Agronomy Journal 95:892-899 (2003)
© 2003 American Society of Agronomy
MODELING
CERES-Maize Predictions of Maize Phenology under Nitrogen-Stressed Conditions in Nigeria
D. T. Gungula*,a,
J. G. Klingb and
A. O. Togunc
a Dep. of Crop Prod., Federal Univ. of Technol., P.M.B. 2076, Yola, Adamawa State, Nigeria
b Int. Inst. of Trop. Agric. (IITA), c/o L.W. Lambourn and Co., 26 Dingwall Rd., Croydon CR9 3EE, UK
c Dep. of Crop Protection and Environ. Biol., Univ. of Ibadan, Ibadan, Nigeria
* Corresponding author (dgungula{at}yahoo.com)
Received for publication March 6, 2002.
 |
ABSTRACT
|
|---|
Simulation models have the potential of greatly enhancing decision-making by farmers and researchers in Nigeria. These models however, need to be adapted before use. This study was conducted to test the phenology module of CERES-Maize model version 3.5 under varying N rates as a step toward adapting the model in the Southern Guinea Savanna of Nigeria. Data on seven late-maturing cultivars of maize (Zea mays L.) grown under 0, 30, 60, 90, and 120 kg N ha-1 in the field for two seasons were used for running the model. There was a linear relationship between N rates and days to silking and maturity with R2 values of > 0.70 for most of the cultivars, indicating that N strongly influenced phenology. Predictions of days to silking at high N rates (90 and 120 kg N ha-1) were close, with most prediction errors of <2 d. The highest deviations in the calibration results were 4 and 2 d for 90 and 120 kg N ha-1, respectively, while in the validation results, they were 1 and 2 d. Similarly, days to maturity were closely predicted by the model at high N rates with <2-d deviations for most predictions. At low N rates, however, there were greater deviations in model predictions. This shows that the CERES-Maize model can be reliably used for predicting maize phenology only under nonlimiting N conditions. Thus, a N stress factor needs to be incorporated into the model for more accurate phenology prediction in low-N tropical soils.
Abbreviations: DTT, daily thermal time
|
INTRODUCTION
|
|---|
SIMULATION MODELS are becoming more accepted by farmers and researchers, primarily because of the availability of faster computers, improved accuracy of simulation, and increased need for simulation. With decreasing resources of farmers and researchers, simulation models are effective tools for better decision-making and enhancing research. Simulation models are cost-effective tools for rapid screening of cultivars. Breeders can use them in stress environments to reduce costs of multilocational trials to evaluate new cultivars (Mutsaers and Wang, 1999). Such applications require models that accurately simulate the complex interactions between the soil and plants as affected by climate and other environmental factors. These models need to be validated before being used in a different environment. Summerfield et al. (1991) reported that crop phenology is the most important aspect of crop adaptation and yield determination. Accurate prediction of phenology, therefore, is essential to predicting physiological responses under varying field conditions (Hodges, 1991). Inaccurate prediction of phenology causes simulated growth processes to occur on the wrong dates. Such differences can cause simulated processes to be erroneously impacted by climatic conditions. Accurate prediction of phenology will enable selection of varieties and planting dates so that stages critical for economic yield will occur during periods of optimum conditions.
Maize phenology is a function of genotype, with strong modulation by temperature and photoperiod as well as by soil N and moisture (Wallance and Yan, 1998; Hodges, 1991; Kiniry and Bonhomme, 1991). Much of the research on phenology has focused on temperature. The influence of N on maize phenology had earlier been reported by Ogunlela et al. (1988) and Dass et al. (1997). CERES-Maize, like most phenology models, does not take soil N into consideration while predicting maize phenology (Kiniry, 1991). Accurate predictions have been obtained (Wallance and Yan, 1998) in environments that are uniformly fertile due to high amounts of fertilizer applied to the soil. Thus, the model assumes nonlimiting conditions of N in predicting maize phenology (Mutsaers and Wang, 1999). Tropical soils are mostly kaolinitic and characterized by low organic matter and N (Agboola, 1990; Jones and Wild, 1975). Where N is limiting, maize phenology may be substantially changed. Moreover, most farmers in the West African savanna use suboptimal N fertilizers due to the high cost of chemical fertilizers and the inability of the soil to retain soil N (Yaro et al., 1999). Hence, under stress conditions of low soil N, there could be more deviations of observed values from values predicted by the model. Although some factors that influence maize phenology like temperature and photoperiod cannot be modified on a large scale, others like soil water and mineral nutrients can be modified (Wallance and Yan, 1998). This shows that it is possible for maize farmers to modify maize phenology while carrying out their normal agronomic practices, thereby increasing grain yield. Accelerating maize phenology can substantially increase maize grain yield in some conditions (Muchow and Carberry, 1993).
This research work was therefore performed to (i) elucidate the effects of N levels on maize phenology and (ii) test CERES-Maize phenology model Version 3.5 in the savanna ecology of Nigeria under adequate N and N stress conditions with late-maturing maize varieties developed for the tropics.
|
MATERIALS AND METHODS
|
|---|
Field experiments were conducted during the 1996 and 1998 cropping seasons in Mokwa (9°18' N, 5°04' E) in the southern Guinea savanna ecological zone of Nigeria during which data used in running the CERES-Maize model Version 3.5 were collected. Mean monthly rainfall distribution, minimum and maximum temperatures, and solar radiation in Mokwa during the two seasons are shown in Table 1. Table 2 contains the initial soil physical and chemical characteristics of soils during the two seasons as used in the model. Seven late-maturing cultivars of maize—S9325-SR, TZLCOMP3C1, TZLCOMP4C1, TZB-SR, IK9129-SR, ACR9222-SR, and 8644-27—were tested under five N rates (0, 30, 60, 90, and 120 kg N ha-1). The experimental design was a split plot with four replicates. Nitrogen rates were assigned to the main plots while maize varieties were the subplots. Each subplot was 5 m wide and 5 m long, consisting of eight rows. The plant spacing was 0.75 m between rows and 0.25 m within rows (53 333 plants ha-1). In the first year, two replications were planted on 21 June while the remaining two replications were planted on 5 July (about 14 d later) to enable differences in photoperiod to be observed. The difference in photoperiod was necessary to calculate the photoperiodic sensitivity constant, P2. In the second year, all four replications were planted on 4 July after the rains commenced.
The N rates were applied in two equal amounts, one application 7 d after planting and the other 28 d after planting. Phosphorus and K were applied once during the first N application at the rate of 26 and 50 kg ha-1 P and K, respectively. Soil water and available N were determined at 14-d intervals from 3 wk after planting until maturity. The maximum root length density was determined 2 wk after 50% silking was attained. Days to 50% silking and maturity and the maximum number of leaves per plant at anthesis were determined. The daily weather data for running the model (minimum and maximum air temperature, rainfall, and solar radiation) were obtained at Mokwa using a minimum data set recorder (Model LI-1200, LI-COR, Lincoln, NE).
Description of Model
CERES-Maize (Jones and Kiniry, 1986) is a process-level, comprehensive model that uses a daily time-step simulation. Growth and rates of change within the soil–plant–atmosphere system are calculated and integrated at daily time intervals. It was developed to simulate plant biomass and final grain yield of maize. It also simulates soil water, N, and P balance in different cropping systems (Tsuji et al., 1994). The model was designed to predict how much grain yield is affected under alternative technologies and for new growing sites by crop variety, soil water, N, and diseases. It was developed to reduce time and cost of agrotechnology transfer of new varieties and management practices. Phenological development is calculated as a function of growing degree days or daily thermal time (DTT) with a base temperature of 8°C. The maize phenological phases used in the model are described in Table 3.
The model assumes that the rate of development increases linearly above the base temperature up to 34°C and then decreases linearly to zero as temperature increases from 34 to 44°C. Similarly, rates of leaf initiation and leaf-tip appearance are assumed to change linearly in these two ranges of temperature. Photoperiodic induction is assumed to decrease with increasing photoperiod for photoperiods >12.5 h. The number of days of tassel initiation delay for each hour increase in photoperiod is assumed to be a constant for any given photoperiod-sensitive cultivar. The total number of leaves is determined from the number of leaf primordia initiated between seedling emergence and tassel initiation. Date of tassel initiation is determined using both DTT with a base temperature of 8°C and photoperiod. Silking or end of leaf growth is determined from total leaf number and the rate of leaf-tip appearance.
Model Calibration
The data collected from the first two replications in the first season were used for calibrating the model, which was a process of adapting some model parameters to the local conditions by adjusting their values. This was necessary because the first two replications performed better than the second two. The second-season data were used for validating the model, which involved using the model with the calibrated values without making any further adjustments of the constants. Calibration was done as suggested by Tsuji et al. (1994). The calibration process involved selecting an existing cultivar in the International Benchmark Sites Network for Agrotechnology Transfer (IBSNAT, 1990) genetics data file similar to the one to be calibrated and using its genetic coefficients as preliminary values. For each variety, the genetic coefficients were determined using 8°C as base temperature: (i) the DTT from seedling emergence to end of the juvenile phase (P1), which was computed using minimum and maximum temperature values only, as described by Alagarswamy and Ritchie (1991); (ii) photoperiodic sensitivity coefficient (P2), which was measured by the number of days tassel initiation is delayed per hour of photoperiod increase; (iii) the DTT from silking to physiological maturity (P5); and (iv) PHINT, phylochron interval, which is the interval in degree days between successive leaf-tip appearance. The P1 and P5 values were calculated as the integral of DTT of the days to 50% silking and physiological maturity and are presented in Table 4. Only temperature values were used in calculating the P1 because photoperiod does not affect the interval from seedling emergence to the end of the juvenile stage. Destructive sampling was employed to observe the end of juvenile stage when tassels were observed. Photoperiodic responses were observed in S9325-SR, ACR9222-SR, and 8644-27. The PHINT values were obtained by tagging five plants per plot and observing the number of leaves per plant at 3-d intervals from 21 d to the beginning of anthesis.
The three genetic coefficients required to run the model are given for each variety in Table 4. The model also requires site-specific inputs, including daily weather data (minimum and maximum temperature, rainfall, and radiation), and physical and chemical properties, planting date, between- and within-row spacing, and amount and timing of fertilizer and irrigation applications. Detailed descriptions of the CERES-Maize phenology model and input requirements are given by Kiniry (1991) and Tsuji et al. (1994).
Prediction deviations were computed by taking the difference of measured and simulated values. Where appropriate, percentage prediction deviations were calculated by dividing the computed deviations by the observed values and multiplying by 100 to convert to percentage. Negative deviations indicate underprediction while positive deviations indicate overprediction
Statistical analyses were performed to compare means of N rates, varieties, and calibrated and validated results using Genstat 5 Release 3.2 (Lawes Agric. Trust, 1995). Means that were significantly different were compared using least significant difference (LSD) values.
|
RESULTS AND DISCUSSION
|
|---|
Days to silking were delayed with increased N stress (Table 5) in all the varieties tested (P = 0.01). This is an indication that maize development and phenology are influenced by N levels in the soil. The differences in silking dates among varieties at particular N rates (P = 0.01) suggest that the effects of N stress levels on maize phenology differ among varieties even when those varieties are adapted to the ecological zone. This was not reflected in the model predictions. There was a linear relationship between N rates and days to silking (Fig. 1). In most cases, the R2 values were >0.7, indicating that N rates accounted for a high percentage in the variation of days to silking. In the few cases where the R2 values were low, other factors in the environment in the particular season might have acted to reduce the effects of N on days to silking. At higher N rates (90 and 120 kg N ha-1), days to silking were closely predicted in the calibrated results, with the highest deviations of 4 and 2 d (7 and 3%, respectively), respectively (Table 5). In the validation results, the predicted values of days to silking were significantly affected by N rates (P = 0.01), such that highest prediction error for silking date was 1 and 2 d (2 and 3% prediction error) for 90 and 120 kg N ha-1. At low N levels, there were greater differences between predicted and observed values, with the highest deviation observed from the 0 kg N ha-1 treatment in both the calibrated and observed results. This shows that silking is affected by N rates, but this has not been incorporated into the model. Hence, the model is not able to predict the effects of N stress on silking. The same values for days after silking were predicted for each variety regardless of the N rates. Silking was delayed with increased N stress. This confirms the report of Kiniry (1991) that the CERES-Maize model assumes optimum N conditions in predicting maize phenology. Under tropical soil conditions where N deficiency is common, the model may not give accurate prediction unless the soil N is high. Although validated values differed significantly from calibrated values (P = 0.01), the standard error of difference value for validation results was lower than the calibration results, showing that there was closer prediction of silking dates after calibration of the model.
View this table:
[in this window]
[in a new window]
|
Table 5. Comparisons of predicted (PR) and observed (OB) mean days to silking under varying N rates for different maize cultivars.
|
|
View larger version (38K):
[in this window]
[in a new window]
|
Fig. 1. Effects of N rates (kg N ha-1) on days to silking of seven maize cultivars taken in 1996 and 1998 cropping seasons.
|
|
Days to maturity significantly decreased with increased N rates (P = 0.01), with the earliest maturity for all of the varieties occurring at 120 kg N ha-1 in both the calibrated and validated results (Table 6). The relationship between N rates and days to silking was linear, with R2 values >0.8 in most cases (Fig. 2). This shows how important N was in determining days to maturity as a high percentage in the variation of days to maturity was accounted for by the N rates. For all of the varieties, maturity was delayed most at 0 kg N ha-1 (P = 0.01) for the two seasons (Table 6). However, the prediction of maturity period did not vary with N rates for all of the varieties. Again, this shows the model assumes that N stress does not affect maize maturity, confirming the report of Kiniry (1991). This suggests that breeders who wish to develop varieties for N stress conditions need a modified version of the model to make more accurate predictions of the performance of their new maize varieties in N stress environments such as those that are common in the West African savanna soils. At higher N rates (90 and 120 kg N ha-1), there were close predictions of maturity dates when compared to observed values in both the calibrated and validated results (P = 0.01). The highest differences in the calibration results were 2 and 1 d for 90 and 120 kg N ha-1, respectively. At 0 kg N ha-1, the highest mean prediction difference in the calibration was 6 d while it was 13 d in the validation results. The results show the model is able to predict maize phenology accurately at higher N levels, but under N stress conditions, there could be greater deviations in model predictions. The model can give better predictions under optimal N conditions in the savanna soils of West Africa.
View this table:
[in this window]
[in a new window]
|
Table 6. Comparisons of predicted (PR) and observed (OB) mean days to maturity under varying N rates for different maize cultivars.
|
|
View larger version (40K):
[in this window]
[in a new window]
|
Fig. 2. Effects of N rates (kg N ha-1) on days to maturity of seven maize cultivars taken in 1996 and 1998 cropping seasons.
|
|
The differences in model predictions of days to silking and maturity between the calibration and validation results without adjustment of model parameters after calibration is an indication that the model responds to the changing environmental conditions. However, the effect of N stress is not well reflected. Hence, it may not be able to predict well under multiple stress factors of N, P, and water (Mutsaers and Wang, 1999).
In the first year, days to silking were delayed more than days to maturity at low N rates. In the second year, days to maturity were more affected by low N rates than days to silking. The longer delay in maturity recorded in the second year than the first year can be attributed to the longer grain-filling period at low N rates in 1998 compared with the 1996 season where there were shorter grain-filling periods at low N rates (Table 7). These differences were not reflected in the model output. There is a need to incorporate other stress factors in the model (like N, P, and water) to make it applicable to conditions in the savanna soils of West Africa. The delay in maturity in the second season must have been due to the further stress caused by the drought as the rainfall amount (Table 1) and distribution were not favorable. Both N and water affect maize phenology and are not adequately accounted for by the model. At 120 kg N ha-1, the differences between predicted and observed grain-filling period were generally between 0 and 1 d, with the highest difference of 2 d in both the calibrated and validated results. However, at higher N stress levels, there were more deviations in model predictions of grain-filling period, with differences of >4 d in most cases and with the highest difference of 12 d in the validation results. This shows that the model at nonlimiting N conditions can reliably predict grain-filling period. Aggarwal et al. (1997) reported grain-filling duration as one of the major yield determinants. Its accurate prediction will be a step toward better prediction of grain yield. Understanding factors affecting grain-filling period will lead to a better understanding of factors limiting grain yield.
View this table:
[in this window]
[in a new window]
|
Table 7. Comparisons of predicted (PR) and observed (OB) mean grain-filling period under varying N rates for different maize cultivars.
|
|
In most varieties, the number of leaves at anthesis was closely predicted by the model at higher N rates in both the calibration and validation results, with mean prediction deviations of not more than two leaves (10% prediction deviations), except in few varieties where differences of up to three leaves (16% deviations) were observed (Table 8). The results show significant effects of N rates on the predictions of leaf number during validation (P = 0.01). At 0 kg N ha-1, there were greater errors as most of the predictions were 
10% errors (2 leaves). This shows that the model can predict leaf appearance and subsequently leaf number more accurately at higher N rates than under high N stress conditions. This confirms the report of Jones and Kiniry (1986) that the model assumes that N stress has no effect on leaf emergence and number.
View this table:
[in this window]
[in a new window]
|
Table 8. Comparisons of predicted (PR) and observed (OB) mean leaf number at anthesis under varying N rates for different maize cultivars.
|
|
|
CONCLUSION
|
|---|
The accurate model prediction of days to silking and maturity, grain-filling period, and maximum number of leaves at higher N rates is an indication that both researchers and farmers can use the model to accurately predict maize phenology under nonlimiting N conditions. But with greater N stress conditions, there are higher prediction deviations and less reliability in predictions of maize phenology by the CERES-Maize model. Predictions below 90 kg N ha-1 were less reliable compared with predictions at 90 kg N ha-1 and above. For accurate phenology predictions in N-deficient tropical soils, a N stress factor needs to be incorporated into the model for farmers and researchers to be able to use it with confidence.
|
ACKNOWLEDGMENTS
|
|---|
International Institute of Tropical Agriculture (IITA), Ibadan, Nigeria, is acknowledged for financing the work.
|
REFERENCES
|
|---|
- Agboola, A.A. 1990. Organic matter and soil fertility management in the humid tropics of Africa. p. 231–243. In E. Pushparajah, M. Latham, and C.R. Elliott (ed.) Organic matter management and tillage in humid and sub-humid Africa. Proc. Regional Workshop of the AFRICALAND Progr., 3rd, Antananarivo, Madagascar. 9–15 Jan. 1990. IBSRAM Proc. 10. Int. Board for Soil Res. and Manage., Bangkok, Thailand.
- Aggarwal, P.K., M.J. Kropff, K.G. Cassman, and H.M.F. ten Berge. 1997. Simulating genotypic strategies for increasing rice yield potential in irrigated, tropical environments. Field Crops Res. 51:5–17.
- Alagarswamy, G., and J.T. Ritchie. 1991. Phasic development in CERES-sorghum model. p. 143–152. In T. Hodges (ed.) Predicting crop phenology. CRC Press, Boca Raton, FL.
- Dass, S., A.K. Dang, N.N. Singh, and S. Kumar. 1997. Morpho-physiological basis of breeding drought and low-N tolerant maize genotypes in India. Combining ability for nitrogen use in maize. p. 107–111. In G.O Edmeades, M. Banziger, H.R. Mickelson and C.B. Pena-Valdivia (ed.) Developing drought- and low N-tolerant maize. Proc. Symp., El Batan, Mexico. 25–29 Mar. 1996. CIMMYT, El Batan, Mexico.
- Lawes Agricultural Trust. 1995. Genstat 5 Release 3.2. Rothamsted Exp. Stn., AFRC Inst. of Arable Crops Res., Harpenden, Hertfordshire AL5 2JQ, UK.
- Hodges, T. 1991. Temperature and water stress effects on phenology. p. 7–13. In T. Hodges (ed.) Predicting crop phenology. CRC Press, Boca Raton, FL.
- [IBSNAT] International Benchmark Sites Network for Agrotechnology Transfer. 1990. Documentation for IBSNAT crop model input and output files, version 1.1. Tech. Rep. 5. Univ. of Honolulu, Honolulu, HI.
- Jones, C.A., and J.R. Kiniry. 1986. CERES-Maize. A simulation model of maize growth and development. Texas A&M Univ. Press, College Station.
- Jones, M.J., and A. Wild. 1975. Soils of West African savanna. Tech. Commun. 55. Commonwealth Bureau of Soils, Harpenden, England.
- Kiniry, J.R. 1991. Maize phasic development. p. 55–70. In J. Hanks and J.T. Ritchie (ed.) Modeling plant and soil systems. Agron. Monogr. 31. ASA, CSSA, and SSSA, Madison, WI.
- Kiniry, J.R., and R. Bonhomme. 1991. Predicting maize phenology. p. 115–131. In T. Hodges (ed.) Predicting crop phenology. CRC Press, Boca Raton, FL.
- Muchow, R.C., and P.S. Carberry. 1993. Designing improved plant types for the semiarid tropics: Agronomist's viewpoint. p. 37–61. In F.W.T. Penning de Vries, P. Teng, and K. Metselaar (ed.) Proc. Int. Symp. on Syst. Approaches for Agric. Dev., Bangkok, Thailand. 2–6 Dec. 1991. Kluwer Academic Publ., Dordrecht, the Netherlands.
- Mutsaers, H.J.W., and Z. Wang. 1999. Are simulation models ready for agricultural research in developing countries? Agron. J. 91:1–4.[Abstract/Free Full Text]
- Ogunlela, V.B., G.M. Amorwa, and O.O. Ologunde. 1988. Growth yield components and micronutrient nutrition of field-grown maize (Zea mays L.) as affected by nitrogen fertilization and plant density. Fert. Res. 17:189–196.
- Summerfield, R.J., E.H. Roberts, R.H. Ellis, and R.J. Lawn. 1991. Towards the reliable prediction of time of flowering in six annual crops: I. The development of simple models for fluctuating field environments. Exp. Agric. 27:11–31.
- Tsuji, G.Y., G. Uechara, and S. Balas. 1994. DSSAT v3. Univ. of Hawaii, Honolulu.
- Wallance, D.H., and W. Yan. 1998. Plant breeding and whole-system crop physiology: Improving adaptation, maturity and yield. CAB Int., Wallingford, UK.
- Yaro, D.T., E.N.O. Iwuafor, V.O. Chude, and B.D. Tarfa. 1999. Use of organic manure and inorganic fertilizer in maize production: A field evaluation. p. 231–239. In B. Badu-Apraku, M.A.B. Fakorede, M. Ouedaogo, and F.M. Quin (ed.) Strategy for sustainable maize production in West Africa. Proc. Regional Maize Workshop, IITA, Cotonou, Benin Republic. 21–25 Apr. 1997. WECAMAN and IITA, Cotonou, Benin Republic.
This article has been cited by other articles:
|
|
|
|
 
Y. Miao, D. J. Mulla, W. D. Batchelor, J. O. Paz, P. C. Robert, and M. Wiebers
Evaluating Management Zone Optimal Nitrogen Rates with a Crop Growth Model
Agron. J.,
April 11, 2006;
98(3):
545 - 553.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
