Abstract

Statistical applications have played a major role in achieving the genetic gain required for the release of improved crop varieties to the agricultural industry. The linear mixed model is the workhorse for analysis of data from plant breeding experiments and can be extended to include variance models for correlated data to improve accuracy of estimation of genetic effects. Spatial correlation structures have been applied to model variation in field trials, and reduced rank variance models have been applied between environments to quantify genotype by environment interaction. More recently, genetic relationships have been incorporated into the model framework and this allows for genomic selection to be undertaken without the need for testing all genetic combinations in field experiments.
More recent advances in statistical applications are in modelling plant growth over time, as an aid for understanding phenotype response to changing temperature and moisture gradients. High throughput phenotyping platforms are developed for the field and under controlled environment conditions to capture images of real-time plant and root growth for many genotypes in response to growth stressors. Statistical methods for correlated repeated measures data over time, in combination with response profile models help plant breeders understand genetic inheritance of plant adaptation to reduced light, increased heat and variability under rainfall extremes.
Two case studies will be presented to demonstrate the role of statistics in genetic improvement of plants. Statistical methods for genetic gain will be showcased through a modernisation approach applied to plant breeding programs in Ethiopia, while analysis of high throughput imaging data will be presented for an Australian application.