Published in Agron J 100:87-97 (2008)
DOI: 10.2134/agrojnl2006.0241
© 2008 American Society of Agronomy
677 S. Segoe Rd., Madison, WI 53711 USA
MODELING
Sequence Analysis of DSSAT to Select Optimum Strategy of Crop Residue and Nitrogen for Sustainable Rice-Wheat Rotation
Reshmi Sarkara,* and
Sandipta Karb
a Dep. of AAHRM, Tarleton State Univ., Stephenville, TX 76402
b Dep. of Agricultural and Food Engineering, Indian Inst. of Technology, Kharagpur, India
* Corresponding author (reshmisa1{at}yahoo.co.in).
 |
ABSTRACT
|
|---|
Weather variability affects the production of most cropping systems, and rice (Oryza sativa L.)-wheat (Triticum aestivum L.) rotation is not an exception. Integrating weather forecasts with soil fertility management options is one way to combat the production decrease by anticipating weather variability along with sustaining the soil environment. Sequence analysis of DSSAT3.5 was used to select the best combination of crop residue and N application rate for sustainable production of rice-wheat rotation under generated weather. CERES-Rice and CERES-Wheat of DSSAT3.5 were calibrated and validated for rice and wheat crops. Weather generator SIMMETEO was used to generate the weather variables of 30 future years. The variables generated by SIMMETEO, which closely matched with actual weather variables, were used for yield prediction by the sequence analysis program driver. The regression analysis showed a strong relationship between generated rainfall values and predicted yield. The different crop residue levels and N rates were compared for transplanted rice-wheat (T) and direct seeded rice-wheat (D) rotation under 10 yr of generated weather scenario. The sequence analysis of DSSAT predicted the combination of wheat crop residue with 100 kg N ha–1 for rice and rice residue with 80 kg N ha–1 for wheat provided stable yield for both T and D rotation. These combinations of crop residues and N rates were also predicted best for stable rotations under 30 yr of generated weather.
Abbreviations: DR, direct seeded rice • DSSAT, Decision Support System for Agrotechnology Transfer • F, flowering • IIT, Indian Institute Technology • M, physiological maturity • PI, panicle initiation • SRAD, solar radiation • TR, transplanted rice
Sequence Analysis of DSSAT to Select Optimum Strategy of Crop Residue and Nitrogen for Sustainable Rice-Wheat Rotation
Reshmi Sarkara,* and
Sandipta Karb
a Dep. of AAHRM, Tarleton State Univ., Stephenville, TX 76402
b Dep. of Agricultural and Food Engineering, Indian Inst. of Technology, Kharagpur, India
* Corresponding author (reshmisa1{at}yahoo.co.in).
Received for publication August 23, 2006.
Weather variability affects the production of most cropping systems, and rice (Oryza sativa L.)-wheat (Triticum aestivum L.) rotation is not an exception. Integrating weather forecasts with soil fertility management options is one way to combat the production decrease by anticipating weather variability along with sustaining the soil environment. Sequence analysis of DSSAT3.5 was used to select the best combination of crop residue and N application rate for sustainable production of rice-wheat rotation under generated weather. CERES-Rice and CERES-Wheat of DSSAT3.5 were calibrated and validated for rice and wheat crops. Weather generator SIMMETEO was used to generate the weather variables of 30 future years. The variables generated by SIMMETEO, which closely matched with actual weather variables, were used for yield prediction by the sequence analysis program driver. The regression analysis showed a strong relationship between generated rainfall values and predicted yield. The different crop residue levels and N rates were compared for transplanted rice-wheat (T) and direct seeded rice-wheat (D) rotation under 10 yr of generated weather scenario. The sequence analysis of DSSAT predicted the combination of wheat crop residue with 100 kg N ha–1 for rice and rice residue with 80 kg N ha–1 for wheat provided stable yield for both T and D rotation. These combinations of crop residues and N rates were also predicted best for stable rotations under 30 yr of generated weather.
Abbreviations: DR, direct seeded rice • DSSAT, Decision Support System for Agrotechnology Transfer • F, flowering • IIT, Indian Institute Technology • M, physiological maturity • PI, panicle initiation • SRAD, solar radiation • TR, transplanted rice
|
INTRODUCTION
|
|---|
THE RICE-WHEAT CROPPING SYSTEM, which covers an area of 10 million ha, is the major cropping sequence in most parts of India and contributes 85% of total cereal production in India. Both transplanted and direct seeded rice are cultivated under rainfed conditions during the monsoon season (June–October). Wheat is generally cultivated during the dry winter season (November–April) and irrigated with groundwater. Growth and development of both rice and wheat crops depend on rainfall and weather variability plays a significant role. The variability in rainfall resulting in either flood or drought acts as the main factor of crop production. The sporadic, spatial, and temporal distribution of precipitation, which rarely coincides with the demand of rice and wheat crops, creates scarcity of water in many regions of India as well as in other countries and ultimately affects the agricultural production, domestic earnings, and economic balance of a country. Thus, it has been a problem for both policymakers and farmers in decision making related to soil and crop management. Integration of crop modeling with weather forecasts can help in these situations by anticipating the severity of weather and helping farmers to choose correct management options in advance.
The sustainability of the rice-wheat system under variable weather conditions can only be maintained through proper soil and crop management. A sustainable farming system not only results in increased production, it also restores soil fertility along with increasing the soil organic carbon content (Pathak et al., 2003). The increase in soil organic carbon level needs balanced fertilization and residue retention (Whitbread et al., 2003). Thus, combined application of inorganic N and crop residue can be an effective sustainable practice. However, selection of proper N application rate is also very important (Timsina et al., 2001). To select the right combination of crop residue and N application rate for sustainable agriculture under variable weather conditions, crop growth models can be employed as reliable and useful tools.
The crop growth models are helpful to assess the impact of climate change on the stability of crop production under different management options (Hoogenboom et al., 1995). Crop models can be used for crop forecasting with potential in forecasting production scenarios (Matthews et al., 2002). Crop models can help researchers, policymakers, and farmers make correct decisions on crop management practices, and also for marketing strategies and food security of a country with a deterministic view on the import-export market. Decision Support System for Agrotechnology Transfer is a package of 16 different crop growth models that access soil and weather data files along with management files of specific crops to predict crop growth and yield (Jones et al., 2003). The DSSAT allows users to input, arrange, and store the crop, soil, and weather data and aids farmers in developing long-term crop rotation strategies. The DSSAT generates future weather scenarios by helping the model to make more reliable predictions, anticipating the variability in weather conditions (Jame and Cutforth, 1996). The effects of one crop on soil, water, and nutrient status are carried over to the next crop in the sequence or rotation. These sequences can be efficiently studied by the sequence analysis program of DSSAT (Thornton et al., 1994).
Crop simulation models can often explain much of the interaction between the environment and crops. Long-term predictions of regional soybean yields in different parts of Georgia were acceptable with different management practices and different soil types (Curry et al., 1995). Crop models successfully assessed the effects of changes in weather on soybean yields (Jagtap and Jones, 2002), the effects of drought on regional wheat yields (Chipanshi, 1995), the effects of changes in temperature and solar radiation (SRAD) on maize yields (Muchow et al., 1990), the trends of national wheat yields with changes of weather (Supit, 1997), the preharvest wheat yields in different regions using minimal data sets of weather variables (Nain et al., 2002), and the forecasts of wheat yields for the central Indo-Gangetic Plains in India using historical weather data (Nain et al., 2004). Earlier studies also focused on the use of a weather generator program to understand the effects of weather variability on agricultural prediction (Muchow et al., 1991; Baffaut et al., 1996; Jones and Thornton, 1997; Luo et al., 2003). However, none of the modeling studies have focused on developing a sustainable practice for a rice-wheat system that incorporates future weather variability. In view of this, the sequence analysis program of DSSAT was used in a rice-wheat-fallow sequence to (i) evaluate the performance of DSSAT in predicting the future weather, (ii) select the best suitable combination of crop residue and rate of fertilizer N, (iii) assess the change in productivity of the rice-wheat system with climatic variability, and (iv) determine stable transplanted rice-wheat and direct seeded rice-wheat rotations for the subhumid and subtropical ecological region in India.
|
MATERIALS AND METHODS
|
|---|
Experimental Site
A 3-yr field experiment was conducted on a rice-wheat system at the experimental farm of the Agricultural & Food Engineering Department, Indian Institute Technology (IIT), Kharagpur, India. Kharagpur (22°19' N; 87°19' E; 48 m.a.s.l) is in the Red and Laterite agroclimatic zone of the West Bengal state in India. The study area is on the drainage basin of the river Kansai.
Regional Climatic Scenario of Kharagpur
The climate of Kharagpur is subhumid and subtropical. Rainfall, SRAD, maximum and minimum temperature, maximum and minimum relative humidity, wind speed, and pan evaporation were recorded at the micro-meteorological observatory at IIT. Distribution of SRAD in three experimental years (2001–2003) showed maximum SRAD (17.8 – 27.5 MJ m–2) in summer (April–May), medium (19–24 MJ m–2) in rainy seasons (June–October) and in the spring (February–March), and minimum (8.1–20.6 MJ m–2) in winter (November–January). The yearly maximum (TMAX) and minimum (TMIN) atmospheric temperature at the experimental site generally ranges from 25 to 36°C and 12 to 26°C, respectively. The temperature in the experimental years was in the medium range in spring (TMAX = 30–35°C; TMIN = 17–23°C) and increased in summer months (TMAX = 36–38°C; TMIN = 23–25°C) as usual. With the arrival of the Monsoon (a seasonal wind in South Asia that blows from the southwest direction, bringing rain to India and neighbor countries), temperature decreased in the rainy season (TMAX = 32–35°C; TMIN = 25°C). Winter temperature showed a sharp fall with TMAX of 16 to 30°C and TMIN of 7 to 17°C. The experimental site generally receives an average annual rainfall of 1350 mm, of which 70 to 75% is received in the monsoon months (June–October). In the experimental period of 2001 to 2003, both the amount and distribution of rainfall varied considerably. Annual rainfall in the year 2003 was less than the average rainfall by 30%, whereas annual rainfall in 2001 and 2002 was higher than the average rainfall by 19 to 20%, respectively. Rainfall was more evenly distributed in 2002 than in 2001 or 2003. The amount of rainfall was maximum in the month of July (412 mm), September (432 mm) and October (338 mm) in the years 2001, 2002, and 2003, respectively. The amount of rainfall was less in winter months and was nil in December in all three experimental years. The maximum and minimum relative humidity in the experimental site generally ranges from 80 to 100 and 38 to 93%, respectively.
Soil
The soil of the experimental site was an acid lateritic sandy loam and taxonomically classified as Kharagpur coarse mixed Hyperthermic Haplustalf (Table 1
). The basic infiltration rate of the soil was 0.6 cm h–1. The soil had a low organic matter content and CEC (8.1 and 9.2 cmol. kg –1 in surface and subsurface soil, respectively). The soil was low in KMnO4–extractable N (36 kg N ha–1), available phosphorus (10.5 kg P2O5 ha–1, extracted by Bray and Kurtz's method) and available potassium (100 kg K2O ha–1). The total contents of N, P, and K in soil were 400 kg N ha–1, 450 kg P2O5 ha–1, and 3500 kg K2O ha–1, respectively.
Field Experiments
Rice cultivar IR36 and wheat cultivar ‘Sonalika’ were grown in a rice-wheat-fallow sequence during June 2001 to April 2004. Both transplanted and direct seeded rice were grown as rainfed crops during monsoon season (June–October). Following rice, wheat was grown as an irrigated crop in winter (November–April). The treatments of crop residue application included no residue (r0w0), with rice residue (r1w0), with wheat residue (r0w1), and with both rice and wheat residues (r1w1). Treatments of N application rate included 0 (N0), 40 (N40), 60 (N60), 80 (N80), 100 (N100), and 120 (N120) kg N ha–1 for rice and wheat.
The land for direct seeded rice was tilled by a tractor-drawn cultivator followed by disc harrow and the land for transplanted rice was prepared by puddling with a tractor-drawn rotavator. After applying the tillage treatments, 50 kg each of P2O5 and K2O in the form of single super phosphate and muriate of potash, respectively, were mixed to the soil. Wheat residue at 0 and 4 Mg ha–1 were also applied. The plots were leveled and laid out in a split-split plot design. Rice was direct seeded in rows 20 cm apart in the east-west direction in the second week of June, and on the same day, rice seeds were sown in a nursery plot. Seedlings were grown until transplanting. There were 21 day-old rice seedlings that were transplanted in hills at a plant-to-plant distance of 15 cm and a row-to-row distance of 20 cm. Fertilizer N (urea) was broadcasted at 0, 40, 60, 80, 100, and 120 kg N ha–1 in two splits. N was applied 29 and 75 d after sowing for rice and 14 and 46 d after transplanting for transplanted rice. Rice was harvested in the second week of October.
Wheat was sown in November. The soil was tilled with a cultivator followed by disc harrow. Either no rice residue or 4 Mg ha–1 of rice residue was applied to the appropriate plots. P2O5 and K2O were both applied at 50 kg ha–1. The field was leveled with a wooden plank, and wheat was sown in 20-cm row spacing. One week after emergence, wheat seedlings were thinned to maintain uniform distribution of plants. Wheat was irrigated with 6 cm water timed at IW/CPE (Irrigation Water/Cumulative Pan Evaporation) ratios of 0.6. Fertilizer N (urea) was broadcasted to soil at 0, 40, 60, 80, 100, and 120 kg N ha–1 split between 24 and 76 d after sowing. Wheat was harvested in the first 14 d of April.
Dates of all of the cultural operations and phenological stages were noted. Biomass weights at the time of panicle initiation (PI), flowering (F), and physiological maturity (M), and grain weight was recorded and used in calibration and validation of the model. Commencement of 50% of plant population to a particular growth stage (e.g., PI, F, and M) was considered as the growth stage of the whole plant population (Singh et al., 1999). For biomass sampling, selected plants were cut just above ground level and dried in an oven for 24 h at 70°C. The procedure described in the manual of DSSAT3.5 was followed for sample collection and calculation of yield components.
Model Description
The DSSAT model can arrange and store the input data of crop, soil, and weather and use them for the prediction of a crop. It predicts growth and development of a crop based on genetic characteristics of the crop, temperature, daylength, and requires daily weather data (SRAD, maximum temperature, minimum temperature, and rainfall) as input. Besides genetic characteristics and weather data, the output data from submodels of water and nitrogen help predict the growth and development of a crop at different crop growth stages.
Once calibrated, the model predicts and analyzes effects of different management options for a particular cropping sequence. In the present study, CERES-Rice (Godwin et al., 1990) and CERES-Wheat (Ritchie, 1985) models integrated in DSSAT were calibrated and validated separately, with the data of weather, soil and crop management. The data used for calibration were 1-yr data; for validation of CERES-Rice and CERES-Wheat, the data used were 2-yr and 1-yr data, respectively. The data related to soil characteristics such as soil moisture, pH, bulk density, soil organic carbon, nitrate-nitrogen and ammonium-nitrogen, and cation exchange capacity were measured and stored in the soil input file. The data sets of weather variables including daily SRAD, maximum and minimum temperature and rainfall were loaded into the model. The input data for calibration and validation of the models CERES-Rice and CERES-Wheat were date of sowing, emergence, flowering, and maturity with detailed data of cropping practice and management options; the output data were biomass at different stages of crop growth, yield parameters, harvest index, grain yield, and biomass N.
The Weather Generator Model: SIMMETEO
The minimum data set of the daily weather variables such as SRAD (MJ m–2), rainfall (mm), and maximum, minimum temperature (°C), of nine consecutive years for the experimental station were stored by the WeatherMan program. The data for SRAD used in WeatherMan were calculated using the daily data of sunshine hours from the Angstrom Formula (Allen et al., 1998).
The weather utility program WeatherMan (Pickering et al., 1994), which is mainly based on PASCAL (a mathematical program language), was used extensively to store, arrange, and produce monthly means of the local weather variables archived for use in the model to predict yields for calibration and validation purpose. However, SIMMETEO (Geng et al., 1986), an integral part of DSSAT3.5, was used for generating the data of future 30 yr from monthly summaries of the weather variables (Soltani and Hoogenboom, 2003). The precipitation data were generated independently by SIMMETEO; whereas other variables such as SRAD or temperature were generated depending on the occurrence of precipitation (Richardson, 1981).
Calibration of CERES-Rice and CERES-Wheat in DSSAT3.5
The model was calibrated using data for agronomic management, soil and crop performance. The data related to agronomic management were crop residue incorporation, rate of fertilizer N application, and irrigation schedule. The data related to crop performance were dates of PI, F, M, panicle number, grain number, biomass yield, and grain yield. The calibration of CERES-Rice and CERES-Wheat involved comparison of simulated and measured crop phenological events (dates of PI, F, and M), biomass and grain N, and grain yields. Calibrations were performed separately for transplanted and direct seeded rice (2002) and wheat (2002–2003). The sensitivity of the models to rate and percentage of incorporation of crop residue and rate of fertilizer N application was also tested before performing the validations. Coefficient of variation (CV) was calculated for all these factors to determine the sensitivity of the model with higher CV values, indicating greater sensitivity (Sarkar, 2005).
Validation of CERES-Rice and CERES-Wheat
To test the accuracy of the model with the cultivars used, the model was run with observed weather data and calibrated cultivar coefficients, and the predicted phenological events, grain N, biomass N, and yield of transplanted and direct seeded rice and wheat were compared with the experimental results. The data of the other years, not used for calibration, were used for validation. The actual and simulated data were drawn in a 1:1 plot to check the validity of the model. Good matching of the field and predicted data would indicate that the model is capable of realistic predictions (Sarkar and Kar, 2006).
Sequence Analysis
A sequence analysis program was run to simulate the combined situation of the experiment involving different crop rotations under different combinations of management options of rice-wheat rotation followed over a period of 10 yr. The effects of one crop on soil, water, and nutrient status were carried over to the next crop in the sequence or rotation (also referred as the germ). This carryover effect was studied by sequence analysis. The results were compared from the different runs to determine the most suitable rotation with the best combination of management options (crop residue and N rate) for the prevailing environment of Kharagpur. To verify the stability of the rice-wheat rotation over a long period, the rotation with the best combination of management options obtained by the previous experiment with 10 yr of predicted weather data, was run further with 30 yr of predicted weather data. The simulation results were analyzed and compared through graphical presentation and the regression menu in the Analyze section of the sequence analysis driver. Five types of different plots used in the study were: (i) box plot, which displayed the trend of average values, standard deviation, maxima and minima against time; (ii) cumulative probability function plot in which case the output variables for each season were ordered from smallest to largest; here the graphs were plotted against equal increments of cumulative probability and the efficiency of a treatment increased if the point of a certain treatment was placed in the northwestern direction (in the upper left corner of the graph); (iii) mean-variance plot, mean of interested treatments was plotted against the variance; this plot gave a better visualization effect and was proved to be useful for the comparison of the treatments; (iv) variance plot, where the variance of sequence was plotted against time; this plot was also very useful to determine the increasing or decreasing pattern of variability of a rotation; and (v) coefficient of variation plot, where CV of a variable (specifically average yield of crops) was plotted against time, indicating the relative change in variability of sequence with time. The results of the present study are illustrated only with the help of covariance plot, mean-standard deviation analysis, and linear regression analysis.
In the present study, transplanted rice-wheat-fallow (TR-Wh-F) and direct seeded rice-wheat-fallow (DR-Wh-F) rotations were simulated separately with all the possible combinations of management options related to six application rates of N and crop residues. Five better combinations of each TR-Wh-F and DR-Wh-F, which showed relatively higher production, were selected.
The treatment combinations for which the DSSAT predicted better productivity of the TR-Wh-F rotation were: (i) transplanted rice (fertilizer N applied at 120 kg ha–1 with wheat crop residue, N120r0w1); wheat (under dry irrigation water regime, fertilizer N applied at 80 kg ha–1 without crop residue; N80r0w0) [T1], (ii) transplanted rice (fertilizer N applied at 120 kg ha–1 with wheat crop residue, N120r0w1); wheat (under dry irrigation water regime, fertilizer N applied at 80 kg ha–1 with rice crop residue, N80r1w0) [T2], (iii) transplanted rice (fertilizer N applied at 120 kg ha–1 without crop residue, N120r0w0); wheat (under dry irrigation water regime, fertilizer N applied at 80 kg ha–1 without crop residue, N80r0w0) [T3], (iv) transplanted rice (fertilizer N applied at 120kg ha–1 without crop residue, N120r0w0); wheat (under dry irrigation water regime, fertilizer N applied at 80 kg ha–1 with rice crop residue, N80r1w0) [T4], and (v) transplanted rice (fertilizer N applied at 100 kg ha–1 with wheat crop residue, N100r0w1); wheat (under dry irrigation water regime, fertilizer N applied at 80 kg ha–1 with rice crop residue, N80r1w0) [T5].
Similarly, the treatment combinations for which the DSSAT predicted better productivity of the DR-Wh-F rotation were: (i) direct seeded rice (fertilizer N applied at 120 kg ha–1 without crop residue, N120r0w0); wheat (under dry irrigation water regime, fertilizer N applied at 80 kg ha–1 without crop residue, N80r0w0) [D1], (ii) direct seeded rice (fertilizer N applied at 120 kg ha–1 with wheat crop residue, N120r0w1); wheat (under dry irrigation water regime, fertilizer N applied at 80 kg ha–1 without crop residue, N80r0w0) [D2], (iii) direct seeded rice (fertilizer N applied at 100 kg ha–1 with wheat crop residue, N100r0w1); wheat (under dry irrigation water regime, fertilizer N applied at 80 kg ha–1 with rice crop residue, N80r1w0) [D3], (iv) direct seeded rice (fertilizer N applied at 120 kg ha–1 with wheat crop residue, N120r0w1); wheat (under dry irrigation water regime, fertilizer N applied at 80 kg ha–1 with rice crop residue, N80r1w0) [D4], and (v) direct seeded rice (fertilizer N applied at 120 kg ha–1 without crop residue, N120r0w0); wheat (under dry irrigation water regime, fertilizer N applied at 80 kg ha–1 with rice crop residue, N80r1w0) [D5].
The results of sequence analysis of these 10 rotations are presented in the results section, and the rotations with the highest stability of production in future years were selected with the help of the Analysis part of the sequence analysis program.
|
RESULTS AND DISCUSSION
|
|---|
Performance Evaluation of Ceres-Rice and Ceres-Wheat
Performance evaluation of the model was performed using root mean square difference (RMSD) and mean bias error (MBE), Spearman correlation, and linear regression analysis. The statistics of comparison of observed and measured phenological events for transplanted rice, direct seeded rice, and wheat clarified that the predictions of CERES-Rice and CERES-Wheat were quite acceptable (Table 2
), except for the somewhat higher values of RMSD for direct seeded rice, where the dates were overpredicted. This variation in predicted phonological events in the case of direct seeded rice was probably due to the thermal variation (Yun, 2002). The higher values of coefficients of the Spearman correlation and linear regression in all the cases revealed that the prediction of phenological events for both transplanted and direct seeded rice by CERES-Rice and wheat by CERES-Wheat were accurate. The average values of predicted and observed N uptake matched well except for slight underprediction in the case of biomass N of rice (Table 3
). The values of predicted N uptake of wheat showed a good match with its counterpart. Savin et al. (1995) also reported good prediction of CERES-Wheat for N uptake by wheat under different N levels and irrigation water regimes. High R2 values and Spearman correlation coefficients clarified the good simulation of N uptake of rice and wheat crops by CERES-Rice and CERES-Wheat of DSSAT.
View this table:
[in this window]
[in a new window]
|
Table 2. Coefficient of linear regression, root mean square difference (RMSD), and Spearman correlation coefficient values of observed and predicted phenological events of transplanted rice (days after transplanting, DAT), direct seeded rice (days after seeding, DAS), and wheat (DAS) crops.
|
|
View this table:
[in this window]
[in a new window]
|
Table 3. Average values, coefficient of linear regression, root mean square difference (RMSD), and Spearman correlation coefficient values of observed and predicted N (grain and biomass) uptake (kg ha–1) of transplanted rice, direct seeded rice, and wheat crops.
|
|
Low RMSD and MBE values and high coefficient values of Spearman correlation and linear regression analysis of rice and wheat yields (Table 4
) proved the high level of accuracy of the models in predicting the transplanted rice, direct seeded rice, and wheat crops.
View this table:
[in this window]
[in a new window]
|
Table 4. Coefficient of linear regression, mean bias error (MBE), root mean square difference (RMSD), and Spearman correlation coefficient values of observed and predicted yields (kg ha–1) of transplanted rice, direct seeded rice, and wheat crops.
|
|
Climate Forecast
Before using the weather data generated by the stochastic weather generator program SIMMETEO of DSSAT for the sequence analysis, the monthly summaries of actual weather data were compared with generated weather data. There was not much variations between generated and actual SRAD data except in the months of July (+0.4 MJ m–2), November (–0.3 MJ m–2), and December (–0.2 MJ m–2). The distribution of generated data showed an increase in the TMAX in future years except in the months of February (–0.2°C), November (0°C), and December (–0.1°C), indicating a realistic approach in the predictability of SIMMETEO. The difference between predicted and actual TMAX varied from 0.0 to 0.9°C, and the largest variations were noted in the month of June (+0.9°C). The generated TMIN did not vary much from the actual data. The generated rainfall in future years showed underprediction by the model in the months of March (–2.8 mm), April (–2.9 mm), July (–8.0 mm), and October (–3.5 mm) and overprediction in the months of August (+12.3 mm) and September (+8.1 mm). The linear regression analysis was performed to test the variability of the monthly summaries of the actual and predicted climatic scenarios. The gradient values near 1 (0.97, 0.94, 1.00, and 0.99, respectively, for SRAD, TMAX, TMIN, and RAIN), low constants (0.67, 1.52, 0.01, and 0.25, respectively, for SRAD, TMAX, TMIN, and RAIN), and high linear regression coefficients of 0.9 (0.995, 0.995, 0.999, and 0.998, respectively, for SRAD, TMAX, TMIN, and RAIN) proved the reliability of the model's prediction and its usability for the simulation of rice and wheat yields for the future years. Hartkamp et al. (2003) also reported SIMMETEO as a robust weather generator and recommended this weather generator program for crop modeling purposes.
Prediction of Yield of Rice-Wheat Crop Rotation in Future 10 Years
Sequence analysis was performed to examine the five better treatment combinations for wheat following transplanted rice and direct seeded rice. The most stable crop rotation or germ was selected with the help of various statistical analyses through the sequence analysis program. Each germ (rice-wheat rotation) was run with the data of 10 predicted weather years (2004–2014), and the sequence of each year rotation was run with 20 replications because 20 is the maximum number of replications that could be used by the model to have the minimum error in the prediction. The predicted average yield trend was interpreted in terms of statistical assessment and the best treatment with stable production under a particular germ was selected, with the help of both model predictions and user's choice.
The predicted average production of the transplanted rice-wheat rotation (Table 5
) under different treatment combinations ranged from 4,229 to 10,599 kg ha–1 in different years with standard deviations ranging from 311 to 630 kg ha–1. The minimum yield was predicted in the year 2004, while the maximum yield in the year 2014. The maximum precipitation was also predicted in the year 2014 (Table 6
). The variation in predicted precipitation was well reflected on the predicted yields of different years. The distribution of precipitation among the years was found to influence the crop yield because the simulation process of DSSAT was mainly based on weather variables, particularly precipitation, which was generated independently (Richardson, 1985). Predicted yield was significantly related with the predicted precipitation under each treatment combination with the Spearman correlation coefficient in the range of 0.986 to 0.993 (Table 7
).
View this table:
[in this window]
[in a new window]
|
Table 7. Spearman correlation coefficient (rs) for the relationships between predicted precipitation and predicted yield under different crop rotations.
|
|
The average yield values of the direct seeded rice-wheat rotation ranged from 2715 to 9194 kg ha–1 under different treatment combinations (Table 8
), and they were closely related with the predicted precipitation during 2004 to 2014 (Table 6). As observed with the transplanted rice-wheat rotation, the strong influence of predicted precipitation on the predicted yields was also found in the case of the direct seeded wheat rotation and this trend was also reflected by the highly significant Spearman correlation coefficient values (Table 7) for the relationships between predicted crop yields and precipitation.
Stability of the Rice-Wheat Crop Rotation under Different Management Options
The trend of predicted average yield of different rotations was interpreted in terms of statistical assessment. Subsequently, the best treatment combination was selected in which the production remained most stable. The best treatment was selected based on both model prediction and user's decision. The interaction of nitrogen and crop residue (N x R) was also tested while studying the effect of climate variability on the rice-wheat rotations. The relationship between production of a rotation and year of production was analyzed to understand the interaction effects between treatments of management combinations and environment. The R2 values of linear regression relationships between yield (Y) and year of production (X) were in close range, but the maximum R2 value was estimated under T5 (100 kg N ha–1 to rice and 80 kg N ha–1 to wheat along with application of wheat and rice crop residues) (Table 9
). Hence, T5 was the most efficient among the five different treatment combinations of N and R.
View this table:
[in this window]
[in a new window]
|
Table 9. Gradient, constant, and R2 values of linear regression relationships between predicted yield of different crop rotation and year of prediction.
|
|
To select the most efficient interaction of N and R amongst the five selected treatments of management combinations (T1, T2, T3, T4, and T5), the values of the CV for different treatments were plotted against the predicted years and compared with each other for the stability range (Fig. 1
). The CV values of each plot (as presented through the output results of DSSAT) were found to be lower than 20%, which indicated better stability of the treatments. Amongst the five different treatments, the CV curve of T5 showed the minimum variability where CV values were in the range of 12.0 to 5.2%. Hence, T5 was selected as the best management combination under which transplanted rice-wheat productivity was more stable. The prediction was reasonable, as N rate was not too high in T5 (in comparison to the recommended application rate of 90 kg N ha–1 to both rice and wheat crops at the experimental site). Application of both rice and wheat crop residues that would definitely enrich and sustain the soil organic matter pool would also help enhance environmental stability for the sustainable production of transplanted rice-wheat. Hartkamp et al. (2004) also reported that DSSAT could simulate the interaction of prevailing soil conditions and crop management, and DSSAT predicted the long-term yield in interaction of crop residue and inorganic N application perfectly.
View larger version (20K):
[in this window]
[in a new window]
|
Fig. 1. Coefficient of variation plots of the transplanted rice-wheat rotation (T1, T2, T3, T4, and T5) and the direct seeded rice-wheat crop rotation (D1, D2, D3, D4, and D5) in 10 yr.
|
|
In the same way, the stability of the predicted average yields of the direct seeded rice-wheat rotation over the future years under the selected treatments of management combinations was also tested by the linear regression analysis. The R2 values (Table 9) for these relationships between yields (Y) and year of production (X) under different treatment combinations indicated that R2 values were distributed in a close range and the predicted yields under different treatments were significantly correlated with the year of prediction. Hence, it was very difficult to select the best combination with the help of only linear regression analysis. Therefore, the CV analysis was adopted to find out the most stable treatment combination for the direct seeded rice-wheat rotation. The plots of CV show that CV was less for the starting year under all treatment combinations (Fig. 1). The figure also shows that the CV decreased gradually after 2008 to the minimum values in 2014 under D1, D3 and D4, but the trend remained different under D2 and D5 where it increased after the year 2013. The smooth decrease of CV in later years under D1, D3, and D4 indicated greater stability under these treatments. However, the comparison of R2 values of linear regression (Table 9) showed that D3 was a better combination since the R2 value for D3 was higher than D1 and D4. So, D3 was selected as the most efficient treatment combination that showed stable productivity of the direct seeded rice-wheat rotation. Moreover, the rate of fertilizer N application under D3 was also lower than other combinations, which would lead to higher N use efficiency. The use of both wheat and rice crop residues was also an added benefit of this treatment combination. In a similar study, Jintrawet (1995) also found that the CERES-Rice model predicted better yield of both transplanted and direct seeded rice when 100 kg of N and organic residue was applied. The CERES-Wheat was also reported to be a well-equipped crop simulation model that was capable of predicting wheat yield reasonably under different N application rates (Ghaffari et al., 2001; Bannayan et al., 2003).
Forecast of Long-Term Stability of the Rice-Wheat Crop Rotation
The simulation study was extended further to examine the long-term stability of the rice-wheat crop rotation as influenced by the interaction between environment and nutrient status (E x Nut) in subtropical and subhumid climates. The simulation was performed for 30 future years with 20 replications with the selected treatment or management combination (T5 and D3). The mean and standard deviation values of the predicted yields and precipitation as presented in Table 10
revealed that DSSAT predicted stable production and precipitation in most years except for a sudden rise in 2034. The CV values under treatment combination T5 ranged between 5 and 9 in future years, with minimum variability (Fig. 2
). The CV values of treatment combination D3 (Fig. 3
) also showed minimum variation. However, an abrupt rise in year 2034 was observed due to abnormally high predicted precipitation (Table 10). Holden et al. (2003) also reported a close link between predicted precipitation and variability in predicted yields. On the whole, results obtained by sequence analysis showed that variability of yield under treatments T5 and D3 was minimum, and stability of yield was maximum.
View this table:
[in this window]
[in a new window]
|
Table 10. Predicted yield of transplanted rice-wheat and direct seeded rice-wheat rotations and precipitation for the future 30 yr from 2004 to 2034.
|
|
View larger version (9K):
[in this window]
[in a new window]
|
Fig. 2. Variation of CV for the maturity yield of the transplanted rice-wheat crop rotation in future 30 yr of simulated weather data under T5
|
|
View larger version (8K):
[in this window]
[in a new window]
|
Fig. 3. Variations of CV for maturity yield of the direct seeded rice-wheat crop rotation in future 30 yr of simulated weather data under D3
|
|
|
CONCLUSIONS
|
|---|
The weather generator program, SIMMETEO of DSSAT3.5, is reliable and can successfully generate the weather variables for a long period. The generated weather scenario is also found acceptable after thorough comparison of actual weather variables with generated ones. The study revealed a strong influence of rainfall data on prediction of rice and wheat yields by DSSAT, which is clarified by the significant correlation between the predicted yields and predicted rainfall. The most remarkable result of this modeling study is the decision-making process and forecasts of long-term stable production of rainfed rice followed by irrigated wheat rotation under subhumid and subtropical climate conditions. The sequence analysis of DSSAT predicted that long-term stable productivity in the rice-wheat system can be attained by applying the residues of both rice and wheat crops along with 100 kg of N ha–1 for rice and 80 kg of N ha–1 for wheat. Indeed, the result seems to be practical as present conditions demand higher soil organic matter levels for sustainable production. Thus, integration of the sequence analysis of DSSAT with SIMMETEO is a suitable and reliable alternative for long-term experiments, especially in developing countries.
|
ACKNOWLEDGMENTS
|
|---|
This research work was supported by a grant (NATP) from the Indian Council of Agricultural Research, New Delhi, India. The first author acknowledges the help and technical support from Prof. G. Hoogenboom, Dep. of Biology and Agricultural Engineering, Univ. of Georgia, Griffin, GA.
|
NOTES
|
|---|
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.
|
REFERENCES
|
|---|
- Allen, R. G., L. S. Pereira, D. Raes, and M. Smith. 1998. Crop evapotranspiration: Guidelines for computing crop water requirements. FAO irrigation and drainage paper 56. FAO, Rome
- Baffaut, C., M.A. Nearing, and A.D. Nicks. 1996. Impact of CLIGEN parameters on WEPP-predicted average annual soil loss. Trans. ASAE 39(3):969–972.
- Bannayan, M, N.M.J. Crout, and G. Hoogenboom. 2003. Application of the CERES-wheat model for within-season prediction of winter wheat yield in the United Kingdom. Agron. J. 95:114–125.[Abstract/Free Full Text]
- Chipanshi. A.C. 1995. Monitoring the effects of drought on wheat yields in Saskatchewan, Saskatoon. Ph.D. thesis. Univ. of Saskatchewan, Saskatoon.
- Curry, R.B., J.W. Jones, K.J. Boote, R.M. Heart, L.H. Allen, and N.B. Pickering. 1995. Response of soybean to predicted climate change in the USA. In C. Rosenzweig et al. (ed.) Climate change and agriculture: Analysis of potential international impacts. ASA Spec. Publ. 59. ASA, Madison, WI.
- Geng, S., F.W.T. Penning de Vries, and I. Supit. 1986. A simple method for generating daily rainfall data. Agric. For. Meteorol. 36:363–376.
- Ghaffari, A., H.F. Cook, and H.C. Lee. 2001. Simulating winter wheat yields under temperate conditions: Exploring different management scenarios. Eur. J. Agron. 15:231–240.
- Godwin, D., J.T. Ritchie, U. Singh, and L. Hunt. 1990. A user's guide to CERES Wheat–V2.10. Int. Fertilizer Dev. Cent., Muscle Shoals, AL.
- Hartkamp, A.D., J.W. White, and G. Hoogenboom. 2003. Comparison of three weather generators for crop modeling: A case study for subtropical environments. Agric. Syst. 76:539–560.[Web of Science]
- Hartkamp, A.D., J.W. White, W.A.H. Rossing, M.K. van Ittersum, E.J. Bakker, and R. Rabbinge. 2004. Regional application of a cropping systems simulation model: Crop residue retention in maize production systems of Jalisco, Mexico. Agric. Syst. 82:117–138.[Web of Science]
- Holden, N.M., A.J. Brereton, R. Fealy, and J. Sweeney. 2003. Possible change in Irish climate and its impact on barley and potato yields. Agric. For. Meteorol. 116(3–4):181–196.
- Hoogenboom, G., G.Y. Tsuji, J.W. Jones, U. Singh, D.C. Godwin, N.B. Pickering, and R.B. Curry. 1995. Decision support system to study climate change impacts on crop production. p. 51–75. In C. Rosenzweig et al. (ed.) Climate change and agriculture: Analysis of potential international impacts. ASA Spec. Publ. No. 59. Am. Soc. of Agron, Madison, WI.
- Jagtap, S.S., and J.W. Jones. 2002. Predicting regional soybean yield and production: Adaptation of the CROPGRO-soybean simulation model and scaling up yields. Agric. Ecosyst. Environ. 93(1):73–85.
- Jame, Y.W., and H.W. Cutforth. 1996. Crop growth models for decision support systems. Plant Sci. 76:9–19.
- Jintrawet, A. 1995. A decision support system for rapid assessment of lowland rice-based cropping alternatives in Thailand. Agric. Syst. 47:245–258.[Web of Science]
- Jones, J.W., G. Hoogenboom, C.H. Porter, K.J. Boote, W.D. Bachelor, L.A. Hunt, P.W. Wilkens, U. Singh, A.J. Gijsman, and J.T. Ritchie. 2003. The DSSAT cropping system model. Eur. J. Agron. 18:235–265.[CrossRef]
- Jones, P.G., and P.K. Thornton. 1997. Spatial and temporal variability of rainfall related to a third-order Markov model. Agric. For. Meteorol. 86(1–2):127–138.
- Luo, Q., M.A.J. Williams, W. Bellotti, and B. Bryan. 2003. Quantitative and visual assessments of climatic change impacts on South Australian wheat production. Agric. Syst. 77(3):173–186.
- Matthews, R., W. Stephens, T. Hess, T. Mason, and A.R. Graves. 2002. Application of crop-soil simulation models in tropical agricultural systems. Adv. Agron. 17:31–123.
- Muchow, R.C., G.L. Hammer, and P.S. Carberry. 1991. Optimizing crop and cultivar selection in response to climatic risks. p. 235–262. In R.C. Muchow et al. (ed.) Climatic risk in crop production: Models and management for the semiarid tropics and subtropics. CAB Int., Wallingford, UK.
- Muchow, R.C., T.R. Sinclair, and J.M. Bennett. 1990. Temperature and solar radiation effects on potential maize yield across locations. Agron. J. 82:338–343.[Abstract/Free Full Text]
- Nain, A.S., V.K. Dadhwal, and T.P. Singh. 2002. Real time wheat yield assessment using technology trend and crop simulation model with minimal data set. Curr. Sci. 82:1255–1258.[Web of Science]
- Nain, A.S., V.K. Dadhwal, and T.P. Singh. 2004. Use of CERES-wheat model for wheat yield forecast in central Indo-Gangetic plains of India. J. Agric. Sci. 142:59–70.
- Pathak, H., J.K. Ladha, P.K. Aggarwal, S. Peng, S. Das, Y. Singh, B. Singh, S.K. Kamra, B. Mishra, A.S.R.A.S. Sastri, H.P. Aggarwal, D.K. Das, and R.K. Gupta. 2003. Trends of climatic potential and on-farm yields of rice and wheat in the Indo-Gangetic Plains. Field Crops Res. 80(3):223–234.
- Pickering, N.B., J.W. Hansen, J.W. Jones, C.M. Wells, V.K. Chan, and D.C. Godwin. 1994. Weatherman: A utility for managing and generating daily weather data. Agron. J. 86(2):332–337.
- Richardson, C.W. 1981. Stochastic simulation of daily precipitation, temperature and solar radiation. Water Resour. Res. 17(1):182–190.
- Richardson, C.W. 1985. Weather simulation for crop management models. Trans. ASAE 28(5):1602–1606.
- Ritchie, J.T. 1985. A user-oriented model of the soil water balance in wheat. p. 293–305. In W. Day et al. (ed.) Wheat growth and modeling. Series A: Life Sciences, Vol. 86. Plenum Press, New York.
- Sarkar, R. 2005. Assessment of sustainability of rice-wheat system under different combinations of tillage, crop residue and fertilizer nitrogen applications using crop simulation model. Ph.D. thesis. Indian Institute of Technology, Kharagpur, West Bengal, India.
- Sarkar, R., and S. Kar, 2006. Evaluation of management strategies for sustainable rice-wheat cropping system, using DSSAT seasonal analysis. J. Agric. Sci. 144(5):421–434.
- Savin, R., E.H. Sattore, A.J. Hall, and G. Slafer. 1995. Assessing strategies for wheat cropping in the monsoonal climate of the Pampas using the CERES-Wheat simulation model. Field Crops Res. 42:81–91.
- Singh, P, G. Alagarswamy, G. Hoogenboom, P. Pathak, S.P. Wani, S.M. Virmani. 1999. Soybean-chickpea rotation on vertic inceptisols II. Long-term simulation of water balance and crop yields. Field Crops Res. 63:225–236.
- Soltani, A., and G. Hoogenboom. 2003. Minimum data requirements for parameter estimation of stochastic weather generators. Clim. Res. 25:109–119.
- Supit, I. 1997. Predicting national wheat yields using a crop simulation and trend models. Agric. For. Meteorol. 88:199–214.
- Thornton, P.K., P.W. Wilkens, G. Hoogenboom, and J.W. Jones. 1994. Sequence analysis. In G.Y. Tsuji et al. (ed.) DSSAT v 3. Vol. 3–2. Univ. of Hawaii, Honolulu.
- Timsina, J., U. Singh, M. Badaruddin, C. Meisner, and M.R. Amin. 2001. Cultivar, nitrogen and water effects on productivity, and nitrogen-use efficiency and balance for rice-wheat sequences of Bangladesh. Field Crops Res. 72:143–161.
- Whitbread, A., G. Blair, Y. Konboon, R. Lefroy, and K. Naklang. 2003. Managing crop residues, fertilizers and leaf litters to improve soil C, nutrient balances, and the grain yield of rice and wheat cropping systems in Thailand and Australia. Agric. Ecosyst. Environ. 100:251–263.
- Yun, J.I. 2002. Predicting regional rice production in South Korea using spatial data and crop-growth modeling. Agric. Syst. 77:23–38.[Web of Science]
TOP
ABSTRACT
INTRODUCTION
