R.W. STONECYPHERROY W. STONECYPHER is associate professor, Forestry Department. Oklahoma State University, United States.
IT WOULD BE WISE to begin by considering an applied tree breeder's definition of multiple-trait breeding as part of the tree breeding concept. This concept can be defined as the aim of the tree breeder to produce a maximum of economically usable wood per tree and per hectare. The tree breeder must decide the definition of economically usable wood for the particular species with which he is working and while being aware that utilization standards may change in the future.
Such a goal implies that the tree breeder will almost certainly consider selection for more than one trait. The main theme in this paper is, therefore, the problems related to multiple-trait breeding in forest trees.
THE NEED FOB MULTIPLE - TRAIT BREEDING
As indicated above, the forest tree breeder is rarely faced with a situation in which improvement in only a single trait is desired. Practically all applied tree breeding programmes now- in operation use multiple traits in evaluating and selecting plus trees (Andersson, 1965; Cech, 1959; Peevy, 1959).
As has been emphasized by Lush (1948), it is economically unwise or even impossible to select for one trait alone. Since the value of an individual tree obviously depends on several characteristics, further justification of the need for considering the whole tree in a selection programme hardly seems necessary.
RELATION OF TREE BREEDING CONCEPT TO SELECTION METHODS
Accepting the premise that several traits should be considered in a selection programme, three methods are available to the breeder in attaining the desired goal (Hazel and Lush, 1942). These three methods and their relevance to forest tree breeding are discussed below.
Tandem selection
The tandem method of selection as described by Hazel and Lush (1942) involves selection for one trait at a time. This is followed by selection for a second trait, third trait, etc., until the desired level of improvement is reached.
The long generation interval of most forest tree species would appear to preclude use of the tandem method by the tree breeder. However, there are circumstances in which the use of this method is justified. Such situations exist in cases where a single characteristic limits the economic usefulness of a species. An example of such a condition occurs in breeding for disease resistance where the disease is a completely limiting factor in the management of a species. gingham et al. (1960) discuss such a situation in breeding for resistance to Cronartium ribicola Fisher in Pinus monticola Dougl.
Selection based on independent culling levels
The selection method of independent culling levels involves the establishment of a level of merit for each trait. Individuals falling below this level for a given trait will be rejected regardless of their superiority in other traits. Many of the present tree breeding programmes select plus trees by scoring systems (Andersson, 1965; Cech, 1959; Stern and Hattemer, 1964), but selection for certain traits such as resistance to disease and bole straightness is based on independent culling levels. This method is at least as efficient as tandem selection. In some cases the method allows selection for given traits before the organism under consideration has matured. For example, disease resistance generally can be evaluated fairly early in the life of the tree, permitting selection for this trait long before other traits can be evaluated.
Selection index
The method of total score (Hazel and Lush, 1942) or selection index (Hazel, 1943) has been shown never to be inferior to the method of independent culling levels. The method of the selection index was initially applied to selection in plants by Smith (1936). He used Fisher's concept of discriminant functions to develop a selection guide for determining plant lines which had the greatest genotypic value.
Numerous selection indices have been developed for crop plants (see Robinson et al., 1951; Johnson et al., 1955; and Brim et al., 1959).
More recently, selection indices have been developed for forest trees by van Buijtenen and van Horn (1960) and Illy (1966).
Considering the three methods of selection available, the tree breeder should give greatest emphasis to the selection index. The methods of tandem selection and independent culling levels will be used in certain special circumstances. Consequently the discussion which follows will be largely concerned with the development of selection indices.
INFORMATION REQUIRED
The theory and methods of deriving selection indices have been adequately developed in the literature (Smith, 1937; Hazel, 1943; Robinson et al., 1951), and Illy (1969) has also given an excellent development for Pinus pinaster Sol. Three basic types of information are required for the development of selection indices. These requirements are discussed below.
Economic values
Perhaps the most troublesome parameters needed for the construction of selection indices are the relative economic weights of the traits. Not only are these weights sometimes difficult to obtain, but they are also subject to fluctuations. For example, Lerner and Donald (1966) have pointed out that the economic weight for a certain trait in selection of Yorkshire pigs changed from positive to negative in a relatively short period of time.
Weights of economic traits in tree breeding programmes will fluctuate. However, products of forest trees are less directly subject to the whims of the producer and consumer. Therefore, their values should remain somewhat more stable than those of animal and crop breeding programmes.
The role of nonlinearity and nonindependence of the economic weights in defining net merit is of more importance in tree breeding programmes. Namkoong (1969) has touched on this problem and offered a possible solution for maximizing genetic gain when limited information is available to the breeder.
Current studies are beginning to furnish fairly reliable data on the biological information necessary for efficient multiple-trait breeding, but very few studies have been designed to determine accurate economic weights for the traits of interest to the tree breeder. Availability of reasonably accurate economic information could well be the limiting factor in applying multiple-trait breeding procedures.
Genotypic and phenotypic variances and covariances
Construction of selection indices requires estimates of the genotypic and phenotypic variances and covariances for traits included in the index. Using these estimates in conjunction with economic weights in construction of indices is rather straightforward.
Procedures for obtaining estimates of genetic and phenotypic variances and covariances have been adequately developed for plant and animal populations. Excellent discussions of these procedures for plant populations are given by Robinson et al. (1951) and Cockerham (1961).
The problems associated with obtaining reliable estimates of genetic and environmental variances and covariances have been discussed by Namkoong (1969) and earlier by Goggans (1962), Bogyo (1964), and Stonecypher (1966) for forest tree populations.
Namkoong (1969) and Illy (1969) have emphasized the importance of obtaining reliable information for constructing indices. Obviously the reliability of the estimates of the parameters will in turn affect the reliability of the calculated indices. Brim et al. (1959) have briefly discussed the problems associated with index reliability.
In the majority of publications dealing with estimates of genetic variances for forest tree populations no attempt was made to calculate genetic and phenotypic correlations between traits. This oversight is most unfortunate. It will not be possible to build up reliable information concerning trait relationships unless forest geneticists perform the necessary calculations as a part of inheritance studies of forest tree populations. Needless to say, the calculation and reporting of trait correlations should be encouraged.
CURRENT INFORMATION AVAILABLE FOR THEE POPULATIONS
Unfortunately, reliable estimates of the parameters necessary for constructing indices for tree breeding programmes are scarce. Namkoong et al. (1966) have recently tabulated interpretable estimates of heritability from forest tree populations. Very few of the publications dealing with estimates of heritability for forest trees report estimates of genetic and phenotypic correlations between traits. It should be further emphasized that of the parameter estimates which are available, many have relatively large standard errors. However, several recent publications have reported estimates of both genetic variances and covariances (Goggans, 1962; Stonecypher et al., 1964; Stonecypher and Zobel, 1966; and Illy, 1966).
Whether the method of tandem, independent culling level, or traditional index selection is used in breeding for the total tree, reliable data concerning the inheritance, economic weights and correlations of the traits which make up net merit are necessary for an efficient multiple-trait breeding programme to operate. Good economic and genetic models using reliable estimates of parameters will always improve any breeding methods used.
As Namkoong et al. (1966) have pointed out, a breeder cannot effectively all traits simultaneously, and he faces several alternative improvement goals. These goals include breeding for a large gain in one trait, breeding for equal gain 'in each of several traits, or weighting the selection emphasis among several traits according to the potential gain and effect on product value of each trait.
Many authors have emphasized the importance of the whole tree concept in optimizing selection within tree breeding populations. There seems to be general agreement concerning the necessity for considering multiple traits in tree breeding. There also appears to be general concurrence that the necessary information required for optimizing multiple-trait selection is generally lacking for tree populations.
Both Namkoong (1969) and Illy (1969) have clearly demonstrated the dangers inherent in using poorly estimated parameters for calculating indices. Although estimates of these parameters are becoming available for some populations, these estimates are in general from young material, and in many cases they lack desirable reliability. The approach suggested by Namkoong (1969) would appear to be an alternative to waiting until more reliable data are available. Such data may never be available in certain situations.
In spite of the difficulty and expense of genetic estimation experiments for tree populations, reliable estimates of parameters from such research are absolutely necessary if maximization of genetic gain, at least by traditional methods, is to be made. In addition, more emphasis should be given to determining economic weights of traits.
Tree breeders should make a concerted effort to gather data necessary for maximizing gain, but they should also give strong consideration as to how accumulated data are to be used in planning tree breeding programmes. Breeders should be encouraged to go beyond the reporting of data, by applying the results to 'practical tree breeding situations.
The success of tree breeding programmes will depend on the breeder's ability to pay attention to several traits. He should avoid those which are unimportant and emphasize those which are most important in producing maximum usable wood per hectare.
ANDERSSON, ENAR. 1965. The selection of plus trees in Sweden. Sumarski List (Zagreb) No 1-2.
BINGHAM, R. T., SQUILLACE, A. E. & WRIGHT, J. W. 1960. Breeding blister rust resistant western white Pine. II. First results of progeny tests including preliminary estimates of heritability and rate of improvements. Silvae Genet. 9: 33-41.
BOGYO, T. D. 1964. Coefficients of variation of heritability estimates obtained from variance analyses. Biometrics, 20: 122-129.
BRIM, C. A., JOHNSON, HERBERT W. & COCKERHAM, C. CLARK. 1959. Multiple selection criteria in soybeans. Agron. J., 51: 42-46.
CECH, FRANKLIN C. 1959. Superior tree selection, slash pine. Proc. 5th South. Conf. Forest Tree Improvement. North Carolina State College, School of Forestry.
COCKERHAM, C. CLARK. 1961. Estimation of genetic variances. In Statistical genetics and plant breeding. Washington, D.C., National Research Council. Publ. 982, p. 53-93.
GOGGANS, JAMES F. 1962. The correlation, variation, and inheritance of wood properties in loblolly pine (Pinus taeda L.). Forest Tree Improv. Prog., North Carolina State Univ., School of Forestry.
HAZEL, L. N. & LUSH, JAY L. 1942. The efficiency of three methods of selection. J. Heredity. 33: 393-399.
HAZEL, L. N. 1943. The genetic basis for constructing selection indexes. Genetics, 28: 476-490.
ILLY, G. 1966. Recherches sur l'amélioration génétique du pin maritime. Ann. Sci. for, 23 (4): 757-948.
ILLY, G. 1969. Les indices de sélection: exemple du pin maritime. Second World Consult. Forest Tree Breeding. FAO-FO-FTB-69-7/3.
JOHNSON, HERBERT W., ROBINSON, H.F. &; COMSTOCK, R.F. 1955. Genotypic and phenotypic correlations in soybeans and their implications in selection. Agron. J., 47: 477-483.
LERNER, I. MICHAEL & DONALD, H. P. 1966. Modern developments in animal breeding. London, Academic Press. 294 p.
LUSH, J. L. 1948. The genetics of population. Iowa City, Iowa State University. (Mimeo)
NAMKOONG, GENE SNYDER, E. B. & STONECYPHER, R. W. 1966. Heritability and gain concepts for evaluating breeding systems such as seedling orchards. Silvae Genet., 15: 76-84.
NAMKOONG, GENE. 1969. Problems of multiple trait breeding. Second World Consult. Forest Tree Breeding. FAO-FO-FTB-69-7/4.
PEEVY, O. J. 1959. Superior tree selection, loblolly pine. Proc. 5th Southern Conf. Forest Tree Improvement, North Carolina State College, School of Forestry.
ROBINSON, H. F., COMSTOCK, R. E. & HARVEY, P. H. 1951. Genotypic and phenotypic correlations in corn and their implications in selection. Agron. Jour., 43: 282-287.
SMITH, H. F. 1937. A discriminant function for plant selection. Ann. Eugenics, 7: 240-250.
STERN, K. & HATTEMER H. H. 1964. Problems involved in some models of selection in forest tree breeding. Silvae Genet., 13: 27-32.
STONECYPHER, R. W. 1966. The loblolly pine heritability study. International Paper Company, Southlands Experiment Forest. Tech. Bull. 5.
STONECYPHER, R. W. 1967. Field testing and data handling in forest genetics research. Proc. 9th Southern Conf. Forest Tree Improvement, p. 15-19.
STONECYPHER, R. W. & ZOBEL, B. J. 1966. Inheritance of specific gravity in five-year-old seedlings of loblolly pine. Tappi, 49: 303-305.
STONECYPHER, R. W., CECH, F. C. & ZOBEL, B. J. 1964. Inheritance of specific gravity in two- and three-year-old seedlings of loblolly pine. Tappi, 47: 405-407.
VAN BUIJTENEN, J. P. & VAN HORN, WILLIS M. A selection index for aspen based on genetic principles. Institute of Paper Chemistry, Progress Rept. to Lake States Aspen Genetics and Tree Improvement Group.
