Wuxiangjing9 (China)

Experimental treatments and design

In experiment I, cultivar Wuxiangjing9 was used, a japonica variety widely grown by farmers in southeast China.  In 2001, the experiment was a random-plot design, with three replicates and plot sizes of 40 m2.  Four N rates (Table 1) were applied as urea, in splits of 55% at transplanting, 5% at tillering, 20% at panicle initiation, and 20% at booting.  We applied 60 kg P and 174 kg K ha–1 at transplanting.  In 2002, the experiment was again a random-plot design, with three replicates and plot size of 30 m2.  Five N rates (Table 1) were applied as urea in splits of 60% at transplanting, 20% at 15 days after the onset of stem elongation, and 20% 25 days later. We applied 60 kg P and 174 kg K ha–1 at transplanting.

In experiment II, Wuxiangjing9 was tested in a random-plot design with three replicates and plot size of 20 m2. In 2001, 100 kg urea-N ha–1 was applied, 40% at transplanting and the remainder in splits at 20-day intervals following transplanting.  All plots received 52 kg P and 100 kg K ha–1 at transplanting.  The experiment was repeated in 2002, with 120 kg urea-N ha–1, 40% at transplanting and the remainder top dressed in four splits at 20-day intervals.

Table 1. Details of N rates and splits for the field experiments at Nanjing, 2001-02.

Cultural practices

For experiment I, in 2001, rice was sown in seedbeds on 11 May and transplanted on 12 June at two seedlings per hill, spaced 13 cm within rows and 25 cm between rows. All treatments were harvested on 21 October. In 2002, rice was sown on 11 May and transplanted on 15 June at the same density and spacing as in 2001. The 0 kg N ha–1 plots were harvested on 5 October, the 75 kg N ha–1 plots on 15 October, and the remaining plots on 21 October.

For experiment II, in 2001, rice was sown on 12 May, transplanted on 13 June, and harvested on 1 October.  In 2002, rice was sown on 11 May, transplanted on 15 June, and harvested on 6 October.

In both experiments, fields were submerged during the entire growing season. The plots were hydrologically separated by plastic film installed at 30 cm below the soil surface to restrict water and N flows between adjacent plots.  All plots were kept free of weeds and received optimal control against pests and diseases.

Plant sampling and data collection

For experiment I, in 2001, dates of emergence, panicle initiation, booting, heading, and maturity were recorded.  Eight hills were harvested from each plot every 7 days before the onset of stem elongation and every 3 days afterward. The roots removed and the samples were partitioned in green leaves, yellow/dead leaves, stems, and panicles (after heading).  The area of the fresh green leaves was measured using a CI-203 Portable Laser Area Meter until 2 weeks before maturity and leaf area index was calculated.  Samples were oven-dried, first for 1 hour at 105oC and subsequently at 80oC for 2 days to determine dry weights of the plant organs. At maturity, two 1-m2 areas in each plot were harvested to determine yield. In 2002, phenological development was recorded at tillering, the onset of stem elongation, panicle initiation, booting, 15 days after full heading, 25 days after full heading, and maturity. At these stages, eight hills were harvested from each plot to determine green leaf area and biomass of plant organs, similar to 2001, but, at 25 days after full heading, only leaf area was measured. After measuring the dry weights, samples were ground and N concentrations determined using the micro-Kjeldahl method, following digestion in  H2SO4–H2O2 solution. As in 2001, two 1-m2 areas in each plot were harvested at maturity to determine yield.

For experiment II, eight hills were harvested from each plot to determine green leaf area, biomass of plant organs, and N concentration, at transplanting, 20 days after transplanting, panicle initiation, and 2 weeks before and after heading.  At maturity, yield was determined from sample areas, similar to experiment I.

Climate data collection

During the experiments, weather data were collected from a weather station 2 km from the experimental site. Historical weather data for the area were obtained from the Meteorological Center of China.

Data analysis

For each measured variable, we calculated mean, standard deviation (SD), and coefficient of variation (CV). All yields were expressed at 14% moisture content. SD and standard error (SE) and CV values of measured crop variables are given in Table 2.

Table 2.  Standard deviation (error) (SD and SE, the same unit as variable) and coefficient of variation (CV, %) for measured crop variables in the field experiments atNanjing, 2001-02.

Simulations and parameterisation

First, ORYZA2000 was parameterized and evaluated using data from field experiments at Nanjing, Jiangsu Province, southeastern China, by following the procedure introduced in Section 2.3 for variety special parameters.  All other common crop parameters were set to the values from ORYZA2000’s standard crop data file for the tropical high-yielding variety IR72 (Bouman et al 2001).

Indigenous soil N supply was first estimated from crop N uptake in the 0 kg N ha–1 treatment divided by crop growth duration, and further fine-tuned by matching simulated and measured values of crop N uptake.

Model evaluation

The performance of ORYZA2000 was evaluated for the calibration data set (experiment I) and for the independent validation data set of experiment II by following the standard procedure described in Section 3.2.

Biomass and leaf area index

A graphical comparison of simulated and measured crop biomass and LAI is given in Figure 1 for the calibration and validation sets. In general, agreements between measured and simulated LAI, dry weight of crop organs, and the whole crop are good. In the calibration set, simulated LAI was slightly overestimated between 80 and 120 days after emergence at 0 and 150 kg N ha–1 (Fig. 1A, and 1B). In the validation set, LAI was overestimated between 80 and 120 days after emergence at 100 kg N ha–1, but overestimated by 1–2 units from 80 days after emergence onward at 120 kg N ha–1 (Fig. 1D, and 1E). Despite these overestimations of LAI, the biomass of crop organs was simulated quite accurately. Goodness-of-fit parameters for the dynamic crop variables are given in Table 3. Student’s test indicates that all simulated values were similar to measured values with 95% confidence in both data sets. Slopes () of the biomass variables are close to 1 and the intercept () values are low, indicating a close fit between simulated and measured data. RMSE values are smaller than the SD of measurements, whereas RMSEn values are 2–3 times higher than CV values. Simulation results are less accurate for LAI, with RMSEn values 3 times higher than the CV of the measured values.

Crop N uptake and leaf N concentration

A graphical comparison between simulated and measured crop N uptake and leaf N concentration is given in Figure 2. The error in measured values of crop N variables is larger than that in the crop biomass variables and LAI (Table 2). Similarly, the discrepancy between simulated and measured crop N variables is larger than for biomass and LAI, as reflected in higher values for RMSEn and in values for and that deviate more from their ideal values (Table 4).  Nevertheless, for total crop N uptake and panicle N uptake, RMSEn values are only 1.5–2.5 times higher than the CV of the measured values in both the calibration and validation data sets. However, for leaf N concentration, RMSEn values are about 4 times higher than the measured CV values. In Figure 2, especially, the overestimation of leaf N concentration in the first 70 days after emergence is striking.  In 2001 of the validation data set, total crop N uptake was overestimated during grain filling, but N uptake in leaves and panicles was simulated accurately.

Another way to compare simulated and measured crop N variables is the so-called “three-quadrant diagram” (van Keulen 1982, Fig. 3). Three-quadrant diagrams show the relationships between N application rate and yield (quadrant I), total N uptake and yield (quadrant II), and N application rate and total N uptake (quadrant III).  In all quadrants, simulated and measured values of yield and crop N uptake fall on the same curve.  In quadrant II, crop N uptake is underestimated in the low-uptake range in the 0-N plots, whereas it is overestimated in the high-uptake range. Seasonal average fertilizer-N recovery, calculated from the slope of the curve in quadrant III, is 40%.

Final biomass and yield

Simulated and measured values of total crop biomass and grain yield at harvest are illustrated in Figure 4 for both the calibration and validation data sets. Because of the low number of data, the goodness-of-fit parameters were calculated for both data sets combined (Table 5). In general, simulated and measured values match quite well. In Figure 4, simulated and measured values are near a 1:1 relationship and within the ±SD ranges of measurements. For yield, RMSE and RMSEn are of the same order of magnitude as the SD and CV values, respectively.

References

Bouman BAM, Kropff MJ, Tuong TP, Woperies MCS, ten Berge HFM, van Laar HH. 2001. ORYZA2000: modeling lowland rice. Los Baños (Philippines): International Rice Research Institute, and Wageningen (Netherlands): Wageningen University and Research Centre. 235 p.

Van Keulen H. 1982. Graphical analysis of annual crop response to fertilizer application. Agric. Syst. 9:113-126.