Procedure for managing large-scale progeny test data: a case study of Scots pine in Finland

Abstract

Large progeny test networks are typical for conventional forest tree breeding programmes. The individual progeny tests differ with respect to age, composition and ability to screen the breeding values of the parent trees. Several approaches have been introduced to manage the unbalanced and diverse nature of the data generated by progeny tests. This report presents a procedure for ranking breeding material on the basis of ‘messy’ data. Plot means were used as input values and missing plots were estimated from least squares. The differences between test means and variances were standardised by the performance level method. The different precision of the tests was quantified through the reliability coefficient. In order to facilitate the selection of plus trees for different purposes, all the available test results were combined into a single variable that was used for ranking. Three different kinds of ranking variable were calculated and each of them proved to be more useful for the selection of plus trees than an arithmetic or weighted mean. One of them, WMEAN, relied on the reliability and number of the progeny tests, while the others, WCONF0.50 and WCONF0.10, relied on the standard error of the plus tree mean, thus emphasising the precision of the values obtained. The analyses were carried out with SAS® procedures, which require only moderate skills in statistics, programming and data processing technology. The procedure has functioned well throughout an eight-year development phase. Nearly three thousand Scots pine (Pinus sylvestris) plus trees have been ranked for various characters, and the results have been used for roguing the seed orchards, to establish new ones, and to select plus trees for breeding populations.