芻料大豆營養成分近紅外光分析檢量線之建立

Calibration equations for determining nutritional components of forage soybean in near-infrared spectroscopy analysis

朱明宏(Ming-Hung Chu) ； 王紓愍(Shu-Min Wang) ； 陳嘉昇(Chia-Sheng Chen)

《畜產研究》 48卷2期 (2015/06) Pp. 107-113

摘要

本研究之目的為建立芻料大豆營養成分近紅外光分析檢量線，以274個青割及青貯樣品進行基礎檢量線的建立。測定之營養成分包含：粗蛋白質（CP）、酸洗纖維（ADF）及中洗纖維（NDF），基礎檢量線中CP、ADF及NDF之決定係數（R^2）分別為0.98、0.83及0.95，檢量線標準偏差（SEC）分別為0.58、1.66及1.31%，交互驗證標準偏差（SECV）分別為0.70、1.77及1.44%。基礎檢量線建立後另以70個青割及青貯樣品進行測試與擴充。測試之結果，CP、ADF及NDF之平均偏差分別為0.09、0.11及1.25%，預測標準偏差（SEP）分別為1.16、2.82及2.08%，斜率（Slope）分別為0.99、1.09及1.08，表示基礎檢量線對不同來源之樣品已具備相當之準確度。其後以測試樣品進行檢量線之擴充，新檢量線中CP、ADF及NDF之R^2分別為0.97、0.84及0.95，SEC分別為0.65、1.72及1.31%，SECV分別為0.71、1.92及1.47%，樣品實測標準偏差與驗證標準機差之比值（RPD）分別為6.7、2.6及4.5，均達可接受之準確度。本檢量線之樣品族群包括不同品種、種植地區、種植期、成熟度、青割與青貯樣品，樣品歧異度大，所建立檢量線經測試結果已具備充分之準確度來作為青割與青貯芻料大豆的品質分析。

關鍵字

芻料大豆 ； 近紅外光分析 ； 快速測定

並列摘要

The purpose of this study was to establish the calibration equations for determining the chemical components of forage soybean by near infrared spectroscopy (NIRS). 274 samples were collected and scanned by NIR spectrums for setting up the basic calibrations. The determined constituents included crude protein (CP), acid detergent fiber (ADF) and neutral detergent fiber (NDF). The regression coefficients (R^2) of CP, ADF and NDF were 0.98, 0.83 and 0.95, the standard error of calibration (SEC) were 0.58, 1.66 and 1.31% and the standard error of cross verification (SECV) were 0.70, 1.77 and 1.44%, respectively. The basic calibration equations were verified with another 51 samples. The predicting results showed that the biases of the means for CP, ADF and NDF were 0.09, 0.11 and 1.25%, the standard error of prediction (SEP) were 1.16, 2.82 and 2.08%, the slope were 0.99, 1.09 and 1.08, respectively. It indicated that the basic calibration equations were accurate for samples from various sources. Then, the new sample sets were added to update the basic calibration. The R^2 of the updated regressions for CP, ADF and NDF were 0.97, 0.84 and 0.95, the SEC were 0.65, 1.72 and 1.31% and the SECV were 0.71, 1.92 and 1.47%, respectively. The RPD (the ratio of the standard deviation of constituent to the standard error of prediction) were 6.7, 2.6 and 4.5 for CP, ADF and NDF, respectively. All of the calibration equations were acceptable for accurately predicting the above-mentioned constituents. The diversified samples of the calibration equations including different sources of variety, cropping area, cropping season, maturity and ensiling. The results showed that the calibration equations established were acceptable and accurate for constituents predicting both in fresh or ensiled forage soybean.

並列關鍵字

Forage soybean ； Near infrared spectroscopy ； Rapid analysis