CERES-Rice model: Calibration, evaluation and application for solar radiation stress assessment on rice production

CERES–Rice model (DSSAT v. 4.0) was calibrated and evaluated for cultivar IR 36 at Cuttack, Orissa using experimental data of wet seasons (June-December) 2001 and 2002. The model accurately predicted phenological events i.e. flowering and maturity date. The simulated grain yield at different N levels was in close agreement with experimental grain yield. Application of the model for solar radiation stress assessment due to Atmospheric Brown Clouds on the same site during dry season (January-May) on historical weather data (1983-2002) revealed a reduction in rice grain yield by 4% with reduction of incident solar radiation by 30% under non-fertilized condition. Compared to non-fertilized condition, grain yield reduction was higher up to 12% with similar solar radiation stress under high rates of N application (120 kg N ha -1 ). The reduction in grain yield is associated with lower grain formation.

A recent international study under the Indian Ocean Experiment (INDOEX) has revealed that a brown haze, a pollutant as a result of biomass burning and industrial emission pervaded particularly during December to April (dry season) over south Asian region and the tropical Indian Ocean, Arabian Sea and Bay of Bengal. This brown haze is named as Atmospheric Brown Cloud. The most direct effect of Atmospheric Brown Cloud is a significant reduction in the solar radiation reaching the surface, which results in reduction of agricultural productivity, reduction in the precipitation efficiency by inhibiting the formation of larger raindrop size particles, and adverse health effect (UNEP, 2002). The reduction in photo synthetically active solar radiation is a major concern to Asia, the largest agricultural continent with 60 -90% of the world's agricultural population (Fu et. al., 1998) et al. (1995) in order to reduce error in sampling. After collection, the plant samples were cleaned to remove sur face contamination, then separated into stems (leaf sheath + stem), leaves and panicles. The samples were oven dried to stop enzymatic reactions and to remove moisture. The dry weight of the samples was recorded till constancy. The total dry matter production at all stages of crop growth was determined. The data recorded from this experiment was used for calibration of genotype coefficients.
The average rainfall of the region is 1421 mm with standard deviation of 262 mm (Directorate of Agriculture and Food Production, 2003). About 81% of the total rainfall is received during monsoon months (June to September). During the wet season 2001, the rice crop was grown without any water stress. The weather for both the years are shown in Fig. 1. Rainfall during the year 2001 was 2250 mm and during 2002, it was only 950 mm. The soil was sandy clay loam CERES-RICE MODEL EVALUATION comparisons that involve predictions of the time of occurrence of discrete events during the life cycle (e.g., flowering date, maturity date), or of yield or biomass at maturity. Regression analysis (Huda, 1988) was applied to a set of simulated and observed data for evaluation purpose. With such an approach, good predictability is indicated if the regression slope is near 1.0, if the intercept is close to 0, and if the R 2 value is high. The experimental data pertaining to phenological events, biomass and grain yield of the year 2001 at 40 kg N level and of the year 2002 at both N levels (40 N and 80 N) were used for model evaluation.
The model was used to simulate the effect of varying incident solar radiation levels on crop behavior of cultivar IR 36 at N application levels 0, 40, 80 and 120 kg N ha -1 during dry season (January-May) at Cuttack, India. Rice yield was estimated with corresponding variation in temperature. The maximum/minimum temperature and solar radiation are normally distributed and the correlation of each variable may be descr ibed by a first-order linear autoregressive model (Richardson, 1982). The relation between incident solar radiation and temperature was obtained by plotting curves (Fig. 2) for mean weather data of past 20 years (1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002). The relation of incident solar radiation was stronger with maximum temperature than with minimum temperature. With reduction of incident radiation by 10 %, not much variation in maximum temperature was noted. Whereas increased reduction of incident radiation by 20% and 30% decreased the maximum temperature considerably. Hence the impact of reduction in incident solar radiation was used to simulate rice grain yield. The crop was planted on 1 st February 2002 using 25 days old seedling with 20 cm row spacing and planting density was 99 plants m -2 . The crop was irrigated as and when required.

Model calibration and evaluation
The calibrated genotype coefficients for the rice cultivar IR 36 are given in Table  2. The genotype coefficients at different phases indicated that the temperature required for vegetative stage is higher than for the reproductive stage. The model was evaluated with reference to phenological occurrence of anthesis and physiological maturity days after transplantation, biomass and grain yield.
A comparison between observed and simulated phenological events (Table 3) shows that the model predicted both anthesis and maturity days almost accurately. The weather of 2002 is highly contrasted from that of 2001, with respect to quantum of rainfall received. A good match between observed and simulated phenological events in varied weather condition reflects the consistency in model performance. These predictions are important because flowering is the most critical stage of rice crop and stresses for moisture or nutrients at this stage cause massive reduction in grain yield (Saseendran et. al., 1998).    et al., 1990). A m varies between 10-50 kg CO 2 ha -1 leaf h -1 for rice depending on leaf N concentration and photo synthetically active radiation (PAR). For lower leaf N concentration (1.5% N), A m reaches 10 kg CO 2 ha -1 leaf h -1 at a level of lower radiation (PAR, 250 J m -2 s -1 ). However, A m reaches 50 kg CO 2 ha -1 leaf h -1 only at a higher radiation level (PAR, 400 J m -2 s -1 ). Further, intensity of reduction in A m due to limitations in PAR is more evident at high N concentration (Ehleringer, and Pearcy, 1983). The most probable effect of limited PAR will be a substantial reduction in grain yield at high leaf N concentration as compared to lower leaf N concentration. We observed reduction in yield to the extent of 12 % at high N application rates with PAR limitations, which was due to interaction of high leaf N concentration and limited PAR. Panicle differentiation begins a 42-day critical sunlight-requiring period (Ronald and Ralph, 2002) and any variation in radiation level beyond this causes significant reduction in grain yield. The present simulation study also reflects lower grain number under incident solar radiation depletion (Fig. 5) in both fertilized and nonfertilized condition.

CONCLUSION
The CERES-Rice (DSSAT v. 4.0) was able to simulate phenological events and grain yield with high accuracy under varied weather condition. The model provides insights about the response mechanism to different N management and various weather conditions. As the CERES-Rice model is found to be capable of predicting the crop yield fairly well, it can serve as a tool in assessing yield potential of alternate technologies and management practices to avoid the yield loss due to reduced solar radiation incidence. The adaptation strategies for climatic change scenarios can be evaluated through the model.