Six mature non-lactating Holstein dairy cows were offered one of three diets with forage to grain ratios of 100:0 (G0), 75:25 (G25), and 50:50 (G50). The forage portion of the diet consisted of 80% grass hay and 20% alfalfa hay (on a Dry Matter (DM) basis). The experiment was a replicated 3x3 Latin Square Design, each animal received each of the three diets over the course of the three 5-week periods.

A statistical model was created combining 23 bacterial members in the feces, chosen due to their significant Variable Influence on Projection (VIP) values, along with ADF, NDF and starch formulated a basic predictive model for overall CH4 production (L d-1). The model had an R2 value of 0.51 and a Q2 value of 0.49. These 23 bacterial members, along with ADF, NDF, and starch can detect an increase or decrease from mean CH4 production levels.