Experimental Design and Analysis for Tree Improvement

A preview of the New Edition of “Experimental Design and Analysis for Tree Improvement”

Chris Harwood

03 January 2024
image_blog

The forthcoming third edition of "Experimental Design and Analysis for Tree Improvement" will shortly be published by CSIRO and CABI.  Originally authored by Emlyn Williams and Colin Matheson, the first edition of this book was published in 1994. I joined them as a co-author for the book’s second edition published in 2002. The new edition delves into significant advancements in experimental design, providing comprehensive insights into areas such as spatial designs and partially replicated designs. 

Tree breeding today 

Today, breeding programs for the world’s most widely planted tree species such as Eucalyptus globulus, Eucalyptus grandis and its hybrids, radiata pine, the southern pines and Douglas Fir are highly sophisticated. “Rolling front” breeding populations for such species include hundreds of thousands of individual trees, spanning multiple generations with complex pedigrees, and genetic trials planted across many different site types. The resulting big data sets require mixed model analysis for prediction of breeding values, and increasingly combine phenotypic and molecular genetic information. Multiple-trait breeding objectives combine desired traits such as growth, wood properties and disease resistance, weighted for their economic importance, making it important to estimate genetic correlations among the different traits. Breeders use ASReml or proprietary software to fit mixed models. 

The majority of trees planted worldwide for timber production, food crops such as cocoa and coffee, agroforestry, carbon sequestration and forest restoration are raised from unimproved planting stock, often of unknown genetic origin and sub-optimal genetic quality. The article by Lillisoe et al. (2017) examines why this is so. Thousands of different tree species and hybrids are widely planted, and it is not feasible to support intensive tree breeding programs for all of them, particularly in developing countries where available human and scientific resources remain limited. Often, growers must choose whether to plant unimproved seed, locally available at low cost, import selected material from other regions, and/or set up their own simple breeding programs. Since plantations of many tree species cover many thousands of or even millions of hectares (for example Vietnam alone now has over 2 million hectares of tropical Acacia plantations), a wrong choice can mean a failed program, foregoing many millions of dollars in net returns. 

Very simple genetic improvement strategies, for example identifying the best provenances (geographic varieties) of a tree species and then establishing unpedigreed seed production areas based on these better provenances to yield improved seed (Harwood et al. 2015), can make a big difference to the productivity and value of planted tree crops. However, even the simplest improvement strategies require a series of field experiments to test and rank different genetic varieties, be they species, provenances, seedlots from seed orchards, progenies of selected trees or clones.   

The scope and role of this book for tree breeding  

Our book takes the reader through the basics of planning, designing and establishing tree improvement trials, collecting and pre-processing data, appropriate statistical analysis of the cleaned-up data sets and the correct interpretation and application of the results.      

Through worked examples, we present simply the fundamental concepts, including the following: 

  • Trial design in relation to tree improvement objectives.  
  • The need for replication and randomization. 
  • The importance of experimental strata in design and analysis. 
  • Determining the significance (or non-significance) of treatment differences.  
  • Across-site analysis to evaluate genotype-by-environment interaction. 
  • Estimation of genetic parameters from progeny trials and their use for predicting genetic gain from breeding. 
  • Nested (e.g. provenance/progeny) and factorial treatment structures.  
  • An introduction to BLUP (Best Linear Unbiased Prediction) breeding or clonal values obtained from mixed model analysis and their use in breeding and deployment.

Chapters 7 and 8 introduce incomplete block designs, which are of great value when comparing large numbers of genetic treatments. We argue that researchers should always employ detailed blocking even when the site looks uniform, as it is often difficult to assess site variation. Some breeding organizations still have not moved beyond using randomized complete blocks; this book will enable readers to make a pain-free transition to better designs that give improved precision and ranking of treatments. 

Major developments in experimental design, including spatial designs and partially replicated designs, are detailed in this new edition. The expanded capabilities of the experimental design software CycDesigN are presented. A nice feature is that this software will set up a spreadsheet for data collection, correctly indexed for blocking factors and treatments.   

All three editions of the book have used Genstat as the primary software for the analysis of experiments. Nowadays, Genstat links easily with spreadsheet programs such as Excel used for data recording and storage, so specialized data pre-processing software advocated in the first and second editions is no longer required. Tips for data checking and pre-processing in Excel are now provided.    

Careful selection of genetic treatment levels for testing is vital in breeding. Often tree breeders test their best, say 50 or 100, families or clones but omit appropriate control treatments such as currently planted commercial sources or clones, and/or the best natural provenances. Including appropriate control treatments makes it easier to demonstrate the achievements and value of breeding programs. Worked examples show how to correctly handle control treatments in the analysis of breeding trials, when calculating genetic parameters such as heritability, and comparing controls with entries from a breeding program. 

We see this book as appropriate for an undergraduate course for tree or plant breeders and as a reference book for experimenters involved in the design and analysis of field, nursery and glasshouse trials. The book is written in the forestry context, but the statistical methodology can easily be applied to other areas such as cereal,, vegetable or fruit tree trials.  

Pre-order 

Experimental Design and Analysis for Tree Improvement is available to pre-order now.  

References and recommended literature for further exploration 

Harwood, C.E., Hardiyanto, E.H., Yong, W.C. (2015). Genetic improvement of tropical acacias: achievements and challenges. Southern Forests 77(1): 11-18.

Lillisoe, J.P. et al. (2017). Why institutional environments for agroforestry seed systems matter. Development Policy Review 36: 89–112 

About the author

Dr Chris Harwood holds a PhD in Environmental Biology at the Australian National University.  He worked with CSIRO forestry divisions from 1987 to 2018, in the areas of forest genetic resources, tree breeding and forest plantation management.  Much of his career has involved collaboration with research agencies in tropical countries.  He has authored/co-authored over 100 scientific papers and several books including Eucalypt Domestication and Breeding (Oxford University Press, 1993). He received a IUFRO (International Union of Forest Research Organizations) Science Achievement Award in 2014 and the Commonwealth Forestry Association’s SE Asia-Pacific Regional Medal in 2018. He still travels regularly to Vietnam to support tree breeding in that country.  He is a Research Associate at the School of Natural Sciences, University of Tasmania.