Crop models have great potential as research tools, for crop system management and policy analysis. One of the most promising future uses of crop models is in crop improvement. The limitation in the use of models for crop improvement has been the inability of crop models to predict variety differences. Currently, the CANEGRO model, a sugarcane crop model developed the South African Sugar Association Experiment Station (Inman-Bamber, 1991a) can only model the performance of the NC0376 variety. Experiments were undertaken in the South East Lowveld of Zimbabwe, which is a hot and dry environment where sugarcane is grown under irrigation, to examine the canopy growth and development of four commercial varieties, ZN6, ZN7, N14 and NC0376. The study aimed at determining variety differences in canopy (tillers and leaves) development, develop parameters that can be used to model variety differences and test the improved CANEGRO canopy model for its ability to predict variety differences in canopy growth and development. For the late season, the numbers of leaves and tillers produced by each variety were counted every fortnight throughout the crop cycle. The total leaf area of the varieties and the individual leaf area on a stalk were determined using a Delta-T leaf area meter every fortnight. The date of emergence of successive leaves on a stalk was recorded daily. The leaf angles of each variety were measured every fortnight. The amount of photosynthetically active radiation (PAR) intercepted by the varieties was measured using a SunScan Ceptometer. Tillering and tiller senescence rates, phyllochron intervals, extinction coefficients and base temperatures were determined for the growth and development processes of varieties ZN6, ZN7, N14 and NC0376. Tiller and leaf population development was varietal. Tillering and leaf emergence were highly correlated to thermal time while tiller and leaf senescence were less correlated to thermal time. The poor correlation of the senescence phases to thermal time could mean that tiller and leaf senescence was driven by other factors other than thermal time. PAR interception could be one of these factors. The data showed that PAR interception could be a trigger of tiller senescence. The study showed that the tiller and leaf population development could be approximated by two linear equations. Tillering will be the first linear phase and tiller senescence the second linear phase. The first linear phase is driven by thermal time. While the second linear phase is triggered by PAR interception, the major driving factors need to be determined. This study proposed the use of two linear equations to model tiller and leaf population development as opposed to the polynomial equations used in the current CANEGRO model. Polynomial equations assume the factors driving tillering and tiller senescence are the same. The green leaf numbers per plant showed that all varieties experienced a decline in green leaf numbers with crop age. Varieties NC0376 and ZN7 had the greatest decline in green leaf numbers per plant while varieties N14 and ZN6 had the least decline. Variety ZN7 had the highest number of green leaves per plant while NC0376 had the least. The tiller growth and development was divided into three phases: the exponential phase during the initiation of stalks, the first linear phase during a period of rapid stalk elongation and the second linear phase during sucrose accumulation and maturation. The first two phases of development were driven by thermal time while the sucrose accumulation was not. There were variety differences in tiller growth and development. There were variety differences in base temperature for the development of various components of the canopy. Internode formation occurred at lower air temperatures than stalk elongation and tillering while canopy heights were correlated with higher air temperatures. This implies that internode formation could occur under conditions unsuitable for stalk elongation and may explain the short internodes frequently observed in stalks exposed to winter during rapid stalk elongation. The basic requirements for physiological parameters are that they should be stable across different environments, have significant differences between varieties and have physiological meaning. The parameters studied were thermal time requirement for shoot emergence, leaf appearance, to reach peak tiller population and to start of stalk elongation; surface area of the youngest biggest leaf, leaf number of the youngest biggest leaf, PAR transmission at the start of tiller senescence, extinction coefficients, and peak and mature tiller population. The difference between varieties in thermal time to shoot emergence was least using a base temperature of 16 QC compared to using 10 QC and therefore 16 QC could be a more appropriate base temperature for shoot emergence. The accumulated soil temperatures were less variable than accumulated air temperature and could therefore be a more reliable driver of shoot emergence. However, the limitations in the use of soil temperature are that it is not a readily available measurement and that it is not easy to measure. The gradual increase in phyllochron intervals appeared to be a better method of predicting leaf appearance compared to using a broken stick model. The phyllochron gradient was proposed, as it is likely to be a more robust way of modelling leaf appearance. The varieties had different phyllochron gradients. Variety ZN7 had highest rate of leaf appearance and produced the highest number of leaves per stalk while NC0376 had the lowest rate and produced the least number of leaves. There were statistically significant differences between varieties (P = 0,05) in PAR transmission at the start of tiller senescence and a base temperature of 16 QC was best at determining accumulated thermal time to the start of tiller senescence. Varieties with higher peak tiller population had higher final tiller population, lower thermal time per tiller and a higher ratio of final to peak tiller population. There were differences between varieties in the youngest leaf number attaining maximum leaf area and the leaf area of the youngest biggest leaf. Variety Nl4 had the biggest leaves and NC0376 had the smallest. Variety Nl4 had the highest leaf area index (LAI) while ZN7 had the lowest. There were significant differences (P = 0,01) in PAR intercepted by the varieties but there were no significant differences in extinction coefficients. Extinction coefficients increased with crop age. The varieties had significantly different (P = 0,01) leaf angles and ellipsoidal leaf angle distribution parameters. The measurement of LAI using SunScan ceptometer provided a better estimate of extinction coefficients than LAI measured using Delta-T leaf area meter. Model evaluation showed that CANEGRO canopy model version 2 was improved compared to than version 1. The model (version 2) was accurate in predicting tiller heights and dead leaf numbers per stalk. It was fairly accurate in predicting green leaf numbers per plant, stalk population and intercepted PAR but was poor in predicting LA!. Version 2 has proved to be a substantial improvement over version 1 in predicting stalk population. Generally, the version 2 model overestimated tiller heights early and underestimated later, overestimated the tiller population and LAI after peak, underestimated green leaf numbers per stalk for varieties ZN6, ZN7 and N14 and overestimated dead leaf numbers per stalk and intercepted PAR. The version 2 model predicted a constant green leaf numbers per plant and LAI from peak to harvest while observed data showed that green leaf numbers per stalk and LAI decreased towards harvest. Version 2 model predicted the tiller population of NC0376 closely but underestimated tiller senescence in N14 and also underestimated final tiller population in varieties ZN6 and ZN7. Future model refinements may need to focus on the prediction of the sigmoid pattern of tiller heights. The model may need to be calibrated to predict the green leaf numbers per stalk accurately, which should possibly improve the prediction of LAI that in turn could improve the prediction of intercepted PAR. The improvement in the timing and rate of tiller senescence should improve the prediction of tiller population particularly in varieties ZN6, ZN7 and N14. The study showed that the broken stick method IS superior in explaining leaf and tiller population development compared to using polynomial equations. The development of variety parameters helped improve the prediction of variety differences in canopy growth and development. A major weakness of most crop models is modelling variety differences in canopy growth and development. The inability of crop models to predict variety differences has limited their use in plant breeding. This study has resulted in an improved version of CANEGRO version 1 that is an initial attempt at modelling variety differences of sugarcane.

Thesis (M.Sc.)-University of Natal, Pietermaritzburg, 2003.