基于混合回归分析的荷斯坦奶牛氮排泄量预测

Prediction of Nitrogen Excretion of Holstein Dairy Cows Based on Mixed Regression Analysis

为了准确预测荷斯坦奶牛的氮排泄量,本研究用27个已发表研究的数据构建数据库。变量包括体重(BW)、干物质采食量(DMI)、粗蛋白质(CP)、氮摄入量(NI)、粗饲料比例(FP)、产奶量(MY)、乳氮(MN)、乳中尿素氮(MUN)、氮的全消化道表观消化率(TTND)、乳氮占氮摄入量的比例(MN/NI)、氮摄入量占干物质采食量的比例(NI/DMI)、饲料利用效率(FE)、氮沉积量(NR)以及尿氮(UN)、粪氮(FN)、总氮(TN)、尿氮占总氮的比例(UN/TN)。采用相关分析的方法,分析了动物或饲粮因子相关的变量与氮排泄量之间的关系。利用混合回归模型建立了预测奶牛尿液和粪便的氮排泄量模型。结果表明,单一变量NI与UN、FN和TN存在极显著的相关关系(P<0.01),TTND与UN/TN存在极显著的相关关系(P<0.01),且相关系数均较高。预测奶牛UN的一元线性模型[UN(g/d)=4.80(±16.99)+0.34NI(±0.03)(R^(2)=0.57,n=54,P<0.01)]和多元线性模型[UN(g/d)=-166.9(±33.10)+2.81CP(±6.25)+10.62DMI(±1.24)+3.50NI/DMI(±3.82)(R^(2)=0.58,n=54,P<0.01)]拟合效果较好。预测奶牛FN的一元线性模型[FN(g/d)=42.91(±13.51)+0.21NI(±0.03)(R^(2)=0.65,n=54,P<0.01)]和多元线性模型[FN(g/d)=34.44(±10.42)+0.11NI(±0.02)+2.43MY(±0.44)(R^(2)=0.70,n=54,P<0.01)]拟合效果较好。预测奶牛TN的一元线性模型[TN(g/d)=37.58(±22.59)+0.57NI(±0.05)(R^(2)=0.79,n=54,P<0.01)]和多元线性模型[TN(g/d)=102.88(±32.11)+0.66NI(±0.06)-4.33NI/DMI(±1.67)(R^(2)=0.81,n=54,P<0.01)]拟合效果好。预测奶牛UN/TN的一元线性模型[UN/TN=-0.09(±0.14)+0.009TTND(±0.002)(R^(2)=0.50,n=54,P<0.01)]和多元线性模型[UN/TN=-0.50(±0.08)+0.00014NI(±0.00005)+0.015TTND(±0.001)-0.0009NR(±0.0002)(R^(2)=0.77,n=54,P<0.01)]拟合效果较好。由此可见,混合回归分析可预测荷斯坦奶牛的氮排泄量。

To accurately predict the nitrogen excretion of Holstein dairy cows,a dataset comprising 27 published studies were developed in this study.Variables included body weight(BW),dry matter intake(DMI),crude protein(CP),nitrogen intake(NI),forage proportion(FP),milk yield(MY),milk nitrogen(MN),milk urea nitrogen(MUN),apparent total tract nitrogen digestibility(TTND),proportion of milk nitrogen in nitrogen intake(MN/NI),proportion of nitrogen intake in dry matter intake(NI/DMI),feed efficiency(FE),nitrogen retention(NR)and urinary nitrogen(UN),fecal nitrogen(FN),total nitrogen(TN),proportion of urinary nitrogen in total nitrogen(UN/TN).Correlation analysis examined relationships between related variables of animal or dietary factors and nitrogen excretion.A mixed model regression analysis was used to develop models to predict nitrogen excretion in urine and feces.The results showed that there were significant correlations between the single variables of NI and UN,FN and TN(P<0.01),there was significant correlation between TTND and UN/TN(P<0.01),and with higher correlation coefficients.The linear prediction models of UN of dairy cows[UN(g/d)=4.80(±16.99)+0.34NI(±0.03)(R^(2)=0.57,n=54,P<0.01)]and the multilinear prediction model[UN(g/d)=-166.9(±33.10)+2.81CP(±6.25)+10.62DMI(±1.24)+3.50NI/DMI(±3.82)(R^(2)=0.58,n=54,P<0.01)]were fit well.The linear prediction models of FN of dairy cows[FN(g/d)=42.91(±13.51)+0.21NI(±0.03)(R^(2)=0.65,n=54,P<0.01)]and the multilinear prediction model[FN(g/d)=34.44(±10.42)+0.11NI(±0.02)+2.43MY(±0.44)(R^(2)=0.70,n=54,P<0.01)]were fit well.The linear prediction models of TN of dairy cows[TN(g/d)=37.58(±22.59)+0.57NI(±0.05)(R^(2)=0.79,n=54,P<0.01)]and the multilinear prediction model[TN(g/d)=102.88(±32.11)+0.66NI(±0.06)-4.33NI/DMI(±1.67)(R^(2)=0.81,n=54,P<0.01)]were fit well.The linear prediction models of UN/TN of dairy cows[UN/TN=-0.09(±0.14)+0.009 TTND(±0.002)(R^(2)=0.50,n=54,P<0.01)]and the multilinear prediction model[UN/TN=-0.50(±0.08)+0.00014 NI(±0.00005)+0.015 TTND(±0.001)-0.0009 NR(±0.0002)(R^(2)=0.77,n=54,P<0.01)]were fit well.In conclusion,the mixed regression analysis can predict nitrogen excretion of Holstein dairy cows.