One of the first reactions of breeders was to look at the wood quality world. There, wood basic density was presented as the canonical characteristic (Zobel and van Buijtenen 1989 is a classical example). It was not only an important variable but, luckily, there was plenty of variation; it had a high degree of genetic control; and it was easy to assess. Some may argue that it became important because it was easy to assess. Pine breeders around the world started assessing, and breeding for, basic density only to find that the intrinsic quality of trees did not change significantly.

A short detour to understand what happened. Tree breeders make, hopefully, a clear distinction between characteristics that are assessed in progeny trials (termed selection criteria) and the characteristics that we want to breed for (called objective traits). Because we are impatient creatures and time is money, we do not wait until rotation age, but we measure trees from around 6 to 8 years of age.

Stiffness and stability can be explained by basic density (DEN) and microfibril angle (MFA), but the relative contributions of these traits change with age. Thus, in the early life of trees MFA is more important than DEN, while DEN becomes more important in older trees. Life becomes interesting because we have been assessing young trees for DEN (our selection criterion) just when stiffness is dominated by MFA. The correlation between MFA and DEN is far from perfect, so making selection for wood quality problematic. This has changed only recently, with the introduction of operational acoustic velocity screening in breeding programs.

Chemical properties are another example of selection criteria that were neglected for a long time. In the last couple of years there has been a high level of interest in the role of galactans in wood longitudinal shrinkage, a relationship first reported by Stan Floyd (2004) of Weyerhaeuser in the USA. There is also a lot of interest in near infrared reflectance analysis (NIR); it has been in use in Australia for several years (Raymond et al. 2001, for example). We are unlikely to find something if we are not looking for it, and we have not payed much attention in the past.

Lesson: we cannot always rely on ‘the official story’. Complacency about our understanding of wood quality delayed progress for at least ten years. We always need to be open to alternative explanations.

A stopover on the way to China

On 24th September 2007, a Chilean newspaper published an opinion piece by Aldo Cerda, forestry manager for Fundación Chile. Mr Cerda made an interesting observation: Chilean foresters now consider New Zealand ‘a stopover on the way to China’, referring to business trips with stopovers in Auckland. His diagnosis: betting too much on the sale of technological services (consulting, software, etc), a lack of investment, poor processing infrastructure, limited ability to pay for raw materials, over regulation and arrogance. The irony of Mr Cerda’s own arrogance is somewhat comforting.

The purpose of this story is not to annoy people, but to point out that today many Chilean foresters are looking to Brazil with admiration and preoccupation. If you think that 22 years rotations on good Chilean sites are a threat, what will you think when Brazilians turn their attention to solid wood? They are already growing pine on rotations shorter than 20 years in Southern Brazil. This raises the question: Can we afford to continue thinking that 30 years is an adequate ’short’ rotation?

Sorensson and Shelbourne (2005) presented a very interesting graph in the clonal forestry chapter of the NZ Forestry Handbook. They plotted value of logs versus wood stiffness, showing a non-linear relationship. There is a big jump of price when moving from industrial grade to structural wood, which then tapers off for better grades. There is an immense amount (around 50% of volume) of poor performing wood, which derives mostly from corewood.

How do we pull together all this information? We can restate our improvement objective to maximise the value of corewood: as outerwood is already good enough. This simple approach has important implications: as shorter rotations become feasible, the assessment of selection criteria is a lot closer to the objective traits (a bonus) and we can look for trees that meet quality thresholds very early in life. Trees do not need to be spectacularly good on average, but good enough very early on.

Will we adopt this perspective? I do not know, but it is one reason for writing this opinion piece. It offers a chance to regain some competitiveness, although it flies on the face of current thinking that 30 years is short enough.

Epilogue

Coming back to the original question, of course genetics is a good investment if we are aiming for the right objective. However, breeding may not make a difference when considered in isolation. Achieving superior forests will depend not only on good genetics but also on site selection and silvicultural regimes aimed to provide decent wood quality in a reasonable time frame.

Breeders’ objectives can be adapted to include not only biological traits but also technical efficiency characteristics, converting them into improvement objectives. In this way we can value the effect of changing trait averages by any means. This makes the comparison of alternative genetics, silviculture and technology ‘projects’ not only possible but desirable.

This essay could be interpreted as an opinion dealing mostly with radiata pine, but this is far from the intention. Most, if not all, issues are applicable to other species. There have been analogies relating the small number of species in agriculture and the dominance of radiata pine in New Zealand. However, when pushing the comparison with the agricultural world one can see that few people are making a lot of money with staple crops (at least when ignoring subsidies). The money is on the specialty crops and I think that one of our mistakes has been to think of alternative species that will have a role as important as radiata pine’s. Probably a reasonable strategy is to provide diversification targeting niche markets.

Breeding is grounded on a vision of the future and my bet is that that future will be based around short rotations. Thus, when I say a reasonable time frame I mean a short, competitive one. We will have to aim for corewood quality: the rest of the tree will sort itself out.

P.S. This is the second part of an opinion piece published in February 2008 in the New Zealand Journal of Forestry. Read the first part here.

The observed properties of a tree can be explained as the result of its genetic make up (the genotype), where it is growing (the environment) and the interaction between genotype and environment. That is, the expression of traits like stem volume, wood stiffness or pulp yield will depend on the genetic value of the trees (often expressed as a GF Plus rating in NZ), the site and silviculture that constitute the growing environment, and the interaction between genetic value and environment.

Genetic value is end-use dependent; the characteristics of a good tree for structural wood may differ, for example, from a good tree for the appearance or pulp markets. This is recognised in breeding programs where — in theory at least — the first step is to define a breeding objective: a list of traits that have an effect on profit and their relative economic importance.

With hindsight it was a really unfortunate idea to call ‘breeding objectives’, well, ‘breeding objectives’. The name implies that such objectives are useful only for breeding; however, their definition makes no reference to breeding at all. Maybe a more appropriate name would be ‘improvement objectives’, because they reflect the marginal benefit of changing a trait by any means. Thus, these objectives can be used to evaluate any silvicultural or technological tool that aims to change the intrinsic characteristics of trees.

The definition of improvement objectives is fraught with difficulties: processors are much more outspoken on what they dislike than on what they want, the relationships between wood characteristics and profit are somewhat opaque, there are asymmetric information problems, to name a few (see Apiolaza and Greaves 2001 for an extended list of issues). However, it is always possible to obtain a very long list of traits that people would like to see improved. The real problem comes when trying to estimate the relative economic importance of each trait.

There have been many attempts to side step the problem of estimating economic values, because it is difficult to obtain them. However, to avoid the estimation process is to tacitly accept unknown values. In other words, ‘if you choose not to decide, you still have made a choice’ (Peart 1980).

A typical economics-free alternative is to come up with ideotypes, an idea introduced in agricultural crops by Colin Donald in 1968. The original concept considered physiological indicators that would allow varieties to be good yield performers in a communal situation, that is, to compete with many individuals of the same variety. The concept was later extended to forestry, where the situation is akin to a Christmas shopping list of desirable traits (see, for example, Martin et al. 2001). It is tempting to base breeding only on such a list but:

• Industry profit depends on multiple traits,
• There are trade offs between traits, and
• The degree of genetic control and association between traits somewhat limits the space for changing traits.

Therefore, we require both a list of characteristics that we want to improve and their relative economic importance.

Those pesky interactions

Considering that environment encompasses both site and silviculture, we can split the interaction between genotype and environment into genotype by site and genotype by silviculture. Most studies in interaction have focused on the former, where results have been encouraging — in that they show little interaction — but they were based on a limited number of sites, families and traits. There is ongoing work to extend the coverage of such research.

But what is the effect of silviculture on the performance of superior material? There are few answers to this question; however, there are indications that silvicultural effects may have been poorly recognised. For example, higher initial stockings induce higher wood stiffness, both for pines (Lasserre et al. 2005) and eucalypts (Warren 2006). The heterogeneity of planted material is also relevant, as the best growing genotypes in a mixed stand are not necessarily the best ones in pure stands (as pointed out by Sharma 2007), which should affect our testing schemes. Similarly, this result should have a profound effect on deployment, particularly of clonal varieties.

From an operational point of view the value of a tree is the aggregated sum of each trait weighted by its relative economic value. I am not aware of any studies in New Zealand that track the interaction of this composite value trait rather than studying single variables, but this is exactly what we should care about from a practical point of view.

While we may be surprised by the changes to silviculture and targeted final product (another form of interaction) in the last twenty years, one thing is clear: more changes are coming. Some of these changes — particularly in processing technology — will fall in the ‘unknown unknows’ type popularised by Donald Rumsfeld . How will a breeding program cope with future unknown objectives, which may involve different economic weights and new traits?

It may well be that we need to maintain large diverse populations (including hybrids and under represented populations) in addition to the current breeding population. A large base population that we can always go back and screen for new traits, even if it implies sacrificing some gain for current objectives.

P.S. This is the first part of an opinion piece published in February 2008 in the New Zealand Journal of Forestry. Read the second part here.

Breeding programs rely on three elements: vision, improvement objectives and breeding strategy. I have covered the second and third ones, but I have not mentioned vision before. Vision relates to our set of believes about what the future brings, including markets, products, technologies, etc. It is our way to deal with this vision that gets ‘mantrified’.

Rather than wasting time on useless and mediocre ‘mission statements’ — which nobody remembers — it is much more powerful to synthesise (or summarise) our objective in a mantra. Plus Tree’s mantra is fairly simple: ‘creating better trees’. This can be easily transformed to an elevator pitch as ‘we create better trees — through the use of genetics, statistics and economic tools — aiming to maximise industry profit’.

I have found that mantras are not only useful to remind me what are we suppose to be doing, but they are great to provide a vision to research clients. For example, one of our projects is being presented as ‘profitable shorter rotations’: short and sweet way for clients to remember where are we going and why they do need to revisit their research portfolio.

Mantras or, better yet, clear visions also provide an advantage over competing research organisations. While they try to communicate a real dog’s breakfast of projects — without much success — one can fit multiple steps in ‘creating better trees’.

A knee-jerk reaction from some breeders has been to recur to the KISS (Keep It Simple Stupid) principle when developing breeding strategies. Unfortunately, the typical reaction has been ‘let’s create this dumb down strategy because it is simple to apply’. Bzzz. Wrong answer! What they have often done is to create a glorified ‘deployment strategy’ that has almost no chance of surviving in the long term: that is, short term gain based on long term disappointment.

Breeders need to realise that what needs to be simple is the interface of the strategy. This means that we need a smooth interaction between the ‘theoretical animal’ and the people that will be implementing it. This does not mean that the strategy is theoretically simple, but that the day to day activities are a breeze to complete.

This type of interface requires the development — either in-house or through contracting the service — of tools that make life easy. For example:

1. Easy access to predicted breeding values, including desktop and online access. In addition, there needs to be an idea of the reliability of those predicted values if we are going to use them for deployment purposes.
2. Tools that make easy deciding what to select and which trees should be mated with each other (mate allocation).
3. Protocols for deployment and tools for keeping control of the availability of genetic material.
4. Easy management of the interaction between improvement and deployment objectives.

In summary, breeders need tools for dealing with the huge amount of data created by breeding and deploment activities, so it can be transformed into information.

• The program was not aligned with corporate needs (read lack of breeding strategy),
• Selections were not superior (poor genetic evaluation system), and
• There was no output from the breeding program (lack of deployment).

I will only discuss the last cause in this article.

Breeders and quantitative geneticists tend to emphasise the breeding side of a breeding program, forgetting that its main objective is to provide organisations with superior genotypes; and lots of them. Ultimately, an excellent breeding population with no material going to operational plantations is a failure, unless you are doing conservation genetics. Thus, a good benchmark to evaluate the success of a breeding program is the improvement (on terms of your breeding objective) that is actually being deployed in the field.

If you are part of a consolidated breeding program you should not have much problem finding genotypes to deploy. However, what are your options if you are just starting a program?

• Buying material from already established programs (normally from landrace selections).
• Buying material from reputable seed merchants collecting the natural distribution of the species. If buying enough seed you probably can establish criteria for collection.
• Slowly include untested genotypes in plantations. This should be a calculated risk; avoid establishing thousands of hectares with untested material. This can provide you with a preview of the results from the progeny trials.

You should keep track of the material released to operations as part of your forest management system. This information will be very valuable from the management point of view in the future, particularly if you are using family or clonal forestry compartments with a single family or clone).

Never underestimate the need for deployment gain (also called ‘realised gain’): it is the only way to keep people paying for breeding work in the long run.

This is no minor point, because it allows us to use breeding objectives to evaluate the effect of any technique (for example, fertilisation and pruning) on traits of economic importance. The main difference with genetics is that these techniques require constant use, while a change through breeding is permanent.

Although a change of names may be seem as trivial, the fact is that an action like that would make life a lot easier. As a start, it would lower the barriers for enlisting the help of other researchers (or from other units of a company) to develop objectives that would be seen as beneficial across the organisation.

Coming back to applied research projects, they require improvement objectives (or profit objectives) to determine the impact of their operational use. This way we can rank applied research by their contribution to enterprise profit, independently of the means to achieve improved plantations.

• It makes the evaluation of simple trials easy and it also makes possible the evaluation of very complex models. Note: I say possible, not easy.
• It is highly optimised for working with genetics data, so it runs orders of magnitude faster than generic software (SAS, for example).
• There is a community of users that has tested the software under a large number of situations. This means that:
  • You may ask for help and probably someone else already solved your problem, and
  • There are resources for using the program in tree breeding. The ASReml cookbook demonstrates how to analyse common trials in tree breeding.
• The program scales very well and — with some care — it is possible to conduct a national breeding evaluation in a desktop PC.
• ASReml is inexpensive, or at least much cheaper than generic statistical software (again, SAS for example).
• The program is under constant development, meaning that:
  • Bugs are quickly fixed.
  • If you have particular needs that would benefit other users they may find their way into the program very soon.

Anyway, let’s say that I didn’t convince you with the previous explanation. What are your options?

• Use generic statistical software (for example, SAS, Genstat, Splus, R, etc). You may find that besides cost — which is not a problem with R — there are two big issues: they are very slow and there is plenty of coding involved for anything relatively complex (for example a diallel).
• Use ‘animal breeding’ software, which in general doesn’t lend itself to designed experiments, making life very miserable. An exception would be VCE by Eildert Groeneveld. I am not very familiar with this program, but you could give it a try, particularly if you are interested in Bayesian estimation using the Gibbs sampler.
• Develop your own software, which requires both plenty of experience and resources. Chances are that if you are reading this article you may not have such experience. An interesting example of this approach is Treeplan.

What to do next? Download the software, browse the cookbook and try a few examples. Have fun ASRemling!

Resources

• ASReml web site. Here you can download the software and request a quote.
• ASReml cookbook, home of all things ASReml for tree breeding. There are templates for typical analyses, documentation and links to other ASReml resources.

Disclosure: I have to make clear that I receive no commission if you buy ASReml. I am only a happy user. I do have a free copy though.

Even if the title question seems to be trivial, avoiding it is one of the main reasons why many breeding programs deliver results way below their potential. For most organisations the answer to the question should be ‘we want to breed for long-term profit maximisation’. Therefore, the biological traits that they would want to change must influence profit.

If profit in your organisation depends on a single trait consider yourself lucky. That is the simplest possible situation and your breeding objective will be, for example, ‘to maximise merchantable volume at rotation age’. Nevertheless, for most of us life is not that simple, and profit will depend on several traits. As an example, if your organisation’s final products are wood chips and payment is based on tons of dry wood, profit will probably depend on volume, wood density and pulp yield.

Once you establish which traits have economic importance and are heritable (so there is hope of changing their average through breeding), the next question is ‘What is the relative economic importance of each trait?’ Another way of presenting this question is ‘how much money is worth increasing volume by one cubic metre per hectare (for example) versus changing wood density by one kilogram per cubic metre (again, for example)?’ This will guide your work as to what should be the emphasis of the breeding program.

You may be thinking that the answer to this question involves a lot of work. The answer is yes and no. A very accurate and precise answer requires an inordinate amount of work. However, a rough answer — which is all that we need — can be obtained with a reasonable amount of work and requires no more than a spreadsheet software to be worked out.

Once we have the list of traits to improve and their respective economic weights we have a formally defined breeding objective. This will be the guide for all your future breeding work and will help you to move your program fast and, more importantly, in the right direction.

Note: This article assumes that you already know a few things, including what species you are working with and what are you planning to do with the trees (pulp them, saw them, etc). If you don’t have these answers you most likely don’t need a breeding program, but actually to define the business you are in first!

Resources

• A much more detailed explanation of this problem can be found in Why are most breeders not using economic breeding objectives? (PDF 150KB).
• This site provides a simple online breeding objective calculator for pulp production.
• Yes, Plus Tree actually develops breeding objectives.