傅里叶变换红外光声光谱法测定土壤中有效磷

以中国科学院封丘生态实验站长期定位实验区的土样为材料(68样),利用傅里叶转换红外光声光谱测定土壤有效磷:以Olsen-P为因变量,通过傅里转换红外光声光谱构建偏最小二乘法和人工神经网络模型,利用模型进行预测。结果表明,偏最小二乘法模型的相关系数(R2)为0.96,校正标准偏差为1.79mg/kg,验证标准偏差为5.25mg/kg;人工神经网络模型的校正系数为0.84,校正标准偏差为2.40mg/kg,验证标准偏差为5.43mg/kg。两种模型均可以用于土壤有效磷的预测,且偏最小二乘模型优于人工神经网络模型。该方法的特点是无需样品前处理,且测定对样品无破坏,为土壤有效磷的快速测定提供新的手段。

Prediction of Soil Available Phosphorus Using Fourier Transform Infrared-Photoacoustic Spectroscopy

Soil samples (n=68) in long-term fertilizer experiment zones in fengqiu ecology experimental station of chinese academy of Sciences were involved in this experiment. Partial least squares (PLS) and artificial neural network (ANN) models were developed to predict soil available P between Olsen-P and fourier transform infrared-photoacoustic (FTIR-PAS) spectra. The calibration error, validation error and calibration coefficient (R2) from partial least square (PLS) model were 1.79 mg/kg, 5.25 mg/kg and 0.96, respec-tively, and from the ANN model were 2.40 mg/kg, 5.43 mg/kg and 0.84, respectively. The calibration statistics show these two modes can be used in the prediction of soil available P, and the prediction from PLS model is better than that from ANN model. This prediction method is non-destructive, and no pretreatment is needed, which make FTIR-PAS a promising method for fast determination of soil available P.