- Estimating breeding values
- Estimation of heritability
- Estimating non-additive effects using full-sib families
Estimating breeding values
The first open-pollinated Sitka spruce progeny-test series planted with the objectives of ranking parent trees for genetic quality relative to unimproved material of the species was planted over eight sites in 1967. Until the late 1970s, families in tests were derived from open-pollinated seed collected from parent trees in situ and stored over a number of years. When all plus trees from which such seed had been obtained had been put into test, a large number, from which no such seed was available, still remained untested. All selections had been systematically archived in clone banks since 1966 and seed of these remaining untested clones was obtained from artificial pollination carried out in the banks.
A mixture of pollen from around 20 known trees was used as the male parental contribution and efforts were made to retain consistency in the composition of this mixture across seasons. These families are referred to as 'polycrossed' and in principle should give a more reliable prediction of breeding values than uncontrolled open-pollinated families. From the late 1970s onwards, therefore, progeny testing was based on polycrossed families.
Data for height at six years from planting started to become available in the mid 1970s, and from this time onwards a breeding population began to be identified. Generally, to qualify for the breeding population, the mean performance across all test-sites of the half-sib progeny collected from a plus-tree had to exceed the Queen Charlotte Island (QCI) control value for 6-year height by 15%. The QCI origin of Sitka spruce is recognised as the most appropriate seed origin for general use across a wide variety of site types in Britain. No similar threshold values were placed on the straightness score, which was assessed according to an objective one (best) to six (worst) system. As a cost saving measure, straightness assessments were always delayed until analysis of height or diameter assessments had revealed the most vigorous families, which were then assessed for straightness (to the exclusion of less vigorous families) together with the controls.
To start with, only height and stem straightness assessments were carried out in progeny tests. Wood density was not measured until around 1986 after which the Pilodyn® was used. This machine measures wood density indirectly as the distance penetration at breast height of a blunt pin fired into the tree with a fixed force of six joules. Initially, as with straightness, only the most vigorous trees were assessed for wood density in this way. However, as it became apparent that there was a strong negative correlation between wood density and 15-year diameter, the practice of measuring all families in test was adopted.
When plus trees were selected, it was assumed that they were all of QCI origin although this was often not confirmed by forest records. No plus-trees of Washington/Oregon, or Alaskan origin were ever deliberately selected during these early selection years although this policy did change in the mid to late 1980s (see later).
Following analysis of data from earlier progeny tests planted on up to eight sites, family mean performance was regressed against site mean. In this way it was possible to investigate how stable the performance of each family was across sites and to use this information to allocate parents trees to breeding populations (Johnstone and Samuel, 1978).
For example, families which performed consistently well on all sites from Wales to North Scotland were assumed to derive from plus trees of QCI origin which were re-selected for the General Breeding Population (GBP). In contrast, if performance was above average on a site in Wales, average on a site in the Borders, and below average on a site in North Scotland, the plus tree was assumed to derive from an origin further south than QCI - perhaps Washington or Oregon and was re-selected for the Southern Breeding Population (SBP). Similarly, above-average performance in the North of Scotland was consistent with an origin further north than QCI - perhaps Alaskan - and the plus tree would be re-selected for the Northern Breeding Population (NBP).
The performance of the large majority of progeny was consistent with parent trees of QCI origin and the GBP soon became the main breeding population upon which most breeding, testing and production was focused. Following a re-analysis of all progeny-test data for 15-year diameter, stem straightness and wood density by Lee (1995), the top 240 parent plus trees were re-selected for the GBP. Selection was carried out using a multi-trait index selection model with the objective of maximising diameter and stem straightness whilst preventing an overall decrease in wood density.
Progress with the Northern and Southern Breeding Populations was originally slower. During the early years of progeny testing, only approximately 20 plus trees had been re-selected for each population based on progeny performance. This changed during 1971-3 when 209 plus-trees thought to be of a more northern origin were selected. Following progeny testing of these, a rudimentary NBP was formed in the late 1980s. Progeny test data for 207 further plus trees selected within material of the same origin have yet to be collected and analysed, but are likely to lead to further re-selections for the NBP.
Similarly, the size of the SBP is expected to increase significantly following analysis of data collected from progeny tests planted in 1990/91 with open-pollinated progeny collected from around 350 plus trees thought to be of Washington origin. Some of these plus-trees had been selected in North Scotland and others in Southern Ireland, but the majority had been selected by Rayonier Forestry in Washington State (USA).
In spring of 1993 it was decided that the first generation of Sitka spruce progeny testing should come to an end. Parent trees, which did not have progeny in test by this time would be discarded; there remained in fact only 22. Between 1967 and 1993 nearly 300 Sitka spruce progeny tests had been established (23 in the 1960s, 122 in the 1970s and 146 in the 1980s). Nearly 100 different series of experiments (involving the same families planted at a range of sites) had been planted; an average of 12 progeny tests or four series per year. Assuming each series contained 50 families, this equated to 200 families per year. The British genetic improvement programme for Sitka spruce was the largest in the world, the improvement programme had been active for 30 years and yet, by the early 1990s, the amount of improved material reaching the forest manager was minimal.
Estimation of heritability
In spring 1972 a 'Population Study' was established over 3 forest sites. One-hundred-and-fifty parent trees which were all flowering had originally been selected in a stand of known QCI origin planted in 1935 and growing in South Strome forest (Fort Augustus) (Samuel and Johnstone, 1979). Selection aimed to represent all dominance classes of trees in the population and included 6 plus trees, 48 dominants, 61 co-dominants and 35 sub-dominants.
Following losses at germination and in the nursery, a maximum of 134 families were planted at one site with different sets of 125 families planted at two other sites; 116 families were common to all three sites. The series was designed to give estimates of additive genetic variation to at least half rotation-age. There were three complete replications at each forest site and plot size varied from 7 x 7 at Garcrogo (Castle Douglas) through 6 x 6 at Wark (Kielder) to 5 x 5 at Tywi (Llanymynyfri). Height, diameter, straightness and, more recently, wood density have been measured at regular intervals in this series of experiments. Analysis of height data up to 6 years from planting was reported by Samuel and Johnstone (1979).
The experiments were intensively studied in the mid 1990s and all height, diameter and wood density data from 1 to 23-years at the Garcrogo site analysed in depth (Lee, 1997a).
Wood density was found to be under strong control of additive genetic variance. Single tree and family heritabilities varied from 0.85 and 0.95 at just nine years old to 0.33 and 0.62 at 23 years old.
Family heritabilities for height and diameter vary little over the first half of a Sitka spruce rotation; commonly around 0.60. Early single tree heritabilities for height (around 0.35) seemed to exceed those for later diameter (0.15) and are clearly under more modest genetic control than wood density.
The optimum selection age for height was 5 years after planting, confirming earlier work (Gill, 1987). More significantly it was found that the optimum selection age for wood density could be lowered to just 9-years from planting and confirmed the strong negative genetic correlation between wood density and vigour (r= -0.70).
Estimating non-additive effects using full-sib families
The first attempt to investigate the amount of non-additive genetic variation within a sample Sitka spruce population was carried out in 1968. All possible pair-crosses were carried out between seven standing trees (49 crosses in total) growing in a small block of Sitka spruce in Roseisle Forest (Moray). These comprise a full diallel crossing design, more commonly used in crop plants at that time as a method of quantifying the importance of several forms of inheritance in the population represented. A small and incomplete set of diallel crosses had previously been made in larch (Matthews et al., 1960), but this was one of the first fully complete diallels ever made in a forest tree species (Samuel et al., 1972).
Trees raised from the harvested seed were planted out in 1970 and 1971 to a site within Bush nursery and Tywi Forest (Llanymynyfri), respectively. Estimation of genetic parameters for height, diameter and stem-form over a 15-year period from planting were reported by Samuel (1991). This was the first information to become available on the amount of non-additive variation in Sitka spruce, although restricted by a small sample size and from an ill-defined population. It was found that at least 50% of the genetic variance for height and stem straightness could be attributed to non-additive effects, but that diameter was under predominately additive control.
Other full-sib crosses were planted from 1985 onwards but these lacked a clearly defined mating design. Although providing information on full-sib families which outperformed expectation based on parental breeding values, they contributed little to the estimation on non-additive variance.