S.S. JAGTAP, R.T. ALABI and O. ADELEYE

International Institute of Tropical Agriculture (IITA), c/o L.W. Lambourn & Co. Carolyn House, 26 Dingwall Road, Croydon CR9 3EE, England

(Received 3 March, 1998; accepted 11 June, 1998)

Code Number:CS98028
Sizes of Files:
      Text: 67K
      Graphics: Line drawings and tables (gif) - 47K

ABSTRACT

Maize productivity relationships were studied at Ibadan in the derived savanna of south-western Nigeria using the CERES-Maize simulation model at six crop densities ranging from 2.96 to 13.3 plants m^-2. The absolute growth rate for all plant components decreased linearly with density. The optimum density was highest for LAI, 9.35 plants m^-2, and lowest for grain yield, 6.95 plants m^-2. The mean absolute error in model predictions were 1.8% for days to silking and maturity, 8% for LAI, 5% for total dry matter and 13% for ear yield. Measured LAI, tops and ear weights across all densities were related to simulation outputs using: LAIM = 1.23 LAIS-0.58, R2= 0.70; TOPM= 0.49 TOPS+3154, R2=0.30; and EARM=1.04 EARS-150, R2=0.90. The model simulation of ear weight at densities greater than 8.89 plants m^-2 was unacceptable. Although simple and easy, the use of the CERES-Maize model in sub-Saharan Africa may be limited due to access to daily weather data, detailed soil data and computers. Using the simulated outputs as inputs, second order polynomials with density as the independent variable were fitted to the data to develop summary models that are more accessible in an intellectual and practical sense. These summary models may be used in the economic analysis of maize production.

Key Words: Crop growth, model predictions, optimal density, yield, Zea mays

Les relations du maïs en matière de productivité ont été étudiées à Ibadan dans la savane dérivée du sud-ouest du Nigéria, à l'aide du modèle de simulation CERES-Maize. Cette étude a porté sur six densités culturales allant de 2,96 à 13,3 plants m^-2. Le taux de croissance absolu, pour l'ensemble des composantes végétales, a chuté de faìon linéaire avec la densité. La densité optimale était la plus forte pour l'indice de surface foliaire (LAI), 9,35 plants m^-2 et la plus faible pour le rendement en grains, 6,95 plants m^-2. L'erreur moyenne absolue dans les prédictions du modèle était de 1,8% pour le nombre de jours à l'épiaison et à la maturité, 8% pour le LAI, 5% pour la matière sèche totale et 13% pour le rendement épis. Le LAI, les poids aériens et épis calculés à toutes les densités, ont été ms en relation avec les résultats simulés suivant la formule: LAIM = 1,23 LAIs-0,58, R2 = 0.70; TOPM = 0,49 TOPs + 3154, R2 = 0,30; et EARM = 1,04 EARs-150, R2 = 0,90. La modélisation du poids épi à des densités supérieures à 8,89 plants m^-2 était inacceptable. Quoique simple et facile, l'utilisation du modèle CERES-Maize en Afrique subsaharienne risque d'étre limitée par le manque de relevés météorologiques quotidiens, d'informations détaillées sur les sols ainsi que d'ordinateurs. En se servant des extrants simulés comme intrants, des polynômes de second degré incluant la densité comme variable indépendante ont été intégrés aux données afin de concevoir des modèles plus accessibles, aux sens intellectuel et pratique. Ces modèles sommaires peuvent étre utilisés dans l'analyse économique de la production du maïs.

Mots Clés: Crop growth, model predictions, optimal density, yield, Zea mays

For maximum biological yield of crops, planting at optimum density is imperative. However, determining such an optimum density depends on a number of factors such as soil fertility, cultivar growth habit and the environment (Bhargava and Saha, 1980; Chambi and Taylor, 1986). Most farmers in the humid tropics grow their crops at wide and random spacings because of the systems of cropping, which often involve growing two or more crops together. These crops are introduced into a piece of land over time, and allowance is usually made for such introductions. However, as management practices improve and more farmers grow their crops as sole plantings, specific plant populations are more likely to be used. Factors influencing the productivity and grain yield of maize are numerous, and include environmental conditions, cultivar characteristics, soil management practices and plant population density (Tetio-Kagho and Gardner, 1988b).

Various authors have reported maximising yields at different optimum densities. For instance, Prior and Russel (1975) reported an optimum density of 5.1 plants m^-2, while Tsai and Chung (1984) obtained maximum maize yield at 6.25 plants m^-2. Moreover, Lang et al. (1956) and Keating et al. (1988; 1990) reported that the optimal density for maximum yield of maize increased as nitrogen supply improved. For sorghum and sunflower, as site yield potential increased, so did optimum plant density (Myers and Foale, 1980; Wade and Foreman, 1988). In Botswana, Jones (1987) concluded that the plant density associated with maximum sorghum yield increased from 2.5 to 6.9 plants m^-2 as growing season rainfall increased from 200 to 700 mm. It is thus clear that spacing and plant density recommendations for maximum yield of maize cannot be easily transposed among diverse ecological environments. This poses a challenge for the development of technical recommendations targeted for diverse environments.

Crop simulation and modeling emerged as tools to integrate knowledge from various disciplines and provide a much needed framework for a technology evaluation approach that simultaneously considers key biological and physical factors that may influence the technology option. CERES-Maize is one such simulation and modeling tool that simulates the growth, development and grain yield of a maize crop under given climatic and cultural inputs. When validated, it provides a simplified method of tailoring recommendations for optimum spatial arrangement and population density to diverse ecological zones. Detailed descriptions of the biological and physical relationships used in CERES-Maize (Jones and Kiniry, 1986), and the ready availability of this software and its documentation have resulted in wide application of the model. Its potentials and constraints (particularly density effects) have been documented under diverse environmental conditions by several researchers in the USA (Hodges et al., 1987), Phillipines, Indonesia (Singh, 1985), Europe (Bonhomme et al., 1991; Plantureux et al., 1991; Lahrouni et al., 1993; de Vos and Mallet, 1987), Kenya (Keating et al., 1988 and 1990), Australia (Carberry et al., 1989), Brazil (Liu et al., 1989) and Nigeria (Jagtap et al., 1993).

The objectives of this study were: (i) to study maize growth and yield response subject to different row arrangement/densities in a tropical environment; (ii) to test the ability of the CERES-Maize model to estimate the development, growth and yield of maize over a range of plant densities; and (iii) to provide summary models that could be used in the absence of computers in an intellectual and practical sense. Highly complex process webs characterise production systems. To capture this variability, we selected the CERES-Maize model as a sufficiently sophisticated model to be sensitive to the interactions of major plant growth processes to the environment, but not requiring excessive computer resources or unorthodox input data. Once validated, the model could serve as a surrogate to field experiments for determining optimum maize density. First, CERES-Maize was validated at two population densities with a specific maize cultivar (TZSR-W). The model was then employed to simulate growth and yield of the same cultivar at six densities to examine the effects of density on production of maize in the transition zone of south-western Nigeria where maize area is expanding rapidly. Although the CERES-Maize model is easy to use, it still requires a computer to simulate effects based on daily weather data and soil data such as water holding capacity. This may limit its use by extension agents and others in sub-Saharan Africa where such information as well as access to computers is lacking. Hence summary models with high predictive value as well as being suitable for instructive purposes were provided.

Crop husbandry. The experiment was carried out at the International Institute of Tropical Agriculture (IITA) main experimental station, Ibadan, Nigeria (7°30' N, 3°54' E, 240 m above sea level) in early season of 1996. The soil of the study site is an Oxic paleustalf (USDA) with a surface of sandy loam texture and the following chemical characteristics: pH (H₂0) 6.2; organic C 1.98%; exchangeable K, Ca, Mg of 0.64, 7.09 and 1.39 ml/100g, respectively; and 40 ppm available P (Bray-I). The soils are characterised by high infiltration rate, low water holding capacity, and high susceptibility to runoff and erosion.

Maize variety TZSRW was planted on May 8, 1996 at six different spatial arrangements of 50 x 15 cm, 30 x 30 cm, 75 x 15 cm, 50 x 30 cm, 75 x 30 cm, and 75 x 45 cm spacings to give 13.3, 11.1, 8.9, 6.7, 4.4 and 2.96 plants m^-2, respectively. The experiment was laid out in three randomised complete blocks with six treatments in each. Each plot measured 4 x 5 m. To provide adequate nutrition, a 50:50 split of fertilizer (120 kg N, 26 kg P2O5 and 26 kg K2O ha^-1) was done at 14 and 42 days after sowing (DAS), in furrows 5cm deep and 10cm away from maize stands. All plots were hand weeded at 14 and 35 DAS.

Phenology and plant sampling. From 27 DAS, plant height, width and number of leaves were recorded twice a week throughout the growing period. Other crucial phenological events such as days to tasseling and silking were monitored and recorded. Plants within 1 m row length were harvested at 10 day intervals from 3 weeks after emergence until maturity for the two densities (4.44 and 6.70 plants m^-2) for which the model performance during all growth stages was to be assessed. Sampling was done twice, at silking and grain filling, for the remaining 4 densities since these are the two critical stages required for systems analysis and crop simulation (IBSNAT, 1988). The total dry weight was obtained after oven-drying for about 3 to 4 days at a temperature of 60 to 70°C. Total dry matter included all oven-dried plant parts except the roots. The leaf area was obtained using an LI-3100 Area Meter (LI-COR, Inc., Lincoln, Nebraska, USA) and the leaf area index was calculated by dividing the leaf area by the sampled ground area.

Radiation. Solar radiation transmitted through the crop canopies was measured by tube solarimeters (Szeicz et al., 1964; Delta-T Devices Ltd., 128 Low Road, Burwell, Cambridge, England). Solarimeters were positioned at 5 to 10 cm above the soil surface 4 weeks after planting and remained in place until crop maturity in 4 treatments. They were calibrated against a LI-COR pyranometer before and after the experiment. Each solarimeter is approximately 1 m long; hence they were placed at an angle with ends touching the middle of adjacent rows. Another solarimeter was mounted above the crop to provide simultaneous measurement of radiation above the canopy. Solarimeters were connected to a 21X micro data logger (Campbell Scientific Inc., Logan, Utah, USA) and solar irradiance through the canopies was sampled every 5 seconds and the half-hourly averages were stored. These were integrated into daily values. Intercepted photosynthetic active radiation (PAR) was calculated from global shortwave radiation by using the relationship given by Monteith (1993):

T_q = T_t^1.35

where T_q = fractional transmission of PAR quanta, and T_t = fractional transmission of short wave energy as measured by solarimeter.

The daily weather data were collected by an automatic weather station (21X micro data logger) located about 1 km from the experimental field. It recorded maximum and minimum temperatures, total solar radiation, wind speed, rainfall, minimum and maximum relative humidity on an hourly basis.

Model validation. To simulate a crop variety, the maize model requires five genetic constants. The growing degree days (base 8°C) from seedling emergence to the end of the juvenile phase (P1), extent to which development is delayed for each hour increase in photoperiod above the longest photoperiod at which development proceeds at maximum rate (P2), degree days (base 8°C) from silking to physiological maturity (P5), potential number of grains per plant (G2), and potential grain growth rate during the linear grain filling stage (G3). The values of the genetic constants used in this study were the same as those employed by Jagtap et al. (1993), since the same variety was used in both experiments. The results of the simulation were regressed against the observed values using SAS (SAS, 1989) linear regression procedure with or without intercept. Analysis of variance (ANOVA) was performed using SAS for data from all six treatments at grain filling stage.

Environment. Figure 1 shows mean weekly weather condition at the study site during the cropping season. Rainfall was erratic and low during the vegetative phase of the crop. Field monitored and simulated soil water balances suggested there was some water stress between 27-40 DAS for all densities. Thereafter, precipitation was sufficient to meet crop water needs for all treatments. The total rainfall received during the whole period of growth was 518 mm and 40% was received prior to silking (57 DAS) and 60% from silking to maturity (97 DAS).

The maximum and minimum temperatures declined from sowing (Day of year [DOY] 129) towards maturity (DOY 225). The mean temperature before silking was 23.59 °C and 22.25 °C thereafter. Total global incoming radiation also declined but with a steeper slope. Lower temperatures during reproductive growth stages implied a delay in crop development and therefore potentially higher yield. However, the rate of dry matter production was reduced as a result of declining radiation. Compared with longterm weather conditions, the recorded temperatures were above normal, rainfall below normal and radiation normal.

Figure 1: Climatic conditions during the 1996 experiment showing rainfall in mm. per week mean minimum and maximum and weekly mean total incoming global radiation (Mj m^-2day^-1).

Stand morphology. The maize plant height declined at higher population density (Fig. 2a). At higher plant density, there is intense interplant competition for growth factors such as nutrients, water and light within the crop stand. The tallest maize plants were obtained at 4.44 plants m^-2 while the shortest plants were observed with the highest population density of 13.3 plants m^-2. Two response groups could be observed. There were no significant differences in height amongst the two highest densities (13.30 and 11.10 plants m^-2), and amongst the four lowest densities (8.89, 6.70, 4.44 and 2.96 plants m^-5), but there were highly significant height differences between these two groups. The maximum height obtained at silking over the six densities was plotted against population density (Fig. 2b) and it showed that height followed a parabolic response to population density. A similar result was obtained by Tetio-Kagho and Gardner (1988a).

Figure 2: Response of maize canopy height (a), height at silking (b), canopy with (c), leaf number per plant (d), and percentage PAR interception (e) to stand density.

Maize plants were not only shorter in stature but were also narrower at high densities, demonstrating ecological adaptation to crowding (Fig. 2c). The widest plant canopies were observed with the density 8.89 and the narrowest with 11.1 plants m^-2. The patterns obtained for canopy width were generally irregular, although there were significant differences in width among the population densities, and they did not follow the same trend as height. The node or leaf number (Fig. 2d) followed a similar trend as height, demonstrating that leaf number increased with canopy height in maize.

Radiation interception. Percentage intercepted photosynthetic active radiation (PAR) was variable over time in all four populations monitored. During the juvenile phase, there were no significant differences (Fig. 2e). From about 50 DAS, however, significant differences existed among the different maize canopy arrangements. The highest percent interception was recorded with 11.1 plants m^-2 and the lowest with 4.44 plants m^-2: PAR interception observed showed a direct proportionality to density. The four maize densities showed that light interception increases with canopy development until about anthesis (DAS 57) when it became relatively constant until dough (R4) stage, and then declined due to senescence. The maximum PAR interception was attained at the same time for 11.1 and 8.89 plants m^-2, while the maximum was reached later for 4.4 and 6.7 plants m^-2.

Simulation seasonal performance. CERES-Maize simulated the six maize density treatments using daily weather data and management inputs. Table 1 summarises the performance for 4.4 and 6.7 plants m^-2 for which detailed measurements were available throughout the season. Days to silking and days to physiological maturity were accurately predicted for both densities to -1 days. Silking date prediction observed in this study was better than the -5 to + 3 days reported by Bonhomme et al. (1991). Better prediction in this study may be due to the previous study with the same cultivar (Jagtap et al., 1993) in which the genetic constants were accurately determined. Grain yield was under-predicted by 10 and 7% for maize at densities of 6.7 and 4.4 plants m^-2, respectively. This result, on the other hand, is consistent with the observation of Lorgeou (1991). Unit grain weight was predicted to an accuracy of 18 and 5% for 6.7 and 4.4 plants m^-2, respectively. Grain number per unit area was over-predicted with an error of 10% in 6.7 plants m^-2, while it was slightly under-predicted at 4.4 plant m^-2 with an error of 3%. Grains per ear was higher in the lower density for both measured and predicted values although it was more accurately predicted (5%) for the higher population. Maximum LAI was under-predicted (18%) for maize at 4.4 plants ha^-1, while it was over-estimated by 14% at 6.7 plants m^-2. Under US conditions, Retta et al., (1991) reported that maximum LAI was under-estimated with CERES-Maize. Biomass at harvest maturity were over-predicted by 2% and 7% for maize at 4.4 and 6.7 plants m^-2, respectively. This result is consistent with the findings of Jagtap et al. (1993) although it differs from that of Entenmann et al. (1989) and Plantureux et al. (1991) .

TABLE 1. Comparison of mean values of some selected field observations and simulated values for the two maize densities of 4.4 and 6.7 plants m^-2

Errors were computed by taking difference of measured and simulated values. Where appropriate, % errors were calculated for computed errors and observed values Hence, negative errors indicate over-prediction and positive errors, under-prediction, respectively

Within season development and growth. Within season simulations of LAI, above-ground biomass, and yield for the two maize densities are shown in Figure 3. The observed values are means of three replicates with a standard error bar indicating the extent of variability among the replications. The model over-predicted LAI in five out of seven data points (Figs. 3a-b), although the standard error bars touch the simulated lines in most cases indicating that the prediction was accurate for at least one replicate. The linear regression for simulated LAI against the observed values without and with the intercept showed that the intercept was not significantly different from zero at 5% and the slope was not significantly different from unity at the same probability level (Table 2). The coefficient of determination (R2) was 0.98 and 0.97 for 6.7 and 4.4 plants m^-2, respectively. This result contrasts with observations of other workers, although with different cultivars. For instance, Lahrouni et al. (1993) reported an under-prediction of LAI under Belgian conditions for cultivars Alarik and Gracia. Plantureux et al. (1991) also found that CERES-Maize underestimated the LAI of cv. Dea at Colmar, France.

Figure 3: CERES-Maize model performance with two population densities: (a) simulating LAI at 6.70 plants m²; (b) simulating LAI at 4.44 plants m^-2; (c) simulating total above ground dry matter at 6.70 plants m^-2; (d) simulating total above ground dry matter at 4.44 plants m^-2; (e) simulating ear yields at 4.44 plants m^-2.

TABLE 2. Results of regression of simulated LAl, biomass and ear weight on actual observed data for maize at densities of 6.7 and 4.4 plants m^-2

NS superscript indicates intercept not significantly differentfrom zero at 5% and slope not significantly different from unity at 5%. SS indicates significance at 5% level. * indicates regression without intercept and ** shows regression with intercept.

Total dry matter was predicted with less than 10% errors (Figs. 3c-d). In most cases, the standard error bars either touched or crossed the simulated line implying that the simulation fell within the range of experimental error. The intercept and the slope of the regression was not significantly different from unity both with and without intercept at 5% probability (Table 2). Excellent prediction of total biomass for this cultivar obtained in this study was reported by Jagtap et al. (1993). The model also predicted ear weight with less than 10% error (Figs. 3e-f). The mean observed ear yield was on the simulated line in three out of four points. For both densities, the regression gave an R2 of 0.99 with intercept not significantly different from zero at the 5% probability level.

Simulated and measured data at silking (57 DAS) and during the linear grain filling period (67 DAS) for all treatments are shown in Figure 4. The final growth analysis samples collected at maturity (97 DAS) had to be discarded as they were accidentally soaked by rain and fungi developed on them. For these sampling dates, LAI had reached its maximum level and ear growth was in the linear phase of growth. The highest LAI of about 4.0 was observed with the highest density of 13.3 plants m^-2 while the lowest was observed with the lowest density of 2.96 plants m^-2. In comparison to mean measured LAI, the simulated LAI over-estimated LAI by 0.2 to 0.5 up to the density of 6.67 and under-estimated by 0.3 to 0.5 thereafter. However, the majority of standard error bars touched the 1:1 line showing good agreement. Linear regression of mean measured LAI against simulated LAI gave LAIM = 1.23 LAIS-0.58, R2= 0.70. There was considerable variability in the measured tops weights with standard errors as high as 1300 kg ha^-1. The majority of the simulations were in agreement with measured values based on the nearness of points to the 1:1 line. A linear model TOPM= 0.49 TOPS+3154 with an R2=0.30 was fitted to the mean tops weight data. The model simulated ear weight in agreement with measured values (EARM=1.04 EARS-150, R2=0.90) for densities 2.96-8.89 plants m^-2. The model essentially became insensitive and over-predicted ear weight at densities greater than 8.89 plants m^-2 (see Fig. 4d). In the field, treatments with 11.10 and 13.30 plants m^-2 had small ears with few grains, which the model did not accurately simulate. This may not be a practical limitation of the model since such high plant densities are perhaps unlikely except if maize is grown for fodder.

Figure 4: Comparison of measured and simulated (a) total dry matter, (b) leaf area index (LAI) and (c) ear weight with CERES-Maize for six densities of maize at two different sampling dates (DAS 57 and 67); and (d) response of simulated and measured ear yield at grain filling (DAS67) to density with fitted second order polynomial lines.

The results of ANOVA are presented in Table 3, and show that there was no significant difference in biomass production (LSD = 2743) among the 6 densities of maize at the grain-filling stage. However, there were significant differences in LAI and ear weight. The highest mean LAI (4.33) was attained with 11.1 plants m^-2, followed by 13.3 plants m^-2 (3.5). The lowest LAI (1.94) was observed with the lowest density of 2.96 plants m^-2. There was an overlap in the mean separation of ear weights (Table 3).

TABLE 3. Results of ANOVA for observed ear weight, total above ground dry matter and LAI on DAS 67 (grain filling stage)

NS = Not significant

TABLE 4. Logistic functions describing response of maize to density in terms of absolute growth rate (fw(D)), optimum density (D_opt) and response function optimum fw(DD_opt)

Notes: fw(D) = absolute growth rate; D_opt optimum value solving fw(D) = 0; fw(D_opt) = Optimum response

Summary models of maize. The CERES-Maize model simulated with a high degree of confidence the field measured phenology as well as LAI, tops biomass, and ear weights for densities from 2.96 to 8.89 plants m^-2. Thus, the model can now be used as a tool for optimising the performance of maize. Although it is an easy and simple model to use, it still requires a computer to run as well as daily weather data, and detailed soil data such as water holding capacity. This may limit its use by extension agents and others in sub-Saharan Africa where such information as well as access to computers is lacking. Using the simulated outputs as dependent variables, simple predictive equations with density as the independent variable were developed for LAI, grains, ears, stem and leaf weight. The objective was to provide summary models more accessible to others in an intellectual and practical sense. An effort was made to maintain its relevance to physiological processes, but the primary aim was simplicity and accuracy of fit. Secondly, the form and degree of polynomials fW(D) chosen ensured ease of estimation of derived quantities such as (1) first order differential fW=(D) to obtain absolute growth rate, (2) the optimum density D_opt by solving fW=(D) =0, and (3) to calculate fW(D) at D_opt = maximum fW(D). The results are given in Table 4.

Second order polynomials with density as the independent variable resulted in data fits with high R2 values. The absolute growth rate or fW=(D) is a linear function whose value decreases with increasing population density. This implies that the sensitivity of yield components to density decreased as density increased, suggesting poor resource use efficiency for densities lower than the optimum and diminishing returns at higher densities. The optimum density was lowest for grain (6.95 plants m^-2) and highest for LAI (9.35 plants m^-2). Therefore, if maize is grown for grain, an ideal optimum density would be 6.95 plants m^-2, whereas if grown for fodder then the density needed is 8.30 to 8.56 plants m^-2. The total dry matter simulations showed that the variability among different population densities was small during the vegetative stages, suggesting that understorey crops may not be adversely affected by stand density. However, there were significant changes in biomass with increasing crop age and population increase. Dry matter yield difference narrowed with increasing density suggesting a nonlinear response with population density. The grain yield displayed a similar trend to that of biomass and LAI. Although Tetio-Kagho and Gardner (1988b) obtained an optimum maize density of 10 plants m^-2, the value of 6.95 plants m^-2 obtained in this study falls within the range of 5 to 7 plants m^-2 previously reported by other workers (Lang et al., 1956; Prior and Russel, 1975; Tsai and Chung, 1984).

The importance of plant density as a factor determining growth and yield of maize has been well established elsewhere but only a few reports are available for Nigeria (Fayemi, 1963; Remison and Kayode, 1982), where no definite recommendations are at present given to farmers. This is particularly critical for improved varieties that would probably not exploit growth factors to the maximum. Information on the effect of plant population on resource use efficiencies is almost non-existent. This study attempted, using growth simulation models, to generate flexible technology options leading to conditional recommendations.

A survey by Bartlett (1980) of farming systems in the forest zone of Nigeria reported that one third of farmers used a spacing of about 90 by 90 cm with 2 to 3 plants per stand; a smaller proportion (24%) reported a modified spacing of 90 by 60 cm; 15% used 75 by 75 cm; none used the recommended 90 by 30 cm with one plant per stand. This shows that in addition to an agronomic optimum, the economics of cultivation influence such variability in management practices.

The logistic function reported may be used in the economic analysis of maize production. For example, is it economical to plant 6.95 plants m^-2 yielding 32% higher compared with 2.96 plants m^-2? Or 12% higher compared to 4.44 plants m^-2? Also, for example, the potential grain yield simulated under no water or nitrogen stress at the optimum density was 7.7 tonnes ha^-1 while the optimum grain yield under experimental conditions was 5.5 tonnes ha^-1 or 29% lower due to less than optimal conditions.

The variety used in this study did not respond to plant density greater than 69,500 plants/ha, though yield responses within the range of 80,000 to 100,000 plants/ha have been recorded for temperate varieties (Milbourn et al., 1978). There are still many unexplained factors responsible for the low response of the maize varieties used in Nigeria, which may well be due to a combination of environmental and varietal factors. For example, there is no response to nitrogen application beyond 75-100 kg N/ha in most ecological zones of Nigeria and most of the currently recommended varieties have lax leaves. Upon the development of high yielding varieties, with upright leaf orientation and greater response to mineral nutrition, responses to higher plant population densities might be expected. There is thus a potential for growing maize at much higher plant populations in the tropics.

Bartlett, C.D.S. 1980. A review of minikits and the economics of recommended practices for rice and maize production. In: Proceedings of the 6th National Rice/Maize NAFRP Workshop. Remisson, S.U. and Olafare, S.O. (Eds.), pp. 8-20. Nigeria, NCRI.

Bhargava, S.C. and Saha, S.N. 1980. Physiological analysis of the growth, development and seed yield of oil-seed sesame. Journal of Agricultural Science Cambridge 95:733-736.

Bonhomme, R., Ruget F. and Derieux, M. 1991. Prise en compte de la photoperiode dans CERES-Maize. In: Phsiologie et Production du Mais. Picard, D. (Ed.), pp. 429-436. Pau, Paris: AGPM, INRA.

Carberry, P.S., Muchow, R.C. and McCown, R.L. 1989. Testing the CERES-Maize simulation model in a semi-arid tropical environment. Field Crops Research 20:297-315.

Chambi, J.Y. and Taylor, B.R. 1986. Sesame agronomy in south-east Tanzania. I. Plant population and sowing method. Experimental Agriculture 22:243-251.

de Vos, R.N. and Mallet, J.B. 1987. Preliminary evaluation of two maize (Zea mays L.) growth simulation models. South Africa Journal of Plant and Soil 4:131-136

Entenmann, T.A., Allison, B.E. and Van der Ploeg, R.R. 1989. Abschatzung der Evapotrans-piration eines Maisbestandes. Mitteilungen des Deutscheu Bodenbundlicheu Gesellschaft 59:129-132.

Fayemi, A.A. 1963. Effect of plant population and spacing on the yield of maize in the humid tropics. Experimental Agriculture 31:371-375.

Hodges, T., Botner, D., Sakamoto, C. and Haug, J. 1987. Using CERES-Maize model to estimate production for the US Cornbelt. Agriculture and Forest Meteorology 40:293-303.

IBSNAT. 1988. Experimental Design and Data Collection Procedures for IBSNAT. The Minimum Data Set for Systems Analysis and Crop Simulation. Technical Report 1. 73pp.

Jagtap, S.S., Mornu, M. and Kang, B.T. 1993. Simulation of growth, development and yield of maize in the transition zone of Nigeria. Agricultural Systems 41:215-229.

Jones, C.A. and Kiniry, J.R. 1986. CERES-Maize: A Simulation Model of Maize Growth and Development. Texas A & M University Press, College Station. 194 pp.

Jones, M.J. 1987. Plant population, rainfall and sorghum production in Botswana. I. Results of experiment station trials. Experimental Agriculture 23:335-347.

Keating, B.A., Godwin, D.C. and Watiki, J.M. 1990. Optimizing nitrogen inputs in response to climatic risk. In: Climatic Risk in Crop Production Models and Management for the semiarid Tropics and Subtropics. R.C. Muchow, and Bellamy, J.A.(Eds.), pp. 329-358. CAB International, Wallingford.

Keating, B.A., Wafula, B.M. and McCown, R.L. 1988. Simulation of plant density effects on maize yield as influenced by water and nitrogen limitations. In: Proceedings of the International Congress of plant Physiology, February 15-20, 1988 New Delhi India. Society for Plant Physiology and Biochemistry, New Delhi, India. 1414 pp

Lahrouni, A., Ledent, J.F., Cazanga Solar, R. and Mourax, D. 1993. Testing the CERES-Maize in Belgian conditions. European Journal of Agronomy 2:193-203.

Lang, A.L., Pendleton, J.W. and Dungan, G.H. 1956. Influence of population and nitrogen levels on yield and protein and oil contents of nine corn hybrids. Agronomy Journal 48:284-289.

Liu, W.T.H., Botner, D.M. and Sakamoto, C.M. 1989. Application of CERES-Maize to yield prediction of a Brazilian maize hybrid. Agriculture and Forest Meteorology 45: 299-312.

Lorgeou, J. 1991. Simulation du nombre de grains par CERES-Maize. In: Phsiologie et Production du Mais. Picard, D. (Ed.), pp. 455-468. Pau, Paris : AGPM, INRA Univ. Paris-Sud, AGPM.

Milbourn, G.M., Tillet, G.E.D. and Carr, M.K.V. 1978. Planting density for grain maize in South east England. Experimental Agriculture 14:261-268.

Monteith, J.L. 1993. Using Tube Solarimeters to Measure Radiation Intercepted by Crop Canopies and to Analyze Stand Growth. Pub. Delta-T Devices, document code TSL-AN-4-1. 11 pp

Myers, R.J.K. and Foale, M.A. 1980. Row spacing and population density in Australian grain sorghum production. Journal of Australian Institute of Agricultural Science 46:214-220.

Plantureux, S., Girardin, P., Fouquet, D. and Chapot, J.Y. 1991. Evaluation et analyze de sensibilite du modele CERES-Mazie en conditions alsaciennes. Agronomie 11:1-8.

Prior, C.L. and Russel, W.A. 1975. Yield performance on non-prolific and prolific maize hybrids at six plant densities. Crop Science 15:482-486.

Remison, S.U. and Kayode, G.O. 1982. Effect of population density and nitrogen on maize in the southern guinea savanna zone of Nigeria. Niger Agric.17:22-31.

Retta, A., Vanderlip, R.L., Higgins, R.A., Moshier, L.J. and Feryerkern, A.M. 1991. Suitability of corn growth models for incorporation of weed and insect stresses. Agronomy Journal 83:757-765.

SAS Institute 1989. SAS Users Guide:Statistics. SAS Institute, Inc., Cary, NC. 1674pp.

Singh, U. 1985. A Crop Growth Model for Predicting Corn (Zea mays L.) Performance in the Tropics. Ph.D. Dissertation, Department of Agronomy and Soil Science, College of Tropical Agriculture and Human Resources, University of Hawaii, Honolulu. 410 pp.

Szeicz, G., Monteith, J.L. and Dos Santos, J.M. 1964. Tube solarimeters to measure radiation among plants. Journal of Applied Ecology 1:169-174.

Tetio-Kagho, F. and Gardner, F.P. 1988a. Responses of maize to plant population density. I. Canopy development, light relationships, and vegetative growth. Agronomy Journal 80:930-935.

Tetio-Kagho, F. and Gardner, F.P. 1988b. Responses of maize to plant population density. II. Reproductive development, yield and yield adjustments. Agronomy Journal 80:935-940.

Tsai, C.L. and Chung, H.W. 1984. Effects of population density and N-fertilizer on the yield and ear quality of super sweet corn. Tainan District Agriculture Improvement Station. Research Bulletin 18:39-45.

Wade, L.J. and Foreman, J.W. 1988. Density x maturity interactions for grain yield in sunflower. Australian Journal of Experimental Agriculture 30:257-267.