主要作物产量分离方法比较

Comparative Study on Main Crop Yield Separation Methods

对作物产量进行分离是分析气象因子对产量影响的重要步骤之一。以1985-2018年江苏省24个县(市)水稻统计产量为基础,分别采用3年滑动平均法、5年滑动平均法、五点二次平滑法、二次指数平滑法、HP滤波法和年际增量法对作物产量进行分离。从趋势产量与气象产量两方面比较6种方法的一致性与差异性,将分离出的气象产量与典型年增减产记录对比,选出更能准确捕获气象因子导致产量变异的方法,利用气象因子与产量关系的合理性对选定的方法进行检验。结果表明:就趋势产量拟合而言,前5种方法(年际增量法不能模拟趋势产量)均能较好地拟合趋势产量,与研究区域的趋势产量的一致性相关系数绝大多数为较好和极好等级范围;就气象产量而言,HP滤波法和年际增量法分离气象产量的合理性较差,标准差明显大于其他方法。综合看,3年滑动平均法与五点二次平滑法更具有普适性,可以捕获整个地区绝大多数典型年份气象因子带来的产量变化。

Crop yield separation is one of the important steps in analyzing the impact of meteorological factors on yield.Statistical rice yield data for 1985—2018 from 24 counties in Jiangsu are used to analyze the rationality of different separation methods.Six separation methods are 3-year moving mean,5-year moving mean,five-point quadratic smoothing,quadratic exponential smoothing,HP filter and year-to-year increment.Consistencies and differences are analyzed from aspects of trend yield and meteorological yield.In order to select better methods that could accurately capture the yield variation caused by meteorological factors,the meteorological yield based on different methods are compared with the typical annual increase and decrease of rice yield records.Finally,as mentioned above,the selected methods are calibrated by the rationality of the relationship between meteorological factors and yield.Results show that the trend yield curves fitted by different methods are in line with the process of social technology development.Compared with the average trend yield,almost all the consistency correlation coefficients are greater than 0.5.It suggests that different methods do not differ much in trend fitting.Characteristics of meteorological yield separated by 3-year moving mean,5-year moving mean,five-point quadratic smoothing and quadratic exponential smoothing in each county are simultaneously increasing or decreasing.And their standard deviation values are significantly smaller than HP filter method and year-to-year increment method.The result suggests that the rationality of separating the meteorological yields by 3-year moving mean,5-year moving mean,five-point quadratic smoothing,and quadratic exponential smoothing is higher than the other two methods.Five-point quadratic smoothing method and 3-year moving mean method can capture almost 100% of typical annual meteorological yield changes in the whole research area.Further verification results show that the positive and negative effects of meteorological factors captured by 3-year moving mean and five-point quadratic smoothing method are more consistent with the response to meteorological factors.Overall,separation methods of five-point quadratic smoothing method and 3-year moving mean method are more suitable for this research area and match well with meteorological factors.