An attainable harvest index in rainfed wheat

Context Yield potential is a benchmark for gap analysis in crops under irrigation or high rainfall. A more relevant benchmark for rainfed crops is the water-limited yield potential, defined as the yield of a crop with no other manageable limitation than water supply. In this context, an attainable harvest index can be more relevant to rainfed crops than the theoretical physiological maximum (∼0.64) calculated on the basis of the largest feasible redistribution of material from stem and sheath, leaf lamina, and chaff to grain. Method We compiled datasets on spring wheat in southern Australia (n = 576) and northeastern Spain (n = 113), and on winter wheat in central USA (n = 2413) to explore a putative upper limit of harvest index in rainfed crops. Sources of variation included soil and weather (mostly rainfall and temperature), cultivar, and management (nitrogen fertilisation, sowing date, foliar fungicide, plant population density), which returned yields from 0.13 to 8.1 t ha −1 in Australia, from 0.3 to 9.3 t ha −1 in Spain, and from 0.7 to 10.3 t ha −1 in the USA. Using yield as a proxy for environmental conditions, we plotted harvest index as a function of yield and used a censored bivariate normal model to objectively tests that a boundary function is plausible, and that any limiting upper bound is exhibited by the data set. Unlike other methods ( e.g., quantile regression), this method does not assume an upper limit. Results In all three data sets, yield correlated strongly with biomass (0.95 ≥ adj r 2 ≥ 0.71, p < 0.0001), and yield correlated with harvest index only in the Australian data set (adj r 2 = 0.34, p < 0.0001). In all three data sets, linear-plateau functions were fitted that estimated robust upper limits of harvest index: 0.44 (s.e. = 0.003) in Australia, 0.53 (s.e. = 0.009) in Spain, and 0.50 (s.e. = 0.005) in USA data sets. The upper limit held for environments with yield above 1.9 t ha −1 (s.e. = 0.029 t ha −1 ) in Australia, 0.9 t ha −1 (s.e. = 0.073 t ha −1 ) in Spain, and 2.8 t ha −1 (s.e. = 0.130 t ha −1 ) in USA. Below these thresholds, the upper limit of harvest index declined linearly. Conclusion An attainable harvest index that captures environmental constraints in rainfed wheat maybe a useful benchmark for modelling, breeding, and agronomic applications, and complements the theoretical physiological maximum.