Genotype x Environmental Interactions of Selected Lines in Water Direct-seeded rice
Jong Gun Won, Tomohiko Yoshida and Yousuke Uchimura*
(Faculty of Agriculture, Kyushu University, Fukuoka 812-8581, Japan;
*Fukuoka Agricultural Research Center, Chikushino, 818-8548, Japan)
To determine the genotype x environment (G x E) interaction effects for yield and yield components and to determine yield adaptability in water direct-seeded rice, five genotypes, including 4 breeding lines and a check cultivar, were grown under different environments (location, plant density, nitrogen and year). There was no G x Y (year) interaction for yield and yield components because of no ranking changes among genotypes across years. The interactions of G x D (plant density) and G x N (nitrogen) were significant only in percentage of filled grains, indicating that the ranking of rice genotypes for yield traits except percentage of filled grains remained constant across plant densities and nitrogen levels. The interaction of G x L (location) was detected in all yield traits except panicles. The difference in adaptability among genotypes evaluated by a regression method was found and one line selected for direct-seeding had high average yield and higher yield in high-yielding conditions.
Key words : Adaptability, Direct seeding, Genotype x environment interaction, Rice.
Two or more genotypes grown in different environments may exhibit different relative performances. This phenomenon is called genotype x environment (G x E) interaction. Growers need cultivars that are dependable and consistent across a wide array of stress conditions yet have high yield potential that may be expressed when production conditions become more favorable (Peterson et al., 1997).
The value of a phenotype depends on the genotype and the environment and their interaction. If no genotype x environment (G x E) interactions is present, the average difference between genotypes evoked by phenotypes in different environments is consistent. Significant G x E interactions results from changes in the magnitude of the differences among genotypes in different environments or from changes in relative ranking of the genotypes (Allard and Braadshaw, 1964; Fernandez, 1991; Cornelius et al., 1993), i.e., (i) a non-crossover G x E interaction, in which case the ranking of genotypes remains constant across environments and the interaction is significant because of changes in the magnitude of the response, or (ii) a crossover G x E interaction, in which case a significant change in rank occurs from one environment to another. In the latter situation, one genotype may be chosen for one environment and a different genotype for another environment (Baker, 1988; Cornelius et al., 1993) When selecting genotypes for wide adaptation, plant breeders look for a non-crossover G x E interaction (Matus et al., 1997).
To evaluate the adaptability of genotypes, Finlay and Wilkinson (1963) reported that the regression technique was particularly effective in emphasizing the actual trend of varietal yield responses to a range of natural environments. Genotypes which are characterized by regression coefficients of the order of 1.0 have average stability over all environments. Regression coefficient greater than 1.0 means below average stability, therefore, the genotypes are sensitive to changes in the environment. Regression coefficient less than 1.0 means above average stability, therefore, the genotypes are very little change in yield despite large changes in the environment.
Several studies of G x E interaction for rice were reported in some countries (Mahajan and Rao, 1982; Brown et al., 1983; Redona and Sebstian, 1996). In Japan, Oosato et al. (1996) and Imabayashi et al. (1997 a, b) reported G x E interaction and adaptability of palatable rice cultivars. We have selected promising lines for direct seeding, reported previously (Won et al., 1998). For further evaluations, they must be tested under several environments. The objectives of this study are (i) to determine the relative importance of the G x E interaction effects for yield and yield components and (ii) to determine yield adaptability by the regression technique for lines selected for direct seeding.
MATERIALS AND METHODS
Five genotypes including four breeding rice lines, line 38, 42, 53 and 76 (Won et al., 1998), and a check cultivar, Hinohikari were used. Seeds were soaked in water at 25℃ for 2 days and the pre-sprouted seeds were directly hand-drill seeded in the plots withstanding water after the plots were thoroughly puddled and leveled. After seeding, the water was irrigated with 2-3 cm water depth continuously. At 10 days after seeding (DAS), the water was once drained for the root anchorage.
1. Genotype x year (G x Y) interaction
The five genotypes were seeded at the experimental field of Kyushu University (Fukuoka, lat. N 33.4, long. E 130.2) in early June in 1997 and early may in 1998. Each plot consisted of two 1m rows with a 25cm space between rows and 50 seeds were sown on each row. There were two replications. The seedlings were thinned to a space of 10cm between plants at 18 and 30 days after seeding (40 plants m-2). The fertilizer, 8, 8 and 8 g m-2 of N, P2O5 and K2O was applied. Basal fertilizer was applied at seeding at a rate of 50% of N, P2O5 and K2O, and top-dressed at a rate of 30% at tillering stage and 20% at panicle initiation stage, respectively.
2. Genotype x location (G x L) interaction
In 1998, these five genotypes were tested at the Fukuoka Agricultural Research Center (Chikushino, lat. N 33.2, long E 130.3). The plot consisted of eight 6 m rows with a 15 cm space between rows with two replications. The fertilizer, 10, 10 and 10 g m-2 of N, P2O5 and K2O was applied and the seeds were seeded on early June as the same method of Kyushu university. The plant density was adjusted to 80 plants m-2. The data of 80 plants m-2 in 1998 obtained in Kyushu Univ. (Fukuoka) described in the next 3 are combined to analyze the G x L interaction.
Palatability was evaluated for breeding lines harvested in Fukuoka by the sensory evaluation test using transplanted Koshihikari as a reference cultivar.
3. Genotype x plant density (G x D) and genotype x nitrogen (G x N) interaction
The five genotypes were seeded at the experimental field of Kyushu University in early May in 1998. Each plot consisted of two 1m rows with a 20cm space between rows and 50 seeds were sown on each row. At 20 DAS and at 30 DAS, the seedlings were thinned to 10 plants (40 plants m-2) and 20 plants (80 plants m-2) per 1 m row length. Two levels of fertilizer, 8, 8 and 8 g m-2 of N, P2O5 and K2O and 16, 16 and 16 g m-2 of N, P2O5 and K2O, were applied with two replications. Basal fertilizer was applied at seeding at a rate of 50% of N, P2O5 and K2O, and top-dressed at a rate of 30% at tillering stage and 20% at panicle initiation stage, respectively.
Prior to harvest, for all tests conducted, 10 hills in center of each row were counted for panicle number and were calculated in terms of square meter. At harvest, ten panicles in center area of each row were randomly chosen for each line, and after one week of drying, the spikelet number per panicle and the weight of panicle were measured. The grains were divided into filled grains and unfilled grains for measuring the percentage of grain filling. After the number of panicles per hill was recorded, ten hills per line were hand-harvested to determine the grain weight. 1000-grain weight was measured with head brown rice and grain weight was calculated in terms of square meter.
4. Yield adaptability
The data of these five genotypes grown under different environments (year, location, plant density and nitrogen) were used for adaptability analysis. Following the method of Finlay and Wilkinson (1963), a linear regression of individual yield on the mean yield of all genotypes over all environments were calculated for each genotype.
RESULTS AND DISCUSSION
1. Genotype x year interaction
Grain yield ranged from 440 (line 76 in 1997) to 740 g m-2 ( line 42 in 1998) (Fig. 1). Panicles ranged from 204 ( line 76 in 1997) to 366 m-2 (Hinohikari in 1998), spikelets ranged from 24578 (line 76 in 1997) to 38096 m-2 ( line 38 in 1998) and 1000-grain weight ranged from 20.3 (line 53 in 1997) to 25.3 g (line 42 in 1997). The years had a significant effect in panicles and 1000-grain weight (Table 1). The effect of genotypes were significant in panicles, spikelets, 1000-grain weight and grain yield. It means that the ranges of variation among genotypes for yield and yield components were wide. However the interaction of G x Y was not detected. For the panicles, 1000-grain weight and grain yield, the ranking among the genotypes was not changed between 1997 and 1998 (Fig. 1), indicating that all genotypes responded similarly. In case of spikelets, although there were some changes of ranking among genotypes across years, the effect of year was not significant and the interaction of G x Y was not detected. Therefore, analysis of these interactions suggests that there are no significant ranking changes and that the means are stable between years.
Often, because of the different weather conditions, significant G x Y interaction was detected for yield of many crops (Sayre et al., 1997; Imabayashi et al., 1997; Baba et al., 1998). During this study in 1997 and 1998, the weather condition was similar and there was not serious damage from typhoon or low temperature during the growing period. Additional tests in different years might be necessary.
2. Genotype x location interaction
Grain yield of two different locations ranged from 466 (line 76 in Fukuoka) to 760 g m-2 ( line 42 in Fukuoka) (Fig. 2.). Panicles ranged from 241 ( line 76 in Chikushino) to 376 m-2 (Hinohikari in Fukuoka), spikelets ranged from 24550 (line 76 in Chikushino) to 36525 m-2 ( line 38 in Chikushino), 1000-grain weight ranged from 19.6 (line 53 in Fukuoka) to 26.2 g (line 42 in Chikushino) and percentage of filled grain ranged from 70.6 (line 53 in Fukuoka) to 93.9 % (line 53 in Chikushino) (Fig. 2.). The effect of genotype was significant for all yield components and yield (Table 2). The effect of location was significant for 1000-grain weight and percentage of filled grains but not for grain yield. The interaction of G x L was detected for yield and all yield components except panicles. Fig. 2. showed that there were significant changes of the ranking among genotypes from Fukuoka to Chikushino for spikelets, percentage of filled grain and grain yield whereas there was no rank-change in panicles. Therefore, the interactions of G x L were caused not by scale-changes but by the rank-changes among genotypes from one location to another location. These rank-changes suggests a need for testing of yield traits at each specific location.
3. Genotype x plant density and genotype x nitrogen interaction
Genotypes differed significantly (p < 0.01) for number of panicles, spikelets, percentage of filled grains, 1000-grain weight and yield (Table 3). Number of panicles and percentage of filled grains exhibited significant differences by different plant densities. Number of panicles, spikelets, 1000-grain weight and yield were significantly different by the different nitrogen levels.
The interactions of G x D and G x N were significant only in percentage of filled grains. Especially line 53 changed the ranking among genotypes for percentage of filled grains from low plant density to high plant density and low nitrogen rate to high nitrogen level (Fig. 3.). However, the G x D and G x N interactions for other yield traits were not significant, indicating that the ranking of rice genotypes for yield traits remained constant across plant densities and nitrogen levels. G x D and G x N interactions in yield were not significant.
In present study, among the environments tested, the interaction of G x E was not detected for year, plant density and nitrogen but detected for the interaction of location in almost all yield traits, showing the importance of yield testing changing the locations.
Across all the environments, some lines did not change the high rank for special yield traits. For example, Hinohikari for panicles, line 38 for spikelets, line 42 for 1000-grain weight and line 38 and 42 for yield, suggesting that these lines may be somewhat stable for these yield traits across the environments.
4. Yield adaptability
The regression lines obtained by the method of Finlay and Wilkinson (1963) for yield of five genotypes are shown in Fig. 4. Genotypes with regression coefficients of less than 1.0 are insensitive to over all environments. Line 38, line 53 and hinohikari had regression coefficient of 0.62, 0.84 and 0.59, respectively, showing little change in yield against changes of environment. They produced relatively above-average yield in low-yielding environments but produced relatively lower yield in high-yielding environments. The order of their yield was line 38 > Hinohikari > line 53. Therefore, line 38 produced relatively higher yield and was not sensitive to environments. Hinohikari and line 53 produced lower yields and were not sensitive to environments. On the contrary, line 42 and line 76 were relatively sensitive to changes in the environments. Line 76 produced below-average yields and the response to environments was relatively high. Line 42 produced higher mean yield than others, and as the environment improved it produced more higher yields. Therefore line 42 can be described that it has general adaptability and is better adapted to high-yielding environments with a regression coefficient greater than 1.0 (line 42 = 1.49). It clearly shows that lines for water direct-seeding with high yield and adaptability can or must be selected under the water direct-seeding condition.
Table 4 showed the means of yield components, yield and palatability across all environments. 1000-grain weight and yield of line 42 was heavier or higher than other genotypes. The palatability of line 42 was similar to Koshihikari, showing that line 42 could be a very promising line for water direct-seeding or could be used as a cross parent of the breeding program.
In conclusion, yield difference among genotypes changed by locations, lines selected for direct seeding showed different adaptability and line 42 had high yield and produced higher yield than average in a high-yielding condition.
Allard, R.W. and Bradshaw, A.D. 1964. Implications of genotype-environment
interactions in applied plant breeding. Crop Sci. 4:503-508.
Baba, T., Yamaguchi, O. and Furusho, M. 1998. Genotype x year interaction in yield
and grain quality of malting barley. Jpn. J. Crop Sci. 67:510-515*.
Baker, R.J. 1988. Test for crossover genotype-environmental interaction. Can. J. Plant
Brown, K.D., Sorrells, M.E. and Coffman, W.R. 1983. A method for classification and
evaluation of testion environments (Rice cultivar, genotype x environment
interaction). Crop Sci. 23:889-893.
Cornelius, P.L., Van Sanford, D.A. and Seyedsadr, M.S. 1993. Clustering cultivars into
groups without rank-change interactions. Crop Sci. 33:1193-1200.
Fernandez, G.C.G. 1991. Analysis of genotype x environment interaction by stability
estimates. Hort Science 26:947-950.
Finlay, K.W. and Wilkinson, G.N. 1963. The analysis of adaptation a plant-breeding
programme. Aust. J. Agric. Res. 14:742-754.
Imabayashi, S., Matsue, Y., Hamachi, Y. and Yoshida, T. 1997a. Genotype x
environment interaction for early-maturing rice cultivars with high palatability in
northern Kyushu. I. Yield. Jap. J. Crop Sci. 66:538-544*.
Imabayashi, S., Matsue, Y., Hamachi, Y. and Yoshida, T. 1997b. Genotype x
environment interaction for early-maturing rice cultivars with high palatability in
northern Kyushu. II. Heading date and culm length. Jap. J. Crop Sci. 66:545-550*.
Mahajan, R.K. and Rao, A.V. 1982. Regression analysis of genotype x environment
interaction in rice. Indian J. Acri. Sci. 52:14-16.
Matus, A., Slinkard, A.E. and Kessel, C. van. 1997. Genotype x environment
interaction for carbon isotope discrimination in spring wheat. Crop Sci. 37:97-102.
Oosato, K.F., Hamachi, Y., Matsue, Y. and Yoshida, T. 1996. Genotype x environment
interaction of palatability in rice. Jap. J. Crop Sci. 65:585-589*.
Peterson, C.J., Moffatt, J.M. and Erickson, J.R. 1997. Yield stability of hybrid vs.
pureline hard winter wheats in regional performance trials. Crop Sci. 37:116-120.
Redona, E.D. and Sebastian, L.S. 1996. Genotype x environment interaction analysis
in national irrigated lowland rice yield trials. Philippine J. Crop Sci. 21:3.
Sayre, K.D., Rajaram, S. and Fischer, R.A. 1997. Yield potential progress in short
wheats in Northwest Mexico. Crop Sci. 37:36-42
Won, J.G., Hirahara, Y., Yoshida, T. and Imabayashi, S. 1998. Selection of rice lines
using SPGP seedling method for direct seeding. Plant Prod. Sci. 1:269-279.
* In Japanese with English summary.
Table 1. Analysis of variance for genotype (G) x year (Y) interaction in water-seeded rice.
Source d.f Panicles m-2 Spikelets m-2 1000-G weight (g) Grain Yield(g m-2)
Treatment 9 4428.3* 25479778 7.88** 22015*
Year(Y) 1 8946.5* 20214572 3.87** 8201
Genotype(G) 4 7405.1** 35391534* 16.50** 45667**
Y x G 4 322.1 16884323 0.26 1817
Error 9 378.6 8277746 0.23 3914
*,** Significant at the 5 % and 1 % level of probability, respectively.
Table 2. Analysis of variance for genotype (G) x location (L) interaction in water-seeded rice.
Source d.f Panicles m-2 Spikelets m-2 1000-G weight (g) Percentage of filled grain(%) Grain Yield (g m-2)
Treatment 9 3917.7* 32801624** 10.85** 132.2** 14099**
Location(L) 1 4.1 4292 5.73** 716.4** 1620
Genotype(G) 4 7589.9** 55256931** 22.54** 72.5* 22439**
L x G 4 1223.9 18545650* 0.45* 45.8* 8979*
Error 9 847.7 4273694 0.11 11.7 2167
*,** Significant at the 5 % and 1 % level of probability, respectively.
Table 3. Analysis of variance for genotype (G) x plant density (D) and genotype (G) x
nitrogen level (N) interaction in water-seeded rice.
Source df Panicles m-2 Spikelets m-2 Percentage of grain filling 1000-G weight (g) Grain yield (g m-2)
Rep (R) 1 1650 1 1.54 0.10 297
Treatment 27 6120** 134170954** 31.27** 6.65** 26148**
Plant density (D) 1 17573** 561922 116.29** 0.18 4375
Nitrogen (N) 1 62578** 2405265573** 1.06 1.51** 93759**
Genotype (G) 6 12445** 167890611** 49.87** 28.71** 83996**
D x N 1 4681 31841722 32.25* 1.45** 2517
D x G 6 345 6350265 16.34* 0.26 2598
N x G 6 244 16217230 28.78** 0.23 6407
Error 27 1179 25164141 6.29 0.12 5443
*,** Significant at the 5% and 1% level of probability, respectively.
Table 4. Mean yield and yield components of five genotypes across all environment.
Genotypes Panicles m-2 Spikelets m-2 1000-G weight(g) % of filled grains Yield (g m-2) Palatability2)
Line 38 320 ab1) 37047 a 20.7 c 79.1 b 605 b - 2.44
Line 42 300 b 33303 ab 25.3 a 81.0 b 744 a - 0.12
Line 53 267 b 30897 ab 19.7 d 82.2 ab 530 c 0.19
Line 76 259 b 29468 ab 21.0 c 85.7 a 501 c - 2.38
Hinohikari 364 a 32205 ab 22.4 b 84.8 a 557 bc -
Mean 302 32584 21.8 82.6 587 -
1) Means followed by the same letter are not significantly different at 0.05 level according to
Duncan`s multiple range test.
2) The reference cultivar for the sensory test was Koshihikari.
Fig. 1. Genotype x year interaction for panicles, spikelets, 1000-grain weight and grain yield from
1997 to 1998. (G1: line 38, G2: line 42, G3: line 53, G4: line 76 and G5: Hinohikari).
Fig. 2. Genotype x location interaction for panicles, spikelets, percentage of filled grains and grain
yield at Fukuoka and Chikushino in 1998 (G1: line 38, G2: line 42, G3: line 53, G4: line 76 and
G5: Hinohikari; L1: Fukuoka, L2: Chikushino).
Fig.3. Genotype x plant density and genotype x nitrogen interactions percentage of filled grains
and grain yield in 1998 (G1: line 38, G2: line 42, G3: line 53, G4: line 76 and G5: Hinohikari;
D1: 40 plants m-2, D2: 80 plants m-2; N1: 8 g m-2, N2: 16 g m-2).
Fig. 4. Regression lines and population mean yield of five genotypes of rice grown under different
environmental conditions (G1: line 38, G2: line 42, G3: line 53, G4: line 76 and G5: Hinohikari).