构建县域早稻氮磷钾施肥的系统聚类方法研究

【目的】 为建立县域早稻氮磷钾施肥类别，探讨多点肥效试验资料的定量分类方法。 【方法】 以仙游县 26 个早稻“3414”设计的氮磷钾肥试验结果为例，探讨不同系统聚类分析方法的适用性及其类别间差异显著性。 【结果】 对生产条件和生产技术水平差异较小的县域肥效试验资料，以能使类别间距离空间扩展范围较大的欧氏距离–离差平方和法为最佳系统聚类分析方法，可将 26 个试验资料清晰地分成 3 类，分别对应于该县的高产、中产和低产稻田类型；两两类别间的空白区产量和平衡施肥产量均达到差异显著水平，且在 95% 置信区间下，稻谷产量水平在 3 个类别间几乎不出现交叉重叠。在此基础上，根据 3 种施肥类别对应的试验点资料，分别建立三元二次多项式类特征肥料效应方程，进而得到这 3 个施肥类别的推荐施肥量。 【结论】 欧氏距离–离差平方和法系统聚类分析是县域多点肥效试验资料的一种有效定量分类方法，可将仙游县早稻分为具有统计显著性差异的 3 个氮磷钾施肥类别。

Building fertilization categories of N,P and K fertilization for early rice using systematic clustering method in county territory

【Objectives】 Quantitative classification methods of multi-point experimental data of fertilizer responses were studied to build fertilization categories of N, P and K fertilizers of early rice in county territory. 【Methods】 Based on results of the 26 field experiments of " 3414”early rice response to N, P and K fertilization in Xianyou County, Fujian Province, China, applicability of different system clustering analysis methods and their significant differences were explored. 【Results】 Based on the county experimental data of fertilizer response to production conditions and the little differences of production technology, and the best system clustering analysis method by Euclidean distance-Ward’s method for larger spatial extension between the categories, the 26 experimental datasets were clearly divided into three categories, which were corresponding to the high yield paddy, middle yield paddy and low yield paddy respectively in the county. There were significant differences between blank area yields and the yields of the balanced fertilization between two categories, and rice yields hardly overlapped among the three categories with the 95% confidence interval. Based on the three fertilization categories corresponding to the experimental data, class feature fertilization response function of quadratic polynomial was developed respectively, and then recommendation rates of fertilizers of the three fertilization categories were gained. 【Conclusions】 System clustering analysis by Euclidean distance-Ward’s method is effective quantitative classification method for the county multi-point fertilization response data, and early rice can be divided into three fertilization categories of N, P and K fertilizers with statistically significant differences in Xianyou county.