Estimating yield potential in temperate high-yielding, direct-seeded US rice production systems

Accurate estimation of a crop's yield potential (Yp) is critical to addressing long-term food security via identification of the exploitable yield gap. Due to lack of field data, efforts to quantify crop yield potential typically rely on crop models. Using the ORYZA rice crop model, we sought to estimate Yp of irrigated rice for two widely used rice varieties (M-206 and CXL745) in three major US rice-producing regions that together represent some of the highest yielding rice regions of the world. Three major issues with the crop model had to be addressed to achieve acceptable simulation of Yp; first, the model simulated leaf area index (LAI) and biomass agreed poorly for all direct-seeded systems using default settings; second, cold-induced sterility and associated yield losses were poorly simulated for environments with a large diurnal temperature variation; lastly, simulated Yp was sensitive to the specified definition of physiological maturity. Except for the simulation of cold-induced sterility, all issues could be remedied within the existing model structure. In contrast, simulation of cold-induced sterility posed a continuing challenge to accurate simulation-one that will likely require changes to ORYZA's formulation. Estimates of Yp from the modified model were validated against large multi-year data sets of experimental yields covering the majority of US rice production areas. Validation showed the adjusted model simulated Yp well, with most top yields falling within 85% of Yp for both varieties (77% and 78% observed yields within 15% of Yp for CXL745 and M-206 respectively). Maximum estimated Yp was 14.3 (range of 8.2-14.5) and 14.5 (range of 8.7-15.3) t ha‐1 for the Southern US and CA, respectively.