【摘要】 土壤水分是直接影响作物产量的重要因素之一,因此科学地预测土壤含水量对充分利用土壤水资源具有十分重要的意义。土壤含水量受到多重气象因素的影响,各气象因子间往往存在多重相关性,从而导致传统的多元回归模型(基于最小二乘法)的失真,丧失稳健性,预测精度降低。为此,采用偏最小二乘回归建模,借助主成分分析与典型相关分析的思路,采用成分提取的方法,有效地解决了各气象因子之间的多重相关问题,建立了土壤含水量预报模型,并对模型进行了辅助分析,得出了各影响因子的评价结果排序,同时模型的拟合和预报精度均较好。更多还原

【Abstract】 Soil moisture is one of key factors affecting plant yield,so it has great significance of scientifically forecasting soil moisture for making full use of soil water. Soil moisture is influenced by many factors and these factors always have multiple correlations with each other that result to distortion,instability and lower prediction accuracy of traditional least square method. Therefore,the partial least-square regression was applied to model base on the idea of principle components analysis and typical correlation analysis. After principle components are extracted,the model effectively solves the multiple correlations among factors. And then we can carry out model-assisted analysis,the sort results of each factor can be derived. At the same time,the simulation and prediction accuracy of the model are all better.更多还原