Comparison of 2-pool and 3-pool digestion kinetic model predictions of neutral detergent fiber digestibility of forages from commercially available data

Authors: BARRY, MICHAEL - Agmodels Foundation, Inc; Hall, Mary Beth

Submitted to: Journal of Dairy Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/1/2024
Publication Date: 12/9/2024
Citation: Barry, M.C., Hall, M. 2024. Comparison of 2-pool and 3-pool digestion kinetic model predictions of neutral detergent fiber digestibility of forages from commercially available data. Journal of Dairy Science. 108:2371-2380. https://doi.org/10.3168/jds.2024-25284.
DOI: https://doi.org/10.3168/jds.2024-25284

Interpretive Summary: Predicting ruminal fiber digestion is essential for calculating nutrient supply to the cow. In vitro fermentation data, models using the data to estimate fermentation rates, lag, and potentially digestible or indigestible fiber pools, plus passage estimates can be used for this purpose. The models vary in complexity, using 3-pools and six parameters, or 2-pools and four parameters. Comparison of models using in vitro fermentation data from two laboratories with analyses from six grasses and six alfalfas showed more variation in the factors included in the 3-pool model, and little difference between 2- or 3-pool models in the predicted rumen fermentation of fiber. The more complex 3-pool model conferred no advantage with these samples. The evidence that a 2-pool model is acceptable for describing fiber fermentation allows feed analysis laboratories, researchers, and diet formulation programs to continue to use a model that is already in common use, and is more easily calculated with less variability than the 3-pool model.

Technical Abstract: The objective of this study was to determine whether a 2-pool (2P) or 3-pool (3P) model most accurately and efficiently characterized amylase-treated neutral detergent fiber (aNDF) fermentation kinetics. Forages (six alfalfas, six grasses) were analyzed for residual aNDF (U) by two laboratories (Lab) which ran each sample and blanks in duplicate in each of two in vitro fermentation runs with mixed ruminal microbes. Sampling hours (t) were 0, 3, 6, 12, 18, 24, 30, 48, 72, 120, and 240 h. Outliers were removed. Pools as decimal proportions of aNDF were: degradable pool B in 2P, B1 rapid and B2 slow pools in 3P, and indigestible pool C in both; B pools have fermentation rates (kd, per h) and lag (L). Models were fit to data for each forage in each fermentation with the equations 2P: Ut = B*e^(-kd*(t-L)) + C, and 3P: Ut = B1*e ^ (-kdB1*(tL)) + B2*e ^ (-kd B2*(t-L))+C, giving 48 curves for each model. The ”optimx” function in base R was used to estimate parameters. Akaike’s Corrected Information Criterion (AICC) was used to select the model with the best fit for each forage in each fermentation: 16 3P, and 32 2P curves were selected. Expressed as the ‘difference between runs’/mean, average deviations between runs for Lab1 and Lab2, respectively, were: for 3P B1 = 0.50, 0.17; B2 = 0.26, 0.33; C = 0.50, 0.06; kdB1 = 0.81, 0.32; kdB2 = 0.93, 0.54; for 2P B = 0.04, 0.01; C = 0.07, 0.01; kdB = 0.17, 0.08. Estimates of percentage of aNDF fermented for 2P and 3P with no L at passage rates (kp) of 2, 3, 4, 5, 6, and 7%/h were calculated. With percentage of aNDF fermented for 2P subtracted from 3P at each kp for each feed, t-tests evaluated if the difference did not equal zero. With differences in aNDF% fermented listed sequentially from kp = 2 to 7%/h, for 16 AICC-selected 3P, values were 0.29, 0.45, 0.65, 0.87, 1.04, and 1.20% (P = 0.03) and for 32 AICC-selected 2P, values were 0.15, 0.13,0.14, 0.17, 0.19, and 0.22% (P = 0.03 for kp 2%/h, P = 0.18 for the rest). Although 2P and 3P differed, differences were small. With little difference between models in predicted rumen-fermented aNDF and numerically smaller variation in parameter estimates between runs for 2P, use of the more complex 3P conferred no advantage in this data set.