科研之友
微信
新浪微博
Facebook
分享链接
摘要
The crop model (PyWOFOST) which coupled remote sensing information and a crop model (WOFOST) with Ensemble Kalman Filter (EnKF) was used to simulate maize growth and yield in Northeastern China with MODIS LAI as the coupling point. The assimilation plan focused on analyzing the impact of uncertainties of remote sensing observations (MODIS LAI) and crop model parameters (thermal time from emergence to anthesis, TSUM1) on the modeling results. First, the PyWOFOST model is used to simulate the maize LAI, yield and growth duration at site's scale; then the impact of remote sensing and crop model uncertainties on crop growth simulation is analyzed; finally, the regional maize yield is estimated with the PyWOFOST model, and the results are verified using the maize statistical yield. Results show that the simulated maize yield with assimilation has significantly improved compared to the one without assimilation. Under a business-as-usual scenario, the modeling results without assimilation has an error of 14.04%. The assimilated results show errors of 12.71%, 11.91%, 10.44%, and 10.48% at different TSUM1 uncertainty levels at 0, 10, 20, and 30 degrees C, respectively. The simulated LAI with assimilation agree better with the field observations than the one without assimilation. Without assimilation, the simulated growth duration has a mean deviation from the observed results at 3.4 days; with assimilation, the deviation would be 3.5, 4.3, 5.0, and 5.5 days respectively at different TSUM1 uncertainty levels. The results show that the errors for 58.82% areas are smaller than 15%. The simulated and statistical yields are highly correlated (R = 0.875), and the determination coefficient is at 0.806. The study shows that it is applicable to simulate crop growth using a crop model assimilated with remote sensing data based on EnKF and it is significant to estimate the remote sensing and crop model uncertainties in crop yield estimation.
