Use of linear models to predict the date of flowering in cultivars of subterranean clover (Trifolium subterraneum L.)

Abstract

Time to flowering (f) of eight phenologically diverse subterranean clover ( Trifolium subterraneum) cultivars was measured on nine sequential outdoor sowings in one season at Launceston, Tasmania, and one sowing date and four seasons in Perth, Western Australia. Linear models were then used to relate the rate of progress towards flowering (l/f) to mean temperature and mean photoperiod prevailing between sowing and flowering. The models accounted for most (70-97%) of the variation observed in time to flowering within cultivars, consistent with all of these cultivars being quantitative long day plants and photoperiodic effects being modulated by temperature. Model parameter estimates varied between cultivars suggesting genotypic variation for sensitivity to both photoperiod and temperature. The models and respective parameter estimates were subsequently used to predict flowering behaviour of the same eight cultivars in seven sequential sowings at Katanning, Western Australia. Across all 56 cultivarxsowing combinations at Katanning, there was a close (1:1) agreement between observed times to flowering and those predicted, with the models accounting for most of the observed variation. The models were least effective where times to flowering were longest, viz. for April sowings of the latest-flowering cultivars, where flowering tended to occur later than predicted. The potential utility of the approach is discussed in relation to cultivar improvement.