Low Input Tree Breeding Strategies Dag Lindgren 1

Low Input Tree Breeding Strategies Dag Lindgren 1 and Run-Peng Wei 2, 3 1 Department of Forest Genetics and Plant Physiology Swedish University of Agricultural Sciences SE-901 83, Umeå, Sweden 2 South China Agricultural University, Wushan, Guangzhou 510642, China 3 Sino-Forest Corporation, Sun Hung Kai Centre, Wanchai, Hong Kong October 9, 2006, Turkey

No strict limit between high input and low! • Thinking low input helps making high input more efficient • Input level can vary between tiers (elite vs main)… • Other factors than budget important…

High-input techniques ® Breeding values estimated from offspring or relatives; ® Test plantations; ® Clone archives; ® Controlled crosses; ® Known pedigrees; ® ® Orchards intensively managed exclusively for seed production; Grafts for seed production.

Low input situations ® Poor; ® Unstable organisation; ® Uncertain continuity; ® No specialists; ® Minor program.

Low-input techniques ® Selection on phenotypes instead of testing of genotypes; ® No records of tree ID or pedigree; ® Wind pollination; ® Seeds derived from stands used for other purposes; ® “Cheap” plantations created for future seed production and long term improvement.

Low-input techniques ® Thin stands rather intense to get better pollen; ® Harvest seeds from best trees for production and long term improvement. But…try to make predictions of inbreeding, coancestry and diversity.

Plantations combining objectives Plantations which look and are managed similar to "normal" plantations: ® Limited need of specialised competence and organisational stability; ® Multiple use (options for seeds, improvement, wood, conservation. . . ); ® Can function as seed collection area; “cheap” trees may be cut for seed collection (climbing often too expensive); ® Not too long rotation time (to keep cones harvestable and to speed up improvement; ® Close to local organisation, enterprise and people = better and cheaper management.

Phenotypic selection ® No tree identities required; ® No computer required; ® No strict objective measures required; ® Transparent ® Can ®A (not black box); be executed immediately in field; type of selection forwards; ® Also called mass-selection; ® Similar to Nature, sustainable and environmental.

Phenotypic selection ® No separate test populations needed!

Testing doubtful low-input ® more complicated; ® more demanding on temporal and organisational stability; ® requires ® Not trust in future; needed if relaying on phenotypic selection.

Phenotypic selection ® Phenotypic selection may be as powerful or more powerful than BLUP (selection for best estimate of breeding values), as I will show. ® For following slides: Combined index selection is a breeding value estimate based on performance of an individual as its sibs. ® There are procedures for finding the most efficient selection at a certain diversity in a population of a large number of large families. I’ll show:

Maximising gain at a given diversity by selection in infinite normal distributions. h 2=0. 25 and P=0. 1 Combined index (maximizes gain) Gain Phenotypic selection (easy) Between family (exhausts diversity) Within family (conserves diversity) 0. 5 0 Mi 1 Diversity Max n Note that phenotypic selection is on the optimising curve, thus no way to get more gain without giving up diversity! Modified From Lindgren and Wei 1993

® This was ”theoretical mathematical”. To make it more realistic a simulator (POPSIM, Mullin) was used. Input close to operative Swedish conifer breeding.

Restricted selection for Phenotypic and Combined index, conciders both individual and family) in a population created by 20 parents with family size 20, h 2=0. 5. Points correspond to restriction intensity. Simulation (POPSIM). Balanced selection means 2 selections per parent 30 Phenotypic Gain 24 Combined index 18 Balanc ed 12 4 6 8 10 12 Effective number (Ns) Andersson 1999 and others 14

Note: ® Phenotypic selection as good as restricted combined index compared at same gene diversity! ® Now let’s consider without restrictions:

30 Phenotypic Gain 24 Combined index 18 Balanc ed 12 4 6 8 10 12 Effective number (Ns) Andersson 1999 and others 14

Comparing Three Selection strategies 30 Phenotypic Gain 24 Combined index Balance d 18 12 4 6 8 10 12 Effective number (Ns) 14

® In the following diversity is measured as loss of gene diversity since tree improvement started. This equivalent to status number as used in earlier figures, but scale and direction on the diversity axis changes; ® Phenotypic selection works with multigenerational programs also:

Restricted selection for Phenotypic and Combined index during multiple generations A population with a family structure, h 2=0. 5, family size 20 Gain 5 generations Selection Combined index criteria: Phenotype 1 generation 0 0. 1 0. 2 0. 3 0. 4 Loss of gene diversity 0. 5 Andersson 1999 and others

® Phenotypic selection is compatible also in a multi-generation program; ® For unrestricted selection genetic variation get exhausted. In the long run phenotypic selection give more gain; ® However, if breeding population large and heritability small, this exhaustion takes long time (next figure).

One and five generations of restricted selection in a population with a family structure, h 2=0. 05, family size 500. Low heritability and large families favor combined index Phenotype Combined index 40 5 generations Gain 30 20 10 1 generation 0 0 0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 Loss of gene diversity

Development of Gain and Gene Diversity over five generations of selection in a population with a family structure, h 2=0. 05, family size = 500, for three selection strategies. 40 Combined index After five generations Gain 30 Phenotypic 20 Balanced 10 After first generation 0 0 0. 1 0. 2 0. 3 0. 4 0. 5 Loss of gene diversity 0. 6 0. 7

® Combined index selection does not seem to give a superior gain and erodes diversity. However, comparisons can be made at variable breeding population size!

Comparison under variable breeding population size ® The size of the breeding population is under the breeders control and could be different (optimized) for different strategies; ® Comparisons were done under a fixed plant number (1280) as fixed resource; ® Simulations by POPSIM similar to earlier. Li and Lindgren 2006

Gain for phenotypic selection compared to combined index selection after first generation under fixed test plant number Combined index selection seems not inferior, and thus gets rehabilitated. Li and Lindgren

Gain for phenotypic selection compared to combined index selection after five generations Phenotypic selection is better at high heritability; The alternatives become similar efficient when the gene diversity is high; Low heritability favours combined index selection. At moderate or high heritabilities phenotypic selection seems equal or slightly superior after some breeding generations

® These comparisons assume the size of the breeding population is a free resource, and that is certainly not the case.

Multigenerational comparison of testing strategies in Swedish conifer breeding • Clonal testing is much superior to progenytesting • Phenotypic testing better than progenytesting at low budget Danusevicius and Lindgren

How may clonal testing look like in practice in low budget? Clone trial of Eucalyptus camaldulensis converted to seed orchard based on clonal performance in the trial Verghese et al 2004

Fertility variation matters for accumulation of coancestry over generations ® It can be predicted ® Female contributions can be controlled

Coancestry (inbreeding) Fertility variation matters for accumulation of relatedness over generations 0. 07 Female and male varies Female constant 0. 06 0. 05 Equal-tree fertility 0. 04 0. 03 0. 02 0. 01 0. 00 1 2 3 4 5 6 7 Generations 8 9 10 Control over female is powerful and easy (count seeds) The development over generations in a closed population of 154 teak trees based on their observed fertility variations (Bila et al. 1999)