The comparison demonstrates
that the global calibration scheme performs satisfactorily in reproducing the spatial pattern of
rice yield, particularly in main rice production areas. Spatial agreement increases substantially when
more parameters are selected and calibrated, but with varying efficiencies. Among the parameters, PHU
and HI exhibit the highest efficiencies in increasing the spatial agreement. Simulations with different
calibration strategies generate a pronounced discrepancy of 5–35% in mean yields across latitude bands,
and a small to moderate difference in estimated yield variability and yield changing trend for the period
of 1981–2000. Present calibration has little effects in improving simulated yield variability and trends at
both regional and global levels, suggesting further works are needed to reproduce temporal variability
of reported yields