Estimating Crop Yield in A Global Hydrological Model

Food and water are essential for human beings. A better understanding of food-water nexus calls for integrated model development that enables the simultaneous simulation of crop yield and water cycle. H08 is a global hydrological model with consideration of human activities (e.g., reservoir operation and crop irrigation). Although a crop growth sub-model was included in H08 to estimate crop-specific calendar globally, the performance as a yield simulator was poor since a default set of parameters has been applied globally. Here, through parameter calibration and algorithm improvement, we enhanced H08 to simulate the yield of four major staple crops (maize, wheat, rice, and soybean). The simulated crop yield was compared to the FAO national yield statistics and the GDHY grided yield estimates with respect to mean bias and time series correlation. We find that the improved simulations showed a good consistency with FAO national yield with the mean bias within ±10% of the FAO statistical yield in the major producer countries. The corresponding determination coefficient (R2) increased from 0.17 to 0.94, 0.24 to 0.98, 0.00 to 1.00, and 0.20 to 0.97, and the root mean square error (RMSE) decreased from 6.5 to 1.8, 1.8 to 0.7, 2.3 to 0.3, and 2.2 to 0.5 t/ha for maize, wheat, rice, and soybean, respectively. Comparison with the reported performance of other mainstream global crop models revealed that the improved simulations have comparable capability in capturing the temporal yield variability. Finally, the gird level analysis showed that the improved simulations have similar spatial pattern and capability in reproducing the temporal variation as GDHY yield did in a wide area although substantial differences exist in other places. Our improvement enables H08 to be one of good tools to investigate global food-water nexus.