Submitted to: Canadian Journal of Plant Science
Publication Type: Peer reviewed journal
Publication Acceptance Date: 10/10/2004
Publication Date: 6/1/2005
Citation: Doehlert, D.C., Mcmullen, M.S., Jannink, J.L., Panigrahi, S., Gu, H., Riveland, N.R. 2005. A bimodal model for oat kernel size distributions. Canadian Journal of Plant Science. Vol. 85:317-326. Interpretive Summary: The uniformity of oat kernel size is important to the oat milling industry because they adjust milling conditions according to the size of the kernels being milled. We have evaluated oat kernel size uniformity by computer assisted analysis of photographs of oat kernels. We find the distribution of oat kernel sizes do not fit the standard bell shaped curve, but instead appears to be composed of a mixture of at least two independent bell-shaped curves. Oats develop on the oat plant in a structure known as a spikelet. A spikelet may contain one, two or three kernels. The most basal kernel, known as the primary kernel, is always the largest. Subsequent kernels, known as secondary and tertiary ordered kernels, decrease in size with increasing order. Our analyses show that double kernel spikelets are the most abundant. Thus, it appears likely that the two putative independent populations that appear in kernel size distribution analyses are derived from primary kernels and secondary kernels. There appears to be some deviations from perfect bimodal distributions, which are partially attributed to the presence of single kernel and triple kernel spikelets. Other variations appear to be due to factors not considered in this study.
Technical Abstract: Oat kernel size uniformity is important to the oat milling industry because size separation of kernels is routine in oat milling. Dehuller rotor speeds are set in order to deliver the optimal mechanical stress to different kernel size streams for dehulling. In this study, size uniformity was evaluated by digital image analysis in ten cultivars grown in eight environments. Observed distributions were compared with quality characteristics and with panicle characteristics and spikelet type frequencies. Size distributions within samples, as evaluated from individual kernel image areas, tended to depart from normal distributions and graphical depictions of data frequently resembled bimodal populations. A statistical test to compare a bimodal distribution with a normal distribution indicated that a bimodal model was more effective at describing the distributions. Panicle analysis indicated that double kernel spikelets were the most abundant spikelet type found. Because double kernel spikelets consist of one larger kernel and one smaller kernel, it is likely that the root of the bimodal distribution can be attributed to these spikelets. Although some departures from perfect bimodal distributions can be attributed to the occurrence of single and triple kernel spikelets, much of the departures from perfect bimodal distributions must be attributed to other sources of variation in oat kernel size.