Parameter estimation for a rice phenology model based on the differential evolution algorithm

Based on the differential evolution algorithm, this study proposes a parameter estimation method for a rice phenology model of ORYZA2000 by using data from the mid-season rice variety regional experiments conducted in field conditions at six representative stations in Middle-Lower Yangtze River Valley of China from 2005 to 2011. Field experiment data during 2005-2007 and 2008-2011 were used for the model calibration and validation, respectively. Dates of three crucial rice development stages (panicle initiation, flowering, and maturity) were estimated by using the estimated phenology parameters. The results show that the normalized root mean square errors (NRMSEs) of the dates in three key development stages during calibration are all less than 5%, and the NRMSE of the dates in panicle initiation, flowering, and maturity stages during validation is 6.53% and 4.4% and 3.29%, respectively. The correlation coefficient between the simulated and observed dates of three development stages during both calibration and validation showed the reliable efficiency of the calibrated model (with the significance of correlations all above the 99% confidence level), indicating that the calibrated model could be used for developing adaptive strategies for rice production in the Middle-Lower Yangtze River Valley of China.