发育模式的理论修订与检验-以中国玉米为例

Modification and validation of the theory of phenology model - a case study of maize in China

中文关键词:  积温理论  响应与适应  发育模式构建  校正验证

英文关键词:Theory of accumulated thermal time  Response and adaptation  Phenology model development  Calibration and validation

中文摘要:

      基于逆境生物学和表观遗传学取得的进展，将现有只考虑发育速率对环境响应机制的发育模式理论框架修订考虑了响应、适应与记忆机制。随后以适应机制为例，选择发育阶段始期的日序(DOY)作为表征适应性的因子，并与线性温度响应函数结合，构建了耦合响应与适应机制的发育模式(RAM)，具体实现形式为：R_i=a+(c+d×DOY_0)×T_i。利用中国194个农业气象站玉米营养生长阶段(VGP)和生殖生长阶段(RGP)观测资料，验证了RAM模式的有效性。结果表明，DOY对发育速率普遍呈正相关关系，且所有站点两个发育阶段的DOY与温度的共线性均在可接受范围内。与线性关系相比，RAM模式在VGP和RGP阶段对观测数据的解释率分别提高0.213和0.274。与经过充分校正的双线性WOFOST和非线性Gao模式相比，在参数校正时RAM可将VGP的RMSE分别降低0.4和0.4 d，RGP阶段分别降低1.6和0.7 d。在参数验证时RAM可将VGP的RMSE分别降低0.1和0.4 d，RGP阶段分别降低1.1和0.4 d。此外，RAM模式还具有参数少、参数化过程简单等优点。修订后的发育模式框架以较少的假设实现了更准确的模拟，具有较好的理论发展和实际应用前景。

英文摘要:

      Based on the progress made in stress biology and epigenetics, the current theoretical framework of phenology model, which only considers the response mechanism of developmental rate to the environment factors, is being revised to incorporate mechanisms of response, adaptation, and memory. Then, taking the adaptation mechanism as an example, the day of year (DOY) of the beginning of the growth period was proposed as the factor to characterize the adaptability, which is then integrated with linear temperature response functions to establish a phenology model coupled with both Response and Adaptation Mechanisms (RAM). Specific implementation of RAM is in the form of: R_i=a+(c+d×DOY_0)×T_i. RAM model was then calibrated and validated using phenology observations of the vegetative growth period (VGP) and reproductive growth period (RGP) of maize at 194 agrometeorological observation sites across China. Results showed that: DOY was generally positively related to developmental rate across sites and development periods. The collinearity of DOY and temperature at the two development periods at all sites was within the acceptable ranges. Compared with the linear temperature response model, the RAM model improves the interpretation rate of the observed data by 0.213 and 0.274 in the VGP and RGP, respectively. Compared with the typical fully calibrated bilinear (WOFOST) and nonlinear (Gao) model, the RAM can reduce the RMSE by 0.4 and 0.4 d during VGP, respectively, and by 1.6 and 0.7 d during RGP, respectively in the calibration process. In the validation process, the RAM can reduce the RMSE by 0.1 and 0.4 d during VGP, respectively, and by 1.1 and 0.4 d during RGP, respectively. In addition, RAM mode has the advantage of a simple parameterization process. Our results show that the modified phenology model framework may have promising prospects in theoretical research and practical application.