Weyerhaeuser Company Timberland Division Forestry Research Center Centralia, Washington Number 57 January 1964 RECOMMENDED TRAITS TO BE IMPROVED IN A BREEDING PROGRAM FOR DOUGLAS FIR by Robert K. Campbell1 SUMMARY Cost of selection limits the number of traits that can be genetically improved in Do11glas fir, Bree<;l:J_ng mu;3t be :re­ stricted to the very :tew t rait s in which fli].pi>ove1TI :fth!1 give the greatest earnings r 1l'ltive to CO$tS of'''{triprove men t By c!on$idering t:raits in relation t& the $everEti factoPs that influerice earnings, the isk bf MakiW 'a wrong ' choice is mihirnized, _' . · this No te , basic information has been gathered from ublished studies of trait heritability and v riabi ity to makE( preqict1qns d:tl geribtid ga1±;1 r6 . 6 Br fts' H1' 1D60g1!a;:J. fit>; ··va:rHibilities f6r1?-ny giveri tra t r sirh:i1a:r,'f':b m',:i s, atid to 'stand' c;tnd spec:Les te 'speb ¥:J'.'' f\1th.otigl\11 1J 1 s d h$ri tal;J'tlities 'are of qu StibniWie vliiue th. pr dictin$1 genetic 'gatri in Dbugl,as '.fir,' they do :ind f ¢t:itE! 'li' l$:t e· ' differences among'', ttlc\ti Bs. ' ' ' :, J. i Fo:r I • ' -, , I ' ' I • ' • c ' I 1' '','1 , i: ·. · '· : ,. ', I , 1' ,': Tentative pr fdii6n bf genetic gain per breeding gener­ ation indicate a low (1-3 per cent) gain in height growth and hollocellulose per cent, a medium (4-7 per cent) gain in the majority of wood-quality traits, and a high (7-15 per cent) gain in branch angle, stem diameter and stem volume, when 1 per cent of a stand are selected as superior parents. Stem volume growth per unit growing space and whole-wood specific gravity are recommended for improvement breeding 1 Forest geneticist, Weyerhaeuser Company, Forestry Research Center, Centralia, Washington Better Timber Crops Through Industrial Forest Management 2 for Douglas fir. Recommendations are based on a comparison of traits for genetic gain, selection costs, and length of breed­ ing generation, all of which influence earnings from gain re­ lative to improvement costs. Both traits promise above aver­ age response to selection, and satisfy the requirement that any trait chosen must retain its relative value over several rotations. INTRODUCTION Most organizations cannot afford the expense of genetically improving a large number of traits simultaneously. This is because many desirable traits are either negatively cor­ related one with another or are inherited independently of one another ( no correlation ) . If two traits are negatively correlated, selection for one will entail concomitant se­ lection against the other. If the traits are not correlated, selection for one carries with it no selection for the other. The problem connected with the simultaneous improvement of negatively correlated traits is clear. On the other hand, it may not be so obvious why lack of correlation between traits affects breeding costs. For example, if a given level of genetic improvement in trait A requires selection of the best individual out of 100, the same level of improvement can be obtained for traits A and B, simultaneously, only by selecting for both traits the one best individual out of 10,000. If a third non-correlated trait were added, the best individual out of 1,000,000 would be required. Cost of selection, therefore, restricts the number of traits to be improved, The choice must be narrowed down to one or a very few traits in which genetic improvement will give the greatest economic benefit. Economic benefits from tree-improvement breeding depend, in (1) The genetic gain per unit large part on three elements: time ( year ) that can be obtained by pract ical breeding methods; (2) The increased economic return directly attributable to genetic gain--this return will vary from trait to trait de­ pending on the monetary value of the separate traits at the end of unit time; (3) The cost of improvement breeding over unit time. By examining each trait in the light of these three elements, the likelihood of making the best cholce of traits should be greatly increased. Tentative predictions of expected genetic gain, as well as estimates of relative costs of improvement, can be made for several traits. This paper includes estimates of genetic gain per breeding generation for each of twenty-six traits in Douglas fir. From these, recommendations are made for the initial 3 phases of a Douglas fir improvement-breeding program for Weyerhaeuser Company, Most of the basic information has been gathered from published studies on heritability or variability. This Note also discusses several items other than genetic gain that affected the recommendations. There remains the difficult task of predicting relative mone­ tary values for each trait at those specified future dates when genetically improved trees are to be harvested. Such predictions require a rather detailed knowledge of organiza­ tional goals, and consequently, they can be made only by Company management. ANALYSIS OF BREEDING POTENTIALS To predict genetic gain three major assumptions are necessary: First, that we are dealing only with metric traits, i.e., traits which vary continuously from one extreme to the other, so that any division into classes is purely arbitrary. Metric traits are amenable to measurement, and are generally con­ trolled by many genes. Secondly, that genetic gain in the initial breeding generation will result primarily from mass selection within naturally regenerated stands. Thirdly, that only the additive component of genetic variation will be used in improvement breeding. This condition follows from the second assumption, but it may not hold true for Douglas fir after the initial breeding generation. Granting these assumptions, the prediction equation described below estimates the genetic gain to be expected per generation. Genetic gain ( R ) of a trait ( t ) results as a response to se­ lection amd is defined as the difference between the trait mean of offspring from selected parents and the trait mean of offspring from the non-selected population. Rt 2 ht S . 2 where ht is trait heritability-in-the-narrow-sense, defined as the proportion of the measurable population variance, in trait t, that can be atttibuted to additi e genetic causes. For exampie, in a hypothetical stand in which the standard deviation ( s ) of tree height has been computed to be 5.0 feet, the measuraple population variance equals s2 or 25. In this stand, each' tree has a set of genes whose average effect on tree height is slightly different from that of the comparable sets of genes in the remaining trees. Assume that the stand­ ard deviation of tree heights caused by additive genetic dif­ ferences between trees has been computed to be 1.6 feet, Variance attributable to additive genetic effects therefore equals ( 1.6 ) or 2.56. Consequently, for tree height of this stand, h2 equals 2.56/25 or 0, 10, i. e., 10 per cent of the measurable variation in tree heigh is due to genetic 2 = 4 differences between trees. The variable conditions of soil, light, competition, etc, that exist within a stand also affect height growth and are responsible for a portion of the vari­ ation in tree heights--in this example, 90 per cent. Selection differential ( S ) is the difference between the trait mean of the selected individuals and the trait mean of the un­ selected population, For example, assume that the average tree height is 50 feet at age 20 in the above stand. By selecting the tallest 5 per cent of trees, the av rage height of the selected individuals is 60 feet, and the resulting selection differential is 10 feet, Under 5 per would stand the conditions hypothesized above, using the tallest cent of trees as parents, their offspring at age 20 be expected to excel average tree he1ght of the original by 0, 10 x 10 feet or 1 foot, Although the above equation clearly demonstrates the reasoning behind genetic-gain prediction, a slight modification of the equation is helpful when comparing relative genetic gains in separate traits or in separate populations. This is because the selection differential ( S ) for any particular trait de­ pends on (1) the proportion of the population selected and (2) the variability of the trait. Thus, genetic gain depends on trait heritability, trait variability and selection intensity. A generalized equation, which takes all factors into account, becomes R= i 0 p h2 where: crp i = = · the standard deviation of the trait, a measure of trait variability. the intensity of selection defined as the selection differential ( S) expressed in standard deviations of the trait, or i S/crp· An estimate of genetic gain results when the appropriate sta­ tistics are substituted in this prediction equation. For the following treatment, heritabilities and standard deviations have.been taken from the literature to provide likely esti­ mates for Douglas fir, Data reliability is discussed, and following this, predicted genetic gains are presented. Heritability For any one trait within a given population it is possible to calculate two general types of heritability--heritability-in­ the-narrow-sense and heritability-in-the-broad-sense. Of the two, heritability-in-the-narrow-sense is the more useful since it can be used most directly to predict response to mass selec­ tion, It has been defined above as 5 2 s = additive genetic variance in population total variance observable in population. For many traits) however) the only available estimates of heritability are reported as heritability-in-the-broad-sense) defined as 2 additive genetic variance + variance due to domi­ hbs nance and interaction total variance observable in population. Heritability-in-the-broad-sense contains non-additive genetic variance components and cannot be used directly to predict response to selection. Nevertheless) broad-sense herit­ abilities are not without value. The numerator of the broad­ sense heritability fraction contains both additive and non­ additive components of genetic variation and henceJ provides an estimate of the upper limit of narrow-sense heritability for a given trait in a given population. The equations indicate that heritability is a ratio of the genetic variability in a population to the total observed variation in that population, Neither the numerator nor the denominator of this ratio is necessarily constant from popu­ lation to population. Genetic variations may differ between populations) affecting the ratio numerator; or environmental variability may differ) affecting the denominator. For example) heritability for growth traits should be higher in plantations where initial stand density and spacing is less variable) than ih wild stands) where density and spacing have not been artificially cbntrolled. From thisJ it is clear that heritability is a property of each stand or population and that heritability estimates have limited utility in pre­ dicting response in one stand from measurements made in an­ other. Since he itability studies for Douglas fir are virtually lack­ ing) the estimates for traits in Douglas fir are based almost completely on those made for other species. Even though such estimates are of questionable value in predicting genetic gain in Douglas firJ they do indicate relative differences among traits in degree of heritability. Published herit­ ability values are presented in Table 1. Variability of traits Variability of a trait in a population is extremely important: it sets practical limits to the magnitude of the selection differential) and consequently) to the genetic gain per gene­ ration. Unfortunately) trait variability in a stand is sub­ ject to little useful control by the breeder) who generally 6 Table 1, Heritability estimates for twenty-six traits from several species as obtained from published liteTature. . Trait Species Total height II II Annual ht, II increment II II Juvenile ht, II II II II II growth II II II Stem volume Diameter breast high II II II Juvenile stem diameter Stem-diameter increment II II II Number branches/whorl Knottiness ratio Branch length Branch angle Blister rust resistance Number female flowers Wood specific gravity II II II II II II II II II II II II II II II II II II Summerwood specific gravity II II Summerwood per cent II II II II Tracheid length II II Slash pine Cryptomeria: Scots pine Douglas fir Larix x eurolepis Western white pine II II II II Age II II Douglas fir Slash pine II II Cryptomeria Larix x eurolepis Scots pine Monterey pine Douglas fir II II Slash pine Scots pine Western white pine Scots pine Monterey pine II II Slash pine Loblolly pine II II II II Monterey pine II II Loblolly pine II II Monterey pine Slash pine Monterey pine Slash pine Summerwood tracheid: Double wall thickness Loblolly pine 11 11 Radial lumen diameter 11 11 Radial width 11 11 Tangential width 11 11 Tangential length 11 11 Ratio of double wall 11 11 Thickness t.o radial width Monterey pine Spiral grain 11 11 Longitudinal shrinkage 11 11 Holocellulose per cent 9-14 20 6- 9 20-35 2 2+ 1 1 14 14 20 2 6- 9 8+ 20-35 20-35 9 6- 9 3- 4 6.L 9 13-1 9 ·6 12-14 2 6 7- 8 8+ 8+ 7- 8 7- 8 12+ 12-14 8+ 6 77777­ 7- Heritability per cent Narrow Broad sense sense 52* 2-30* 28* 50-63 1-39* 16-36* 71"*: 38* 50-70 73 45-75 45-54 45-72 48 73-86 56 8-14 6-14 8-20 0-10 18-35 29-58 26 0-10 12-48 69 20 21-56 37-49 64-100 76-87 50-76 86-92 8-26 0-84 31-A4 0-81 27-59 85-97 0-72 8 8 8 8 8 8 8+ 8+ 12+ 5-16 26 Refer­ ence 18) 20) i 17 10 24 18 18 20 14 1 5 '3 1 i ) 1 2 1 8 8 19 21 21 l § 5 9 9 5 19 ) 5) 27) 9 9 9 9 9 9 66 54-74 29 *Repeatability, which is measured by intraclass correlation, was determined from data presented by the cited author. Repeatability is the upper limit of heritability-in­ the-broad-sense for populations sampled by the above asterisked s.tudies. 7 wishes a large amount of genetic variation in the stands from which he makes selections. Greater genetic variation tends to give rise to greater heritability and variability. This, in turn, promises greater response to selection. Variability can be easily decreased in traits such as height or stem diameter by cutting procedures which eliminate extremes of the population. On the other hand, variability can be increased only at con­ siderable expense, by hybridization or by special mating sys­ tems, in long-term improvement programs. For these reasons, trait variation may be considered as being capable of only slight change. Coefficients of variation are reported in Table 2 for many traits. For a few of these, variability values are reported for several species. Variabilities for any given trait are sur­ prisingly similar from stand to stand and species to species. Intensity of sel·ect-ion Intensity of selection remains the only factor in the genetic­ gain equation over which the breeder has direct control. He will generally attempt to select as intensively as possible with­ in limits set by: (1) the percentage of the population needed to reproduce the populationj (2) the number of families or clones needed to minimize inbreeding in orchardsj and (3) the diffi­ culty and consequently expense of trait measurement. Fortunately, most coniferous forest trees have relatively high reproduct ve potentials, each tree being capable of producing many offspring per year, In Douglas fir less than 0.1 per cent of trees cut yearly from a given acreage could produce numbers of seedlings sufficient to replant the resulting cutover acre­ age. ( Kozak, et al. 1962). Thus reproductive potential is not likely to restrict selection intensity in this or similar species, Since large populations from which selections can be made are usually available, the second restriction is also re­ latively unimportant. On the other hand, selection intensity probably will be limited by the difficulty of measuring or scoring traits. For some traits selection intensity will be much more limited than for others. For this paper a selection intensity (i) of 2.67 has been chosen, meaning that the selected parents are superior to the original population by 2.67 standard deviations from the mean when one per cent of the original population are selected as superior parents. This will generally approximate the upper limit of selection intensity practicable for most traits. It is also assumed that the trait values are distributed normally within the populations, and that population size is infinite. Neither of these assumptions is strictly true, but the result­ ing error will be negligible for purposes of this Note. 8 Table 2, Trait variability estimates derived from published and unpublished data for several traits for several species, Trait Total height Annual height increment Juvenile ht, growth Diameter breast high Stem volume Stem diameter increment Number branches/whorl Knottiness ratio Branch length Branch angle Number of female flowers Wood specific gravity Species Age of stand Slash pine Cryptomeria Douglas fir a,Dominants b,Dominants & codominants Douglas fir a,Dominants b,Dominants & codominants Scots pine Douglas fir 8 20 II b,Dominant & codominant c,Dominants d,Dominants & codominant e,Dominants f,Dominant & codominant Douglas fir a,Dominants b,Dominant & codominant c,Dominant d,Dominant & codominant e,Dominants f,Dominant & codominant - - - - --- Scots pine Western hemlock Douglas fir II Tracheid length II Scots pine II II Douglas fir Scots pine Slash .pine Longleaf pine Loblolly pine Shortleaf pine Douglas fir II Summerwood specific gravity Summerwood per cent II Slash pine Cryptomeria Douglas fir a.Dominants II Western hemlock II II Loblolly pine II II Douglas fir Loblolly pine Douglas fir Western hemlock Summerwood tracheid: a.Double wall thickness Loblolly pine b,Radial lumen diameter c,Radial width d,Tangential width e,Length f.Ratio of double wall thickness to radial width Loblolly pine Holocellulose per cent Western hemlock Douglas fir 59-78 Coefficient of variation (CV= s/ within stand x Type of stand ) Plantation 9.2 23.2 II Natural II 59-78 l 14 20 33 Plantation Natural Nursery Plantation II Natural-medium site II 33 59-78 Natural-low site 59-78 56 Natural-high site II (11) 11.0 (11) 2,9 (12) 9.8 11.6 8,4 30.0 38.6 32.7 12) 1 20 15.3 (12) 21,2 15.8 gn g 24.7 19.6 II 56 (12) (12) Natural 32.5 33 45 II 46.0 ' 5.4 45 56 56 6- 9 80+ 80+ 20-35 20-35 20-35 6 6- 9 20-35 6- 9 30+ II Plantation Natural II II II Plantation II Natural Plantation II Natural II II Second growth­ 77777- 8 8 8 8 8 7- 8 15-35 80+ 80+ 40,2 34.5 ?g g 48,0 9.6' ----- i) -- -- 11.8 13.5 22,2 20.7 11.1 21.1 24.7 11.5 53.1 11.3 11,1 10.4 10.4 9.9 18.3 4.8 80+ 80+ 7- 8 7- 8 80+ 25 80+ 80+ -- Plantation II II II Natural II Plantation II II II II Natural II {) 22,6 33 II ?6 8,0 56 56 6- 9 20-35 Refer­ ence 5.0 20,0 5.5 9.3 5.5 8,8 (12) 23) 23) 3 3 3 1 1 3 1 16 15 15 15 6) ? §j j 23 26 23 23 5 5 6 2 6 9 9 9 9 9 6 1.4 1.5 1.2 9) §{) 23 - 9 Predicted genetic gain Reliability of predicted genetic gain depends, of course, upon the reliability of the statistics used in the prediction equa­ tion. For present purposes, a single selection intensity of 2.67 has been used for all traits even though this is not com­ pletely realistic for allj selection intensity for the more readily measured traits such as stem diameter and stem height will likely approximate 2. 67, whereas selection intensity for traits requiring more costly measurement will undoubtedly be considerably lower. Another source of predictive error lies in the heritability figures extracted from the literature. For most of the traits considered here, heritability figures are not available for Douglas fir, or even for other species having stand conditions similar to those common to Douglas fir, Furthermore, reported heritability figures are generally based on few individuals, or upon family averages from young plantation-grown progeny testsj such conditions, respectively, tend to make estimates unreliable, or too high to be applied under conditions of mass selection in natural stands, In this Note, therefore, for purposes of prediction, an arbitrary range of herit­ abilities has been selected based on the data given in Table l. In general, the heritability used is conservative when compared to reported heritabilities and the bias will likely show up as underestimation of response. Means and standard deviations used to compute the coefficients of variation of Table 2 also vary in reliability. For many of the traits considered here, standard deviations and means were not reported, as such, by the authors cited in Table 2, and it was necessary to work with data originally presented otherwise, For example, standard deviations for a trait were often der i ved from reported ranges for the trait, or from variance analysis tables, Data for variability estimates were taken from stands of diverse origin and condition. In spite of these approximations, coefficients of variation for individual traits do not vary greatly between species and stand conditions and it seems unlikely that variability esti­ mates will unduly bias the estimates of genetic gain. Gain to be expected per breeding generation is presented in Tables 3 and 4. In Table 3, gains are based upon herit­ abilities, chosen as discussed above, combined with estimates of variability for traits in Douglas fir, as given in Table 2. In Table 4, provisional predictions of gain have been made for traits for which estimates are lacking, either for herit­ ability in any species, or for variability in Douglas fir. Estimates are given in two ways, In column 6 of Tables 3 and 10 Table 3. Estimated genetic gain following one·generation of selection for the bet i er cent of the population from naturally regenerated stands of Douglas fir·, Selection under stand conditions Trait - Total· height · Annual ht, increment Juvenile ht, growth Diameter breast ht, Age Low site Dominant only Dominant & codominant High site-­ Dominant only Dominant & codominant Dominant & codominant Nursery Low site -­ Dominant only Dominant & codominant High site Dominant only Dominant & co dominant Medium site -Dominant only Dominant & codominant High site Dominant only Dominant & codominant Medium site Dominant only Dominant & codominant Dominant only Dominant & codominant Dominant & codominant Dominant & codominant Dominant & codominant Dominant & codominant Dominant & codominant Dominant & codominant Dominant & codominant Dominant & codominant Dominant & codominant Dominant & codominant -- stem.volume -- - stem diameter increment Number branches per whorl Branch length Branch angle Knottiness ratio Wood specific gravity Summerwood per cent Tracheid length Holocellulose Estimated heritability per cent Mean for. trait Genetic gain 1 per Percentage genetic gain 59-78 10 100 2,1 ft. 2 59-78 10 89 2,6 ft, 3 56 10 148 1.2 ft. 1 56 10 137 3.6 ft. 3 20-35 1 10 10 2.5 6.5 ft. em 0.06 .52 ft. em 2 8 59-78 20 13.2 in. 1.12 in, 8 59-78 20 10.5 in, 1.38 in, 13 56 20 21.8 in, 2,29 in. 10 56 20 18.4 in, 2,24 in. 12 33 20 12,7 in, 1.04 in, 8 33 20 10,8 in, 1. 22 in, 11 56 20 145 cu,ft. 27 cu,ft, 18 56 20 102 cu,ft. 11 cu.ft. 26 45 20 48 cu,ft, 1.4 cu,ft. 3 45 33 20 20 34 27 cu,ft. cu,ft, 33 20 20 cu,ft. 80+ 10 3.3 20-35 10 8.1 20...:35 5 9.5 20-35 30 71° 20-35 10 Second­ growth mm 5 7 cu,ft, cu,ft, 22 17 5 cu,ft, 25 0,134 mm ,48 ft. 106 4 6 .14 ft. 2 6,6° 9 5.9 6 20 .413 ,022 5 80+ 20 .448 .o44 8 80+ 30 44% 80+ 30 3.64 80+ 10 80.3% 2% mm .16 4 mm ,26% 4 <1 Table 4, Estimated genetic gain following one generation of selection for the better 1 per cent of the population from naturally regenerated stands of Douglas fir. Selection Estimated Mean Percentage under stand Trait heritability for Genetic genetic conditions trait per cent gain gain Surnmerwood Specific gravity Surnmerwood tracheid: a.Double wall thickness b.Radial lumen diameter c.Radial width d.Tangential width e.Length f.Ratio of double wall thickness to radial width Wood: Modulus of rupture 11 11 Modulus of elasticity 11 11 Compression perpendicular to grain 11 11 Compression parallel to grain 11 Shear 11 11 Plantation 30 y 1/ 2 Plantation 30 30 30 30 30 1/ I/ I/ 1/ I/ I/ I/ 1/ 4 4 4 2 5 30 y y 5 11 11 11 11 11 1/ 1/ Dominant & codominant Slow grown Rapid grown Slow grown Rapid grown 10 10 10 10 2/ "2/ "2/ "2/ 7473 psi 6950 psi l503Mpsi 1387 psi 71 psi 326 psi 60Mpsi 78 psi 3 5 4 6 Slow grown Rapid grown 10 2/ 10 "Z/ 367 psi 375 psi 28 psi 30 psi 8 8 Slow grown Rapid grown Slow grown Rapid grown 10 10 10 10 2/ 2/ "Z/ 2/ 3717 3480 931 905 psi psi psi psi 138 190 26 33 psi psi psi psi 4 5 3 3 Variability y No estimate of variability available for these traits in Douglas fir. estimates for loblolly pine were used in genetic gain predictions. 2/ No heritability estimates based on experimental evidence are available for these traits. I-' I-' 4J response to selection is presented in actual units of gain. For exampleJ the first line of Table 3 indicates that height growth of Douglas fir stands 60-80 years old can be genetically improved by about 2 feet in one breeding generation. It is very likelyJ however, that one may wish also to compare the re­ lative response to selection of two or more traits. This can best be done by comparing percentage genetic gain per gener­ ation. This value is presented in column 7 of Tables 3 and 4. The prediction formula here is: Percentage genetic gain per h2 x i x CV where h2 is heritability in breeding generation the narrow senseJ i is the selection intensityJ and CV is the coefficient of variation of a trait. = In Table 5 the traits have been placed in percentage gain classes. It can be seen that a majority of the estimates fall in the 4-6 per cent class. Factors to consider other than genetic gain per generation The most realistic criterion for choosing traits for improve­ ment-breeding is earnings from genetic gain relative to costs of improvement. Since expenditures are typically levied from year to yearJ it is clear that calculation of earnings must also be adjusted to a yearly basis. This in turn requires an estimate of genetic gain per yearJ which can be easily com­ puted by dividing genetic gain per generation by years per breeding-generation interval. From the definition of genetic gain per year it is apparent that traits requiring shorter breeding-generation intervals will provide more genetic gain per year than will comparable traits with longer breeding generations. A generation in­ terval is the interval between corresponding stages of the life-cycle in successive generations ( FalconerJ 1961). A breeding-generation interval may be similarly definedJ but it must include the stage of the life cycle at which a given trait may be evaluated. In agricultural crop breedingJ trait evalu­ ation and sexual maturity usually occur at about the same ageJ but in forest trees, trait evaluation may follow sexual ma­ turity by several decades for some traits. ThereforeJ breeding­ generation interval is not strictly controlled by the length of the sexual cycle, It depends to a large extent on the age at which a trait may be evaluated. This means that in any one species of forest treesJ different traits may have widely different breeding-generation intervals. Choice of traits to improve should consequently consider breeding-generation in­ terval. This interval will greatly affect genetic gain per year. The importance of genetic gain per year brings up another point which should be examined: the importance of choosing 13 Table 5. Estimated percentage-genetic-gain per generation for twenty-six traits for Douglas fir. Estimated percentage­ genetic-gain per generation Trait 1- 3 Total height, annual height increment, branch length, holocellulose percehtage, shear strength, springwood specific gravity, summerwood specific gravity, tangential width of tracheids. 4- 6 Diameter increment per year, number branches per whorl, whole-wood specific gravity, per cent summerwood, knotti­ ness ratio, modulus of rupture, modulus of elasticity, compression parallel to grain, tracheid double-wall thickness, tracheid radial lumen-diameter, tracheid radial-width, tracheid length, ratio of tracheid double-wall thickness to tracheid width. 7-10 Diameter breast height of individual trees, branch angle, compression perpen­ dicular to grain. ll-15 Stem volume of individual trees. 16-20 Spiral grain. 14 traits having a high probability of retaining their relative value over more than one economic rotation. This interval will also include several breeding generations. With every added breeding generation, within limits, genetic gain per generation is likely to increase, because more sophisticated breeding methods will likely be used in each succeeding generation. For example, in a long-term breeding program, heritabilities can be increased by reducing environmental variability through better stand management, Additionally, it will be possible to employ selection methods more accurate and more appropriate to forest trees than is now possible, including selection for general and specific combining ability, Inbreeding to collect favorable genes will also be possible, Each improvement in technique will tend to increase genetic gain per year. Since improvement in every generation is added to improvement in the previous generations, a long-term breeding program per­ mits truly noteworthy advances in genetic improvement. Hence, every effort must be made to insure that the relative economic weights now placed on traits are likely to be valid over several generations. RECOMMENDATIONS CONCERNING CHOICE OF TRAITS Branching traits Although number of branches per whorl, knottiness ratio, and branch angle all promise a fairly rapid response to selection ( Table 5), none of these traits can be recommended for first­ generation selection from native stands. There are three reasons for this conclusion: First, the traits are expensive to measure in large trees, with the result that selection differentials are likely to be low and improvement is likely to be less rapid than predicted in Table 3. Secondly, the effect of these traits on wood quality has not yet been pro­ perly evaluated for Douglas fir. Thirdly, preliminary evidence indicates a positive genetic correlation between high stem volume and greater knottiness, or between high stem volume and greater numbers of branches per whorl ( Campbell, 1963). Selection for desirable branching traits may, there­ fore, entail a concomitant selection for lower stem volume. Wood-quality traits Choice of wood-quality traits to improve may well be made on some basis other than genetic-gain per generation, since there appears to be little basis for differentiating between traits in respect to their relative response to selection. Most wood­ quality traits show moderately rapid response to selection ( Table 5). For several reasons wood-quality improvement should be re­ stricted to whole wood specific gravity during the initial breeding generation. First, in an operative breeding program, specific gravity probably can be improved more rapidly than other wood traits within the gain category of 4-6 per cent. Specific gravity is somewhat more easily and more cheaply measured than other traits. Consequently, selection intensity can be greater than for other wood-quality traits. Secondly, specific gravity may be evaluated early in a tree's life and Van Buijtenan the breeding-generation interval shortened Although the same may be true for some of the other wood-quality traits, such has not yet been dem onstrated. ( 1962). Whole-wood specific gravity is chosen for two reasons. First, it is important in all phases of wood utilization. In lumber, in poles and piling, and in veneer manufacture, it affects strength, workability, preservative intake, etc. It affects pulp yield because it is a measure of actual wood substance in a given volume of wood, and it also strongly affects the quality of paper produced from pulp. S econdly, by choosing whole-wood specific gravity over other wood-quality traits the risk of making a wrong choice is minimized. Specific gravity is known to be phenotypically closely correlated with other wood-quality traits Wangaard Goggans Fairly strong genetic correlations between specific gravity and Goggans We can other wood traits also appear to exist expect, therefore, that selection for high or low specific gravity will affect the other associated wood-quality traits. Hence, if two separate strains are developed, one for high and the other for low specific-gravity wood, progress will also be made in one or the other toward improving the associated and more fundamental properties such as tracheid length, diameter, etc. Once the value of these fundamental traits in wood uti­ lization has been determined, they can be easily introduced into selection indexes for succeeding breeding generations. Since previous selectioh for specific gravity would have brought with it some improvement in these other properties, there ap­ pears to be little risk in using specific gravity as the in­ itial wood-quality criterion. ( 1950, ( 1962). 1962). Growth traits It Height growth is not a desirable trait for improvement. shows relatively poor response to selection whereas, on the basis of available information, both stem diameter breast-high and stem volume growth of individual trees may respond well. Since stem diameter is a component of stem volume, stem volume might appear to be the logical choice for improvement--yet stem volume may be prohibitively difficult to measure. Then 16' selection for greater stem diameter would probably result in gain in stem volume almost equivalent to that obtainable by direct selection for volume, By so doing, however, the breeder would risk a change in stem form brought about by improvement in stem diameter breast high at the expense of stem form and, eventually at the expense of volume, If the breeder defines the trait "stem volume" as volume growth per unit growing space per year, the choice of stem volume, as a trait to improve, also satisfies the requirement that any trait chosen must retain its relative value over several gener­ ations, By selecting for volume growth, as defined above, the breeder is essentially attempting to improve land productivity by increasing the efficiency with which a tree utilizes grow­ ing space, It is difficult to visualize a situation where de­ creased productivity of land would be desired by a forest manager, Unfortunately the information that permits us to predict re..,. sponse to selection for volume growth, permits us to predict only the genetic gain in volume growth for individual trees. This may be a trait completely separate from volume growth per unit growing space per year. If so, there would be no basis for extrapolating genetic-gain predictions from one trait to the other, Disease, insect and animal-damage resistance In this Note, no previous mention has -been made concerning selection for pest resistance, Although little has been pub­ lished in this aspect of tree-improvement breeding, the avail­ able information indicates the possibility of moderate to rapid response to selection, Nevertheless, it is question­ able whether first-generation selection should include pest resistance unless pests are the major limitation to growth in the species concerned, 17 LITERATURE CITED I,, ArnborgJ T. and G. Radder. 1957. Studies of some forestry qualities in clones of Pinus silvestris. Meddel. nr, 87 f. s&llskapet f r praktisk skogsf8r#dling. Uppsala. pp 125­ 157. 2, BinghamJ R. T.J A. E. SquillaceJ and J, W. Wright. 1960. Breeding blister rust resistant western white pine II. First results of progeny tests including preliminary esti­ mates of heritability and rate of improvement. Silvae Genetica 9(4): 33-41. 3. CampbellJ R. K, 1961. Phenotypic variation and some esti­ mates of repeatability in branching characteristics of Douglas-fir. Silvae Genetica 10(4): 109-118. 4. CampbellJ R. K. 1963. Phenotypic correlation among branch and upper-crown stem attributes in Douglas-fir. Forest Science 9(4): 444-451. 5. DadswellJ H. E.J J. M. FieldingJ J. W. P. NichollsJ and A. G. Brown. 1962. Tree to tree variations and the gross heritability of wood characteristics of Pinus radiata. Tappi 44(3): 174-179. 6. DrowJ J. T. 1957. Relationship of locality and rate of growth to density and strength of Douglas fir. u. s. D. A. Forest Products Laboratory Rep. No. 2078. 56 pp. 7. FalconerJ D. S. 1961. Introduction to quantitative ge­ netics, Oliver and BoydJ Edinburgh and London. 365 pp. 8. FieldingJ J. M. and A. G. Brown. 1960. Variations in the density of the wood of Monter y pine from tree to tree Forestry and Timber Bureau. Comm. of Australia. Leaflet No. 77. ,I 9. Goggans J J. F. 1962', The correlationJ variationJ and inheritance of wood properties in loblolly pine (Pinus taeda L.). N. C. State CollegeJ School of ForestryJ Technical Report No. l4J 155 pp. 10. HanoverJ J. W. and B. V. Barnes. 1962. Heritability of height growth in year-old western white pine, Southern Forest Tree Imp. Comm,J Sponsored Publ. No. 22J pp 71-75. 11. KerJ J. W. 1952. An evaluation of' several methods of estimating site index of immature stands. Forestry Chronicle, 28(3): 63-74. 19 24. WilsonJ B. C. 1962. Methods of selection in juvenile populations of Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco.) of known parentage, Master's ThesisJ Univ. of Wash. :73 pp. 25. ZobelJ B. J. and R. L. McElwee lulose in loblolly pine. Tappi 1958. Variation of cel­ 41(4): 167-170. 26. ZobelJ B. J.J F. GoggansJ T. E. Maki and F. Henson. 1961. Some effects of fertilizers on wood properties of loblolly pine, Tappi 44(3): 186-192. 27. ZobelJ B. J.J D. Cole and R. Stonecypher. 1962. Wood properties of clones of slash pine. Southern Forest Tree Imp. Comm.J Sponsored Publ, No. 22J pp 32-39. l-22-64 rcc