基于主成分回归法的长江中下游双季早稻相对气象产量模拟模型

Simulation Model of Relative Meteorological Yield of Double-cropping Early Rice in Middle and Lower Reaches of Yangtze River Based on Principal Component Regression

以长江中下游稻区为研究对象,构建适于早稻产量估算的相对气象产量模拟模型。首先,选用研究区1981-2010年30个农业气象站早稻观测资料和逐日气象资料,对双季早稻3个关键生育时段内影响相对气象产量的主导气候因子进行分析,再运用主成分回归建立包含主导气候因子的相对气象产量模拟模型,并与由5点滑动平均法估算的趋势产量计算得到模拟产量,与实际产量比较对模型进行验证。结果表明:对双季早稻相对气象产量影响最大的因子为降水量,二者在整个研究时段内均呈极显著负相关(P<0.01);其次是呈极显著正相关的抽穗开花-灌浆末期日照时数;最后是呈极显著正相关的幼穗分化-抽穗开花期的3个温度指标,包括总热效应、平均最高温度和平均日较差。模型检验结果表明,模拟各站早稻产量与实际产量吻合,且在典型年份(1996年)两者的相关系数达0.93,平均相对模拟误差为7%。由此可见,该模型在长江中下游地区早稻相对气象产量的模拟和预测上具有较强的适用性。

Taking the middle and lower reaches of Yangtze River as a study area,a simulation model related to meteorological yield of early rice was developed.The dominant climatic factors in three key phenological phases were analyzed based on field experimental data and daily meteorological data from 30 agro-meteorological stations during 1981-2010,then an empirical model used to simulate relative meteorological yields of early rice was established by using of the principal component regression method.The model was verified by comparing actual yields and simulated yields of early rice,which calculated with five-point moving average method.The results showed that the most important climatic factor for the relative meteorological yields of double-cropping early rice was the total precipitation,which showed highly significant negative correlation with the relative meteorological yields of early rice in all three phases （P 〈0.01）.The second important factor was sunshine hours from heading stage to the end of grain-filling,which showed a significant positive correlation with early rice yields.Furthermore,the three temperature indicators,total thermal effect,average daily maximum temperature and average daily temperature range during the panicle initiation to heading stage,also showed significant positive correlation with early rice yields.The model simulation results showed that the simulated early rice yields were generally consistent with the actual yields,their correlation coefficient reached 0.93 in the typical year （1996）,and the average relative error was 7%.It indicated that the model had strong applicability on simulating and predicting the relative meteorological yield of early rice in the middle and lower reaches of Yangtze River.