There have been few practical ways of measuring physiological determinants of rice yield. Rapid evaluation of yield determination traits may expedite breeding of high-yielding rice. Here, we report a new remote-sensing technique for the evaluation of canopy ecophysiological status under field conditions developed based on simultaneous measurements of sunlit and suddenly shaded canopy temperatures. This technique has the advantage of instantaneous estimation of aerodynamic resistance (ra) and canopy diffusive resistance (rc) without measuring wind velocity. Canopy diffusive conductance (1 / rc) estimated by the remote sensing method was closely related to leaf stomatal conductance (gs) measured with a portable gas exchange system. This result supported the validity of this new method for quantitative estimation of canopy physiological characteristics. Significant genotypic differences were obtained in canopy–air temperature difference (Tc − Ta), rc and 1 / rc during the 2-week period preceding full heading for two years, and 1 / rc was highly correlated with crop growth rate (CGR), which was closely related to the final yield. These results suggest that 1 / rc can be an effective criterion for the selection of high-yielding rice genotypes, and the remote sensing technique proposed here can be a powerful tool for the rapid evaluation of 1 / rc under field conditions.
Although breeding of rice genotypes with increased yield potential is urgently required to meet the increasing rice demand in Asia and Africa, rice yield potential has not substantially increased especially in the tropics in the last decades (Peng et al. 1999). The remarkable increase in rice yield potential during the Green Revolution (mid-1960s to mid-1970s) was achieved mainly through breeding of genotypes with short plant height, high tillering ability and elected leaves (Peng & Khush 2003). However, further increase of rice yield potential through modification of such visual morphological traits may not be feasible because most of the high-yielding rice cultivars already have an ideal canopy structure (Richards 2000).
Crop physiology has developed as a powerful discipline during the 20th century and has contributed to our understanding of the mechanisms underlying crop growth and development (Miflin 2000). Nevertheless, physiological traits have rarely been used in crop breeding for higher yield potential due to lack of the distinct physiological traits for higher yield and/or difficulties in measuring these traits on a large number of plants (Hall et al. 1994). Thus, there is a need to identify the physiological traits that lead to better yield and develop effective and useful screening methods based on these traits.
Previous studies indicated that rice genotypic difference in grain yield of rice was most closely related to that in crop growth rate (CGR) during the 2-week period preceding full heading (Horie et al. 2003; Takai et al. 2005). Horie et al. (2003) also showed that rice genotypes having a higher CGR during this period had higher leaf stomatal conductance (gs). A similar result has also been reported in wheat (Fischer et al. 1998) and cotton (Lu et al. 1994), in which a higher yield potential was associated with a higher gs. These previous studies suggest that plant gs can be an effective criterion for the selection of crop genotypes with a higher yield potential. However, determination of a representative gs value for each genotype requires measurements on several leaves by the leaf chamber method. This suggests that canopy diffusive conductance (1 / rc), which can reflect the stomatal status of leaves located at the top of the canopy, may be a more practical and useful selection criterion than leaf gs in the breeding programs.
The objectives of the present study were: (1) to develop a rapid remote sensing method to measure canopy diffusive resistance (rc) that is applicable to rice breeding; (2) to examine the validity of this method in comparison with leaf gs measured by the usual leaf chamber method; and (3) to examine if rc values obtained for different rice genotypes are related to their yield potentials. The new remote sensing method for measuring rc involves simultaneous measurements of sunlit and suddenly shaded canopy temperatures, net radiation flux density and dry and wet bulb temperatures. This method enables us to determine rc and aerodynamic resistance (ra) without measurements of wind profiles. Two-year field experiments were conducted using widely different rice genotypes. The new remote sensing method was applied to measure the rice genotypes’rc values during the 2-week period preceding full heading, the most critical stage in rice yield determination. We also measured CGR during that stage and final grain yield of those genotypes was obtained. This paper describes the theory of the newly developed remote sensing method for measuring rc and the relationship of rc with CGR and grain yield. The applicability of this method for plant breeding to improve rice yield potential is also discussed.
On the basis of the energy balance theory (Monteith 1973), simultaneous measurements of sunlit and suddenly shaded canopy surface temperatures (Tc and Tcs, respectively) and microclimate data allow us to estimate the evapotranspiration rate (E; g m−2 s−1), aerodynamic resistance (ra; s m−1) and canopy diffusive resistance (rc; s m−1) under field conditions. The energy balance on a canopy surface can be given as
Rn = H + λE + G(1)
where Rn is the net radiation (W m−2), H is the sensible heat flux density(W m−2), λ is the latent heat of vapourization (2442 J g−1) and G is the soil heat flux density (W m−2). In well-developed rice canopies that completely cover ground [leaf area index (LAI > 4)], as in the present study, it is assumed that G ≈ 0 and E nearly equal the canopy transpiration rate (Sakuratani & Horie 1985). Then H and the latent heat flux density (λE) between canopy surface and atmosphere are expressed as
where Cp is the heat capacity of air (1.006 J g−1°C−1), ρ is the density of air (1200 g m−3), Ta is the air temperature (°C), ec* is the saturated vapour pressure at intercellular spaces of leaves given as a function of Ta (hPa), ea is the vapour pressure of air (hPa) and γ is the psychrometric constant (0.662 hPa °C−1). The ra for heat transfer was assumed to be equivalent to that for vapour transfer in turbulent boundary layers on the rice canopy. By comparing Eqns 1 and 2, E can be given by
Because all the terms, except for ra, in Eqns 4 and 5 are measurable or already known, determination of ra enables us to estimate E and rc. The present method of estimating ra is based on the principle that while leaf temperature responds quickly to a sudden change in solar radiation, stomata need some time before they respond to the change. Horie (1978) reported that the time lag in the closing response of rice stomata to a sudden darkening was about 6 min at 30 °C. Therefore, energy balance on the canopy surface with sudden shading under full sunlight can be expressed as follows (Fig. 1a):
Rns = Hs + λEs + Gs(6)
where the subscript s denotes the respective quantities for the shaded part. For the same reason as for G, Gs in Eqn 6 can also be neglected. Then, in the Eqns 1–3 and 6–8, there are six unknown variables (E, H, ra, rc, Es and Hs), and the other terms are all measurable or already known. Hence, we can solve the six simultaneous equations to yield ra as follows:
Equation 9 indicates that simultaneous measurements of canopy temperatures and microclimate under full sunlight and sudden shade enable us to estimate ra, leading to the subsequent determination of E and rc by Eqns 4 and 5.
MATERIALS AND METHODS
Field experiments were conducted in the paddy field at the Graduate School of Agriculture, Kyoto University, Kyoto, Japan (35°2′N, 135°47′E, 65 m altitude) in 1999 and 2000. The soil was grey lowland soil classified into Haplaquept. Six rice (Oryza sativa L.) genotypes having a wide genetic background were used in the experiments (Table 1). These included Indica, Japonica, Javanica (Tropical Japonica), first generation of new plant type (NPT) released by the International Rice Research Institute (IRRI) and Indica–Japonica crossbred. Takanari, Milyang 23 and Nipponbare were used in both years. IR65564-44-2-2 (hereafter, NPT1), IR65600-127-6-2-3 (NPT2) and Kamenoo were used only in 1999, while the genotypes of IR72, Shanguichao and Takenari, were used only in 2000.
Table 1. Genotypic difference in canopy-air temperature difference (Tc − Ta), aerodynamic resistance (ra), canopy diffusive resistance (rc), canopy diffusive conductance (1 / rc) and crop growth rate (CGR) during the 2-week period preceding full heading
Tc − Ta (°C)
ra (s m−1)
rc (s m−1)
1 / rc (m s−1)
CGR (g m−1 day−1)
Values are shown as mean ± standard error in two replications of experimental plot.
Means followed by same letter within each term and year are not significantly different at 5% level among genotypes with Duncan's multiple range test.
NPT1 and 2 stand for IR65564-44-2-2 and IR65600-127-6-2-3, respectively.
In 1990, seeds were sown in the seedling nursery on April 22 and then transplanted with one seedling per hill on 20 May. In 2000, seeds were sown on 27 April and then transplanted on 24 May. Planting densities (hill m−2) were 50 for NPTs and 22.5 for other genotypes in 1999, and 25 for all genotypes in 2000. Each plot in 1999 and 2000 was 40 m−2 and 26 m−2, respectively. Rice genotypes were arranged in a randomized block design with two replications in both years. The amounts of fertilizer applied as basal dressing were 40 kg N ha−1, 100 kg P ha−1 and 100 kg K ha−1 in 1999, and 30 kg N ha−1, 100 kg P ha−1 and 100 kg K ha−1 in 2000. N was top-dressed at a rate of 30 kg ha−1 at tillering, 60 kg ha−1 at panicle initiation and 20 kg ha−1 at full heading in 1999, but at 20, 50 and 20 kg N ha−1, respectively, in 2000. The time of full heading was defined as the time when panicles emerged from 90% stems of each hill. Weeds, diseases and insects were strictly controlled throughout the entire rice growth periods in both years.
Twelve hills of NPTs and eight hills of other genotypes were sampled from each plot 2 weeks before full heading and during the full heading stage. Whole plant samples except roots were dried in the oven at 80 °C to a constant weight to determine the dry weight. The CGR during the 2-week period preceding full heading was calculated from the biomass increase during that period. At maturity, plants on 1 m2 of each plot were harvested to determine grain yield, but Kamenoo plants were not harvested in 1999 because they had lodged down after full heading. [For the detailed results on yield and growth analysis, see Takai et al. (2005).]
The instrument used for remote sensing of canopy surface temperature was a portable infrared thermometer (Thermotracer, TH5104, NEC San-ei Co. Ltd., Japan) with a resolution temperature of 0.1 °C and view angle of 21.2°. Measurements of Tc and Tcs were carried out from the south side with depression angles of 15–25°, and the canopy surface emissivity was regarded as 0.96 because these conditions kept the measured error to a minimum (Inoue 1986). After the canopy was suddenly shaded with a 1 m2styrofoam plate placed at about 1 m above the canopy perpendicular to the sunlight (Fig. 1a), Tc and Tcs included in the image were recorded two or three times within 1–3 min by the thermometer set approximately 2 m away from the measured spot (Fig. 1b). The recorded data were transferred to a computer for image analysis, and the average Tc and Tcs per measurement were calculated from two or three records. Wet and dry bulb thermometers and a net radiometer (CN-11, Eiko-Seiki Co. Ltd., Japan) were set at 0.1–0.2 m and 1.5 m above the canopy and the ground, respectively, in the Nipponbare plot. A quantum sensor (LI-190SH, Li-Cor, Lincoln, Nebraska, USA) to measure photosynthetically active photon flux density was placed near the paddy field. Because it was impracticable to measure Rn and Rns simultaneously with a net radiometer, we determined Rns during measurements of Tc and Tcs with the following method. In a preliminary experiment, a net radiometer was shaded with a styrofoam board for several minutes until Rns showed stable values. A close relationship was observed between Rn just before shading and Rns / Rn under various solar radiations (Fig. 2).
(r2 = 0.881, P < 0.001)
This relationship was available only for the fully established canopy. Since each rice genotype had already established a full canopy (LAI > 4) at the time Tc and Tcs were measured, we estimated Rns from the measured value of Rn using Eqn 10. Each microclimate datum was automatically collected every 6 s with a data logger (CR-10X, Campbell, USA) and recorded as the average for 1 min. Tc and Tcs were measured intensively more than 10 times in 1999 and 5 times in 2000 from 9:00 to 15:00 during the 2-week period preceding full heading under the condition where the shaded area on the canopy surface could be clearly recognized.
The data obtained by our new remote sensing method was compared with those obtained by a portable photosynthesis system (LI-6400, Li-Cor) in 2001. Energy balance terms on canopy and gas exchange-associated traits of single leaves were measured in eight rice genotypes during the 2-week period preceding full heading. Figure 3 shows the genotypes used for this comparative measurement. Plants were adequately cultivated as in 1999 and 2000. Fully expanded leaves were used for the leaf gas exchange measurements. Mean data in each genotype during the period were used for the comparison of the two methods.
Genotypic difference in energy balance terms on canopy surface
Figure 4 shows the change of Tcs continuously measured with an infrared thermometer after shading the canopy surface with styrofoam board in Nipponbare plot. Tcs reached a minimum after 1 min of shading, maintained mainly the same temperature with small fluctuation for a while and then rose after 5 min. A previous study reported that leaf temperature of citrus reached a steady state by several tens of seconds when solar radiation changed (Takechi et al. 1962). Furthermore, increase of Tcs after 5 min of shading may be attributed to stomatal closure. Horie (1978) showed that the rice stomata closed about 6 min after a sudden darkening. Therefore, in this study, Tc and Tcs were measured within 1–3 min after shading.
One of the notable points in the present remote sensing technique is that nothing can be estimated in the case of Tc = Tcs as Eqn 9 indicates. The reliability of this method increases with the increase in the difference between Tc and Tcs, so that a clear sky condition is ideal. Horie (1978) and Homma et al. (1999) reported that stomatal resistance in rice reached the minimum at a photosynthetically active photon flux density of 930 and 1265 µmol m−2 s−1. For the genotypic comparisons of estimated terms such as rc, the conditions of the light intensity above 1200 µmol m−2 s−1 were used in this study. Figure 5 shows an example of diurnal changes in energy balance terms observed on Takanari and NPT2 canopy surface. The figure shows that Ta increased about 1.8 °C from 9:00 to 15:00, and Tc also increased in both genotypes. The Tc of NPT2 was approximately 1.0 °C higher than that of Takanari in the daytime. It is of interest that both genotypes appeared to maintain each specific canopy-air temperature difference (Tc − Ta) throughout the daytime. Using Tc − Ta and other meteorological data, the present remote sensing method could successfully estimate ra between the atmosphere and canopy surface of both genotypes. The obtained ra was stable around 4.0 s m−1Through the daytime and lower than the estimated rc in both genotypes. The rc of NPT2 was twice as large as that of Takanari, and the rise of rc was observed in NPT2 late in the afternoon. The estimated E in Takanari was about 60% higher than that in NPT2 throughout the daytime (Fig. 5d). The lower Tc in Takanari canopy may have resulted from this higher E.
Table 1 shows the mean values of the estimated energy balance terms of rice genotypes obtained at photosynthetically active photon flux density above 1200 µmol m−2 s−1 from 9:00 to 15:00 during the 2-week period preceding full heading for 1999 and 2000. During the 2-week period, measurements were done more than 10 times in 1999 and 5 times in 2000 in each cultivar. The mean value of each experimental plot was calculated from those data at first, and then the mean value and standard error of each cultivar were determined for the two plot values. Tc − Ta was lowest in Kamenoo and Milyang 23 among the genotypes in 1999 and 2000, respectively. Milyang 23 and Takanari showed lower TcThan Ta across the two years. There was no significant genotypic difference in ra in either year. The ra ranged from 3.5 to 9.4 s m−1 among the genotypes and was clearly lower than the rc in each genotype. These results suggest that the influence of ra on heat and mass transfer was less than that of rc. The remote-sensing technique detected a significant difference in rc among the genotypes in both years. The lowest rc was observed in Takanari and Milyang 23 in 1999 and 2000, respectively. On the other hand, Nipponbare and Takenari showed the highest rc in 1999 and 2000, respectively. Table 1 also shows the canopy diffusive conductance obtained as the reciprocal of rc. As rc was significantly different among the genotypes, a genotypic difference in 1 / rc was also observed in both years. Takanari and Milyang 23 showed a significantly higher 1 / rcThan other genotypes in both years. The lowest 1 / rc was observed in NPT2 & Takenari in 1999 and 2000, respectively.
Correlation of 1 / rc with CGR and yield in different genotypes
A significant genotypic difference in CGR during the 2-week period preceding full heading and the close correlation between the CGR and grain yield were reported previously (Horie et al. 2003; Takai et al. 2005). These results suggested that the difference in CGR may be derived from that in canopy photosynthesis. Figure 6 shows the relationship between 1 / rc and CGR during the 2-week period preceding full heading in the two years. Interestingly, 1 / rc was significantly correlated with CGR in both years (r = 0.874, P < 0.05 for 1999; r = 0.824, P < 0.05 for 2000, respectively), and genotypes having higher 1 / rc had consistently higher CGR and vice versa. Furthermore, 1 / rc showed a significantly high correlation with grain yield in both years (r = 0.903, P < 0.05 for 1999; r = 0.868, P < 0.05 for 2000, respectively) (Fig. 7).
Relationship between 1 / rc measured by the remote-sensing method and gs measured by leaf gas exchange method
The values of 1 / rc estimated by the remote sensing technique in eight rice genotypes was compared with the gs of fully expanded leaves measured with a leaf gas exchange system (Fig. 3). Although the unit of each conductance was different, a significantly close relationship existed between 1 / rc and gs (r = 0.889, P < 0.01), and genotypes having higher 1 / rc had consistently higher gs. The present results suggest that the new remote sensing method can provide reasonable estimates for gas exchange characteristics of rice genotypes.
One of the problems in monitoring plant ecophysiological status based on the resistance-type energy balance model is how to estimate ra precisely under field conditions. In most cases, ra is estimated depending on measurements of wind profile above the canopy. This is often time-consuming and requires a greater fetch with a uniform canopy surface (Paw U et al. 1995). Furthermore, Tolk et al. (1995) observed the serious shortcoming that the correlation between wind velocity and measured ra disappeared under neutral (Ta = Tc) and unstable (Ta < Tc) atmosphere conditions although they were closely correlated under stable (Ta > Tc) conditions. These results imply the difficulty of estimating ra using anemometers under natural field conditions.
The present remote sensing technique enabled us to estimate ra within 1–3 min without measuring wind profile even on the small canopy surface where a sufficient fetch could not be expected. The estimated ra ranged between 3.5 and 9.4 s m−1 among the genotypes (Table 1). The ra is very sensitive to wind velocity, and thus it changes every moment. Therefore, even if ra varied by more than twofold among genotypes, no significant genotypic difference was observed in ra due to the large standard errors. These ra were slightly lower than those of 11.7 s m−1 estimated by Homma et al. (1999) at constant wind velocity of 0.71 m s−1 in a temperature gradient chamber. Since it was impossible to validate the present remote sensing model through the evaluation of ra, 1 / rc estimated from this model was compared with gs measured with a leaf gas exchange system. A linear relationship (r = 0.889, P < 0.01) did exist between them (Fig. 3). There is a report that gs calculated from the energy balance on a single leaf surface fairly corresponded with gs measured directly with a leaf porometer (Jones 1999). Although estimates of 1 / rc by the present remote sensing method were not on a single leaf but on a canopy surface, they were closely correlated with gs measured on fully expanded leaves. These results suggest that 1 / rc reflects well the stomatal status of leaves located at the top of canopy. Furthermore, rc estimated at a photosynthetically active photon flux density of above 1200 µmol m−2 s−1 ranged between 16.2 and 51.8 s m−1 among genotypes in the present study (Table 1). These rc were close to those of about 30 and 28–44 s m−1 measured on rice canopies under a strong solar radiation by Adachi et al. (1995) and Homma et al. (1999), respectively. These results support the validity of our new remote sensing technique for precise estimation of canopy physiological characteristics under field conditions. Moreover, the precise and rapid estimation may be useful for the evaluation of hundreds of genotypes or breeding lines.
The infrared thermometer detected significant genotypic difference in Tc − Ta during the 2-week period preceding full heading. Takanari and Milyang 23, having lower Tc − Ta, consistently showed significantly higher 1 / rcThan others in both years. Interestingly, 1 / rc was significantly correlated with CGR in both years although regression lines differed in the two years due to different genotypes used and the different environments (Fig. 6). Since 1 / rc reflects the stomatal status of leaves located at the top of the canopy, the linear relationship between 1 / rc and CGR implies that single-leaf photosynthetic rate (Pn) may also differ significantly among the genotypes. Indeed, a linear relationship between gs and Pn has been reported for several crops including rice (Yoshida & Coronel 1976; Wong, Cowan & Farquhar 1985; Kuroda & Kumura 1990). The linear relationship between gs and Pn on a single leaf may have caused the linear relationship between 1 / rc and CGR on the canopy during the 2-week period preceding full heading in this study. Thus, Takanari and Milyang 23 could achieve larger CGR than other genotypes by higher PnThrough the higher gs.
Since the CGR during the 2-week period preceding full heading was highly correlated with the final yield (Horie et al. 2003; Takai et al. 2005), 1 / rc was also significantly correlated with the yield (Fig. 7). This agrees well with the results on wheat (Fischer et al. 1998) and cotton (Lu et al. 1994) in which yields of different cultivars were highly correlated with their gs. Therefore, 1 / rc could be an effective criterion for the selection of high-yielding rice genotypes, and the remote sensing method we proposed here could be a powerful tool for rapid evaluation of 1 / rc under field conditions.
In addition to the effect of rc on rice productivity, the lower Tc − Ta may have other important effects on rice growth under field conditions. The lower Tc − Ta results from a higher EThat is prompted by a lower rc, provided that other things are equal (Jackson et al. 1981). The cooler leaves can prevent deleterious heat effects on growth and biomass production by high irradiance and temperatures (Lu et al. 1994). In addition, it can decrease canopy respiration rate since respiration rate is proportional to temperature (Amthor 1989). These may be other reasons why rice genotypes having higher 1 / rc had higher CGR during the 2-week period preceding full heading. There are some reports showing a close relationship between yield and canopy temperature depression (CTD, Ta − Tc) or Tc itself under hot and drought environments (Garrity & O’Toole 1995; Amani, Fischer & Reynolds 1996). These results suggest that Tc − Ta or CTD may also be an important physiological trait for higher CGR and final rice yield. In fact, CTD was used for the pre-screening of physiological potential prior to the execution of yield trials (Araus et al. 2002). However, many factors including vapour pressure deficit (VPD), Ta itself, Rn, and ra can influence CTD or Tc − Ta. In particular, CTD is dominantly sensitive to the change in VPD (Amani et al. 1996). Araus et al. (2002) claimed that CTD was useful for predicting differences in yield only under invariable conditions such as the absence of wind and clouds in hot and dry environments (i.e. high VPD). Recently, Tc in a set of wheat lines was simultaneously measured with an infrared thermometer mounted on a light aircraft flying at a height of 800 m above the plots (Reynolds, Rajaram & Sayre 1999). The simultaneous measurements of Tc among the genotypes can eliminate the meteorological influences caused by the time lag. The trial was quite successful in detecting the relationship between CTD and yield. Nevertheless, the use of an aircraft is definitely high-cost and impractical. On this account, 1 / rc estimated by the remote sensing technique proposed here would be a more promising criterion for the rapid selection of high-yielding plants under field conditions because meteorological factors such as VPD and ra are taken into consideration for the estimation.
In this paper, we described a new remote sensing technique for the evaluation of canopy ecophysiological status under field conditions developed on the basis of simultaneous measurements of sunlit and suddenly shaded canopy temperatures. The technique had the advantage of rapid estimation of ra over the traditional methods that require measurement of wind profiles. Estimates of 1 / rc by this method were closely related with gs measured with a portable gas exchange system in rice. This result confers the validity of the new method for quantitative estimation of rice canopy physiological characteristics under field conditions. Significant genotypic differences were obtained in Tc − Ta, rc and 1 / rc during the 2-week period preceding full heading in the two years studied, and 1 / rc was highly correlated with CGR during the period and in the final yield. We emphasize that 1 / rc could be an effective criterion for the selection of high-yielding rice genotypes, and the remote sensing technique proposed here can be a powerful tool for the rapid evaluation of 1 / rc in the hundreds of breeding lines under field conditions. Selection by 1 / rc may help improve the stagnant yield potential in rice.
We thank the staff of Laboratory of Crop Science, Graduate School of Agriculture, Kyoto University for their support in the experiments. The present work was supported by a Grant-in-Aid (Scientific Research B(2) 13556006) from the Ministry of Education, Culture, Sports, Science and Technology, Japan.