Hayirli A, Grummer R R, Nordheim E V, Crump P M
Department of Animal Nutrition and Nutritional Diseases, Ataturk University, Erzurum, 25700 Turkey.
J Dairy Sci. 2003 May;86(5):1771-9. doi: 10.3168/jds.S0022-0302(03)73762-X.
The objectives of this study were to develop and validate a model for predicting dry matter intake (DMI) of Holsteins during the prefresh transition period. The original database (ODB) for model development was established by compiling parity, body condition score (BCS), and DMI data during the final 3 wk of gestation from 366 Holsteins fed 24 different diets that were used in eight experiments conducted at three universities. For model validation, a validation database (VDB) was established by compiling data from 333 prefresh transition Holsteins fed 25 different diets that were used in eight experiments conducted at five universities. Dry matter intake during the prefresh transition period was fitted to an exponential function: DMI(t) = a + pe(kt), where DMI(t) = DMI as a percentage of body weight (BW) at time t, a = asymptotic intercept at time--infinity, p = change in intake (kg) from the asymptotic intercept until parturition, k = rate constant influencing the shape of the curve, and t = day relative to parturition expressed as days pregnant--280. The model developed from the ODB predicted DMI of heifers in the VDB with satisfactory accuracy and precision. However, this was not true for cows, probably due to differences in BCS of cows and diets fed to cows from the two data sets. When a subset of cows was selected from each data set that had similar BCS (> 4.0) and were fed similar diets, accuracy and precision of the model predicting DMI was improved. Finally, both databases were combined to develop final models for predicting DMI of heifers and cows. Proposed models for predicting mean daily DMI of heifers and cows during the prefresh transition period were DMI(t) = 1.713-0.688e(0.344t) (R2 = 0.96) and DMI(t) = 1.979-0.756e(0.154t) (R2 = 0.97), respectively. Adjustment factors for animal and dietary factors were generated to demonstrate the plausibility of adaptive fitting of the prediction. The regression coefficients of prediction models (a, p, and k) were affected by BCS and dietary organic macronutrient concentrations.
本研究的目的是开发并验证一个用于预测荷斯坦奶牛围产前期干物质采食量（DMI）的模型。用于模型开发的原始数据库（ODB）是通过汇总366头荷斯坦奶牛在妊娠最后3周的胎次、体况评分（BCS）和DMI数据建立的，这些奶牛饲喂了24种不同日粮，数据来自于在三所大学进行的8项试验。为了进行模型验证，通过汇总333头围产前期奶牛的DMI数据建立了一个验证数据库（VDB），这些奶牛饲喂了25种不同日粮，数据来自于在五所大学进行的8项试验。围产前期的干物质采食量与一个指数函数拟合：DMI(t) = a + pe(kt)，其中DMI(t)是时间t时以体重（BW）百分比表示的DMI，a是时间趋于负无穷时的渐近截距，p是从渐近截距到分娩时采食量的变化量（kg），k是影响曲线形状的速率常数，t是相对于分娩的天数，以怀孕天数 - 280表示。从ODB开发的模型对VDB中后备奶牛的DMI进行预测时，具有令人满意的准确性和精确性。然而，对于经产奶牛并非如此，这可能是由于两组数据中经产奶牛的BCS和所饲喂日粮存在差异。当从每个数据集中选择BCS相似（> 4.0）且饲喂相似日粮的经产奶牛子集时，预测DMI的模型的准确性和精确性得到了提高。最后，将两个数据库合并以开发用于预测后备奶牛和经产奶牛DMI的最终模型。预测围产前期后备奶牛和经产奶牛平均每日DMI的建议模型分别为DMI(t) = 1.713 - 0.688e(0.344t)（R2 = 0.96）和DMI(t) = 1.979 - 0.756e(0.154t)（R2 = 0.97）。生成了动物和日粮因素的调整因子，以证明预测的适应性拟合的合理性。预测模型的回归系数（a、p和k）受BCS和日粮有机常量营养素浓度的影响。
