Results and discussion Directory UMM :Data Elmu:jurnal:E:European Journal of Agronomy:Vol13.Issue4.Oct2000:

raswamy 1989. Potential kernel number G2 and growth rate G3 were found by calibration of kernel weight and kernel number. The duration of grain-filling P5 was the value observed in the data-set used for calibration IRR 96. Values for the lower, drained upper and saturated limits LL, DUL, SAT were based on those measured in soil samples collected in an independent trial Castri- gnano` et al., 1994. 2 . 3 . Model 6alidation The accuracy of model predictions of crop de- velopment and growth LAI and above-ground biomass and grain yield was estimated using two statistical procedures. The first consisted of a linear regression between measured and predicted values on all the observation dates for each of the three variables under study. Two Student’s t-tests were then applied to verify the following two ‘null’ hypotheses: intercept = 0 and slope = 1. The second procedure followed the methodol- ogy proposed by Addiscott and Whitmore 1987 and Whitmore 1991 and consisted in partition- ing the sum of the squares of the residuals into pure error i.e. random variation and lack of fit i.e. systematic variation. The accuracy of maize evapotranspiration sim- ulations was tested by linear regression between ET values provided by the model and ET esti- mates from the simplified soil water balance, pre- sented above, and based on soil water measurements with the TDR. Also for the ET validation, estimates from IRR 96 were not considered.

3. Results and discussion

3 . 1 . Experimental obser6ations Fig. 2 shows the temporal variation in pre- dawn leaf water potential for each irrigation treat- ment during the two cropping seasons. Differentiation between treatments began 49 and 40 days after sowing, in 1996 and 1997, respec- tively. Owing to the warmer climate in 1997, development was speeded up and crop cycle dura- tion reduced, whereas pre-dawn leaf water poten- tial remained almost constant, ranging from − 0.3 to − 0.1 MPa for the well-watered treat- ment. Since, in maize the optimum for photosyn- thesis and stomatal conductance coincides with pre-dawn leaf water potential values less negative than − 0.3 MPa Katerji and Bethenod, 1997, IRR treatments may be considered as controls. For the other water regimes the plant experienced several water-shortage periods of intensity, which increased with the imposed stress level. Rainfall distribution in the 2 years differed in intensity and timing. The situation was usually different late in the crop season, when the extended rainless pe- riod in 1997 caused an increased water stress in the maize plants. The time variations in soil water availability for each treatment during 1996 and 1997 are shown Fig. 2. Time evolution of observed pre-dawn leaf water poten- tial and S.D. vertical bars. Fig. 3. Time evolution of water stored mm into soil profile calculated on the basis of TDR measurements. precipitation in 1996 partially compensated for water depletion in the soil. In addition, treatments were more differentiated towards the end of the crop season 1997. Plant water consumption was affected by water stress level, which caused a reduction in both the daily evapotranspiration Fig. 4 and seasonal evapotranspiration. The effects of water regime on the development of leaf area and the accumulation of above- ground dry biomass during the crop cycle were examined. Fig. 5 shows the green LAI evolution for all the treatments during 1996 and 1997. There were no significant differences among the treat- ments until the 56th and 42nd day after sowing in 1996 and 1997, respectively. Thereafter, LAI dif- fered significantly among water regimes; the lower irrigation volume gave the smaller LAI. In 1996 maximum LAI values were 16 and 33 lower Fig. 4. Time evolution of ET estimated independently from the CERES-Maize model. in Fig. 3. Soil water changes were similar to those in pre-dawn leaf water potential for all irrigation treatments and during both the years. This confi- rms the accuracy of pre-dawn leaf water potential for diagnosing soil water status Mastrorilli et al., 1999. Daily crop evapotranspiration, estimated from the soil water balance for all treatments and dur- ing 1996 and 1997, is shown in Fig. 4. Close agreement exists between the evapotranspiration and pre-dawn leaf water potential through time. During the first 30 days after sowing, most of the ET consisted of soil evaporation, controlled mainly by soil hydraulic properties and solar radi- ation. This period is characterised by a mean value of daily ET of about 3 mm. As the crop canopy grew, ET increased and reached its highest values about 9 mm per day at flowering after irrigation. The main differences between the 2 years are observed at the end of the crop season, when late Fig. 5. Evolution of measured LAI; vertical bars represent S.D. Fig. 7. The evolution in time of observed grain yield growth kg ha − 1 . Evolution of cumulative above-ground dry biomass Fig. 6 shows that the above-given considerations for LAI also hold for cumulative biomass. The accumulation of biomass differed in 1996 and 1997. This was due to a different rainfall pattern between the 2 years. Water stress intensity was higher in 1997 during the waxy maturity stage, mainly for the STR2 treat- ment. In the stressed treatments, the extent of reduction in the total biomass depended on the degree of water deficit; losses in percent of con- trol fully irrigated treatment were 19 and 34 in 1996, and 16 and 38 in 1997 for STRI and STR2, respectively. Grain growth for all the treatments and both the seasons is shown in Fig. 7. Grain yield was highest in the controls, approaching 9200 kg ha − 1 at 15 moisture content in the first year. However, the greater growth in leaf area and biomass that year caused a higher sensitivity to water stress; consequently the reduction in grain yield relatively to IRR treatments was greater in 1996 26 and 45 for STRI and STR2, re- spectively than in 1997 17 and 40. 3 . 2 . Calibration Observed values for the well-watered treat- ment in the year 1996 IRR96 were used to approach by trial and error Godwin et al., 1989 the ‘genetic coefficients’ that gave the most realistic estimates of phenology and pro- duction. These coefficients were used in all the subsequent simulation runs validation over a range of meteorological and water stress condi- tions. The genetic coefficient values estimated for Maltus cultivar are listed in Table 2. In the same table there are also the genetic coefficients for the cultivar A632 × WI 17 which were re- ported in the user’s guide to CERES model. The latter cultivar has been chosen because it belongs to the same maturity class as Maltus. Fig. 6. Evolution in time of cumulative measured above- ground dry biomass kg ha − 1 . Vertical bars represent S.D. of each measurement. Table 2 Genetic coefficients estimated for the cultivar Maltus and those reported for the cultivar A632×W117, and soil data of Rutigliano Italy Genetic P1 Cultivar P2 P5 G2 G3 220 780 Maltus 781.5 6.18 187 685 A632× W117 825.4 10 U SWCON Soil CN2 SALB 8.5 0.6 81 0.13 DUL SAT Layer cm WR LL 0.330 0.470 1 0–20 0.155 0.345 0.470 0.7 0.159 20–40 0.349 0.476 0.1 40–60 0.159 3 . 3 . Validation Validation was based on all the data set except IRR96, which was used for calibration. We focused on two kinds of variables given as output by the CERES-Maize model; a final simulations of biomass, grain yield, yield components, maximum LAI and seasonal evapotranspiration at maturity, and b daily simulations of LAI, above-ground biomass, grain yield, evapotranspiration and plant available soil water content during the crop cycle. Table 3 shows the comparison between observa- tions and simulations. In the case of treatment IRR in 1997, the simulated values and observations generally matched well and differences were less than 10, with the exception of simulated kernel weight, which was higher. As regards the stress treatments STR1 and STR2 in both the years, simulated and observed maturity dates coincide. On the contrary, a delay of 1 – 2 days STR1 or 5 – 6 days STR2 was observed between the recorded and simulated an- thesis dates. The final simulations were also satisfactory for actual cumulative evapotranspiration, whereas grain yield, above-ground biomass and maximum LAI were generally underestimated. The differ- ences between observations and simulations were more severe for the STR2 treatment than for STR1. 3 . 3 . 1 . Daily simulations Fig. 8 shows the time trend of available soil water content. In the case of the IRR treatment in the 1997 season, the overall simulation is satisfactory, but during the early stages of crop growth, the model overestimated the changes in soil water content. Regarding the stress treatments, the model overestimated the observations in two cases. Fig. 8. Time variation of available soil water mm stored into soil profiles. B . Ben Nouna et al . Europ . J . Agronomy 13 2000 309 – 322 317 Table 3 Comparison between predicted and observed data for each treatment and percent difference D Units STR2 D IRR D STR1 Variable Predicted Observed Predicted Observed Predicted Observed 1996 214 − 0.94 212 218 − 2.7 d.o.y a Anthesis date 212 270 – 270 270 Maturity date – d.o.y 270 6836 − 7.27 4441 5030 − 11.71 6339 Grain yield kg ha − 1 0.2276 g 0.2100 + 8.38 0.2235 0.2000 + 11.75 Kernel weight 3255 − 14.44 1987 2515 − 20.66 Grains m 2 – 2785 651 − 14.44 397.45 503 − 20.99 556.93 – Grains per ear 2.6 – 3.52 − 26.14 1.67 2.81 − 40.57 Max LAI 16 134 Biomass − 22.96 kg ha − 1 9266 13 099 − 29.26 12 429 425.86 − 2.55 359 372.06 − 3.51 415 mm Seasonal ET 1997 – 213 214 − 0.47 213 217 − 1.8 213 Anthesis date 213 d.o.y a 262 – 262 Maturity date 262 d.o.y – 262 262 – 262 7212 − 15.42 3975 5180 − 23.26 6100 kg ha − 1 Grain yield − 7.26 8676 8046 0.1938 g 0.2100 − 7.71 0.1561 0.2000 − 21.95 0.2059 0.2400 − 14.20 Kernel weight 3147 – 3434 − 8.35 2547 2590 − 1.66 3908 3615 + 8.11 Grains m 2 686 − 8.26 509.33 518 − 1.67 629.34 723 + 8.09 Grains ear – 781.5 2.95 – 3.47 − 14.98 1.61 2.99 − 46.15 4.14 4.1 + 0.97 Max LAI Biomass 15 665 kg ha − 1 − 14.74 8717 11 591 − 24.80 17 638 18 674 − 5.55 13 355 442.07 313 338.9 − 7.64 440 498.17 Seasonal ET mm 498 a Day of the year. Fig. 9. Time variation of evapotranspiration mm per day estimated by TDR measurements and simulated by CERES- Maize. Fig. 11 shows the change in LAI through time. For the IRR treatment in 1997, a slight overesti- mation early in the season and an underestima- tion from flowering to maturity was observed, so the model seems to overestimate leaf senescence. These results apply also to the treatments STR1 and STR2. The underestimation from flowering to physiological maturity appears to increase with the intensity of water stress. The model underestimates above-ground dry biomass for the treatment Fig. 12 before flower- ing, but later on the matching with observations becomes better. For STR1 and STR2 in both the seasons, the simulated values of biomass are sys- Fig. 10. Cumulative evapotranspiration SET in mm per season estimated by TDR measurements fine lines and simulated by CERES model bold lines. “ Early in the crop cycle when the crop is vegeta- tive, for both STRI and STR2 treatments dur- ing the season 1997. “ When the total soil water reserve became less than 110 mm for the STR2 treatment. The time trend of actual evapotranspiration is reported in Fig. 9. An overall agreement between estimation based on daily TDR measurements and simulation can be observed; the greatest dif- ferences occur mostly after each irrigation because the model underestimates daily ET. Nevertheless if we consider the accumulated evapotranspiration SET during the whole cycle Fig. 10, the differ- ences between the estimated and simulated values are less than 8 in the five studied treatments. Fig. 11. Time variation of LAI simulated by the model con- tinuous lines. Dots are the measured data and the vertical bars represent their S.D. different from 0. Also dry biomass was systemat- ically underestimated and ET overestimated, as indicated by the slope significantly greater and smaller than 1, respectively. For the LAI, grain yield, available water, the correspondence be- tween observation and simulation was 1:1, once the bias was taken into account, as indicated by a slope not significantly different from 1. Lack of fit Table 5 was significant for all the studied variables and for both the years P B 0.05. This implies that the model could be im- proved, mainly for dry above ground biomass and LAI, under soil water shortage.

4. Conclusions