基于偏最小二乘回归的土壤碱解氮含量估测

Estimation of soil alkali-hydrolyzed nitrogen content based on partial least squares regression

构建基于室内高光谱数据的土壤碱解氮含量估测模型,为快速、准确获取土壤中养分信息提供新的方法。对新疆乌鲁木齐市106个风干的土壤样品进行研磨过筛,在室内进行反射率光谱数据的采集,对采集的光谱数据进行Savitzky-Golay滤波、一阶微分(FDR)、连续统去除(CR)、多元散射校正(MSC)4种预处理,在此基础上利用连续投影算法对预处理后的数据进行特征波段的筛选,用偏最小二乘回归建立不同预处理后土壤碱解氮含量预测的高光谱分析模型。模型评价指标采用决定系数(R^(2))、均方根误差(RMSE)、相对分析误差(RPD)平均相对误差(MAE)。结果显示:4种预处理方法中以连续统去除处理的预测精度最为突出,其模型R^(2)、RMSE、RPD、MAE分别为0.90、13.0、2.26、0.13;模型的线性回归方程为:y=0.9316x+8.763。因此,利用连续统去除结合偏最小二乘回归,能够较好地估测乌鲁木齐市土壤中碱解氮的含量。该结果可为室内高光谱数据快速反演土壤碱解氮含量提供理论依据。

The estimation model of soil alkali-hydrolyzable nitrogen content based on indoor hyperspectral data was constructed,which provided a new method for rapid and accurate acquisition of nutrient information in soil.106 soil samples collected from Urumqi,Xinjiang were air-dried,ground and sifted,and the reflectance spectral data were collected indoors.The collected spectral data were pretreated by Savitzky-Golay filtering,first-order differentiation(FDR),continuum removal(CR)and multiple scattering correction(MSC).On this basis,continuous projection algorithm was used to screen the characteristic bands of the pre-treated data,and partial least squares regression was used to establish a hyperspectral analysis model for predicting soil alkali-hydrolyzed nitrogen content after different pretreatment.Coefficient of determination(R^(2)),root mean square error(RMSE),relative analysis error(RPD)and mean relative error(MAE)were used as evaluation indexes of the model.The results showed that the prediction accuracy of the continuous removal treatment was the most prominent among the four pretreatment methods.R^(2),RMSE,RPD and MAE were 0.90,13.0,2.26 and 0.13,respectively.The linear regression equation of the model was y=0.9316x+8.763.Therefore,the continuous projection algorithm combined with partial least squares regression could be used to estimate the content of alkali-hydrolyzed nitrogen in soil in Urumqi.The results could provide a theoretical basis for rapid inversion of soil alkali-hydrolyzed nitrogen content with indoor hyperspectral data.