after herbivory, was determined by measuring the leaves
themselves in the area meter. To improve accuracy since the meter had
a faulty 0.I cm2 display, the leaf samples from each plot were measured
as a group. Control plots were sampled in Year 1, all plots in Year 2.
Cowpea aphids. Percent of cowpea plants on which Aphis fabae
(Homoptera: Aphididae) were found was measured in control plots on day
88, Year 1 and in all plots on day 80, Year 2. The samples contained
23-121 (Year 1) and 28-122 (Year 2) systematically selected plants.
Cowpea leaf diseases. Three cowpea leaf diseases could be
distinguished well enough to be reliably scored. Top necrosis, a
common viral disease in the Morogoro area, causes characteristic
yellow mottling and leaf shrivelling very similar to that caused by
cowpea mosaic virus. Occurrence of top necrosis was recorded for
systematic samples of 139-349 plants for control plots on day 39,
Year 1, and 28-71 plants for all plots on day 53, Year 2. Three
degrees of severity of top necrosis were recorded: 0 (no symptoms),
1 (fewer than five leaves with symptoms), and 2 (five or more leaves
with symptoms). These data were later reduced to percent of plants
having any symptoms of top necrosis.
The pseudorust Synchitrium dolichi produced obvious orange
patches on leaves and stems, and powdery mildew (Erysiphe polygoni)
created easily distinguished white patches on leaves. Presence or
absence of these two diseases was recorded in the same sample in which
aphids were counted (above).
Cowpea seed and pod discoloration. Many cowpea seeds and pods
were disformed and discolored as a result of numerous pests and diseases.
Discoloration seemed to be primarily caused by unidentified fungal
diseases but may also have been caused to some extent by heteropteran
feeding. Pod discoloration was scored as 0 (no discoloration), 1 (less
than 25 percent of the pod surface discolored), or 2 (more than 25
percent of the pod surface discolored) for random samples of 59-120
pods per plot from the first harvest in control plots, Year 1, and
5-20 pods per plot from the three harvests from all plots, Year 2.
Most samples contained the maximum number of pods in this range;
occasional smaller samples were from low-yielding plots, where all
pods were sampled. The data were later reduced to percent of pods
having any fungal discoloration. In Year 2, a three-harvest average
was calculated by weighting percent discolored pods by pod mass for
the three harvests.
A distinctive oval, off-white growth on the pods thought to be
a scab disease was also monitored in Year 2 in the same pod samples
that were scored for pod discoloration. Percent of scab-infected pods
was determined for all three harvests and the weighted average calcu-
lated as above.
Cowpea seeds were inspected for several types of damage, including
seed discoloration, in samples from the first harvest of control plots
(Year 1) and all three harvests from all plots (Year 2). Seeds were
scored as discolored or not discolored; browning around the edges of
holes was not included. In Year 1 a small cup was used to remove a
random sample of seeds from each subplot's well-mixed bag of seed
yield. In Year 2 all seeds were sampled. Sample size ranged from
648-865 (sum of six subplots) in Year 1 and from 0-1196 in Year 2.
In Year 2, a three-harvest average was calculated by weighting the
percent of discolored seeds by the number of seeds in each harvest.
Cowpea Maruca damage. Maruca testiculatis (Lepidoptera) larvae
feed on cowpea flowers and seeds, causing considerable yield reduction.
The population of Maruca larvae in closed cowpea flowers was sampled on
day 81, Year 2. Presence or absence of larvae was scored in 1-40
flowers from the border strips surrounding the sampling areas of cowpea
plots. Samples containing fewer than five flowers were later discarded.
In addition, percent of pods having one or more Maruca exit holes was
recorded and the three-harvest average calculated for the pod samples
described above. Percent of seeds partially eaten by Maruca larvae was
also recorded and the three-harvest average calculated for the seed
samples described above.
Other cowpea seed damage. Seed shrivelling and seed bruchid
beetle holes were recorded and the three-harvest average calculated for
the seed samples. In addition, the percent of seeds without any visible
damage was recorded as a composite measure of susceptibility to pests
and diseases. Shrivelling or wrinkling of seeds was due primarily to
piercing-sucking feeding by Acanthomia horrida but may also be attributable
in part to thrips. Acanthoscelides obtectus (Coleoptera: Bruchidae)
produced small holes or characteristic transparent "windows" in the
seed coat that could be easily distinguished and counted.
Methods of Analysis
Some of the above response measures are related to whole-system
function, while others pertain to a given species' performance. For
convenience, successional monocots and successional dicots as well as
the four crop species are referred to as "species." Some parameters,
such as LAI and yield, were measured both on the species and system
Three Types of Analysis
Three general types of analysis were performed: (1) absolute
comparisons of system performance (system-level measures only), (2)
comparisons of intercrop performance with that of a weighted mean of
corresponding monocultures (system-level measures), and (3) comparison
of each species' performance in intercrop and monoculture systems.
Absolute comparison of system performance
The response measures were compared among systems and treatments
by analysis of variance (ANOVA). The measures were not adjusted for
planting density. This type of analysis tests for absolute differences
among systems and treatments, without regard for crop composition or
Comparison of intercrops and corresponding monocultures
Whole-system performance of the intercrops was compared with a
weighted mean of corresponding monocultures by the Contrast procedure
of the Statistical Analysis System (SAS) statistical computer program
package (Helwig 1978). In this procedure, the given variable is compared
with several other variables that are weighted as specified. The
weighting factors were the fractions of the intercrop system allocated
to each species. Thus, system LAI of the maize-sorghum intercrop
could be compared with maize and sorghum LAI's, weighted equally. In
the maize-cowpea system the weighting factors were also 1/2 + 1/2; in
maize-pumpkin, 2/3 + 1/3; and in the four-crop system, 1/4 + 1/4 +
1/4 + 1/4. Intercrop performance was also compared with that of
corresponding monocultures using Yield Equivalent Ratio (YER).
Comparison of species performance in intercrops and monocultures
Measures of species performance (e.g., sorghum LAI) were corrected
for planting density, creating a new variable ("adjusted sorghum LAI")
on which the ANOVA was performed. In this example, adjusted sorghum
LAI equals sorghum LAI x 2 for the maize-sorghum system, and sorghum
LAI x 4 for the four-crop system. Adjusted variables equal the unadjusted
value for the monoculture systems (and successional monocots and dicots).
Analyses of variance performed on density-adjusted species variables test
for differences in a species' performance in different systems or
treatments, and are roughly equivalent to testing for differences in
per-plant performance. Variables expressed as frequencies were not
adjusted for planting density. Species performance was also evaluated
by the species' Yield Equivalent Ratio (YER), the ratio of species'
yield in intercrop to its expected yield based on planting density and
Analyses of Variance
Data from the three experimental data sets (Year 1 main control
plots; Year 1 small control and defoliated plots; Year 2 plots of all
treatments) were analyzed separately due to inherent differences in-
cluding plot size, weather, and sampling dates. One-way ANOVA by
system was performed on the data from the main control plots, Year 1,
and two-way ANOVA by system and treatment for the other two data sets.
Equality of variance among groups being compared is one of the
assumptions of ANOVA. Levene's test, in Biomedical Data Programs
program P7D (Dixon and Brown 1979) was used to test this assumption for
each variable. Species variables were tested for equality of variance
both before and after adjustment for planting density. Normality was
not tested for because ANOVA is known to be relatively insensitive to
deviations from this assumption (Sokal and Rohlf 1969, Steele and
Almost all variables met the equal variance assumption as tested
by the BMDP program. This may have been due partly to small sample
size and partly to the fact that many variables were taken as means,
and should therefore be normally distributed with reduced variances.
Four variables had unequal variance (p< .05): system edible yield
(Year 2), system total yield (Year 2), pumpkin edible yield (Year 2),
and adjusted pumpkin edible yield (Year 2). Log transformation would
be expected to reduce variance inequality due to differences in growth
rate of difference species, but log transformation did not equalize the
variance of the above variables (except adjusted pumpkin edible yield,
and that because the program eliminated variances of zero, from the
four-crop system, from the calculation). On closer inspection it was
evident that the unequal variance of all four variables was due to
great fluctuation in pumpkin edible and total yield in the pumpkin
monoculture plots. I ignored this relatively minor departure from the
ANOVA assumptions, and did not transform any of the response variables.
The spatial isolation of the watered and pesticide-sprayed plots
in Year 2 raises the question of whether the ANOVA assumption of
randomness was seriously violated. This was unlikely since all plots
were located within approximately 250 m of each other, and soil analysis
had revealed no clear trends in soil pH or nutrients across the site.
As a further check the test of equality of variance was performed on all
variables without the water and pesticide plots, and little or no
improvement in variance equality resulted. Plot-to-plot variability by
system and treatment was no greater when water and pesticide plots were
included than when they were omitted.
The ANOVA was conducted with the General Linear Model (GLM)
procedure of the SAS package, due to unequal replication, followed by
the Duncan procedure (Duncan's multiple range test) to detect specific
differences among systems and treatments. In two-way ANOVAs where
system-by-treatment interacttonwas significant, Duncan's tests of system
differences were performed for each system. When system-by-treatment
interaction was nonsignificant, tests for differences among systems were
performed on data from plots of all treatments, and tests for differences
among treatments were performed on data from plots of all systems. In
the figures, the composition of the data sets or "samples" on which the
Duncan's tests were performed is described where any ambiguity exists.
A significance level of p < .05 was used throughout the analyses unless
In many cases, striking differences in mean levels of response variables
were not found to be statistically significant by the ANOVA and Duncan's
tests. This was especially true of differences among systems in Year 1,
where the number of systems was high and the number of replicates low.
Duncan's tests performed by system or treatment tended to give lower signif-
icance than those performed for all systems or treatments combined, due
to reduced sample size. In interpretation of results, therefore, some weight
has been given to consistency of trends (among several variables, across
treatments, and in the two study years) as well as the results of the statis-
Analysis of Stability
Several approaches were taken to assess the stability of the experi-
mental systems and the species comprising them. Levels of naturally
occurring stressors were measured (e.g., pest frequencies) and system
comparisons made by ANOVA. Pest levels were also correlated with pro-
ductivity; strong negative correlations suggest that either high produc-
tivity reduces pest levels or pests reduce productivity. The effects of the
pesticide treatment on both pest levels and productivity indirectly indicate
the importance of pests as productivity drains in the agricultural
The second measure of stability was the productivity change of
the systems and their component species in the fertilization, pesticide,
defoliation, and watering treatments. Both absolute magnitude of
change and percent change from controls were calculated; absolute change
is here called "response," and percent change is called "responsiveness."
The sign of the change is ignored. Differing system responses to change
in stress level indicate differences in resource use. Responsiveness
is a measure of instability; those systems or species least responsive
to changes in levels of stressors should also be the most constant.
Responsiveness of species to increases in diversity (reduced competition
stress) was also used as a measure of species stability.
Finally, the coefficient of variation (standard deviation/mean)
of systems or species over the range of stress treatments was determined
as a measure of overall constancy. Year 1 and Year 2 data were not
combined for this measure due to differing plot size. Temporal
(year-to-year) variability was also assessed, by the coefficient of
variation of the Year 1 and Year 2 control plot means of a number of
productivity variables (analyzed by-system and by-species).
GROWTH AND PRODUCTIVITY
This chapter is divided into three results sections, that cover the
three approaches taken to analysis of the productivity data, and a
discussion of these results as a whole. The three approaches to the data
analysis are (1) comparison of whole-system productivity of the ten
experimental system monoculturess, intercrops, and successional
vegetation), (2) comparison of the productivity of each intercrop
system with a weighted mean of its component crops grown as monocultures,
and (3) comparison of each species' productivity in various crop systems.
The first approach compares systems on an absolute basis, without regard
for species composition; the second evaluates system response to changes
in diversity (spatial mixing of species); the third evaluates individual
species' responses to changes in diversity.
The terms "growth" and "productivity" are used in the general
sense of biomass accretion per unit time and biomass distribution, rather
than as net primary productivity (NPP), which includes drains to
herbivores, death, and decomposition (see Methods chapter). The
variables used to evaluate productivity of systems and species were
divided into those that relate directly to survival and biomass accretion
(called "direct" measures) and those that contribute information on the
distribution of biomass ("indirect" measures). The direct variables
include edible and total aboveground biomass (called "edible biomass"
and "total biomass"); aboveground biomass at flowering (called "biomass
at flowering"); root biomass (by two different methods in the two study
years); LAI (by the plumb-bob method in both years and also by harvest
at flowering in Year 2); a measure of leaf biomass (determined at the
time the systems were experimentally defoliated); canopy cover; stem
length agronomicc species only); mortality (for each agronomic species);
and fullstandedness (percent of full stand count, for each agronomic
system as a whole). The indirect measures include allocation ratio (the
ratio of edible to total aboveground biomass); root/shoot ratio (at
flowering); specific leaf mass (mass per unit leaf area); and a number
of "miscellaneous yield measures" that were measured for the four
agronomic species, such as maize cob production.
For clarity, most measures of productivity are first examined in
control plots to establish baseline trends. Data from the four stress
treatments (fertilization, pesticide spraying, defoliation, and watering)
are then added to evaluate effects of higher or lower levels of stress
compared with the controls, which represent local farmers' conditions.
In Year 1, two separate experiments were performed, and the data from
them are not mixed. Discussion of Year 1 control plots refers to the
large main control plots, while discussion of effects of defoliation
refers to the small defoliated plots (and their own controls).
Whenever interaction between cropping system and treatment effects
was low (Pinteraction > .05) tests for differences among systems and among
treatments were performed on complete samples of all treatments or systems
combined to increase the power of the test. When interaction was high
(Pinteraction < .05), Duncan's tests for differences among systems and
among treatments were performed separately for each treatment and system.
For the sake of expediency in the field, not all measures were
taken in all systems and treatments. Such missing blocks of data are
omitted from tables and figures or may appear as blanks in tables.
Results: Productivity Differences Among Systems
Measures of Biomass Accretion
Edible and total aboveground biomass at harvest
Control treatment. The mean total biomass of the systems studied
ranged from 10-350 g/m2, and the mean edible biomass ranged from 0-122
g/m2 (Figures 5 and 6). Edible and total yields were generally the same
or slightly lower in Year 2 than Year 1. Edible and total biomass of
crop systems containing either maize or sorghum was consistently higher
than that of either cowpea or pumpkin monocultures. The difference was
significant in all cases except one; pumpkin edible yield was not
significantly lower than that of the maize and sorghum monocultures in
Year 2, although it was less than half the naize and sorghum edible
yields. Total biomass of the successional system was approximately equal
to that of the maize- and sorghum-containing crop systems and was also
significantly higher than cowpes and pumpkin monocultures. Edible yield
of the succession systan was assumed to be zero.
Among the crop systems containing monoc.ts, there were no significant
differences in total biomass, but there were significant differences in
edible biomass in Year 1. Edible yield was significantly lower in the
recommended-density maize monoculture than in the sorghum, maize-sorghum,
sparse maize, and maize-cowpea systems. The four-crop system had
SORGHUM FOUR SUCCES-
MAIZE PUMPKIN COWPEA
Figure 5. Total aboveground biomass in the control treatment,
Years 1 and 2. Systems not sharing a common line are
significantly different by Duncan's tests.
MAIZE- SOR- SUCCES-SPARSE MAIZE- MAIZE- MAIZE FOUR
SORGHUM GHUM SION MAIZE PUMPKIN COWPEA CROP
SORGHUM MAIZE- SPARSE MAIZE- MAIZE- FOUR- MAIZE COWPEA PUMPKIN SUCCES-
0 SORGHUM MAIZE COWPEA PUMPKIN CROP SION
FOUR- MAIZE SORGHUM PUMP- COWPEA SUCCES-
CROP KIN SION
Figure 6. Edible biomass in the control treatment, Years 1 and 2.
Systems not sharing a common line are significantly different
by Duncan's tests.
the second-lowest edible yield, significantly lower than that of the
Stress Treatments. Defoliation had no consistent or significant
effect on whole-system edible and total yield compared with controls
(Figures 7-10), except significantly higher total biomass in control
than defoliated maize-pumpkin plots in Year 1. When the defoliated
leaf biomass was added to the final biomass, this difference was still
significant. After adding the defoliated biomass, total production
was significantly higher in defoliated than control plots in the maize-
cowpea (Year 1) and succession (Years 1 and 2) systems. Thus, all
systems except maize-pumpkin recovered from defoliation in terms of
edible and total biomass. Significant stimulation of production (when
defoliated leaf mass is included in the calculation) was demonstrated in
the maize-cowpea and succession systems.
Fertilization significantly increased total aboveground biomass
production compared with controls in all systems except cowpea monoculture,
and significantly increased edible biomass in all systems except sorghum
and cowpea. Edible biomass was, however, more than 50 percent greater
in fertilized than control sorghum plots. Watering had no consistent or
significant effect on edible and total yield. Pesticide spraying had
no effect on total biomass in the maize and sorghum monoculture, but
did cause a slight (nonsignificant) increase in the succession system,
a large (nonsignificant) increase in cowpea monoculture, and a decrease
in the four-crop and pumpkin systems. Spraying decreased edible yield
(nonsignificantly) in the maize, pumpkin, and four-crop systems, and
substantially (but nonsignificantly) increased it in the sorghum and
I I~ -
30 0 0
30 0 0
(ZW/B) SSVV0o8I 7VIOl
1 4-1 r o
r-l 4-J 4J 0
0 0) w u
0 ) a
CO p C-i 0)
a 4 C D- N
En *c 4- *
4i0 40) (
C -H) On) H
U- 04 0
CO r-lr-WWl -
4J 0 *40C 0 C
Pt 0 -
a n 4p 4-
4-I -H 0 0 )-
0> 0 w 04
14 > C 0 44 -H
w) w) u 0) 1
d 0) *rH P
C a) 0) 4-l |
m 41 c 0 w
COl 4J CO 4-4 N
,-l r 0) U 0W *-r
0o a -Hl -H
S0 4-4 -
4. 44 W *-H
O (0 01 0U UJ,
0 0 o 0 4
4-i I- c 41
000) 0 r
4 4 4-4 0 4 .
U) CO ') CO4
m ( *cH ow
" 01 rl 41 3
LO 10 CM 0) r
* rei *ri u a
S0UM -Hl 0
)1 1 H w 4J-
W 4.3J 0 l -
o Enlr-l 0a
> a) u a m u
0 >4i *H
-4 Ua A
r- *H 0) E
0rl (fl (0 c
o4 Co EC *
4 0o 0 0O 0
CO-I C- C
E- rl 4 E 4(
- 0) I-l CO .0 Co
4J 4J 4- -) C
ca En o
r'- g fCO C 4J -LH
0 (f 0
0) p 0) o -H1 Co
p- 44- CO 0)-
:3 to O 0 u
M 0) C r4-4 u
r 1-a Erl 3m 0)
4 *H 0
r,-W- QO .j
-a. o a
0 P4 1 H
-1i 4-J UW
a.( 4o e44
O o a- 14
I 0 2mu
C '- 0
N 0 N e t* N a
0-0L0-0 0 w
0 WCO 0
0 0 0 0 0 0 i- wno
~~~~ ~ ~ ** ~Sx ^
(~~~~~ mU/ )8VuM1 "ii-a
0 F D W P C
0 F PC W
0 PD WF C
F D C W P
Figure 9. Total aboveground biomass in five treatments,
Year 2. C = control treatment, F = fertilized, P =
pesticide, D = defoliated, W = watered. Mass of defoliated
leaves is not included in total biomass. Treatments not
sharing a common line are significantly different by
F C W P
P W D C F
Figure 10. Edible biomass in five treatments, Year 2.
C = control treatment, F = fertilized, P = pesticide,
D = defoliated, W = watered. Treatments not sharing a
common line are significantly different by Duncan's
D W C P
The ordering of whole-system edible and total biomass varied some-
what among the five stress treatments (Figures 11 and 12). Maize mono-
culture responded most to fertilization while sorghum monoculture
responded little; as a result, maize had the greatest edible and total
biomass production in the fertilization treatment. Maize total biomass
was significantly higher than all other systems in the fertilization
treatment. Pumpkin total biomass increased with fertilization to a
level significantly higher than that of cowpea monoculture; pumpkin
edible biomass increased to a level not significantly different from
that of the four-crop and sorghum systems. Pesticide spraying raised
the total biomass of the succession system to first place (significantly
higher than that of the four-crop and maize systems), and raised sorghum
and cowpea edible biomass relative to the other crop systems. The ordering
of systems by total biomass was the same in the watering treatment as in
controls, but there were minor changes in the ordering of systems by
edible biomass due to slight increases in maize and cowpea edible bio-
mass and slight decreases in the four-crop and pumpkin systems. The
defoliated maize-cowpea system was nearly as productive as the highest-
yielding system, sorghum monoculture. The four-crop system was adversely
affected by defoliation, and as a result its total biomass in the de-
foliated treatment was significantly lower than that of defoliated
succession and not significantly different from that of the two lowest-
yielding systems. Edible biomass of the four-crop system was also
reduced by defoliation to a level not significantly higher than that
of the cowpea and pumpkin monocultures.
S Ma pa
C U U
1 e w
w Cz 4 44
ui 0 CO t
0 H Q) 0H
00 (0 4
) Ia ; u -
rM a P *
O C uci)
n -4 0 0
SS r -
U) 4 H r : U
OU 0 0 0 C
A 0 r0) 0 C
U) I IZ H 0 W CO
M MM MM CD < a <
NxJ 0C 0 0
& cn t4 1 r.
M c Oc S U
M=4C ) a) i
0 P 6 0 0C a
SI to *>
N M 0 CO *O
u 0 c a 0 co
N N D 0 0 0 Mu "-I
p0 M 0 r o0
S-1 *0 u)
U U W :3 ca 410
1H V) E E o ri m
S1 U ) MpCU
O NU i NU 0 S CO 0 0 0
q 00 0 * Sz C/ En En 4 a
P4 V ( 3 E E S rc
3L vl 4 M
Suo u o 0o
0 ~ 40 C-
MH 0U- 0C
n I.4 .r
0 r4-1 -4 >
w -1 0
0 a 0 cn d
SN U .r
C 0 >
M M *ra
O 0 4 W aw i i
P O WaC
H H H H H cU
NU UP. U 4U U( *J41
I 0 j 0 .W
*a d n a)
S0 0n 0 oC
Q- H aI w < < < H H
0 40 P-4 W O
0 O *
M 44 00 1
>- I-1 C
EnN N NZ 2 S-l
P4 0 a c o -
MM M NL C4
) nNP. i N O v
30 (i<. P a a -
N 0 0B 0
0 0 O O 04
SU) C 4 C
0 p *a (0
^< a s HN N HN u h QN *
1 ; P. M 4- *
O En O4 Q)
P-4 F9 pr4
0 0 H)
Total aboveground biomass at flowering
Aboveground biomass at flowering (Figure 13) corresponded closely
with the final harvest biomass values (r = .91 for all systems and
treatments sampled). Biomass at flowering was 50-80 percent of final
biomass in the control and fertilized maize, sorghum, four-crop and
succession systems. In pumpkin monoculture, biomass did not increase
between flowering and the final harvest in control plots, but it did
increase dramatically in the fertilization treatment, where biomass at
flowering was only 36 percent of final biomass. In cowpea monoculture,
biomass increased from flowering to final harvest in control plots,
where biomass at flowering was 42 percent of final biomass, but declined
by almost 50 percent in the fertilized plots.
In both the control and fertilization treatments, highest biomass
at flowering was found in the sorghum monoculture, successional vegetation,
and maize monoculture, followed by the four-crop system (Figure 13).
Lowest biomass was found in the pumpkin and cowpea monocultures. Bio-
mass at flowering was significantly greater in fertilized than control
plots in the sorghum, maize, four-crop, and succession systems.
LAI and canopy cover
Temporal patterns of leaf area development varied from system to
system (Figures 14 and 15). Early development of LAI was most rapid in
successional vegetation, maize monoculture, maize-sorghum intercrop,
and pumpkin monoculture. In the second half of the growing season (after
day 46 and day 40 in Years 1 and 2, respectively) successional LAI increased
only slightly or declined, whereas LAI of systems containing maize continued
Figure 13. Total aboveground biomass at flowering, Year 2. Entries are
x s, in g/m2. Systems not sharing a common vertical line are signif-
icantly different by Duncan's tests. Asterisks mark systems in which
biomass was significantly greater in fertilized than control plots.
1.0 A FERTILIZED
a PESTICIDE --
SPARSE MAIZE .--
X 0- -----------
< O 5
0 -- **.-,
0 10 20 30 40 50 60 70 80
DAYS AFTER PLANTING
Figure 14. Temporal development of LAI in three treatments
in the monoculture systems, Years 1 and 2. Dashed lines
are Year 1; solid lines are Year 2.
0 10 20
30 40 50 60 70
Figure 15. Temporal development of LAI in three treatments
in the succession and intercrop systems, Years 1 and 2. Dashed
lines are Year 1; solid lines are Year 2. Successional system
data were set back seven days by linear interpolation to
correct for growth during planting of the crop systems.
1 I 1 1 I