冬小麦水分利用效率相关驱动因子及其预测模型构建

The Related Driving Factors of Water Use Efficiency and Its Prediction Model Construction in Winter Wheat

【目的】水分利用效率能够综合反映冬小麦生长适宜度和能量转化效率,筛选并探究冬小麦各生育时期响应标准化水分利用效率(WP*)的驱动因子,构建相关驱动因子的WP*预测模型,对黄淮海平原冬小麦水分利用效率监测及水资源高效利用具有重要意义。【方法】以冬小麦为研究对象,设置3个水分处理,包括水分亏缺处理(W1):35 mm和(W2):48 mm,对照处理(W3):68 mm,获取冬小麦拔节期、孕穗期和灌浆期的冠层温度参数、生理指标和标准化水分利用效率(WP*)。通过逐步回归和通径分析筛选各生育时期响应WP*变化的主要驱动因子,探究WP*与相关驱动因子间关系,最后采用偏最小二乘回归(PLSR)和支持向量机(SVM)方法构建各生育时期基于驱动因子的WP*预测模型。【结果】较对照处理(常规灌溉),水分亏缺处理下冬小麦冠层温度参数、生理指标和WP*均表现出显著差异。基于逐步回归方法筛选出了各生育时期响应WP*的主要驱动因子,并采用通径分析得到各驱动因子响应WP*敏感程度排序,即拔节期依次为冠层温度极差(MTD)、气孔导度(Gs)、叶片含水量(LWC)和POD;孕穗期依次为冠层相对温差(CRTD)、等效水厚度(EWT)、可溶性糖含量(SSC)和作物水分胁迫指数(CWSI);灌浆期依次为SSC、冠层温度标准差(CTSD)、LWC和Gs。最后,采用偏最小二乘回归(PLSR)和支持向量机(SVM)基于筛选后的驱动因子构建了各生育时期WP*预测模型,其中以SVM构建的孕穗期WP*预测模型精度最优,R2cal(R^(2)val)、RMSEcal(RMSEval)和n RMSEcal(n RMSE_(val))分别为0.945(0.926)、0.533 g·m^(-2)(0.580 g·m^(-2))和2.844%(3.075%)。【结论】通过筛选冬小麦各生育时期响应WP*的相关驱动因子信息及构建冬小麦水分利用效率预测模型,为黄淮海平原冬小麦水分精准监测与管理提供了理论基础。

【Objective】The water use efficiency can comprehensively reflect the growth suitability and energy conversion efficiency of winter wheat.The driving factors of winter wheat in response to standardized water use efficiency(WP*)at different growth stages were screened and explored,and the WP*prediction model of related driving factors was constructed,which was of great significance for the monitoring of water use efficiency and efficient use of water resources in winter wheat in the Huang-Huai-Hai Plain.【Method】Three water treatments were set up,including water deficit treatments(W1:35 mm,and W2:48 mm)and control treatment(W3:68 mm),and the canopy temperature parameters,physiological indexes and standardized WP*of winter wheat at the jointing,booting and filling stages were measured.Stepwise regression and pathway analysis were used to screen the main driving factors in response to WP*changes at each growth stage,the relationship between WP*and related drivers was explored,and finally the partial least squares regression(PLSR)and support vector machine(SVM)methods were used to construct a driver-based WP*prediction model in each growth stage.【Result】Compared with W3,the canopy temperature parameters,physiological indexes and WP*of winter wheat under the water deficit treatments showed significant differences.Based on the stepwise regression method,the main driving factors in response to WP*at each growth stage were screened,and the sensitivity of each driving factor in response to WP*was ranked by pathway analysis,that is,maximum temperature difference(MTD),stomatal conductance(Gs),leaf water content(LWC)and POD were selected at the jointing stage;canopy relative temperature difference(CRTD),equivalent water thickness(EWT),soluble sugar content(SSC)and crop water stress index(CWSI)were selected at the booting stage;SSC,standard deviation of canopy temperature(CTSD),LWC and Gs were selected at the filling stage.Finally,the driver-based WP*prediction model for each growth stage was construct by using PLSR and SVM.Among them,the prediction model of WP*at booting stage constructed by SVM had the best accuracy,with R2cal(R^(2)val),RMSEcal(RMSEval)and nRMSEcal(nRMSE_(val))of 0.945(0.926),0.533 g·m^(-2)(0.580 g·m^(-2))and 2.844%(3.075%),respectively.【Conclusion】By screening the relevant driving factors of WP*at each growth stage of winter wheat and constructing a prediction model of winter wheat water use efficiency,this paper provided a theoretical basis for accurate monitoring and management of winter wheat moisture in the Huang-Huai-Hai Plain.