Directory UMM :Data Elmu:jurnal:S:Scientia Horticulturae:Vol85.Issue1-2.July2000:
Scientia Horticulturae 85 (2000) 75±83
Optimal spatial and temporal measurement repetition
for reducing environmental variation of berry
traits in grape breeding
Akihiko Sato*, Masahiko Yamada, Hiroshi Iwanami,
Nobuyuki Hirakawa1
Persimmon and Grape Research Center, National Institute of Fruit Tree Science,
Ministry of Agriculture, Forestry and Fisheries, Akitsu, Hiroshima, 729-2494 Japan
Accepted 8 November 1999
Abstract
Environmental variance components were estimated for berry ripening time (BRT), berry weight
(BW), soluble solids concentration (SSC), titratable acidity (TA), and deformation at the ®rst major
peak (DFP) and maximum force (MF) in penetration tests of berry ¯esh in grapes (Vitis vinifera L.
and V. labruscana Bailey). The variance among berries within clusters was largest among
environmental variance components for BW, SSC, DFP and MF. The variance among clusters
within vines was smaller than the variance among berries within clusters for all traits. The sum of
the variance among years and the variance associated with the genotype year interaction was
generally larger than that among vines within genotypes. Consequently, increasing the number of
yearly repetitions was more ef®cient than increasing vine replications in evaluating the genetic
potential in breeding. # 2000 Elsevier Science B.V. All rights reserved.
Keywords: Vitis vinifera; Vitis labruscana; Fruit breeding; Environmental variation
1. Introduction
Fruit breeding requires testing of many seedlings to increase the probability of
selecting superior genotypes. However, the number of seedlings used in fruit tree
*
Corresponding author. Tel.: 81-846-45-1260; fax: 81-846-45-5370.
E-mail address: [email protected] (A. Sato)
1
Present address: Fukuoka Agricultural Research Center, Chikushino, Fukuoka, 818-0011 Japan.
0304-4238/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 4 - 4 2 3 8 ( 9 9 ) 0 0 1 4 4 - 2
76
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
breeding is restricted compared to annual crop breeding, because perennial fruit
crops occupy a large area and have a long juvenile phase. Early evaluation and
selection allow an ef®cient use of testing facilities, and enable breeders to test a
large number of seedlings in a given period.
Most commercially important traits of grapes, such as berry ripening time
(BRT), berry weight (BW), soluble solids concentration (SSC), titratable acidity
(TA), and berry ¯esh texture are quantitative, and ¯uctuate depending on
environmental factors. Therefore, it is important for grape breeders to obtain
information on the contribution of genetic and environmental factors to
phenotypic expression as reported for other fruit crops (Hansche and Brooks,
1965; Hansche and Beres, 1966; Machida and Kozaki, 1975; Kester et al., 1977;
Yamada et al., 1993).
Increasing the number of vine replications, yearly repetitions, cluster and berry
samples, increase the precision of discriminating genetic properties. Repeating
measurements too much, however, decreases the number of offspring evaluated in
a given period. Estimates of genetic and environmental variance components for
traits provide information on the optimal sample size, yearly repetition, and vine
replication. Estimated values of environmental variance components have been
utilized to predict accurately the proportion of genotypes exceeding a critical
value in an offspring population (Yamada et al., 1994a,b, 1995, 1997; Yamada
and Yamane, 1997).
Generally, adding more locations, years, or both have been reported to be more
ef®cient for estimating genetic properties than adding replications (Rasmusson
and Lambert, 1961; Kaltsikes, 1970; Sekioka and Lauer, 1970; Patterson et al.,
1977; Shorter and Norman, 1983; Swallow and Wehner, 1989). Yamada et al.
(1993) reported that a broad-sense heritability for fruit ripening time, fruit weight,
and SSC in Japanese persimmon (Diospyros kaki Thunb.) was increased more by
adding yearly repetition than adding tree replication.
Nesbitt and Kirk (1972) also showed that adding plot replication was more
ef®cient than increasing the plot size to estimate the yield of muscadine grape.
The variance component among vines within genotypes was generally larger than
that within vines for SSC in grape (Rankine et al., 1962; Wolpert et al., 1980;
Hagiwara et al., 1987). However, no information on the year effect, genotype
year interaction, and vine year interaction was provided in grapes.
A grape breeding project has been carried out at the Persimmon and Grape
Research Center (PGRC) of the National Institute of Fruit Tree Science (NIFTS),
Akitsu, Hiroshima, Japan, since 1968. Environmental variance normally ¯uctuate,
depending on location, climate, and cultural management. Therefore, it is
important to obtain information on the environmental variances under our
cultural management and climate condition at Akitsu. The objectives of this study
were: (1) to estimate environmental variance components speci®c to grape
breeding at Akitsu and (2) to determine the effect of adding yearly repetition
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
77
and adding vine replication for reducing the environmental variation of six berry
traits in grapes.
2. Materials and methods
This study was conducted at the PGRC of the NIFTS, using genotypes of seven
diploid (`July Muscat', `Hiro Humburg', `Italia', `North Red', `Buffalo',
`Steuben', and `Muscat Bailey A'), and two tetraploid (`Ryuho' and `Fujiminori)'
cultivars of V. vinifera L. and V. labruscana Bailey. The genotypes evaluated were
used as cross-parents in a grape breeding program at the PGRC. Current-scion of
these genotypes were grafted onto `Teleki 5BB' rootstock in May or June of
1992. Two vines per genotype were planted randomly into an offspring
population grown in the same way in a test ®eld in June 1992, at 0.6 3.2 m
spacing. Four year-old plants were evaluated for berry traits in 1995. Berry traits
were evaluated for 2 years (1995 and 1996).
Four clusters were sampled per vine, using two vines for each of nine
genotypes. Each cluster on a vine was harvested at its respective BRT. BRT was
judged when berries had fully ripened and reached the best eating quality, and
was rated on a scale of 1±8; a rating of 1 (late July), 2 (early August), 3 (midAugust), 4 (late August), 5 (early September), 6 (mid-September), 7 (late
September), and 8 (early October). BW, SSC, TA, and the deformation at the ®rst
major peak (DFP) and the maximum force (MF) as the textural properties in the
berry penetration test (Sato et al., 1997) were also evaluated. SSC and TA were
determined with a calibrated refractometer (model N1; Atago, Tokyo, Japan) and
continuous titrator (model BK-D; Mitamura Riken Kogyo, Tokyo, Japan),
respectively. TA was expressed as grams of tartaric acid in 100 ml of juice.
There are high correlation coef®cients between the sensory rating of dif®culty
of breakdown in mastication and DFP, and between the sensory ¯esh ®rmness and
MF (Sato et al., 1997). To measure DFP and MF, an 8 mm thick ¯esh section was
cut longitudinally from each berry and was subjected to a penetration test. The
sample was mounted on the stage of the rheometer (model NRM-2010J-CW;
Fudoh, Tokyo, Japan). A 3 mm diameter plunger was used at a penetrating rate of
50.0 mm minÿ1 according to Sato et al. (1997). The force for compression and
the distance from the surface of the sample was recorded as the force±
deformation curve to an X±Y recorder. The value of DFP and MF were obtained
from the force-deformation curve.
The measurements from three berries taken randomly from each of four
clusters per vine were subjected to analysis of variance (ANOVA) (Table 1).
Before statistical analysis, the values of BW, DFP, and MF were log-transformed
to improve the normality of the distribution of the residual estimates. The
distribution of the residual was not signi®cant at P 0.05 with a Kolmogorov±
78
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
Table 1
Expected mean squares in analysis of variance using nine genotypes with two vines per genotype
for 2 years
Source of variation
DF
Genotype
8
Year
1
Genotype year
8
Among vines within genotypes
9
Vine year
9
Among clusters within vines
108
Among berries within clusters
288
Expected mean squares
s2b 3s2c 12s2vy 24s2gy 24s2v 48s2g
s2b 3s2c 12s2vy 24s2gy 216s2y
s2b 3s2c 12s2vy 24s2gy
s2b 3s2c 12s2vy 24s2v
s2b 3s2c 12s2vy
s2b 3s2c
s2b
Smirnov one sample test, and approached a normal distribution for all traits.
Therefore, the model of ANOVA was assumed to be applicable to the data.
The model adopted here to express the phenotypic value is Pijklm m gi
vij cijkm bijklm ym (gy)im (vy)ijm, where Pijklm is the phenotypic value of
the lth berry of the kth cluster of the jth vine of the ith genotype in the mth year, m
the overall mean, gi a random effect contributed by the ith genotype, vij a random
effect of the jth vine of the ith genotype, ym a random effect of the mth year, cijkm
a random effect of kth cluster of the jth vine of the ith genotype in the mth year,
bijklm a random effect of the lth berry of the kth cluster of the jth vine of the ith
genotype in the mth year, (gy)im the interaction between ith genotype and the mth
year, and (vy)ijm the interaction between the jth vine of the ith genotype and the
mth year.
The ANOVA provided the variance associated with genotype (s2g ) among vines
within genotypes (s2v ), among clusters within vines (s2c ), among berries within
clusters (s2b ), among years (s2y ), the genotype year interaction (s2gy ), and the
vine year interaction (s2vy ). The sum of these variance components was
regarded as the total variance (s2T ). The total environmental variance (s2E ) can be
expressed as s2E s2y =y s2gy =y s2v =v s2vy = yv s2c = yvc s2b = yvcb,
where y is the number of yearly repetitions, v the number of vine replications
per genotype, c the number of cluster samples per year and vine, b the number of
berries sampled per cluster, respectively.
3. Results and discussion
The mean performance of the evaluated genotypes was 5.2 for BRT, 6.4 g for
BW, 20.3% for SSC, 0.48 g100 mlÿ1 for TA, 4.09 10ÿ3 m for DFP, and
0.75 N for MF, respectively (Table 2).
Table 2
Grape cultivars evaluated and their performance in berry traitsa
Cultivars
BRT
BW (g)
b
V. vinifera
July Muscat
Hiro Humburg
Italia
4.0 a
5.5 bc
5.8 cd
V. labruscana
North Red
Buffalo
Steuben
Muscat Bailey A
Ryuho
Fujiminori
4.0
4.5
5.6
6.9
5.3
5.0
Mean
a
5.2
a
ab
bc
d
bc
a±c
5.2 b
5.1 b
8.1 c
4.5
3.5
4.6
4.2
9.5
13.0
6.4
b
a
b
ab
c
d
SSC (%)
TA (g100 mlÿ1)
Texture
DFP 10ÿ3 m
MF (N)
20.7 bc
22.3 c
19.3 ab
0.32 a
0.36 ab
0.45 a±c
2.18 a
2.90 b
1.76 a
0.861 cd
0.547 ab
0.555 ab
20.4
22.3
20.3
20.5
l8.0
18.7
0.46
0.65
0.50
0.64
0.56
0.39
4.51
6.10
5.97
5.40
5.04
2.92
0.441
1.047
0.838
1.269
0.699
0.494
20.3
bc
c
bc
bc
a
ab
0.48
a±c
e
b±e
de
c±e
a±c
4.09
c
d
d
cd
cd
b
a
cd
c
d
bc
a
0.750
The abbreviations used for the column heading have already been de®ned in text.
Mean separation
p within columns by LSD test at P < 0.05. The log-transformed values were used in ANOVA for BW, DFP and MF. LSD was
calculated as 1:96 2sE , where s2E s2y =2 s2gy =2 s2v =2 s2vy =4 s2c =16 s2b =48.
b
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
Species
79
80
Source of variation
Genotype
Year
Genotype year
Among vines within genotypes
Vine year
Among clusters within vines
Among berries within clusters
a
Mean squares (MS)b
BRT 10
BW 102
SSC
TA 102
DFP 102
MF 102
422.61**b
222.32*
40.70*
30.21NS
10.23**
0.71**
0.44
168.86**
17.43*
3.15NS
3.88NS
2.51**
0.74**
0.40
102.66*
35.31NS
19.88NS
14.23NS
6.84**
1.80**
0.60
67.03**
5.92NS
6.93NS
9.42NS
4.54**
0.72**
0.29
191.75**
20.78NS
5.46NS
1.77NS
1.96NS
1.55NS
1.12
117.23**
36.18NS
13.67*
6.34NS
3.83*
1.66NS
1.75
The abbreviations used for the column heading have already been de®ned in text. BW, DFP and MF were log-transformed.
NS, * and ** referred to nonsigni®cant, signi®cant at P < 0.05 and 0.01, respectively, using F-test. F-value for the effect of genotype was
calculated as F (MS for genotype MS for vine year)/(MS for among vines MS for genotype year), according to Snedecor and Cochran
(1972).
b
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
Table 3
Mean square for six berry traits evaluated using nine genotypes with two vines per genotype for 2 yearsa
81
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
Table 4
Estimates of variance components and their percentage to the total variance obtained from the
analysis of variance for berry traits using nine genotypes with two vines per genotype for 2 yearsa
Variance
BRT 10
components
BW 102
SSC
TA 102
DFP 102
MF 102
s2g
s2y
s2gy
s2v
s2vy
s2c
s2b
3.42(80.7%)
0.07(1.7%)
0.03(0.7%)
0.06(1.4%)
0.15(3.5%)
0.11(2.6%)
0.40(9.4%)
1.57(40.1%)
0.07(1.8%)
0.54(13.9%)
0.31(7.9%)
0.42(10.7%)
0.40(10.2%)
0.60(15.4%)
1.15(52.3%)
0b(0%)
0.10(4.5%)
0.20(9.1%)
0.32(14.5%)
0.14(6.4%)
0.29(13.2%)
3.00(66.5%)
0.07(1.6%)
0.15(3.3%)
0b(0%)
0.03(0.7%)
0.14(3.1%)
1.12(24.8%)
2.11(45.4%)
0.10(2.2%)
0.41(8.8%)
0.10(2.2%)
0.18(3.9%)
0b(0%)
1.75(37.5%)
4.24(100%) 3.91(100%) 2.20(100%) 4.51(100%)
4.65(100%)
7.54(63.9%)
0.84(7.1%)
1.27(10.8%)
0.83(7.1%)
0.79(6.7%)
0.09(0.8%)
0.44(3.6%)
s2T
11.80(100%)
a
The abbreviations used for the column heading have already been de®ned in text. BW, DFP
and MF were log-transformed.
b
Negative values were assumed to be zero.
The result of ANOVA showed that the effect of genotype was signi®cant at
P < 0.05 for SSC and at P < 0.01 for the other traits (Table 3). The effect of the
vines within genotypes was nonsigni®cant for all traits.
The contribution of the environmental variance components to the total
variance varied with the traits (Table 4). For BRT, the variances associated with
year and vine (s2y ; s2gy ; s2v and s2vy ) were larger than s2c and s2b . For BW, as the
environmental variance components were generally small, it was easy to identify
the genetic property. The variance among berries within clusters was the largest
among the environmental variance components for all traits with the exception of
BRT. Especially, the s2b of DFP and MF were very large, indicating that
increasing the number of berries evaluated per cluster increases the measurement
accuracy effectively.
The cost and time to increase berry and cluster samples are generally smaller
than yearly repetitions or vine replications. So, ®rst choice of grape breeders
should be to measure more berries by increasing cluster or berry samples. Yearly
repetitions or vine replications, however, take much cost or time. So, comparison
between yearly repetitions and vine replications should be considered primarily.
To compare the in¯uence of adding yearly repetitions and adding vine
replications on the decrease of the s2E , the ratio of s2v = s2y s2gy shows the
ef®ciency of the two factors in decreasing the s2E , because the s2vy ; s2c and s2b
decrease equally both by adding years and vine replications. Our results showed
that the ratio was
Optimal spatial and temporal measurement repetition
for reducing environmental variation of berry
traits in grape breeding
Akihiko Sato*, Masahiko Yamada, Hiroshi Iwanami,
Nobuyuki Hirakawa1
Persimmon and Grape Research Center, National Institute of Fruit Tree Science,
Ministry of Agriculture, Forestry and Fisheries, Akitsu, Hiroshima, 729-2494 Japan
Accepted 8 November 1999
Abstract
Environmental variance components were estimated for berry ripening time (BRT), berry weight
(BW), soluble solids concentration (SSC), titratable acidity (TA), and deformation at the ®rst major
peak (DFP) and maximum force (MF) in penetration tests of berry ¯esh in grapes (Vitis vinifera L.
and V. labruscana Bailey). The variance among berries within clusters was largest among
environmental variance components for BW, SSC, DFP and MF. The variance among clusters
within vines was smaller than the variance among berries within clusters for all traits. The sum of
the variance among years and the variance associated with the genotype year interaction was
generally larger than that among vines within genotypes. Consequently, increasing the number of
yearly repetitions was more ef®cient than increasing vine replications in evaluating the genetic
potential in breeding. # 2000 Elsevier Science B.V. All rights reserved.
Keywords: Vitis vinifera; Vitis labruscana; Fruit breeding; Environmental variation
1. Introduction
Fruit breeding requires testing of many seedlings to increase the probability of
selecting superior genotypes. However, the number of seedlings used in fruit tree
*
Corresponding author. Tel.: 81-846-45-1260; fax: 81-846-45-5370.
E-mail address: [email protected] (A. Sato)
1
Present address: Fukuoka Agricultural Research Center, Chikushino, Fukuoka, 818-0011 Japan.
0304-4238/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 4 - 4 2 3 8 ( 9 9 ) 0 0 1 4 4 - 2
76
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
breeding is restricted compared to annual crop breeding, because perennial fruit
crops occupy a large area and have a long juvenile phase. Early evaluation and
selection allow an ef®cient use of testing facilities, and enable breeders to test a
large number of seedlings in a given period.
Most commercially important traits of grapes, such as berry ripening time
(BRT), berry weight (BW), soluble solids concentration (SSC), titratable acidity
(TA), and berry ¯esh texture are quantitative, and ¯uctuate depending on
environmental factors. Therefore, it is important for grape breeders to obtain
information on the contribution of genetic and environmental factors to
phenotypic expression as reported for other fruit crops (Hansche and Brooks,
1965; Hansche and Beres, 1966; Machida and Kozaki, 1975; Kester et al., 1977;
Yamada et al., 1993).
Increasing the number of vine replications, yearly repetitions, cluster and berry
samples, increase the precision of discriminating genetic properties. Repeating
measurements too much, however, decreases the number of offspring evaluated in
a given period. Estimates of genetic and environmental variance components for
traits provide information on the optimal sample size, yearly repetition, and vine
replication. Estimated values of environmental variance components have been
utilized to predict accurately the proportion of genotypes exceeding a critical
value in an offspring population (Yamada et al., 1994a,b, 1995, 1997; Yamada
and Yamane, 1997).
Generally, adding more locations, years, or both have been reported to be more
ef®cient for estimating genetic properties than adding replications (Rasmusson
and Lambert, 1961; Kaltsikes, 1970; Sekioka and Lauer, 1970; Patterson et al.,
1977; Shorter and Norman, 1983; Swallow and Wehner, 1989). Yamada et al.
(1993) reported that a broad-sense heritability for fruit ripening time, fruit weight,
and SSC in Japanese persimmon (Diospyros kaki Thunb.) was increased more by
adding yearly repetition than adding tree replication.
Nesbitt and Kirk (1972) also showed that adding plot replication was more
ef®cient than increasing the plot size to estimate the yield of muscadine grape.
The variance component among vines within genotypes was generally larger than
that within vines for SSC in grape (Rankine et al., 1962; Wolpert et al., 1980;
Hagiwara et al., 1987). However, no information on the year effect, genotype
year interaction, and vine year interaction was provided in grapes.
A grape breeding project has been carried out at the Persimmon and Grape
Research Center (PGRC) of the National Institute of Fruit Tree Science (NIFTS),
Akitsu, Hiroshima, Japan, since 1968. Environmental variance normally ¯uctuate,
depending on location, climate, and cultural management. Therefore, it is
important to obtain information on the environmental variances under our
cultural management and climate condition at Akitsu. The objectives of this study
were: (1) to estimate environmental variance components speci®c to grape
breeding at Akitsu and (2) to determine the effect of adding yearly repetition
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
77
and adding vine replication for reducing the environmental variation of six berry
traits in grapes.
2. Materials and methods
This study was conducted at the PGRC of the NIFTS, using genotypes of seven
diploid (`July Muscat', `Hiro Humburg', `Italia', `North Red', `Buffalo',
`Steuben', and `Muscat Bailey A'), and two tetraploid (`Ryuho' and `Fujiminori)'
cultivars of V. vinifera L. and V. labruscana Bailey. The genotypes evaluated were
used as cross-parents in a grape breeding program at the PGRC. Current-scion of
these genotypes were grafted onto `Teleki 5BB' rootstock in May or June of
1992. Two vines per genotype were planted randomly into an offspring
population grown in the same way in a test ®eld in June 1992, at 0.6 3.2 m
spacing. Four year-old plants were evaluated for berry traits in 1995. Berry traits
were evaluated for 2 years (1995 and 1996).
Four clusters were sampled per vine, using two vines for each of nine
genotypes. Each cluster on a vine was harvested at its respective BRT. BRT was
judged when berries had fully ripened and reached the best eating quality, and
was rated on a scale of 1±8; a rating of 1 (late July), 2 (early August), 3 (midAugust), 4 (late August), 5 (early September), 6 (mid-September), 7 (late
September), and 8 (early October). BW, SSC, TA, and the deformation at the ®rst
major peak (DFP) and the maximum force (MF) as the textural properties in the
berry penetration test (Sato et al., 1997) were also evaluated. SSC and TA were
determined with a calibrated refractometer (model N1; Atago, Tokyo, Japan) and
continuous titrator (model BK-D; Mitamura Riken Kogyo, Tokyo, Japan),
respectively. TA was expressed as grams of tartaric acid in 100 ml of juice.
There are high correlation coef®cients between the sensory rating of dif®culty
of breakdown in mastication and DFP, and between the sensory ¯esh ®rmness and
MF (Sato et al., 1997). To measure DFP and MF, an 8 mm thick ¯esh section was
cut longitudinally from each berry and was subjected to a penetration test. The
sample was mounted on the stage of the rheometer (model NRM-2010J-CW;
Fudoh, Tokyo, Japan). A 3 mm diameter plunger was used at a penetrating rate of
50.0 mm minÿ1 according to Sato et al. (1997). The force for compression and
the distance from the surface of the sample was recorded as the force±
deformation curve to an X±Y recorder. The value of DFP and MF were obtained
from the force-deformation curve.
The measurements from three berries taken randomly from each of four
clusters per vine were subjected to analysis of variance (ANOVA) (Table 1).
Before statistical analysis, the values of BW, DFP, and MF were log-transformed
to improve the normality of the distribution of the residual estimates. The
distribution of the residual was not signi®cant at P 0.05 with a Kolmogorov±
78
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
Table 1
Expected mean squares in analysis of variance using nine genotypes with two vines per genotype
for 2 years
Source of variation
DF
Genotype
8
Year
1
Genotype year
8
Among vines within genotypes
9
Vine year
9
Among clusters within vines
108
Among berries within clusters
288
Expected mean squares
s2b 3s2c 12s2vy 24s2gy 24s2v 48s2g
s2b 3s2c 12s2vy 24s2gy 216s2y
s2b 3s2c 12s2vy 24s2gy
s2b 3s2c 12s2vy 24s2v
s2b 3s2c 12s2vy
s2b 3s2c
s2b
Smirnov one sample test, and approached a normal distribution for all traits.
Therefore, the model of ANOVA was assumed to be applicable to the data.
The model adopted here to express the phenotypic value is Pijklm m gi
vij cijkm bijklm ym (gy)im (vy)ijm, where Pijklm is the phenotypic value of
the lth berry of the kth cluster of the jth vine of the ith genotype in the mth year, m
the overall mean, gi a random effect contributed by the ith genotype, vij a random
effect of the jth vine of the ith genotype, ym a random effect of the mth year, cijkm
a random effect of kth cluster of the jth vine of the ith genotype in the mth year,
bijklm a random effect of the lth berry of the kth cluster of the jth vine of the ith
genotype in the mth year, (gy)im the interaction between ith genotype and the mth
year, and (vy)ijm the interaction between the jth vine of the ith genotype and the
mth year.
The ANOVA provided the variance associated with genotype (s2g ) among vines
within genotypes (s2v ), among clusters within vines (s2c ), among berries within
clusters (s2b ), among years (s2y ), the genotype year interaction (s2gy ), and the
vine year interaction (s2vy ). The sum of these variance components was
regarded as the total variance (s2T ). The total environmental variance (s2E ) can be
expressed as s2E s2y =y s2gy =y s2v =v s2vy = yv s2c = yvc s2b = yvcb,
where y is the number of yearly repetitions, v the number of vine replications
per genotype, c the number of cluster samples per year and vine, b the number of
berries sampled per cluster, respectively.
3. Results and discussion
The mean performance of the evaluated genotypes was 5.2 for BRT, 6.4 g for
BW, 20.3% for SSC, 0.48 g100 mlÿ1 for TA, 4.09 10ÿ3 m for DFP, and
0.75 N for MF, respectively (Table 2).
Table 2
Grape cultivars evaluated and their performance in berry traitsa
Cultivars
BRT
BW (g)
b
V. vinifera
July Muscat
Hiro Humburg
Italia
4.0 a
5.5 bc
5.8 cd
V. labruscana
North Red
Buffalo
Steuben
Muscat Bailey A
Ryuho
Fujiminori
4.0
4.5
5.6
6.9
5.3
5.0
Mean
a
5.2
a
ab
bc
d
bc
a±c
5.2 b
5.1 b
8.1 c
4.5
3.5
4.6
4.2
9.5
13.0
6.4
b
a
b
ab
c
d
SSC (%)
TA (g100 mlÿ1)
Texture
DFP 10ÿ3 m
MF (N)
20.7 bc
22.3 c
19.3 ab
0.32 a
0.36 ab
0.45 a±c
2.18 a
2.90 b
1.76 a
0.861 cd
0.547 ab
0.555 ab
20.4
22.3
20.3
20.5
l8.0
18.7
0.46
0.65
0.50
0.64
0.56
0.39
4.51
6.10
5.97
5.40
5.04
2.92
0.441
1.047
0.838
1.269
0.699
0.494
20.3
bc
c
bc
bc
a
ab
0.48
a±c
e
b±e
de
c±e
a±c
4.09
c
d
d
cd
cd
b
a
cd
c
d
bc
a
0.750
The abbreviations used for the column heading have already been de®ned in text.
Mean separation
p within columns by LSD test at P < 0.05. The log-transformed values were used in ANOVA for BW, DFP and MF. LSD was
calculated as 1:96 2sE , where s2E s2y =2 s2gy =2 s2v =2 s2vy =4 s2c =16 s2b =48.
b
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
Species
79
80
Source of variation
Genotype
Year
Genotype year
Among vines within genotypes
Vine year
Among clusters within vines
Among berries within clusters
a
Mean squares (MS)b
BRT 10
BW 102
SSC
TA 102
DFP 102
MF 102
422.61**b
222.32*
40.70*
30.21NS
10.23**
0.71**
0.44
168.86**
17.43*
3.15NS
3.88NS
2.51**
0.74**
0.40
102.66*
35.31NS
19.88NS
14.23NS
6.84**
1.80**
0.60
67.03**
5.92NS
6.93NS
9.42NS
4.54**
0.72**
0.29
191.75**
20.78NS
5.46NS
1.77NS
1.96NS
1.55NS
1.12
117.23**
36.18NS
13.67*
6.34NS
3.83*
1.66NS
1.75
The abbreviations used for the column heading have already been de®ned in text. BW, DFP and MF were log-transformed.
NS, * and ** referred to nonsigni®cant, signi®cant at P < 0.05 and 0.01, respectively, using F-test. F-value for the effect of genotype was
calculated as F (MS for genotype MS for vine year)/(MS for among vines MS for genotype year), according to Snedecor and Cochran
(1972).
b
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
Table 3
Mean square for six berry traits evaluated using nine genotypes with two vines per genotype for 2 yearsa
81
A. Sato et al. / Scientia Horticulturae 85 (2000) 75±83
Table 4
Estimates of variance components and their percentage to the total variance obtained from the
analysis of variance for berry traits using nine genotypes with two vines per genotype for 2 yearsa
Variance
BRT 10
components
BW 102
SSC
TA 102
DFP 102
MF 102
s2g
s2y
s2gy
s2v
s2vy
s2c
s2b
3.42(80.7%)
0.07(1.7%)
0.03(0.7%)
0.06(1.4%)
0.15(3.5%)
0.11(2.6%)
0.40(9.4%)
1.57(40.1%)
0.07(1.8%)
0.54(13.9%)
0.31(7.9%)
0.42(10.7%)
0.40(10.2%)
0.60(15.4%)
1.15(52.3%)
0b(0%)
0.10(4.5%)
0.20(9.1%)
0.32(14.5%)
0.14(6.4%)
0.29(13.2%)
3.00(66.5%)
0.07(1.6%)
0.15(3.3%)
0b(0%)
0.03(0.7%)
0.14(3.1%)
1.12(24.8%)
2.11(45.4%)
0.10(2.2%)
0.41(8.8%)
0.10(2.2%)
0.18(3.9%)
0b(0%)
1.75(37.5%)
4.24(100%) 3.91(100%) 2.20(100%) 4.51(100%)
4.65(100%)
7.54(63.9%)
0.84(7.1%)
1.27(10.8%)
0.83(7.1%)
0.79(6.7%)
0.09(0.8%)
0.44(3.6%)
s2T
11.80(100%)
a
The abbreviations used for the column heading have already been de®ned in text. BW, DFP
and MF were log-transformed.
b
Negative values were assumed to be zero.
The result of ANOVA showed that the effect of genotype was signi®cant at
P < 0.05 for SSC and at P < 0.01 for the other traits (Table 3). The effect of the
vines within genotypes was nonsigni®cant for all traits.
The contribution of the environmental variance components to the total
variance varied with the traits (Table 4). For BRT, the variances associated with
year and vine (s2y ; s2gy ; s2v and s2vy ) were larger than s2c and s2b . For BW, as the
environmental variance components were generally small, it was easy to identify
the genetic property. The variance among berries within clusters was the largest
among the environmental variance components for all traits with the exception of
BRT. Especially, the s2b of DFP and MF were very large, indicating that
increasing the number of berries evaluated per cluster increases the measurement
accuracy effectively.
The cost and time to increase berry and cluster samples are generally smaller
than yearly repetitions or vine replications. So, ®rst choice of grape breeders
should be to measure more berries by increasing cluster or berry samples. Yearly
repetitions or vine replications, however, take much cost or time. So, comparison
between yearly repetitions and vine replications should be considered primarily.
To compare the in¯uence of adding yearly repetitions and adding vine
replications on the decrease of the s2E , the ratio of s2v = s2y s2gy shows the
ef®ciency of the two factors in decreasing the s2E , because the s2vy ; s2c and s2b
decrease equally both by adding years and vine replications. Our results showed
that the ratio was