Weight of individual wheat grains estimated from high-throughput digital images of grain area

Jinwook Kim,Roxana Savin,Gustavo A Slafer

doi:10.1016/j.eja.2021.126237

Abstract

Average grain weight (AGW) is a major component of wheat yield. When attempting to elucidate mechanisms behind treatments effects on AGW, the distribution of the weight of individual grains may be critical. Determining the individual weight of thousands of grains in each sample would be unmanageable. Then, when individual sizes must be considered, researchers either weigh individually a very minor proportion of the grains or determine for the complete sample individual linear dimensions (length, width, area) through an image processing equipment. We aimed to generate a single model equation to trustworthily convert grain linear dimensions to grain weights. Firstly, we used a set of data to build and calibrate a model for the relationship between weight and linear dimensions of individual grains. Then, we validated the model calibrated with independent data. Grain area was a better predictor of grain weight than length and width of grains. Initially, we generated a single linear model but (i) the intercept was incongruently negative and therefore (ii) we forced the linear regression through the origin, but that consistently overestimated the weight of small grains and underestimated large grains. Finally, we fitted the data again with a power curve model and forced the intercept to zero (with the log-transformed data) obtaining the model (ŷ = x1.32) to estimate individual grain weight from grain area. The model was validated with (i) independent data from the same studies used to build the model, (ii) data from other completely independent experiments, and (iii) data from the literature. Considering the diversity of genotypes and environments in the model generation and validation, the proposed power curve model could be trustworthily used to estimate grain weights from measured areas.

Highlights

Wheat yield comprises two major components: the number of grains per m2 (GN) and their average grain weight (AGW) (Frederick and Bauer, 1999)
For researchers attempting to offer a mechanistic explanation for differences in AGW, the level of detail of having the weight of each in dividual grain, with which to analyse the grain weight distribution in particular treatments, may be more relevant than having only their average weight
The percentage of variation in grain size was better explained by the area than by either the length or the width of grains, and the magnitude of variation in grain area was more proportional to that in grain weight than variation in grain length or width (Fig. 1; Table 1)

Summary

Introduction

Wheat yield comprises two major components: the number of grains per m2 (GN) and their average grain weight (AGW) (Frederick and Bauer, 1999). The most frequent procedures are (i) to estimate yield through weighing the threshed grains from the sample, counting them and estimating AGW as the yield-to-GN ratio, or (ii) to determine yield directly from the combine-harvested plot and measuring AGW (from sub-samples of a certain number of grains) and estimating grain number as the yield-to-AGW ratio. In many cases, these two components are enough to reach sound interpretations for the particular aim of the studies. Among a longer list of examples, it was necessary to determine AGW and the individual weight of particular grains for a mechanistic explanation of (i) a negative relationship between AGW and GN in response to yielding condition (e.g. Acreche and Slafer, 2006), (ii) the effects of awns on potential grain size (Sanchez-Bragado et al, 2020), (iii) the relevance of pericarp characteristics in determining grain size (Hasan et al, 2011; Brinton et al, 2017; Herrera and Calderini, 2020), and (iv) the genetic and molecular basis of trade-offs between

Objectives

Methods

Results

Conclusion