Forest Genetics Projects

Quantitative Genetics (Alvin Yanchuk, Chang-Yi Xie)

'Backward' and 'Forward' Selection Approaches

We currently use and develop analytical methods to deal with both 'backward' (parental breeding value prediction) and 'forward selection' (within family or 'animal model' predictions) based on best linear (BLP) or best linear unbiased prediction (BLUP) approaches. Backward selection software is developed in SAS based on the approaches described by White and Hodge (1989). Forward selection approaches are developed and being developed on a case by case basis, but primarily based on the approach described by Bin and Li (2003)(Can. J. For. Res. 33:2036-2043). Programs and software are modified to meet specific needs of each analyses, such as progeny being cloned, use of standard normal deviates in some cases, two-trait selection etc.

Breeding Values for Selected Orchard Parents in B.C.

The development of breeding values and a system of rating seedlots [http://www.for.gov.bc.ca/hre/pubs/pubs/0256.htm, and Stoehr et al 2004] based on their 'genetic worth' (GW) has evolved in B.C. over the last several years, to assist seed orchard managers in optimizing the potential genetic gain from seed orchard seed, as well as moving levels of genetic improvement in reforestation stock into growth and yield predictions.

As mentioned above in the section on 'backward' and 'forward' selection approaches, BLP procedures have been developed and used as standard techniques for predicting our breeding values (BV) of the first-generation tests. For the most part, BV's are predicted to an 'average site' in the seed planning zone (i.e., the elements of the 'C' matrix are genetic covariances of a test site with the average of the others). A recent publication by Xie and Yanchuk describes the analytical approach used to derive and apply breeding values for operational use in B.C.

As in animal breeding, we are increasing the use of techniques that more accurately identify and predict the genetic value or worth of selected parent trees. The use of a mother tree's offspring, whether they are cuttings (clones) or seedling offspring, are powerful techniques for examining the genetic value (i.e., breeding value) of a parent tree. With many offspring trees per parent, planted across the test site(s), some trees will sample better than average sites (i.e., micro environments), and some worse, but on average in general the statistical average or mean of the family will reflect the true genetic mean of the family.

This is not only occurring with more advanced statistical procedures but also in improving field designs we use. Over the past few decades tree breeders relied on Randomised Complete Block designs (RCB's), usually with a small number of trees per row plot (e.g., 4), and typically we laid out eight blocks or replications with the hope of 'blocking' out patches of similar micro-sites. The inherent weakness with the design has typically been that the blocks are too large and they still sample across similar 'environmental patches'. With fewer numbers of trees per family in each block, and therefore more blocks on the test site, the efficiency of the design becomes greater, and over the past decade or so most tests have been with single-tree plots. Over the past few years, however, we have moved to the use of Incomplete Block Designs (ICB's), which have been shown to be more efficient than RCB's for most situations in forest tree breeding (Williams and Matheson 1994, and a series of papers by Fu et al -- see below). In general, we believe our efficiency has increased by approximately 20%, depending on many factors and conditions of any particular experiment, with the use of these designs.

One significant problem with progeny testing is that as the test matures, competitive effects among trees take place. This will cause a bias in our estimates of family differences over time, as slower growing families will continually be outcompeted and biases among families set in. Hence, predicting genetic gain from older progeny tests has some problems, as the trees from the best families are not in a 'true plantation' situation. Over the last few decades, many realised genetic gain trials in many programs have been established to help corroborate estimates from progeny test designs to volume yields on a per hectare basis. Many institutions working with fast growing species, such as Douglas-fir and radiata pine, now have very good measures of yield per hectare, with commercially well known clones or families. In B.C., where growth rates are slower, we will be relying on estimates of yield from our GW predictions and the use of TIPSY.

However, in the mid-1990's a design for RGGT's in B.C. was developed and several installations have been put in for several species (http://www.for.gov.bc.ca/hre/pubs/pubs/0352.htm). More recently, RGGT's have been put in for lodgepole pine (Project Report 2002/2003 pages 15-16: http://www.for.gov.bc.ca/hfd/library/documents/bib91340.pdf) and several are going in for our western larch and interior spruce programs. Results from these trials will be used to confirm and 'fine-tune' per hectare yields with genetically improved materials being used for reforestation in B.C.