Response of rice yield and yield components to elevated [CO2]: A synthesis of updated data from FACE experiments

Rice is the most widely consumed staple food for more than half of the world's population. Rising atmospheric carbon dioxide concentration [CO2] is expected to improve crop yields in the future. Rice responds to elevated [CO2] through photosynthesis improving yield components. This response depends on rice types, climate and fertilizers. However, the determinants of rice yield and the contribution of yield components at elevated [CO2] are far from certain. We extracted data from articles published before the end of 2018. These articles reported the responses of rice yield and yield components to elevated [CO2] at FACE conditions across four locations in China and Japan. Using CART (Classification and regression tree, a nonparametric modeling approach to recursively partition predictor variables) and regression models, we identified the principal determinants and the contribution of yield components to yield at elevated [CO2]. Elevated [CO2] (similar to 200 mu mol mol(-)(1) above ambient) increased rice yields by 13.5% (n = 93), 22,6% (n = 10) and 32.8% (n = 17) for japonica, indica and hybrid cultivars. The type of rice cultivars dominantly determined the response of spikelets per panicle, while temperature is of greatest importance in determining the response of filled spikelets percentage to elevated [CO2]. Optimal nitrogen rates at elevated [CO2] are site-specific, depending on local soil fertility and temperature. The contribution of post-heading elevated [CO2] to yield is higher for indica and hybrid (24%) than for japonica cultivars (13%). Lower benefit of japonica cultivars from post-heading elevated [CO2] is likely attributed to an intensive photosynthetic acclimation. Our findings highlight the importance of pre- and post-heading CO2 fertilization effect and nitrogen management in yield benefits from elevated [CO2], and the necessity for crop modelers to incorporate the knowledge from FACE studies into models so that models become more accurate, rigorous and robust.