Euphytica

, Volume 191, Issue 3, pp 365–373

Rubber tree early selection for yield stability in time and among locations

Authors

  • Lígia Regina Lima Gouvêa
    • Programa Seringueira, Instituto Agronômico
  • Guilherme Augusto Peres Silva
    • Programa Seringueira, Instituto Agronômico
  • Cecília Khusala Verardi
    • Programa Seringueira, Instituto Agronômico
  • André Luis Bombonato de Oliveira
    • Programa Seringueira, Instituto Agronômico
  • Elaine Cristine Piffer Gonçalves
    • Apta Regional Alta Mogiana
  • Erivaldo José Scaloppi-Junior
    • Apta Regional Noroeste Paulista
  • Mário Luiz Teixeira de Moraes
    • Universidade Estadual Paulista ‘Júlio de Mesquita Filho’
    • Programa Seringueira, Instituto Agronômico
Article

DOI: 10.1007/s10681-013-0874-6

Cite this article as:
Gouvêa, L.R.L., Silva, G.A.P., Verardi, C.K. et al. Euphytica (2013) 191: 365. doi:10.1007/s10681-013-0874-6

Abstract

Rubber production in the rubber tree [Hevea brasiliensis (Willd. ex Adr. de Juss.) Muell. Arg.] can be expressed differently in different environments. Thus the objective of the present study was to select productive progenies, stable and responsive in time and among locations. Thirty progenies were assessed by early yield tests at three ages and in three locations. A randomized block design was used with three replications and ten plants per plot, in 3 × 3 m spacing. The procedure of the mixed linear Reml/Blup model—restricted maximum likelihood/best non-biased linear prediction was used in the genetic statistical analyses. In all the individual analyses, the values observed for the progeny average heritability (\( \hat{h}_{pa}^{2} \)) were greater than those of the additive effect based on single individuals (\( \hat{h}_{a}^{2} \)) and within plot additive (\( \hat{h}_{ad}^{2} \)). In the joint analyses in time, there was genotype × test interaction in the three locations. When 20 % of the best progenies were selected the predicted genetic gains were: Colina GG = 24.63 %, Selvíria GG = 13.63 %, and Votuporanga GG = 25.39 %. Two progenies were among the best in the analyses in the time and between locations. In the joint analysis among locations there was only genotype × location interaction in the first early test. In this test, selecting 20 %, the general predicted genetic gain was GG = 25.10 %. Identifying progenies with high and stable yield over time and among locations contributes to the efficiency of the genetic breeding program. The relative performance of the progenies varies depending of the age of early selection test.

Keywords

Hevea brasiliensisReml/BlupRepeated meansGenetic parametersEarly yield tests

Introduction

The convergence of many factors, including uncertainty about the planet energy reserves and global warming, have given a new impulse to the development of crops to produce renewable energy and materials, including natural rubber, one of the most important polymers used in global society (Beilen and Poirier 2007). The rubber tree [Hevea brasiliensis (Willd. ex Adr. de Juss.) Muell. Arg.] is found in this context as the main source of rubber in nature.

The species is cropped in several tropical countries, most of which have active plant breeding programs (Sedgley and Attanayake 1988). The main objective of breeding the rubber tree is to increase yield and vigor by methods that shorten the crop breeding cycle (Goncalves et al. 2006; Costa et al. 2002). The cycle is long and many selection and assessment steps are necessary (Chandrasekhar et al. 2007; Priyadarshan and Gonçalves 2003), forming three phases. This justifies the standardization of early selection methods to optimize and reduce the cycle whenever possible (Gonçalves et al. 2009b).

Breeders search for genotypes that show vigor, high yield and stability over years and locations (Gonçalves et al. 2009a) because traits such as rubber yield can be expressed differently over time and among locations. Early yield tests are usually used to reduce the long cycle in rubber tree breeding programs and these tests have been carried out at a single age to assess the yield stability of young rubber tree progenies among locations (Gonçalves et al. 2009a; Verardi et al. 2009). In these studies the HMMm early selection test has been used on 3 year old progenies. In perennial crops, assessing traits by repeated means in time can also lead to more efficient selection.

Several approaches can be used to model longitudinal data or repeated means resulting from perennial individuals, such as regression, repeatability, multivariate analysis, random regression and ante-dependence models (Resende et al. 2006). In the analysis of means repeated in time, the use of mixed models that allow more complex analyses to be incorporated has been recommended (Resende et al. 2006; De Ketelaere et al. 2003; Littell et al. 1998). In the mixed model context, statistics of the harmonic mean of relative performance of genetic values—HMRPGV, predicted by Blup (best linear unbiased prediction) is an option for simultaneous selection for genetic values of yield, adaptability (responsiveness) and stability (Resende 2007a). Proposed by Resende (2004), this method considers the genotypic affects as random and therefore gives genotypic stability and adaptability. Its advantages include that it allows heterogeneity of variances to be dealt with, with imbalances and non-orthogonal designs; it can be applied to any number of environments; it eliminates noise from the genotype × environment interaction and generates results in the size itself of the trait assessed.

The objective of the present study was to select productive progenies stable and responsive in time and among locations by the mixed linear model procedure.

Materials and methods

The genetic material used consisted of a population of 30 open pollination progenies from the clones identified as: 1—RRIM 600; 2—IAC 40; 3—IAC 41; 4—PB 235; 5—IAC 35; 6—64B 850; 7—PB 260; 8—PB 252; 9—IRCA 111; 10—RRIM 606; 11—RRIM 701; 12—IAC 15; 13—PB 28/59; 14—IAC 311; 15—IAC 301; 16—ROI 110; 17—IAC 307; 18—PR 255; 19—RO I 35; 20—IAN 873; 21—IAC 44; 22—PR 261; 23—PB 217; 24—GT1; 25—PB 330; 26—Fx 3864; 27—MT 45; 28—Fx 2261; 29—1-12-56-77; 30—P 595/89.

Progeny tests were set up in Selvíria in Mato Grosso do Sul state, Colina and Votuporanga in São Paulo state (Table 1). A randomized block design was used with three replications and linear plots of ten plants, in 3 × 3 m spacing.
Table 1

Localization of the rubber tree progeny tests carried out in Brazil

Remarks

Colina-SP

Selvíria_MS

Votuporanga-SP

Latitude (S)

20º43′05′′

20º20′

20°27′24.7″

Longitude (EE)

48º32′38′′

51º23′

50°03′54.1″

Elevation (m)

568

370

482

Temperature annual mean (°C)

22.6

24.5

22.3

Annual rainfall (mm) (mean annual)

1,255

1,232

1,480

The progeny yield was assessed by early yield tests (EYT) at the age of three (EYT 3), four (EYT 4) and five (EYT 5) years of age, respectively. The Hamaker Morris–Mann test was used, developed originally to select 5 year old trees in Indonesia (Dijkman 1951) and later in Malaysia by the rubber research institute of Malaysia. In this study it was also used for three and 4 year old progenies. The mean of the dry rubber yield from 30 cuts per plant was used. The tapping panel was opened 20 cm above the ground, by the S/2 d/3 system in a total of 35 cuts, discarding the first five samples that corresponded to the running in phase of the panel. The S/2 nomenclature refers to the half spirals cuts and the d/3 nomenclature expresses the interval between tappings, that is, one tapping every 3 days. The results were used in the form of grams tapping−1 tree−1.

The genetic parameters were obtained by the Reml/Blup method. The statistical models used were reported by Resende (2007b). In the individual analyses, the model used was applied to open pollination progenies, with a mixed reproduction system, 22 % self-pollination rate (Costa et al. 2000b; Furlani et al. 2005). The statistical model used was: Y = Xr + Za + Wc + ε, where: Y is the data vector, r is the repetition effects vector, assumed fixed, added to the general mean, a is the vector of the individual additive genetic effects presumed as random, c is the plot effects vector, ε is the random residue vector. The uppercase letters represent the incidence matrix for the effect. The significance of the progenies effect was verified on the values of the likelihood ratio test (LRT), obtained from the analysis of deviance.

The following genetic parameters were estimated: narrow sense individual heritability, with 22 % self-pollination rate: \( \hat{h}_{a}^{2} \) = \( [4\,\hat{\sigma }_{g}^{2} /(1 + s)^{2} ]/(\hat{\sigma }_{g}^{2} + \hat{\sigma }_{c}^{2} \, + \hat{\sigma }_{e}^{2} ) \); progeny average heritability, considering complete survival: \( \hat{h}_{pa}^{2} \) = \( (\hat{\sigma }_{g}^{2} )/[\hat{\sigma }_{g}^{2} + \hat{\sigma }_{c}^{2} /b + \hat{\sigma }_{e}^{2} /(nb)] \); within plot additive heritability; \( \hat{h}_{ad}^{2} \) = \( \{ 1 - [(1 + s)^{2 \, } /4]\} [4/(1 + s)^{2} ]\hat{\sigma }_{g}^{2} /\hat{\sigma }_{e}^{2} \); coefficient of determination of the plot effects;\( (\hat{c}_{plot}^{2} = \hat{\sigma }_{c}^{2} )/(\hat{\sigma }_{g}^{2} + \hat{\sigma }_{c}^{2} + \hat{\sigma }_{e}^{2} ) \); accuracy of the progenies selected, when considering complete survival: \( \hat{A}c = (\hat{h}_{mp}^{2} )^{0.5} \); coefficient of genetic variation; CVg(%) = 100 × \( (\hat{\sigma }_{g}^{2} )^{0.5} /\mu \); coefficient experimental variation; CVe(%) = \( \{ [(0.75.\hat{\sigma }_{g}^{2} + \hat{\sigma }_{e}^{2} )/n] + \hat{\sigma }_{c}^{2} \}^{0.5} /\mu \times 100 \), where n is the number of plants per plot; coefficient of relative variation : CVr = CVg/CVe; selection genetic in percent Gg(%) = (GG/\( \bar{x} \)) × 100.

The method of the Blup-predicted harmonic mean of the relative performance of the genetic values (HMRPGV) was used in the simultaneous analysis of yield, stability and temporal adaptability of the rubber tree progenies. Three early yield tests were used, at the ages of three (EYT3), four (EYT4) and five (EYT5) years, respectively. An analysis was made including the three production tests for each location by the statistic model Y = Xm + Za + Wp + Qi + Ts + ε, where Y is the data vector, m is the vector of the effects of the measuring-repetition combinations (presumed fixed) added to the general mean, a is the vector of the individual additive genetic effects (presumed random), p is the plot effects vector (random), i is the effects of the genotype measuring interaction vector (random), s is the vector of the permanent effects (random) and ε is the vector of residues (random). The uppercase letters represent the incident matrixes for the referred effects. The Blup-predicted HMRPGV method was used in the simultaneous analysis of years, stability and adaptability among locations. The three progeny tests were set up in Colina and Votuporanga in São Paulo state and Selvíria in Mato Grosso do Sul state. For each early yield test, analysis was carried out including the three locations. These statistical model used was Y = Xr + Zg + Wp + Ti + ε, where Y is the data vector, r is the replication effect vector (presumed fixed) added to the general mean, g′ is the genotype effects vector (presumed random), p is the plot effects vector (random), i is the genotype × environment interaction vector (random) and ε is the residue vector (random). The uppercase letters represent the incidence matrixes for the referred effects.

The significance of the progenies × test and progenies × location interactions was verified by LRT obtained in the analysis of deviation reported by Resende (2007a).

In all statistical analyzes it was used the statistical-genetic program selegen REML/BLUP.

Results and discussion

The progeny average heritability (\( \hat{h}_{pa}^{2} \)) ranged from 0.459 to 0.917 (Table 2). The additive effect heritability based on single individuals (\( \hat{h}_{a}^{2} \)) ranged from 0.214 a 0.750. The within plot additive heritability (\( \hat{h}_{ad}^{2} \)) ranged from 0.190 to 0.657. In all the assessments the \( \hat{h}_{pa}^{2} \) were greater than the \( \hat{h}_{a}^{2} \) and \( \hat{h}_{ad}^{2} \). This indicated that higher genetic gains can be obtained by selecting the best progenies than by mass selection or in within progeny selection. This result was similar to that observed for early production in rubber tree progenies by Silva et al. (2012) and Goncalves et al. (2006) and showed that \( \hat{h}_{pa}^{2} \) has been a parameter indicated in early progeny selection for rubber yield. Analyzing the progeny average heritability (\( \hat{h}_{pa}^{2} \)) within each location, in Colina the highest value was observed in the mean of the three early yield tests (\( \hat{h}_{pa}^{2} \) = 0.814) in Selvíria on EYT3 (\( \hat{h}_{pa}^{2} \) = 0.917) and in Votuporanga in EYT4 (\( \hat{h}_{pa}^{2} \) = 0.827). These results showed the high genetic control of the studied train in level of progeny. Falconer (1964) explained that heritability is a property not only of the traits, but also of the population and of the environmental circumstances to which the individuals are subject and that the heritability value can be affected if there is alteration in any of the components of variance.
Table 2

Estimate of genetic parameters of a population of 30 rubber tree open population progenies assessed in three early yield tests (EYT) and in three progeny tests in Colina and Votuporanga in São Paulo state and Selvíria in Mato Grosso do Sul State

. Parameters

Colina

Selvíria

Votuporanga

 

EYT 3

EYT 4

EYT 5

\( \bar{x} \)

EYT 3

EYT 4

EYT

\( \bar{x} \)

EYT 3

EYT 4

EYT 5

\( \bar{x} \)

\( \hat{h}_{a}^{2} \)

0.214 + 0.091

0.278 + 0.10

0.366 + 0.2

0.362 + 0.12

0.750 + 0.21

0.475 + 0.17

0.418 + 0.16

0.600 + 0.19

0.386 + 0.12

0.420 + 0.13

0.338 + 0.11

0.429 + 0.13

\( \hat{h}_{ad}^{2} \)

0.190

0.195

0.276

0.264

0.657

0.365

0.314

0.488

0.286

0.318

0.258

0.333

\( \hat{h}_{pa}^{2} \)

0.459

0.769

0.785

0.814

0.917

0.859

0.837

0.892

0.821

0.827

0.740

0.808

\( \hat{c}_{plot}^{2} \)

0.210

0.004

0.028

0.006

0.004

0.006

0.008

0.005

0.010

0.014

0.050

0.032

\( \hat{A}c_{{}}^{{}} \)

0.678

0.877

0.886

0.902

0.957

0.927

0.915

0.944

0.905

0.910

0.860

0.899

CVg%

19.588

27.001

30.661

28.619

24.970

18.082

22.272

21.630

22.728

38.273

34.538

33.760

CVe%

37.332

26.650

29.029

24.920

14.720

13.687

18.110

14.373

19.470

31.993

36.693

29.923

CVr

0.525

1.013

1.056

1.148

1.696

1.321

1.230

1.505

1.167

1.196

0.941

1.128

μ

1.540

11.081

17.203

9.886

9.370

23.928

29.458

20.754

2.265

5.418

7.559

4.975

LRT(x2)

2.904ns

14.448**

13.447**

56.136**

32.870**

16.090**

14.28**

24.550**

17.735**

18.579**

12.090**

22.480**

\( \hat{h}_{a}^{2} \) Additive effect heritability based on single individuals; \( \hat{h}_{ad}^{2} \) among plot additive heritability; \( \hat{h}_{pa}^{2} \) progeny average heritability; \( \hat{c}_{plot}^{2} \) Coefficient of determination of the plot effects; \( \hat{A}c_{{}}^{{}} \) Accuracy; CVg% coefficient of genetic variation; CVe% coefficients of experimental variation; CVr; coefficient of relative variation, μ general mean; LRT (x2) Chi square of the likelihood ratio test for progeny effect, x2 tabled: 3.84 and 6.63 for the 5 and 1 % levels of significance respectively

Considerable genetic variability was estimated for rubber yield in this population by the early yield tests. The coefficient of genetic variation (CVg %) ranged from 19.58 to 38.27 %. This is a favorable situation for the genetic breeding program. Costa et al. (2000a) assessed early yield of an Asian rubber tree population introduced in Brazil and reported similar values ranging from 15.64 to 40.92 %. The coefficient of experimental variation (CVe %) ranged from 14.27 to 37.33 %. These values were in line with those observed for this variable by Gonçalves et al. (2005) and Costa et al. (2000a).

The coefficient of relative variation (CVr), a parameter that helps to detect the genetic variability of the population, was greater than the unit in almost all the analyses. According to (Vencovsky 1987), corn progenies are selected when the CVr is equal or greater than the unit, since the condition is highly favorable for selection.

Accuracy (correlation between the predicted genetic values and the true genetic values) was high and very high in almost all the assessments and was moderate only in Colina in the EYT3. Resende and Duarte (2007) ranked the accuracies as: very high (\( \hat{A}c \) ≤ 0.90), high (0.70 ≤ \( \hat{A}c \) ≤ 0.90), moderate (0.50 ≤ \( \hat{A}c \) ≤ 0.70) and low (\( \hat{A}c \) ≤ 0.50). Accuracies of 70 % or more are desirable in the initial and intermediate breeding stages (Resende 2007a).

The coefficients of determination for the plot effects (\( \hat{c}_{plot}^{2} \)) were lower in almost all the assessments, except in Colina EYT 3. According to Fogaça et al. (2012) this parameter indicates low environmental variability among plots within the block.

The progeny effect, observed by the analysis of deviance, was significant in almost all the assessments (Table 2), that indicated among progeny genetic variability and consequently the possibility of gain with selection. Analysis of deviance is reported in Resende (2007b) and was used to verify the significance of the progeny effect on the rubber yield trait for the first time by Arantes et al. (2010). The progeny effect was not significant only in Colina EYT 3 where the coefficient of experimental variation CVe % was greater than the CVg that could explain the non-significance only in this assessment.

In the analysis of joint deviance of the repeated means in time (Table 3) the x2 of the likelihood ratio was significant for the progeny effect in the three locations. These results confirm the progeny genetic variability observed in Table 2. The x2 of the likelihood ratio test was also significant for the progeny test interaction. This showed that the relative performance of the progenies was not the same in the three early yield tests. The best progenies in one test (age) might not be the best in the other tests. This occurred in the three assessment locations. In a situation such as this, stability and/or adaptability analyses are indicated to identify the best progenies in the three tests simultaneously and are shown in Table 5.
Table 3

Analysis of joint deviance (ANADEV) by means repeated in time, corresponding to the three rubber tree early yield tests (EYT) assessed by location in Colina and Votuporanga in São Paulo state and Selvíria in Mato Grosso do Sul State

Locations/Effect

Deviance

LRT(x2)

Colina

  

 Progenies

9131.040a

5.840*

 Progenies × EYT

9181.880a

56.680**

 Plot

9125.280a

0.008ns

 Complet Model

9125.200

Selvíria

  

 Progenies

5711.250a

16.333**

 Progenies × EYT

5708.984a

14.134**

 Plot

5694.917a

0.067ns

 Complet Model

5694.950

Votuporanga

  

 Progenies

5811.329a

6.549*

 Progenies × EYT

5813.79a

9.010**

 Plot

5805.08a

0.300ns

 Complet Model

5804.788

x2 of the likelihood ratio test (LRT) tabled: 3.84 and 6.63 for the 5 and 1 % levels of significance respectively

aDeviance from the fitted model without the referred effects

In the analysis of joint deviance for locations (Table 4) the x2 of the likelihood ratio test was significant for the progeny effect in the three tests that confirmed the genetic variability for progenies observed in Tables 2 and 3. The x2 of the progeny × location interaction was significant only in EYT 3 and this interaction did not occur in EYT 4 and EYT 5. Therefore both in EYT 4 and EYT 5 the progenies presented a constant relative performance among locations. The genetic correlations among locations (Table 4) were coherent with the significance observed in the progeny location interaction. The following values were observed: EYT3 (rgloc = 0.658), EYT4 (rgloc = 0.939), EYT5(rgloc = 0.967) and mean (rgloc = 0.896). There was significant interaction only in EYT3. For this test the genetic correlation among locations was the lowest indicating complex type interaction. Robertson (1959) explained that when the value of the genetic correlation among the environments is the closer to one, the variance of the interaction referred to the single part of the interaction. The author partitioned the genotype × environment interaction into the single part of the interaction explained by the change in genetic variance in the materials and the different environments and the complex part of the interaction came from the lack of genetic correlation between the performance of the genetic material in one environment to the other.
Table 4

Joint analysis of deviance (ANADEV) among the locations in Colina and Votuporanga in São Paulo State and Selvíria in Mato Grosso do Sul, assessing each early yield test (EYT) and their mean

Test/effect

Deviance

LRT(x2)

EYT3

  

 Progenies

3858.903a

73.837**

 Progenies × local

3961.565a

28.825**

 Plot

3932.844a

0.104ns

 Complet model

3932.740

EYT4

  

 Progenies

6870.015a

40.995**

 Progenies × local

6829.910a

0.890ns

 Plot

6828.949a

0.071ns

 Complet model

6829.020

EYT5

  

 Progenies

8036.403a

44.563**

 Progenies × local

7992.060a

0.220ns

 Plot

7992.363a

0.523ns

 Complet model

7991.840

Mean

  

 Progenies

6650.552a

45.80**

 Progenies × local

6608.189a

3.44ns

 Plot

6604.760a

0.010ns

 Complet model

6604.750

x2 of the likelihood ratio test (LRT) tabled: 3.84 and 6.63 for the 5 and 1 % levels of significance respectively. RGLOC: genetic correlation across locations: EYT3 (gcloc = 0.658), EYT4 (gcloc = 0.939), EYT5(gcloc = 0.967) and mean(gcloc = 0.896)

aDeviance from the fitted model without the referred effects

In the analysis by the HMRPGV method for means repeated over time (Table 5), 20 % of the best progeny were selected. This analysis selects simultaneously for stability, adaptability (responsiveness) and yield. The best performances were for progenies (5, 23, 24, 1, and 6) in Colina, progenies (13, 24, 23, 10, 5, and 25) in Selvíria and progenies (5, 15, 13, 24, 9, and 4) in Votuporanga. Identifying genotypes with high yield, high stability and wide adaptability for several environments is one of the main objectives in forest species genetic breeding programs (Verardi et al. 2009). With the selection of these progenies the predicted genetic gains were: Colina (GG = 24.63 %), Selvíria (GG = 13.63 %) and Votuporanga (GG = 25.39 %). Progenies (5 and 24) appeared among the best in the three locations. They are high yielding progenies, stable in time and among locations. Lin and Binns (1991) assessed stability parameters in “smooth bromegrass”, a perennial forage species and showed that stability can be inherited, and selection can be made for high and stable yield. Progenies 27, 29, 16, and 26 had the poorest performances in the three locations. There were also progenies with good performance only in one location, such as (1, 6) in Colina, (10 and 25) in Selvíria and (15) in Votuporanga that show the effect of the genotype × environment interaction. Data of this nature can help in the genetic breeding program for high and stable yield.
Table 5

Result of the simultaneous analysis for yield, stability and adaptability by harmonic mean of the relative performance of the genetic values (HMRPGV) assessing means repeated in time, corresponding to the three rubber early yield tests per location in Colina and Votuporanga in São Paulo state and Selvíria in Mato Grosso do Sul state

Rank

Temporal analysis of the three early yield tests

 

Colina

Selvíria

Votuporanga

 

HMRPGV*GM

Progenie

HMRPGV*GM

Progenie

HMRPGV*GM

Progenie

1

14.722

5

27.614

13

7.687

5

2

13.616

23

26.199

24

7.525

15

3

12.255

24

24.612

23

7.060

13

4

12.253

1

23.948

10

6.305

24

5

12.163

6

23.937

5

6.079

9

6

11.962

2

23.815

25

5.963

4

7

11.760

13

23.588

9

5.787

18

8

10.890

12

23.227

22

5.717

20

9

10.660

15

23.204

11

5.696

10

10

10.218

18

22.535

4

5.645

6

11

10.204

20

22.385

3

5.582

21

12

10.129

9

22.014

2

5.53

22

13

10.086

4

21.829

18

5.28

23

14

9.997

3

21.731

1

5.211

1

15

9.757

22

20.934

20

5.206

3

16

9.726

28

20.661

8

5.025

11

17

9.643

11

20.510

7

4.924

30

18

9.589

17

20.240

12

4.907

8

19

9.480

8

20.077

17

4.756

12

20

9.386

7

19.985

15

4.640

14

21

9.209

10

19.125

28

4.410

28

22

8.484

21

18.991

6

4.115

2

23

8.480

30

18.710

21

4.050

19

24

8.254

19

18.104

14

3.929

17

25

8.122

14

18.001

16

3.886

7

26

7.369

27

17.548

29

3.772

25

27

7.347

25

14.606

26

3.760

27

28

6.759

29

10.733

19

3.706

26

29

6.713

16

9.898

30

3.694

29

30

6.149

26

9.711

27

2.958

16

GM general mean

The best performances were observed for progenies 24, 13, 3, 5, 22, and 10 in the HMRPGV among locations analysis (Table 6) for the first early yield test (EYT 3). Selection by a criterion that uses estimates of the harmonic means of the genotype values is an excellent strategy that permits safe inferences on the prediction of the genetic values and has the advantages of joining both yield and stability in a single selection criterion (Ferreira et al. 2012). With the selection of these progenies the predicted genetic gain was (GG = 25.10 %) a higher value than that estimated by Verardi et al. (2009). These authors assessed the rubber tree early yield trait in 22 progenies in three locations, also in the third year, and selected the best five and reported predicted genetic gain (GG = 8.16 %), that showed that the population assessed in the present study had greater genetic potential for rubber yield probably because the female parent plants had already been selected. Progenies (24 and 5) are among the best in the among locations analysis and also among the best in the analysis of means repeated in time (Table 5). They are high yielding progenies stable in time and among locations and are therefore of great interest for genetic breeding (Gonçalves et al. 2009a).
Table 6

Result of the simultaneous analysis for yield, stability and adaptability by harmonic mean of the relative performance of the genetic values (HMRPGV) in the joint analysis by locations in the first early yield test (EYT3), carried out at three years of age, referring to the locations in Colina and Votuporanga in São Paulo state and Selvíria in Mato Grosso do Sul state

Location analysis

EYT3

Rank

HMRPGV*GM

Progeniess

1

10.001

24

2

9.176

13

3

8.907

3

4

8.838

5

5

8.440

22

6

8.325

10

7

8.203

23

8

8.127

1

9

7.778

17

10

7.716

18

11

7.554

15

12

7.467

7

13

7.353

9

14

7.190

8

15

7.143

4

16

7.141

12

17

6.797

20

18

6.664

6

19

6.537

11

20

6.522

2

21

6.234

25

22

6.172

14

23

5.953

28

24

5.716

21

25

4.821

29

26

4.759

16

27

2.639

30

28

2.603

26

29

1.832

19

30

0.056

27

This study contributed to the knowledge that identifying progenies with high and stable yield over time and among locations increased the efficiency of the genetic breeding program. In this sense, studies correlating the ortete-ramete yield with the age at the early test will bring contributions. For this population, selection of the best progenies was more efficient than mass selection and within progeny selection. The progeny performance varied with the age at which the early yield tests were carried out.

Acknowledgments

To Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), to Conselho Nacional de Desenvolvimento Científico e Tecnológico(CNPq) and to Secretaria de Agricultura e Abastecimento do Estado de São Paulo(SAA), for grants and support.

Copyright information

© Springer Science+Business Media Dordrecht 2013