Rice is the second major food crop in central Asia. Climate change may greatly affect the rice production in the region. This study quantifies the effects of projected increases in temperature and atmospheric CO2 concentration on the phenological development and grain yield of rice using the “ORYZA2000” simulation model. The model was parameterized and validated on the basis of datasets from three field experiments with three widely cultivated rice varieties under various seeding dates in the 2008–09 growing seasons in the Khorezm region of Uzbekistan. The selected rice varieties represent short-duration (SD), medium-duration (MD), and long-duration (LD) maturity types. The model was linked with historical climate data (1970–99) and temperatures and CO2 concentrations projected by the Intergovernmental Panel on Climate Change for the B1 and A1F1 scenarios for the period 2040–69 to explore rice growth and yield formation at eight emergence dates from early May to mid-July. Simulation results with historical daily weather data reveal a close relationship between seeding date and rice grain yield. Optimal emergence dates were 25 June for SD, 5 June for MD, and 26 May for LD varieties. Under both climate change scenarios, the seeding dates could be delayed by 10 days. Increased temperature and CO2 concentration resulted in higher rice grain yields. However, seeding rice before and after the optimal seeding dates reduced crop yield and yield stability significantly because of spikelet sterility induced by both high and low temperatures. As the grain yield of SD varieties could be adversely affected by climate change, rice breeding programs for central Asia should focus on developing appropriate heat-tolerant MD and LD varieties.
Climate change has become an important global issue. Predictions for central Asia show that by the end of the twenty-first century temperatures are likely to increase by 3°–4°C, and the atmospheric CO2 concentration will increase from the current 380 ppm to 485–1000 ppm. Under such scenarios, crop yields are likely to decrease by as much as 30% in the region even when the direct positive physiological effects of increased CO2 are accounted for (Parry et al. 2007).
Rice growth, development, and yield formation are very sensitive to temperature. Currently, most of the rice production occurs in regions where temperatures are already above the optimum for crop growth (daytime maximum 28°C and nighttime minimum 22°C) (Krishnan et al. 2011). It is estimated that each 1°C increase in the daytime maximum–nighttime minimum temperatures within the 28°–21° to 34°–27°C range can decrease rice yields by about 7%–8% (Baker et al. 1992).
In central Asia, rice is the second major food crop and is grown on an area of 0.18 million ha (FAOSTAT 2010) in irrigated lowlands in the Amu Darya and Syr Darya river basins. A number of studies have already been conducted to examine the effect of increased temperature and CO2 concentration on East and South Asian rice cultivars (Wassmann et al. 2009; Krishnan et al. 2011), while far less attention has been devoted to rice cultivars grown in central Asia. Rice is a C3 plant and generally responds favorably to CO2 enrichment. However, several studies have shown that high air temperatures can reduce grain yield even under CO2 enrichment. Each 1°C increase in the minimum temperature during the growing season could decrease yields by 10%, whereas the effect of an increase in the maximum temperature on crop yield is insignificant (Peng et al. 2004); rice yield could decrease by (2%–6%) °C−1 with an average mean daily temperature of 26°C (Baker and Allen 1993; Matthews et al. 1995; Sheehy et al. 2006). However, Krishnan et al. (2007) reported that every 1°C increase in temperature decreases rice yield by 7.2% at the current CO2 concentration (380 ppm), but increases in CO2 enrichment up to 700 ppm will lead to an average yield increase of about 31% in India. Similarly, Baker (2004) reported a 46%–71% increase in rice yield at an ambient temperature of 28°C with CO2 enrichment in U.S. cultivars. Furthermore, modeling studies from Bangladesh (Karim et al. 1994), Japan (Horie et al. 2000), China (Bachelet et al. 1995), and India (Mall and Aggarwal 2002) reported country-specific variations in future rice production due to climate change; the greatest decline in crop yields will likely occur between the latitudes 10° and 35°N (Penning de Vries 1993; Krishnan et al. 2011). CO2 enrichment is likely to increase the photosynthetic rate, and thus biomass production, which in turn may positively affect assimilated allocation to reproductive organs (Wassmann et al. 2009). However, the yield decline under increased temperature conditions is the result of spikelet sterility due to the negative effect on pollination processes (Krishnan et al. 2011).
The climate in central Asia is continental and arid (Kottek et al. 2006), with short, hot summers and long, cold, dry winters. Under such climatic conditions, rice cultivation is only possible for around 140 days during the period May–October (Christmann et al. 2009). The seeding time of rice, therefore, is very crucial, as the flowering period with early seeding may coincide with peak maximum temperatures, while late seeding may result in low-temperature stress during grain filling (Devkota 2011). Given the sensitivity of rice to temperature, optimizing the seeding date and using adapted varieties will be of central importance for enhancing yields under climate change scenarios (Matthews et al. 1997; Blanche and Linscombe 2009). Simulation studies on various seeding dates with different growth-duration rice varieties can contribute to identifying the optimal seeding date and appropriate rice variety for specific geographical regions under both current and predicted climate change scenarios (Krishnan et al. 2011). This can indirectly overcome the predicted adverse effects of climate change on rice production in a particular region and may contribute to the development of suitable adaptation strategies through agronomic and plant breeding practices (Matthews et al. 1995).
The rice simulation model ORYZA2000, version 2.13, is capable of simulating phenology, growth, spikelet sterility, and grain yield of indica and japonica rice ecotypes in response to temperature, CO2, solar radiation, and cultivar-specific genetic characteristics (Matthews et al. 1997; Jing et al. 2007; Krishnan et al. 2007; Shen et al. 2011; Zhang and Tao 2013). It can simulate the response of rice phenology to climate change and variability in different climatic zones equally well or better than other rice phenology models such as the Crop Estimation through Resource and Environment Synthesis (CERES) rice model, regional climate model (RCM), Beta model, and Simulation Model for Rice–Weather Relationships (SIMRIW) (Zhang and Tao 2013). The objective of this study was to explore the potential effect of climate change on rice phenology and grain yield in central Asia by (i) parameterizing and validating the rice growth model ORYZA2000 for local rice varieties and (ii) assessing the impact of climate change as projected by the Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions Scenarios (SRES) under lowest future emission trajectory (SRES B1) and highest future emission trajectory (SRES A1F1) (Parry et al. 2007) at different emergence dates.
2. Materials and methods
a. Study area
The field experiments were conducted in the 2008 and 2009 rice-growing seasons in the Urgench–Khorezm region (41°32′12′′N, 60°40′44′′E) located in northwestern Uzbekistan on the left bank of the Amu Darya River. The climate of the area is arid with a long-term average annual rainfall of less than 100 mm. The soil at the experimental site is an irrigated alluvial meadow (Russian classification), that is, arenosol, gleyic, calcaric, sodic [Food and Agriculture Organization (FAO) classification], sandy loam to loamy sand with high soil salinity [2.7 dS m−1, electrical conductivity of a saturated soil extract (ECe) 1:1 in 0–15-cm soil depth], shallow (0.5–2 m) and saline (2–4 dS m−1) groundwater table, and low soil organic matter (0.4%–0.8%) (Table 1).
b. Field experiments
1) Experimental design and treatments
Three field experiments (one in 2008 and two in 2009) were conducted to evaluate phenology and growth of a set of widely cultivated Uzbek rice varieties seeded at different dates. Rice varieties of short duration (SD; Shoternboy-1, 85 days), medium duration (MD; Allanga-3, 105 days), and long duration (LD; Mustakillik, 125 days) were evaluated. The first experiment (experiment I) conducted in 2008 included evaluation of these varieties in a randomized complete block design with eight replications in a 600-m2 plot at the Cotton Research Institute, Urgench, Uzbekistan. Rice was seeded on 16 June in this experiment, and the final yield and yield attributes were recorded in a 7.5-m2 area in each plot. At a similar site, a second experiment (experiment II) was conducted in 2009 to evaluate the three rice varieties seeded on 28 May and 19 June in an unreplicated 30-m2 plot. For each variety, final yield and biomass were measured on three subplots of 6 m2. In the same year (2009), another unreplicated experiment (experiment III) was conducted at Urgench State University, Urgench, Uzbekistan (5 km from the Cotton Research Institute), to evaluate the effect of six seeding dates starting on 5 May–15 July at 15–20-day intervals. The plot size in this experiment was 20 m2, and final yield and biomass were measured in a 6-m2 plot for all seeding dates.
2) Crop establishment and management
Field preparation, sowing, seed rate, and irrigation water management were managed according to the recommended practices in the region. In all experiments, the field was dry ploughed 3–4 times, leveled, and irrigation water was applied. Pregerminated rice seeds were then uniformly directly broadcast into the standing water using the recommended seed rate of 80 kg ha−1. From seeding to 10 days of emergence, 1–2 cm of standing water was maintained in the fields. After 10 days of emergence, similar to the farmers' practice in the region, 5–15 cm irrigation water was applied after the disappearance of the standing water. In all experiments, a fertilizer dose of 250 kg ha−1 nitrogen (N), 120 kg ha−1 phosphorus (P2O5), and 80 kg ha−1 potash (K2O) was applied. Phosphorus and K2O fertilizers and 50% of the N were applied as a basal application during field preparation. The remaining N was top-dressed in two equal splits, that is, at panicle initiation and flowering. The experimental fields were kept weed-free during the entire crop-growing period through the combined use of the postemergence herbicide Gulliver (Azimsulfuron 50 WG) at 25 g ha−1 and hand weeding (performed on two occasions). There was no visible nutrient or water stress, and the crop was kept free of insect, pest, and disease infestation; thus, the experiment was considered as a potential production system (Bouman et al. 2001).
The phenological development of the rice was recorded in all experiments through visual observation using the standard evaluation system for rice (IRRI 2002). For each seeding date of experiments II and III, plant samples were collected from a 50 cm × 50 cm area at 15–20-day intervals for determination of biomass production and leaf area. The leaf area of green leaves was measured with a leaf area meter (Li-Cor, Inc., LI-3100; cm2) and converted to leaf area index (LAI; m2 m−2). Stem, green leaves, dead leaves, and panicles were separated and oven-dried separately for 72 h at 65°C until constant weight.
c. ORYZA2000 model
1) Model description
ORYZA2000 simulates rice growth, development, and water balance under potential production, water-limited, and N-limited conditions (Bouman et al. 2001). The model calculates the daily rate of biomass production as a function of solar radiation, LAI, temperature, leaf N content, and atmospheric CO2 concentration. The phenological development is simulated based on daily ambient temperature and photoperiod. The key development stages (DVSs) for rice are emergence, panicle initiation, flowering, and physiological maturity. Consequently, the life cycle of rice is divided into four phenological phases: (i) juvenile phase from emergence (DVS = 0) to start of photoperiod-sensitive phase (DVS = 0.4), (ii) photoperiod-sensitive phase from DVS = 0.4 until panicle initiation (DVS = 0.65), (iii) panicle formation phase from panicle initiation to 50% flowering (DVS = 1.0), and (iv) grain-filling phase from flowering to physiological maturity (DVS = 2.0). The duration of each of these four phases is calculated based on a cultivar-specific development rate constant, daily increment in heat units expressed in degree-days, and photoperiod (Table 2).
The ORYZA2000 model also accounts for the effect of CO2 enrichment by introducing a corrected coefficient to the initial light-use efficiency of a single leaf [ɛ; kgCO2 ha−1 h−1 (J m−2 s−1)−1]. The calculation of this value uses the formula by Jansen (1990):
where ɛ is the value of CO2 effect, represents the reference effect value of CO2 with the concentration of 340 ppm (defined as 1), and is the CO2 concentration in the actual simulation environment.
Effect of temperature on grain formation and spikelet fertility.
Rice grain yield is determined by carbohydrate production (source size) during grain filling and the storage capacity of grains (sink size). Sink size is a function of the number and maximum growth rate of spikelets. The number of spikelets at flowering is calculated from the total biomass accumulated from panicle initiation until first flowering (Kropff et al. 1994). In ORYZA2000, the rate of grain growth from panicle initiation to 50% flowering is tracked, and the number of spikelets formed (Si; number of spikelets per hectare per day) is calculated as the product of biomass accumulation from panicle initiation to 50% flowering (G; kg dry matter ha−1 day−1) and spikelet formation factor (Y; number per kilogram). The spikelet formation factor Y is the slope of the relationship between the effect of solar radiation, temperature, nitrogen, competition, and water on spikelet formation. Spikelets turn into grains during crop growth. However, some spikelets can become sterile because of either too high or too low temperatures and do not fill (Horie et al. 1992). Sterility caused by cold temperatures is based on the cooling degree-day (SQt) and is calculated as follows:
where Td is the average temperature (corrected for temperature increase caused by drought). The summation of SQt is done for the period of highest sensitivity of the rice panicle to low temperatures (0.75 ≤ DVS ≤ 1.2). The relation between the percentage sterility caused by cold (Sc, SF1) and the sum of the cooling degree-day is
Rice spikelets are also sensitive to high temperature, particularly at anthesis. Damage to the pollen occurs when the temperature at flowering is above approximately 35°C (Satake and Yoshida 1978). In ORYZA2000, the fraction of fertile spikelets caused by high temperatures (Sh, SF2) is calculated as (Horie 1993):
where Tm,a is average daily maximum temperature over the growing period (0.96 ≤ DVS ≤ 1.22) with elevated and ambient CO2 concentrations.
2) Model parameterization
The model was parameterized for SD (Shoternboy-1, 85 days), MD (Allanga-3, 105 days), and LD (Mustakillik, 125 days) varieties starting with the standard crop parameters for cultivar IR72 and following the procedures set out by Bouman et al. (2001). The data from two seeding dates (from experiment II) were used for model parameterization.
3) Model evaluation
Following model parameterization, the data for phenology, biomass partitioning, and yield from experiments I and III (seven seeding dates) were used for evaluating model performance. Following the procedures set out by Bouman and van Laar (2006), a combination of graphical presentation and various statistical measures was used to evaluate the performance of ORYZA2000. The graphs of the simulated and measured grain yield, biomass, green leaf dry weight, dead leaf dry weight, and phenological stages were compared. For the same variables, we computed the slope α, intercept β, and coefficient of determination R2 of the linear regression between measured X and simulated Y values. The model was also evaluated using the Student's t test of means assuming unequal variance P(t*). The variation in measured data is represented by mean standard deviation. The absolute root-mean-square errors (RMSEa) and normalized root-mean-square errors (RMSEn) were calculated as
where Yi and Xi are simulated and measured values, respectively; Xi is the mean of all measured values; and n is the number of measurements.
It is assumed that the model reproduces experimental data best when α is close to 0, β is close to 1, R2 is close to 1, P(t*) is larger than 0.05, RMSEa is similar to the standard errors of measured values, and RMSEn is similar to the coefficient of variation of measured values.
4) Climate change scenario analysis
Historical data on rainfall, minimum and maximum temperature, solar radiation, relative humidity, and vapor pressure (as required by ORYZA2000) were collected for a 29-yr period (1970–99) from the Urgench airport (3 km from the experimental site). The projected changes in surface air temperature under the SRES B1 and A1F1 scenarios for central Asia (Table 3) for 2040–69 were collected from the IPCC Fourth Assessment Report (Parry et al. 2007). The projected increase in temperature was added to the daily minimum and maximum temperatures, and two climate change scenarios were generated. The ambient CO2 concentration of 340 ppm in historical data, 540 ppm in the B1 scenario, and 960 ppm in the A1F1 scenario as projected by Parry et al. (2007) were used in ORYZA2000 for scenario analysis.
In the climate change simulations, ORYZA2000 was used to simulate the impact of climate change on phenological development (days to flowering and physiological maturity), grain yield, spikelet sterility factor due to low temperature (Sc, SF1), and spikelet fertility factor due to high temperature (Sh, SF2) in SD, MD, and LD rice varieties at eight emergence dates from early May to mid-July (6 May, 16 May, 26 May, 5 June, 15 June, 25 June, 5 July, and 15 July) over 29 years under current historical weather data and for 2040–69 under the SRES B1 and A1F1 scenarios. As the model is not capable of predicting emergence dates, the simulation treatments were planned with emergence dates instead of seeding dates.
a. Parameterization and validation of ORYZA2000
The details seeding date, days to emergence, panicle initiation, flowering, and physiological maturity were recorded (Table 4). The goodness-of-fit parameters (Tables 5, 6) show that the observed and simulated phenological stages of all rice varieties at seeding dates and years matched well. Furthermore, the observed and simulated dates for panicle initiation, flowering, and physiological maturity stages did not differ by more than 4 days at all seeding dates and in all rice varieties. Phenological stages were not affected under different seeding dates in the SD variety, while the MD and LD varieties did not reach flowering and physiological maturity stages when seeded in July.
The dynamics in biomass of green leaves, stems, dead leaves, grain, and LAI (Figs. 1, 2) and periodic and final grain yield and total aboveground biomass (Fig. 3) were simulated quite well throughout the growing season. The simulated LAI generally exceeded the measured LAI in all varieties.
b. Impact of climate change on rice phenology, grain yield and spikelet sterility
1) Rice phenology
Days to flowering varied among varieties and emergence dates in both historical weather data and climate change scenarios (Fig. 4). With historical weather data, the simulated days to flowering at the farmers' current seeding date in the region (5–15 June emergence) were 59, 76, and 91 days after emergence in the SD, MD and LD varieties, respectively. In the climate change scenarios, the predicted flowering dates for 5–15 June emergence showed that flowering could be delayed by 4 days under the B1 and by 8 days under the A1F1 scenario relative to the historical data.
In the SD variety, flowering was delayed under climate change scenarios relative to the historical data for all eight emergence dates. However, in the MD and LD varieties, flowering could be delayed by 1.5 days per 1°C increase in temperature with early emergence dates, while under late (July) emergence conditions, flowering was later in historical data than in the climate change scenarios (Fig. 4).
Under both current and climate change scenarios at all emergence dates, the SD variety reached the flowering stage (Table 7). However, the MD variety emerging on 15 July did not flower in 37% of the years in the historical data and in 7% of the years under the B1 scenario. Similarly, the flowering date of the LD variety was affected when emergence was after 5 July. Physiological maturity was also affected by the emergence date in all varieties (Table 7). The SD variety emerging on 5 July did not physiologically mature in 19% of the years in the historical data and in 7% of the years under the B1 scenario, while it was not affected under the A1F1 scenario. Likewise, the MD variety at 15 July emergence and the LD variety at 5 July and at 15 July emergence did not reach maturity in all years in any of the scenarios.
2) Grain yield
With the historical weather data, the simulated grain yield was highest at 25 June emergence in the SD variety (4.9 t ha−1), at 5 June in the MD variety (7.1 t ha−1), and at 26 May emergence in the LD variety (6.8 t ha−1). For these emergence dates, relative to historical data, grain yield of the SD variety was increased by 9% in the B1 and by 27% in the A1F1 scenario. In the MD variety, grain yield was not different between B1 and historical data, while it was higher by 14% under A1F1. The simulated grain yields of the LD variety were 3% higher in the B1 and 16% higher in the A1F1 scenario relative to historical data. Overall, relative to historical data, rice yield could be increased by 3% (185 kg ha−1) under the B1 scenario and by 18% (1122 kg ha−1) under the A1F1 scenario, which corresponds to an increase of about 187 kg °C−1 or 3% °C−1 in grain yield in 2040–69 relative to 1970–99 (Fig. 5). Under both climate change scenarios, grain yield was significantly higher at 10–20 days after emergence than at the abovementioned highest-yielding emergence dates in historical data, that is, highest yields of SD rice at 5–15 July emergence, MD rice at 15–25 June emergence, and LD rice at 5–15 June emergence.
Under earlier emergence than the highest-yielding emergence date in the historical data, grain yield was reduced by 4% under B1 and by 10% under A1F1 in SD rice, by 14% under B1 and 9% under A1F1 in MD rice, and by 6% under B1 and 5% under A1F1 scenario in LD rice. Despite higher grain yield under later emergence than at the highest-yielding emergence date in the historical data, rice plants in most years could not reach maturity (not shown in Fig. 5). Under early emergence dates in the climate change scenarios, grain yield was more variable in the SD variety, while the LD variety had consistently higher yields followed by the MD and SD varieties.
3) Spikelet sterility
In the historical weather data under best emergence dates, that is, 15–25 June, the average spikelet sterility for the MD rice variety was 10% and 5% due to high and low temperatures, respectively (Fig. 6). For this emergence period, spikelet sterility due to high temperature was higher under climate change scenarios (30% in B1 and 35% in A1F1) than in the historical data, while spikelet sterility due to low temperature was at par in the historical data and the climate change scenarios. The spikelet fertility factor due to high temperature at early emergence (compared to above-mentioned best emergence dates) was significantly low under the climate change scenarios, where it was lowest under A1F1 followed by B1. In contrast, spikelet sterility due to low temperature at later emergence than the above-mentioned best emergence date was highest in historical data followed by the B1 and A1F1 scenarios. The same trend was also observed in the SD and LD varieties (data not shown). The grain filling process and spikelet fertility due to high temperature on one of the early emergence dates (16 May; yearday 136) in three rice varieties is shown in Fig. 7. In all rice varieties, the spikelet fertility factor was lowest in A1F1 followed by B1 and historical data.
In the current version of ORYZA2000, v2.13, development rates, partitioning factors, relative leaf growth rate, specific leaf area, leaf death rate, and fraction of stem reserve coefficients are genotype × environment × management (G × E × M) parameters, and therefore, the model needs to be calibrated against treatments for accurate simulation (Bouman et al. 2001). Also, the model does not produce yield components such as panicle density, which limits its applicability to diagnose the causes of different treatment responses. Therefore, model improvements that incorporate development rates and other coefficients into genetic parameters would be highly desirable to enable application of the calibrated model across a wider range of environmental and management conditions.
The modeled biomass, leaf area index, and phenological development matched well with the observed values (Figs. 1–3). The slightly lower measured LAI than the simulated values could be due to a lower leaf production (5–8 leaves) (Devkota 2011). In general, rice varieties have 6–14 green leaves during flowering (De Datta 1981). The results of the statistical evaluation of model parameterization (Table 5) and validation (Table 6) are comparable in terms of accuracy with previous studies (Bouman and van Laar 2006; Belder et al. 2007), where ORYZA2000 was evaluated in potential production and water- and nitrogen-limited conditions, mostly in the humid, tropic region of Asia. Furthermore, the evaluation results of this study are comparable with those reported for semiarid regions by Amiri and Rezaei (2010). This suggests that the model is able to simulate phenological development and grain yield of rice accurately for arid climate conditions in central Asia. However, under the climate change scenarios, the temperature during early May (before 20 May) was higher than the threshold temperature required for emergence (20°C), but as the current version of ORYZA2000 starts simulation only after emergence (Bouman et al. 2001), the required emergence days under climate change conditions could not be simulated.
The projected average temperature of the rice seasons in 2040–69 would be raised by 2.9°C under the B1 and 4.1°C under the A1F1 scenarios in central Asia (Table 3). This would potentially extend the length of the rice-growing season by around 1 month. The historical data for 1970–2009 show that, except for June to 10 August, where the daily maximum temperature exceeds 35°C in 70% of the days (50 days out of 70), the daily maximum temperature remains lower than 35°C in other days of a year. As the study region has an arid climate, minimum and night temperatures, which generally cause spikelet sterility and yield reduction in rice (Peng et al. 2004), are generally low (Table 3). Thus, the projected future climatic conditions (increased temperatures and CO2) could provide greater opportunities for rice cultivation and yield increases in the region.
The long duration of seeding to emergence (10–13 days; Table 4) before 20 May seeding could be related to temperatures that were lower than required for emergence. In the 29-yr average (1970–99), the minimum and average temperatures before 20 May were 14° and 20°C, respectively (Table 3), and were thus lower than the critical threshold temperatures required for rice emergence (Yoshida 1981). Rice requires an average temperature of more than 20°C for emergence (Basnayake et al. 2003).
Previous modeling studies have generally shown that phenological development rate of rice would be accelerated and that the growing period would be shortened in the future as a result of climatic warming (Chen et al. 2005; Karlsen et al. 2009). However, in agreement with recent findings (Zhang and Tao 2013), our results reveal increased growing duration and delayed flowering and maturity under both B1 and A1F1 scenarios compared to the historical data. The delayed flowering under the climate change scenarios (Fig. 4) could be associated with higher temperatures. For normal heading, rice requires a daily mean temperature of 21°–30°C (Krishnan et al. 2011). Under the B1 and A1F1 scenarios, the predicted mean daily temperature in June, July, and August could be at or higher than the threshold level (Table 3). The base, optimum, and maximum temperatures for rice are 8°, 30°, and 42°C, respectively (Gao et al. 1992). The development rate of rice increases linearly above the base temperature to the optimum temperature. Beyond the optimum temperature, the development rate decreases linearly until a maximum temperature is reached (Kiniry et al. 1991). Below the base temperature or above the maximum temperature, the development rate is zero. Furthermore, flowering is longer at a mean temperature of 33°C in comparison to 29°C (Matthews et al. 1995). The delayed physiological maturity of rice (Table 7) under late seeding conditions with historical data could be related to low temperature. Generally, low temperature delays heading and physiological maturity (Krishnan et al. 2011). For physiological maturity, rice requires a minimum temperature of 12°–18°C and an optimum temperature of 30°C (Yoshida 1981). Because of the extreme aridity of the climate, even under the climate change scenarios (increased temperature), the minimum and average temperatures required for maturity during rice maturing months (September, and October) are lower than the threshold temperature for proper maturity (Table 3).
With the best seeding dates with historical data, unlike many earlier findings (Baker and Allen 1993; Matthews et al. 1995; Peng et al. 2004; Sheehy et al. 2006), our findings show no yield reduction in rice under climate change scenarios in central Asia (0%–9% and 14%–27% yield increase under B1 and A1F1 scenarios, respectively). However, this is in agreement with observations in India (Krishnan et al. 2007) and in the United States (Baker 2004). To date, rice cultivation is mostly concentrated in tropical and subtropical regions. In such environments, temperatures are already above the optimum for rice growth (28°–22°C), and high temperature is already one of the major environmental stresses limiting rice productivity (Krishnan et al. 2011). Increasing CO2 may influence rice yield positively by increasing the amount of carbon available for photosynthesis and negatively by increasing the air temperature due to the greenhouse effect (Krishnan et al. 2011). However, in contrast to other earlier findings, higher simulated yields under climate change scenarios in our study indicate that an increase in photosynthesis due to enhanced CO2 could surpass the negative effect of increased temperature on yield reduction in the higher latitudes zones where the ambient temperature is at a low level (Table 3). The critical air temperature for spikelet sterility could be reduced by 1°C at elevated CO2 because of low transpiration cooling driven by stomata closure (Matsui et al. 1997). Further, rice yields in the existing cropping areas could be zero if climate predications are correct (Matsui et al. 2001). As central Asia has arid climatic conditions with extensive irrigation facilities, the higher predicted rice yield under climate change conditions could be due to optimal temperatures and ample solar radiation throughout the growing season and sufficient irrigation water supply. However, the projected climate change could greatly affect the supply of water for irrigation in central Asia (Parry et al. 2007; Christmann et al. 2009). The predicted higher temperatures due to climate change will reduce yields under rain-fed conditions but may not have a strong effect under irrigation conditions (Krishnan et al. 2011). Thus, the predicted increase in grain yield of rice under climate change scenarios could depend on the future water supply situation in the region.
With early rice seeding, the lower grain yield under climate change scenarios was due to the fact that the flowering and grain-filling stages coincided with the high temperatures during the hot months (June and July; Table 3), which resulted in a significant reduction in spikelet fertility (Figs. 6, 7). Increased daily average and maximum temperatures shorten the length of the grain-filling phase (Bachelet et al. 1993) and reduce the seed-setting rate in rice (Fu et al. 2008). Daily average temperatures higher than 35°C for more than 1 h during flowering lead to a high spikelet sterility (Yoshida 1981; Jagadish et al. 2007). Flowering in rice occurs over an extended time period of 7–10 days (Yoshida 1981), and the high temperatures during flowering could cause significant spikelet sterility (Defeng and Shaokai 1995). Furthermore, in ORYZA2000, respiration is modeled explicitly as a function of temperature (Matthews et al. 1995; Bouman et al. 2001). Thus, yield reduction under early emergence conditions could also be related to higher respiration rates. Under late seeding conditions, the decreased yields (Fig. 2) were due to slow development rate (Table 7), delayed and incomplete grain filling, poor physiological maturity, and increased spikelet sterility due to cold temperatures (Fig. 6) as a result of the onset of the cold winter (Satake and Hayase 1970; Farrell et al. 2001; Lee 2001).
Short-duration varieties had lower and more variable grain yields than MD and LD varieties (Fig. 5). Thus, our simulation results suggest that selection of appropriate seeding time and proper rice varieties are crucial for adaptation to climate change. In the historical data, the lowest grain yield variability and highest yield at 25 June emergence in SD (Fig. 5), at 5 June emergence in MD, and at 26 May emergence in LD varieties compared to the other emergence dates suggest that these are the best emergence dates for a consistently higher yield. However, under both climate change scenarios, consistently higher grain yield (Fig. 5) and phenological development (Table 7, Fig. 4) at 5 July emergence in the SD, 15–25 June in the MD, and 5–15 June in the LD rice varieties indicates that the effect of increased temperature can be minimized through seeding 10 days later than the dates in the historical data.
Similar findings on phenology and productivity with respect to shifting the planting date of rice have been reported (Zhang and Tao 2013). Similarly, comparatively higher grain yields of LD and MD varieties than of SD varieties under the climate change scenarios (especially with early emergence conditions) suggests that the adoption of LD and MD varieties could be an alternative adaptation strategy under climate change scenarios. The use of LD varieties has also been suggested by Matthews et al. (1997) for Southeast Asian countries.
Besides increased rice yield, climate change may have other beneficial effects in central Asia. Increased temperatures in the region would increase the number of frost-free days. In the historical weather data, the frost-free period is approximately from April to September, while under climate change scenarios it may last 1 month longer, that is, from mid-March to mid-October. This could provide an opportunity for intensification of the cropping system by allowing timely sowing of a second crop.
ORYZA2000 is capable of simulating rice growth and development under different seeding dates in arid, irrigated drylands of central Asia. Simulation studies with the parameterized and evaluated model suggest that the length of the rice-growing season and the crop grain yield may increase under climate change scenarios because of more favorable temperature regimes. However, selection and use of adapted rice varieties and optimal seeding dates are crucial. The appropriate seeding dates under the current climate conditions are 25 June for SD, 5 June for MD, and 26 May for LD varieties. Under climate change scenarios, delaying rice seeding by 10 days is likely to result in comparatively higher rice yields. As the SD varieties could be more negatively affected by climate change, breeding programs should focus on developing heat- and cold-tolerant MD and LD rice varieties for the irrigated lowlands of central Asia.
This study was funded by the German Ministry for Education and Research (BMBF; project number 0339970A). Field work was conducted in the ZEF/UNESCO project “Economic and Ecological Restructuring of Land and Water Use in the Khorezm Region (Uzbekistan): A Pilot Project in Development Research.” Further comments and suggestions from two anonymous reviewers substantially improved the quality of this manuscript.
Current affiliation: South Asia Regional Office, International Maize and Wheat Improvement Center (CIMMYT), Kathmandu, Nepal.