Enhancement of rice canopy carbon gain by elevated CO2 is sensitive to growth stage and leaf nitrogen concentration


Author for correspondence: Hidemitsu Sakai Tel.: +81 29 838 8205 Fax: +81 29 838 8211 Email: hsakai@niaes.affrc.go.jp


  • • Increasing our understanding of the factors regulating seasonal changes in rice canopy carbon gain (Cgain: daily net photosynthesis – night respiration) under elevated CO2 concentrations ([CO2]) will reduce our uncertainty in predicting future rice yields and assist in the development of adaptation strategies.
  • • In this study we measured CO2 exchange from rice (Oryza sativa) canopies grown at c. 360 and 690 µmol mol−1[CO2] in growth chambers continuously over three growing seasons.
  • • Stimulation of Cgain by elevated [CO2] was 22–79% during vegetative growth, but decreased to between −12 and 5% after the grain-filling stage, resulting in a 7–22% net enhancement for the whole season.
  • • The decreased stimulation of Cgain resulted mainly from decreased canopy net photosynthesis and partially from increased respiration. A decrease in canopy photosynthetic capacity was noted where leaf nitrogen (N) decreased. The effect of elevated [CO2] on leaf area was generally small, but most dramatic under ample N conditions; this increased the stimulation of whole-season Cgain. These results suggest that a decrease in Cgain enhancement following elevated CO2 levels is difficult to avoid, but that careful management of nitrogen levels can alter the whole-season Cgain enhancement.


Crop yields will be affected by projected changes in the global climate, which result in changes such as increases in atmospheric CO2 concentrations ([CO2]), increased surface temperatures, and more frequent occurrences of extreme climate events (IPCC, 2001). Many experimental and modelling studies have been conducted in order to predict the impacts of these changes and to explore potential technologies for adapting various food crops to possible future environments (Amthor & Loomis, 1996; Kimball et al., 2002). These studies have included rice, which is the staple food of the largest proportion of the world's population (Maclean et al., 2002). In general, negative impacts on crop production are projected, due to increases in temperatures and in rainfall variability, but increasing [CO2] will have a positive impact, by what is known as the CO2 fertilization effect. An important adaptation strategy will be to take full advantage of the CO2 fertilization effect, but there are concerns about whether this positive effect can be fully realized (Long et al., 2004).

The effect of elevated [CO2] on crop yield largely occurs through increased photosynthetic rates, and the magnitudes and mechanisms of photosynthetic enhancement due to increased [CO2] (typically of the order of 200–300 µmol mol−1) have been reported in a number of studies at organ, tissue and molecular levels (Stitt, 1991; Drake et al., 1997; Stitt & Krapp, 1999). In rice, controlled-environment chamber experiments have shown that leaf photosynthetic rates were greatly increased (30% to 70%) by doubling [CO2] (Imai & Murata, 1978; Imai & Okamoto-Sato, 1991; Lin et al., 1997). This is because under current CO2 levels, its supply is often a limit to growth and increased [CO2] accelerates carboxylation processes. However, the CO2 fertilization effect often decreases with increasing exposure to elevated [CO2]. This down-regulation of leaf photosynthesis has often been observed in pot-grown rice (Imai & Murata, 1978), but has also been seen under field conditions (Seneweera et al., 2002). These reduced responses of leaf photosynthesis to elevated [CO2] are often associated with decreases in leaf nitrogen and the amount of Rubisco (ribulose 1,5-bisphosphate carboxylase/oxygenase) present (Drake et al., 1997).

Only a few studies have reported on canopy-scale photosynthetic responses of rice to elevated [CO2], but controlled-environment chamber experiments to measure the carbon budget of the canopy have indicated that doubling [CO2] increased the photosynthesis of the rice canopy by 36% at the panicle initiation stage (Baker & Allen, 1993). Lin et al. (1999) also reported that elevated [CO2] significantly increased the photosynthesis of the rice canopy at the tillering and flowering stages (by 51% and 38%, respectively), but that this stimulation diminished at the grain-filling stage. Sakai et al. (2001) measured the whole-season CO2 exchange of the rice canopy and showed an almost continuous decrease in the stimulation of carbon gain (Cgain), defined as the difference between daytime net photosynthesis and night respiration, by the rice canopy in response to elevated [CO2], as the crop aged. Thus, a decreased enhancement following elevated [CO2] was also evident at the canopy level.

Stimulation of the canopy Cgain due to elevated [CO2] results from several processes, which can broadly be divided into canopy photosynthesis and respiration. The canopy photosynthesis processes involve the aforementioned single-leaf photosynthetic response to elevated [CO2], but a simple integration of single-leaf measurements, generally done using only the top leaves of the canopy, does not adequately explain the canopy responses. In fact, Lin et al. (1999) reported a decreased enhancement of canopy photosynthesis during the later stages of growth without any decreased enhancement in the uppermost leaves; this suggests that leaves of different ages respond differently to elevated [CO2]. The enhancement of canopy photosynthesis due to elevated [CO2] may also involve an increased light interception by the canopy, but how much this increased light interception contributes to canopy photosynthesis remains poorly understood.

Both photosynthetic capacity and canopy light interception depend strongly on nitrogen (N) nutrition, because the leaf photosynthetic system demands a large amount of N (Evans, 1989). The interactive effects of leaf N and elevated [CO2] have therefore been intensively studied, particularly in terms of the photosynthetic apparatus of individual leaves, where decreased amounts of N and Rubisco often reduce the stimulation of photosynthesis by elevated [CO2] (Drake et al., 1997). The association between canopy-scale photosynthetic stimulation following elevated [CO2] and crop N status has yet to be analyzed, and this information will be necessary to provide us with appropriate fertilizer strategies in the future.

Canopy night respiration is another important element in determining the whole-canopy Cgain, and elevated [CO2] has been shown to increase the respiration of the rice canopy per unit of ground area (Baker et al., 1992). In general, this increased canopy respiration could be attributed to the increased amount of plant biomass that results from elevated [CO2] (Amthor, 1991), but the extent to which canopy respiration is involved in the decreased response of Cgain to elevated [CO2] remains unclear.

The importance of identifying the factors responsible for the decreased canopy-scale responses to elevated [CO2] is that canopy-scale responses can be directly related to the responses of growth and yield to elevated [CO2], which is not the case for organ- or tissue-scale responses. Findings at this level can therefore assist us in developing adaptation strategies.

We have previously shown an almost continuous decrease in the stimulation of rice canopy Cgain during the growing season (Sakai et al., 2001). The objectives of the present study were to further examine the extent to which down-regulation of Cgain in elevated [CO2] occurs under different environmental conditions and to identify the factors regulating this down-regulation at the canopy level. We hypothesized that N availability and crop N status can alter the seasonal pattern of Cgain stimulation under elevated [CO2]. For this purpose, we directly measured CO2 exchange rates and the daily Cgain of rice canopies throughout the growing season in naturally sunlit semi-closed controlled-environment chambers for 3 years under different weather conditions; in the last year we supplied a larger amount of N than in the other two. To identify the factors regulating the decreased enhancement of the canopy's Cgain by elevated [CO2], we analyzed two major components of the canopy's C budget: daytime net photosynthesis and night respiration. Daytime net photosynthesis was further divided into canopy light interception and photosynthetic capacity. We tested whether the crop's N status affected photosynthetic stimulation at the canopy scale.

Materials and methods

Experimental design

We conducted our study using a growth chamber system (Sakai et al., 2001) at the National Institute for Agro-Environmental Sciences, Tsukuba, Japan (36°01′ N, 140°07′ E) over 3 years (1998–2000). The growth chamber system consisted of six naturally sunlit, controlled-environment chambers (4 × 3 × 2 m). These chambers are semi-closed, where CO2 is supplied to compensate for plant CO2 uptake, and are capable of measuring canopy CO2 exchange. All the controlled-environment chambers were glazed with 5 mm-thick tempered glass, whose transmittance of photosynthetically active radiation was > 80%, but the rear (north) wall and the floor were made of stainless steel. Within each controlled-environment chamber, air temperature and relative humidity were controlled using electrical resistance heaters (with a bubbling system for humidification) and cold-water heat exchangers using proportional-integral-derivative (PID) controllers (DB1000; Chino, Tokyo, Japan). Daytime [CO2] was maintained at a set-point concentration using a computer-controlled pure-CO2 injection system with PID controllers, which compensated for CO2 uptake by the plant canopy. During the night, [CO2] increased, mainly due to plant respiration, but was kept within 100 µmol mol−1 of the daytime [CO2] by a computer-controlled air ventilation system which introduced ambient air when necessary to reduce [CO2]. Further details of this chamber system are described in Sakai et al. (2001).

In all 3 years, we grew a Japonica type rice (Oryza sativa L., Nipponbare cultivar) throughout the growing season (c. 4.5 months) under ambient [CO2] (363 : 413 µmol mol−1; day : night) and elevated [CO2] (695 : 744 µmol mol−1) in each of three controlled-environment chambers, with the treatments randomly assigned to these chambers in each year (see Table 1 for details). The concentrations used are the current ambient level and that projected for the end of the 21st century (IPCC, 2001).

Table 1.  Summary of the environmental conditions during the study
YearTreatment[CO2] (µmol mol−1)Mean incident PAR (mol m−2)Mean air temp. (°C)Mean r.h. (%)Basal Na applied (g m−2)Top-dressed NGrowth durationc (days)
DayNightDayNight(g m−2)
  • Data represent seasonal average values (±SD) for daily mean CO2 concentrations during the day and night, daily total incident photosynthetically active radiation (PAR), daily mean air temperature (T) during the day and night, relative humidity (RH), level of nitrogen (N) application, and growth duration for each year of the experiment. Rice (Oryza sativa) plants were grown under two CO2 treatments (ambient and elevated CO2) in each of three naturally sunlit semi-closed growth chambers.

  • a

    N was applied as ammonium sulphate.

  • b

    Values in parentheses are days after transplanting when nitrogen was top-dressed.

  • c

    Rice plants were harvested at the day of maturity under ambient CO2 concentration simultaneously in both CO2 treatments.

1998Ambient354 ± 8397 ± 2116.4 ± 8.224.7 ± 3.921.2 ± 3.379.6 ± (56)b151
Elevated670 ± 18703 ± 32        
1999Ambient366 ± 13409 ± 2021.0 ± 9.426.6 ± 4.022.2 ± 3.478.6 ± (47)1.5 (57)142
Elevated695 ± 14750 ± 33        
2000Ambient370 ± 7435 ± 1920.4 ± 9.027.2 ± 4.122.1 ± 3.479.6 ± (31)2.5 (60)136
Elevated721 ± 8779 ± 37        

Germinated rice seeds were sown in seedling trays on 20 April in each year, and the seedlings were grown in the controlled-environment chambers under one of the two CO2 concentrations, with air temperature and relative humidity controlled at 23°C and 80%, respectively. In each controlled-environment chamber, seedlings were transplanted into two stainless-steel containers (1.5 × 1.5 × 0.3 m each) filled with alluvial paddy soil, in groups of three (referred to as a hill) at a 20 × 20 cm spacing, on 15 May in each year. The containers were flooded with water to a depth of 1–5 cm throughout the growing season.

Nitrogen was applied as ammonium sulphate just before transplanting (basal) and again at around the panicle-initiation stage (top-dressing). The amount and timing of the application were similar to those followed by local farmers in 1998 and 1999, but we applied a larger amount of N in 2000 (Table 1). In all years, phosphorous (P, 6.5 g m−2) and potassium (K, 12.5 g m−2) were applied as calcium superphosphate and potassium chloride, respectively, just before transplanting. Air temperature inside the chambers was controlled to track ambient air temperature with a daily mean difference of ±0.1°C. Relative humidity was kept at 78–80%. When the rice canopy closed, ∼60 days after transplanting (DAT), we installed shade nets (50% light transmittance) along the side of the canopy to make the light environment similar to that in fields in 1998 and 1999, but not in 2000. When rice plants in the ambient [CO2] chambers completed the grain-filling stage (which always occurred later than in the elevated [CO2] treatment), we harvested the rice plants simultaneously in both [CO2] treatments (at 151, 142 and 136 DAT in 1998, 1999 and 2000, respectively).

Canopy CO2 exchange measurement

We monitored [CO2] in each chamber at 10-s intervals using an infrared CO2 controller (ZFP9GD11; Fuji-denki, Tokyo, Japan) and recorded the results every 5 min as a 5-min average. The rates of pure CO2 injection required to keep [CO2] in each chamber constant at the target level were controlled and measured with a mass-flow controller (SEC-400MARK3; Stec, Kyoto, Japan), and recorded every 5 min for each chamber. To obtain a more precise measurement of [CO2] in the chambers than was possible with the CO2 controllers, we extracted sample air from each of the six chambers through a sampling line to the control house, and determined [CO2] using an infrared gas analyzer (IRA-107; Shimadzu, Kyoto, Japan) that was automatically calibrated three times per day against CO2-free nitrogen gas and a standard CO2 gas (700 µmol mol−1). It took 5 min for the infrared gas analyzer to determine [CO2] in each chamber, for a total of 30 min to scan all the six chambers. The [CO2] thus determined was recorded to describe the CO2 regimes in the chambers.

The canopy Cgain per unit ground area was determined as follows:

Cgain = Pnet − Rnight,(Eqn 1)

where Pnet is the net photosynthetic rate of the rice canopy per unit ground area, and Rnight is the canopy dark respiration rate per unit ground area at night. Pnet was determined as follows:

Pnet = Cin − L + Rsoil,(Eqn 2)

where Cin is the CO2 injection rate into the chamber, L is the CO2 leakage rate from the chamber, and Rsoil is the CO2 flux from the paddy water and soil. Rnight was determined as follows:

Rnight = ΔC + L − Rsoil,(Eqn 3)

where ΔC is the [CO2] increase in a chamber while the air-ventilation system was closed. The air-ventilation system was programmed to allow for the measurement of ΔC while maintaining the night-time [CO2] at the desired level.

The rate of CO2 leakage (L) out of a chamber was estimated by modifying the methods of Acock & Acock (1989) and Kimball (1990). Every few weeks, we injected pure N2O into each chamber, and measured the decay of its concentration using the air sampling system described above and an infrared gas analyzer for N2O (ZRC1ZC11 Fuji-denki, Tokyo, Japan). We calculated L from the measured leakage rate, and calculated the [CO2] gradient between the ambient air and the chambers using a mass-balance method (Baker et al., 2000, 2004; Sakai et al., 2001).

We derived Rsoil from soil temperature and an empirical temperature function for Rsoil, which was obtained from the CO2 flux measurements at five air temperatures between 15 and 35°C, at 5°C intervals under flooded conditions after all above-ground plant material had been removed at the end of the growing season. Here, we assumed that the temperature function could be applied to the whole growing season.

Intercepted photosynthetically active radiation and radiation-use efficiency

We measured photosynthetically active radiation (PAR) in each chamber using PAR sensors (IKS-25 & IKS-225; Koito, Tokyo, Japan) mounted just above the canopy and just above the water surface in the canopy. We used the difference between these two PAR readings to calculate the PAR interception of the canopy. In addition, we calculated the light interception of the side of the canopy following the methods of Luo et al. (2000).

To evaluate stage-dependent changes in the canopy's photosynthetic capacity, we selected four 20-day periods (at around the tillering, panicle-formation, heading and grain-filling stages) each year; the corresponding DAT is listed in Table 4. For each period, we regressed daily Pnet against the daily PAR intercepted by the canopy for each chamber. All regressions were highly significant (P < 0.001) and there were no clear indications of non-linearity because we used daily integral values. These regression slopes represent the radiation-use efficiency (RUE).

Table 4.  Periodic mean values of daily intercepted photosynthetically active radiation (PAR) by the rice (Oryza sativa) canopy and radiation-use efficiency (RUE) under ambient (A) and elevated (E) [CO2] and the respective enhancement ratios (E/A) at different stages (1998–2000)
YearStageDATDaily incident PAR (mol m−2 d−1)Daily intercepted PAR (mol m−2 d−1)RUE (gCO2 mol−1)
  • DAT = days after transplanting. Mean values of daily incident PAR on the rice canopy during each stage are also shown. Rice plants were grown under A (360 : 410 µmol mol−1; day : night) and E (695 : 744 µmol mol−1) concentrations in each of three naturally sunlit semi-closed growth chambers. Each value represents the average for 20 d of the destructive sampling date ± 10 d at each stage.

  • a


  • **

    P < 0.01;

  • *

    P < 0.05; ns, no significant difference.

1998Tillering 31–5020.314.414.31.001.341.741.30
Panicle formation 56–7516.416.616.40.991.461.841.26
Heading 87–10615.517.417.41.001.822.011.10
Grain filling111–13016.117.917.41.000.961.021.06
1999Tillering 31–5015.013.513.30.981.571.971.25
Panicle formation 53–7220.621.220.90.981.421.611.13
Heading 79–9826.128.428.00.991.371.531.12
Grain filling101–12019.121.921.70.990.870.871.00
2000Tillering 30–4919.214.418.41.281.762.191.24
Panicle formation 50–6924.425.927.31.061.381.521.10
Heading 78–9724.
Grain filling100–11917.720.520.
anovaa d.f.       
Year 2 **  ns  
Main plot error 6       
CO2 1 ns  **  
Year × CO2 2 ns  ns  
Split plot error 6       
Stage 3 **  **  
Stage × CO2 3 *  **  
Stage × year 6 **  **  
Year × stage × CO2 6 ns  ns  
Split-split plot error36       

Destructive sampling

We destructively sampled three hills in each chamber at tillering, panicle formation, heading and grain filling stage in each growing season. At harvest, 12 hills were harvested to determine whole season C accumulation in the crop. At each hill, we excavated a 20 × 20 × 15 cm (L × W × D) block of soil and gently washed away the soil from the roots with running water. We then separated the plants into leaf blade, leaf sheath + stem, root, ear, and dead leaf blade, and measured the leaf area (excluding dead leaf blades) of all the sampled plants using an area meter (AAM-7; Hayashi denko, Tokyo, Japan). The plant parts were then oven-dried at 80°C for 3 days and the dry weights were determined. After grinding the samples, we determined the C and N concentrations using a CN coder (MT-700; Yanako, Kyoto, Japan).

Statistical analysis

We tested the effects of [CO2] on Pnet, Rnight, Cgain and the ratio of Rnight : Pnet (the R/P ratio) for significance by means of analysis of variance (anova) of the daily averages for each of the three years during four growth periods: from transplanting to the late vegetative growth stage (tillering, hereafter), from the late vegetative growth stage to the mid-reproductive growth stage (panicle-formation), from the mid-reproductive growth stage to the early grain-filling stage (heading) and the early grain-filling stage to maturity (grain-filling). We applied a split–split plot design in this analysis, with year as the main factor, [CO2] as the split-plot factor, and growth stage as the split–split plot factor. We analyzed the daily total PAR intercepted by the canopy and the RUE of the canopy in the same manner. Calculations were done using SPSS for Windows 7.5 J statistical software (SPSS Inc., Chicago, IL).


Meteorology and growth duration

The seasonal mean air temperature ranged between 24.7 and 27.2°C during the day and 21.2 and 22.2°C during the night (Table 1) and daily incident PAR (inside the chamber) ranged between 16.4 and 21.0 mol m−2 d−1 (Table 1). The 1998 crop experienced cooler temperatures and lower PAR than the other two crops for most of the growing period, and the difference was particularly large in the summer (60–120 DAT, data not shown). Mean temperatures and PAR were similar between the 1999 and 2000 crops. The standard deviations of [CO2] for the daytime and night-time measurements were 18 and 37 µmol mol−1, respectively. Crop duration varied widely, from 136 to 151 d (Table 1), reflecting the temperature differences between years; the longest duration was recorded in the coolest year (1998) and the shortest in the warmest year (2000).

Estimated and measured Cgain

We compared the total measured crop Cgain at harvesting with the estimated cumulative Cgain based on the chamber gas exchange for the whole growing period to determine whether intrinsic errors existed in the gas-exchange measurements (Fig. 1). The estimated Cgain values corresponded well to the measured Cgain values, with a root-mean-square error of 37.3 g m−2 and a mean difference between the estimated and measured values (bias) of 3.4 g m−2. The differences between the estimated and measured Cgain were mostly less than 5% of the measured Cgain in all years and for all [CO2] treatments, despite measured and estimated Cgain values that ranged widely, from 670 g m−2 to 1065 g m−2, depending on the year and the [CO2] conditions.

Figure 1.

Comparison between the measured and estimated total carbon gain (Cgain) of rice (Oryza sativa) canopies grown under ambient [CO2] (A, 363 : 413 µmol mol−1; day : night) and elevated [CO2] (E, 695 : 744 µmol mol−1) in each of three naturally sunlit semi-closed controlled-environment chambers in the 1998, 1999 and 2000 growing seasons. The measured total Cgain of the rice canopy was calculated from the total biomass at harvest multiplied by the corresponding C concentration; estimated total C was obtained from daily CO2-exchange measurements integrated over the whole growing period. Each data point represents a value obtained for each chamber (n = 3 for each treatment and each season). RMSE represents the root-mean-square error, R2 represents the coefficient of determination, and bias represents the mean difference between the measured and the estimated values. The solid line represents the 1 : 1 relationship; dotted and dashed lines show the ±5% and ±10% ranges for the measured C values, respectively.

Canopy CO2 exchange

The seasonal averages for Cgain, Pnet and Rnight varied significantly across the 3 years (Table 2). All three variables were largest in 2000, the year with the shortest time to harvest, and smallest in 1998, the year with the longest time to harvest. Elevated [CO2] significantly increased the seasonal averages of Cgain, Pnet, Rnight in all 3 years, but the [CO2] effects on Cgain and Pnet were larger in 2000 than in the other years, resulting in a significant year × [CO2] interaction. In all 3 years, the overall mean enhancement ratio (E/A; enhancement ratio for a parameter (value under elevated [CO2] divided by value under ambient [CO2])) for Pnet due to elevated [CO2] was generally similar to that of Rnight, resulting in a non-significant effect of [CO2] on the overall average Rnight to Pnet ratio (R/P).

Table 2.  Daily mean values of carbon gain (Cgain), canopy net photosynthesis (Pnet), night respiration (Rnight), and the ratio of night respiration to net photosynthesis (R/P ratio) under ambient (A) and elevated (E) [CO2], and the enhancement ratio (E/A) of each parameter during different stages and for the growing season as a whole (1998–2000)
YearStageDATCgain (gCO2 m−2 d−1)Pnet (gCO2 m−2 d−1)Rnight (gCO2 m−2 d−1)R/P ratio
  • DAT, days after transplanting. Rice (Oryza sativa) plants were grown under A (363 : 413 µmol mol−1; day : night) and E (695 : 744 µmol mol−1) [CO2] in each of three naturally sunlit semi-closed growth chambers.

  • a

    The products of the overall mean in this table and growth duration in Table 1 are seasonal carbon balance (Cgain, Pnet and Rnight) of each year.

  • b


  • **

    P < 0.01;

  • *

    P < 0.05; ns, no significant difference.

  • c

    The P-value is shown for the probability between 0.05 and 0.10 level.

1998Tillering  1–45 7.1 9.11.28 9.511.91.262.3
Panicle formation 46–9026.432.11.2234.
Heading 91–11023.024.21.0530.932.81.067.8
Grain filling111–15112.711.20.8817.115.80.934.3
 Overall meana16.518.51.1221.724.21.125.2
1999Tillering  1–4212.515.21.2215.318.61.222.8
Panicle formation 43–8332.135.01.0939.643.21.097.5
Heading 84–10334.034.21.0142.
Grain filling104–14215.915.70.9920.420.61.014.5
 Overall mean22.123.71.0727.529.91.085.4
2000Tillering  1–38 7.914.21.7910.017.61.762.0 3.41.670.2040.1930.95
Panicle formation 39–7634.642.81.2443.653.
Heading 77–9642.345.61.0851.958.31.129.612.71.320.1850.2171.17
Grain filling 97–13618.219.21.0524.
 Overall mean23.528.61.2229.736.01.216.3
anovab d.f.            
Year  2**  **   **  ** 
Main plot error  6            
CO2  1**  **   **  ns 
Year × CO2  2*  **   ns  ns 
Split plot error  6            
Stage  3**  **   **  ** 
Stage × CO2  3**  **   **  P = 0.074c 
Stage × year  6**  **   **  ns 
Year × stage × CO2  6ns  ns   ns  ns 
Split-split plot error 36            

The effects of [CO2] on Cgain, Pnet and Rnight differed significantly between growth stages. The E/A ratios for Cgain and Pnet were highest during the tillering stage, then decreased during subsequent rice development in all 3 years. The significant [CO2] × stage interaction for Cgain and Pnet confirmed the clear trend: decreased stimulation of these traits due to elevated [CO2].

The main effect of [CO2] on the R/P ratio was not significant, but the R/P ratio generally increased significantly with increasing growth stage. In addition, the E/A ratio was initially slightly below 1 in each year (i.e. a negative effect of elevated [CO2]), but gradually increased during the later growth stages, though this was only marginally significant for the [CO2] × stage interaction (P = 0.074).

N content of the rice plants

The total N uptake by the rice plants differed significantly among the years due to the differences in N application (P < 0.01, Table 3): the largest N uptake at harvest was observed in 2000, when the amount of N applied was greater than in the other years. Of the two crops that received the same amount of N, N uptake was smaller in 1998 when the crop grew under cooler temperatures and lower PAR. N uptake was slightly but significantly larger under elevated [CO2] than under ambient [CO2] during the growing season in all 3 years (P < 0.01, Table 3). There was no significant interaction between [CO2] and growth stage in terms of N uptake, but at the tillering stage in 2000, N uptake increased by 36% under elevated [CO2].

Table 3.  Total nitrogen (N) content and mean leaf N content per unit leaf area of the rice (Oryza sativa) plants grown under ambient (A) and elevated (E) CO2 concentration and the respective enhancement ratios (E/A) at different stages (1998–2000)
YearStageDATTotal N (g m−2)Leaf N (g m−2)
  • DAT, days after transplanting. Rice plants were grown under A (363 : 413 µmol mol−1; day : night) and E (695 : 744 µmol mol−1) in three naturally sunlit semi-closed controlled-environment chambers.

  • a


  • **

    P < 0.01;

  • *

    P < 0.05; ns, no significant difference.

1998Tillering 40 4.4
Panicle formation 65 8.5
Heading 9612.211.90.980.840.790.94
Grain filling12012.
1999Tillering 40 5.4
Panicle initiation 62 9.0
Heading 8811.712.91.100.880.830.94
Grain filling11014.814.70.990.630.630.99
2000Tillering 39 5.9 8.01.361.681.440.86
Panicle initiation 59 8.7
Heading 8715.
Grain filling10917.618.81.060.650.580.89
anovaa d.f.      
Year  2**  *  
Main plot error  6      
CO2  1*  *  
Year × CO2  2ns  ns  
Split plot error  6      
Stage  3**  **  
Stage × CO2  3ns  ns  
Stage × year  6**  **  
Year × stage × CO2  6ns  ns  
Split-split plot error 36      

Leaf N content per unit leaf area also differed significantly over the 3 years, but a noticeable difference was only observed at the tillering stage (Table 3): leaf N at this stage was similar in 1998 and 2000, despite the higher level of N fertilization in 2000, but was lower in 1999, when air temperatures were warmer than in 1998 and the level of N fertilization was lower than in 2000. From then on, leaf N decreased with increasing development, to on average around 0.6 g m−2 at the grain-filling stage for the 3 years. Leaf N was slightly and significantly lower than in the plants grown under ambient [CO2] during most of the remainder of the growing period in all 3 years (P < 0.05, Table 3). There were neither significant stage × [CO2] nor year × [CO2] interactions, suggesting the effect of [CO2] on leaf N was generally consistent across the years and the stages.

RUE and intercepted PAR

The main effect of [CO2] on RUE was highly significant (Table 4). Similarly to the trends for Cgain and Pnet, the enhancement ratio for RUE was highest at the tillering stage and decreased during subsequent rice development; this resulted in a significant [CO2] × stage interaction. Unlike Cgain and Pnet, however, the main effect of the year and the [CO2] × year interaction were not significant, and the enhancement ratio at each stage was similar over the 3 years.

The variation in RUE as a function of growth stage and year was closely associated with leaf N contents in both [CO2] treatments, and the relationships were well fitted by a quadratic function for each [CO2] level (Fig. 2). RUE was substantially higher under elevated [CO2] than under ambient [CO2] when leaf N content was high, but the difference between the quadratic curves decreased as leaf N decreased. The two quadratic functions intersected at leaf N = 0.39 g m−2.

Figure 2.

The relationship between radiation-use efficiency (RUE, daily net photosynthesis divided by the daily photosynthetically active radiation (PAR) intercepted by the rice (Oryza sativa) canopy) and mean leaf nitrogen (N) content per unit leaf area of the rice canopies grown under ambient (A, 363 : 413 µmol mol−1; day : night) and elevated (E, 695 : 744 µmol mol−1) [CO2] in three naturally sunlit semi-closed controlled-environment chambers in the 1998, 1999 and 2000 growing seasons. Data points represent values obtained at the tillering, panicle-initiation, heading and grain-filling stages (corresponding dates are listed in Table 4). The leaf N content was obtained by destructive sampling at each stage and RUE was obtained by regressing daily net canopy photosynthesis against daily PAR for the 20-d period at each stage (±10 days from the sampling date). The quadratic function: Y =−1.09 + 4.68X− 1.76X2 (E), Y =−0.72 + 3.58X– 1.33X2 (A).

The main effect of [CO2] on the daily amount of PAR intercepted by the rice canopy was not significant (Table 4). In fact, there was little response of intercepted PAR to elevated [CO2] (i.e. an E/A ratio of approximately 1) during the first 2 years. As indicated by the significant interactions between [CO2] and growth stage and between [CO2] and year, elevated [CO2] substantially increased daily intercepted PAR (by 28% at the tillering stage in 2000). The main reason for this increase for the 2000 crop was that its leaf area index (LAI) was enhanced by elevated [CO2] until c. 10 days before heading, whereas no such stimulation of LAI due to elevated [CO2] before the heading stage was recorded in 1998 and 1999 (data not shown). The enhancement of LAI (i.e. the E/A ratio) for all growth stages combined over the 3 years was significantly and closely correlated with the enhancement of crop N uptake; the slope of the regression line (1.17) was significantly different from 0 (P < 0.001), but was not significantly different from to 1 (Fig. 3).

Figure 3.

The relationship between the enhancement ratios resulting from elevated [CO2] (Elevated (E, 695 : 744 µmol mol−1; day : night)/Ambient (A, 363 : 413 µmol mol−1)) for leaf area index (LAI) and crop nitrogen (N) uptake for rice (Oryza sativa) grown under ambient and elevated [CO2] in naturally sunlit semiclosed controlled-environment chambers in the 1998, 1999 and 2000 growing seasons. Data points represent values measured by destructive sampling at the tillering, panicle-initiation, heading and grain-filling stage.


The whole-season canopy gas-exchange measurements conducted over the 3 years revealed two major features of the responses of rice to elevated [CO2]. First, the enhancement of canopy Cgain due to elevated [CO2] (by about 300 µmol mol−1) differed in the 3 years. Note that there is no simple interpretation of the effects of climatic conditions during these years because the 2000 crop received a higher amount of N. However, a comparison of the results for the first two crops, which were exposed to contrasting air temperatures but received the same N application, revealed that both the stimulation of Cgain and its pattern of change during crop development were similar between years. Baker & Allen (1993) reported that photosynthesis by the rice canopy measured at about 2 months after planting was relatively insensitive to air temperatures ranging from 25 : 18°C (day : night) to 37 : 30°C, under elevated [CO2]. Nakagawa & Horie (2000) reviewed the effects of the interaction between temperature and [CO2] on rice and showed a significant positive relationship between air temperature and biomass stimulation by elevated [CO2]; however, this temperature dependence was generally weak. The present results suggest that a difference of about 2°C in seasonal mean temperature had no significant impact on the stimulation of canopy Cgain by elevated [CO2], or on its time course with respect to the plant's developmental stages.

The second important feature is that the stimulation of canopy Cgain decreased continuously with increasing crop development in all 3 years. The enhancement ratios (E/A) for both net canopy photosynthesis and night respiration, which are the two main components of canopy Cgain, decreased with increasing crop age, but not in parallel, suggesting that both components were involved in the decreasing stimulation of canopy Cgain under elevated [CO2]. In general, high photosynthetic rates (and therefore, heavier crops) lead to increases in subsequent growth and maintenance respiration rates (Amthor, 1989), so both components can change concomitantly. As shown in Table 2, however, the R/P ratio was unaffected by elevated [CO2] before the heading stage, but from then on, the R/P ratio increased under elevated [CO2] (P = 0.074). A larger growth and maintenance requirement resulting from the increased biomass produced in response to elevated [CO2] (Amthor, 1991) could be a major reason for this observation. In fact, Baker et al. (1992) reported that canopy respiration per unit ground area increased in response to elevated [CO2] as a result of the concomitant biomass increase. This in turn accelerated the decreased stimulation of Cgain due to elevated [CO2], but the magnitude of its contribution was much smaller than that of the photosynthetic enhancements, which decreased by 20–70% between the tillering and grain-filling stages.

The major determinants of canopy photosynthesis are PAR interception by the canopy and the efficiency of using this received light for CO2 assimilation (i.e. RUE; Loomis & Connor, 1992). Of these two determinants, we observed a consistent decrease in stimulation by elevated [CO2] for RUE in all 3 years; this was the main factor responsible for the decreased stimulation of canopy photosynthesis. Several factors can influence RUE, including incident radiation, canopy architecture and photosynthetic capacity (Sinclair & Muchow, 1999), but incident PAR in the present study was identical between [CO2] treatments, and the effects of the [CO2] treatment on leaf area were generally small, except in 2000. We did not measure canopy structure by stratified sampling, but the architecture of a rice canopy is typically unaffected by elevated [CO2] (Anten et al., 2003). These data suggest that differences in RUE most likely reflected changes in the canopy's photosynthetic capacity. At the single-leaf level, photosynthetic capacity is known to depend strongly on the leaf N status (e.g. Evans, 1989). At the canopy level, the links between photosynthetic capacity and leaf N status is less well understood. Ample evidence exists to show that decreases in the stimulation of single-leaf photosynthesis due to elevated [CO2] are more dramatic in N-limited plants than in well-fertilized ones (Stitt & Krapp, 1999). Only a few studies have addressed the possible relationship between a plant's N status and the enhancement of canopy photosynthesis due to elevated [CO2] in rice. Anten et al. (2003) measured leaf photosynthetic parameters and their relationships with leaf N contents, and scaled-up the results from the level of individual leaves to whole-canopy photosynthesis using a simple model. Their theoretical results predicted an N × [CO2] interaction that determines canopy Cgain, and that the stimulation of canopy Cgain decreases with decreasing leaf N content. Our results provided direct evidence of the decreased stimulation of canopy photosynthetic capacity with decreasing leaf N (Fig. 2).

Leaf N content generally declines as the crop ages, which is accelerated by elevated [CO2] through the dilution effect: the stimulation of biomass accumulation under elevated [CO2] exceeds that of N uptake by the plants. This would lead to the observed decreased stimulation of canopy photosynthetic capacity. In other words, plants under elevated [CO2] will need progressively more N as the growing season progresses to avoid a dilution of leaf N. The ontogenetic decline in leaf N is difficult to avoid, but the present results suggest that appropriate N management to keep leaf N at high levels may alleviate the acclimation of canopy photosynthetic capacity to elevated [CO2].

The amount of PAR intercepted by the canopy is mainly determined by LAI, but for major crops, including rice, LAI is generally insensitive to elevated [CO2] (Ainsworth & Long, 2005). In the present study, LAI (hence, the amount of intercepted PAR) was also insensitive to elevated [CO2] during the first 2 years, but was strongly stimulated by elevated [CO2] at the tillering stage in 2000, when a larger amount of N was applied (Table 4). This stimulation of PAR interception, along with the accompanying RUE enhancement, resulted in a substantial enhancement of canopy Cgain (about 80%) at the tillering stage, and the cumulative whole-season enhancement in 2000 was also larger than in other years. These multiplicative enhancements in Cgain or growth have also been reported in young, vigorously growing woody species (Ferris et al., 2001; Taylor et al., 2003).

The enhancements of LAI and of crop N uptake under enhanced [CO2] were strongly correlated (Fig. 3), with the slope of the regression being close to and not significantly different from 1. This clearly indicates the strong dependence of leaf area development on N availability and that elevated [CO2] can promote crop N uptake when ample N is supplied. The enhancement of canopy Cgain due to increased N uptake and LAI development during the early growth stages in 2000 was the major reason for the significant [CO2] × year interaction for the whole-season Cgain, because RUE did not differ at the tillering stage across all 3 years. A [CO2] × N interaction was also found for rice growth in the field when N application was sufficient (Kim et al., 2001). Promoting N uptake under elevated [CO2] will therefore increase the overall stimulation of Cgain under elevated [CO2].

Like many other chamber studies, there is some need for caution in interpreting and extrapolating the present results in the context of climate change impacts on crop production. First, any component of the gas exchange measurements will be subject to error because some components, such as soil respiration and CO2 leakage rates, were not measured continuously and we used temperature functions to estimate these components. For soil respiration, the variation in such respiration–temperature curves over the course of the growing season is generally small (Yamagishi et al., 1980). In addition, the CO2 flux from a submerged paddy field is typically much smaller than that from an upland field (Baker et al., 2000) and the canopy CO2 exchange rate, so that the estimation errors for canopy CO2 exchange associated with Rsoil can be considered insignificant. In fact, a 10% overestimate of soil respiration at any soil temperature is estimated to influence Pnet and Rnitht by less than 2% and 5%, respectively. CO2 leakage could be a bigger source of error in the elevated [CO2] chambers because L was larger than Rsoil. However, possible changes in L with time could be detected by N2O decay measurements every few weeks during the growing season, and the high correlation between estimates of canopy Cgain from gas exchange and harvested plants (Fig. 1) provides evidence for a proper estimation of both soil respiration and leakage. Second, the plants were rooted in relatively shallow containers (1.5 × 1.5 × 0.3 m) and ample evidence exists that rooting volume influences the growth and yield responses to elevated [CO2] (Arp, 1991). In fact, we cannot totally exclude the possibility that the restricted rooting volume may account for a part of the down-regulation observed in the present study. However, we planted rice seedlings at a 0.2 × 0.2 m spacing, so that the rooting volume per hill was slightly larger than 10 L, which is the threshold below which a down-regulation of photosynthesis tends to occur (Arp, 1991). It is also worth noting that the rooting zone, even in paddy fields, is generally shallow due to the existence of a hardpan layer, the typical depth of the ploughed layer being 0.1–0.15 m, where the majority of roots exist (Yoshida, 1981). Third, differences exist in the environmental conditions between the inside and outside of the chamber including light quality and quantity, air movement, variability of temperatures and humidity; these pose concerns regarding the applicability of enclosure-based experiments to future open fields (Kimball et al., 1997). To reduce these concerns and any uncertainty in predicting future crop production, we will use the current canopy-scale gas-exchange measurements to improve the existing rice growth model and to test it against the data obtained in the open field experiments such as free-air CO2 enrichment.


Our multi-year CO2-exchange measurements for rice canopies grown under two [CO2] levels in naturally sunlit semi-closed chambers confirmed our previous results showing that the stimulation of canopy Cgain under elevated [CO2] decreased as rice development progresses (Sakai et al., 2001); this occurred in all three growing seasons, despite differences in environmental conditions. The present results have revealed that the decreased canopy Cgain response was mainly attributable to the decreased enhancement of canopy photosynthesis, and partially to the increased night respiration after heading. Furthermore, this study shows that the decrease in the photosynthetic response to elevated [CO2] resulted mainly from the decreased stimulation of RUE, which was in turn closely associated with ontogenetic decreases in leaf N. The whole-season canopy Cgain was largest where a positive LAI response to elevated [CO2] was observed as a result of higher N supply and higher crop N uptake. These results suggest that the decreased stimulation of canopy Cgain can be alleviated by the management of N application through a stimulation of LAI and RUE.


This work was supported by the Rice-FACE project under the CREST (Core Research for Evolution Science and Technology) program of the Japan Science and Technology Corporation.