Across-study evaluation of enteric methane emissions from dairy cattle for spot sampling schemes using simulation approaches Henk J. van Lingen1,*, G. Foggi2, U. Arshad2, A. Wang2, M. Niu2, S. van Gastelen3, A. Bannink3, J. Dijkstra4, J.L. Ellis1 1Department of Animal Biosciences, University of Guelph, Canada; 2Animal Nutrition Group, Department of Environmental Systems Science, ETH Zurich, Switzerland; 3Wageningen Livestock Research, Wageningen University & Research, The Netherlands; 4Animal Nutrition Group, Wageningen University & Research, The Netherlands. * Correspondence:
[email protected] INTRODUCTION CONCLUSIONS •Multiple error-included samplings from penalized smoothing splines may mimic multiple numbers of days of GreenFeed gas spot sampling MATERIALS and METHODS RESULTS and DISCUSSION REFERENCES •Measuring enteric methane (CH4) emitted from dairy cattle using spot sampling (e.g. for GreenFeed systems) requires a minimum sampling frequency and timing, particularly for restricted feeding (Van Lingen et al., 2023) •Measurement accuracy of enteric CH4from dairy cattle increases with sampling frequency (Lee et al., 2022) •More data/studies will solidify evaluation of sampling frequency and timing for enteric CH4collection from cattle ➢AIM: a) to assess the accuracy of preset sampling schemes for dairy cattle enteric CH4emissions across various studies, and b) to simulate the measurement uncertainty relative to sampling schemes and sample sizes •Lee, C., K.A. Beauchemin, J. Dijkstra, D.L. Morris, K. Nichols. P.J. Kononoff, D. Vyas. 2022. Estimates of daily oxygen consumption, carbon dioxide and methane emissions, and heat production for beef and dairy cattle using spot gas sampling. Journal of Dairy Science 105,9623-9638.https://doi.org/10.3168/jds.2022-22213 •Van Lingen, H.J., J.G. Fadel, E. Kebreab, A. Bannink, J. Dijkstra, S. van Gastelen. 2023. Smoothing spline assessment of accuracy of enteric hydrogen and methane production measurements from dairy cattle using various sampling schemes. Journal of Dairy Science 106,6834-3848.https://doi.org/10.3168/jds.2022-23207 •Data references: A) Van Lingen et al.(2017; doi: 10.3389/fmicb.2017.00425) + Van Gastelen et al.(2017; doi: 10.3168/jds.2016-12367), B) Van Gastelen et al.(2020; doi: 10.3168/jds.2019-17936), C) Warner et al.(2015; doi: 10.3168/jds.2014-9068), D) Van Gastelen et al.(2015; doi: 10.3168/jds.2014-8552), E) Hatew et al.(2016; doi: 10.3168/jds.2015-10047), F) Klop et al.(2017; doi: 10.3168/jds.2016-12033), G) Ma et al.(2024; doi: 10.3168/jds.2024-25246) General additive model, i.e. penalized smoothing spline, for diurnal CH4emission rates: •𝑦𝑖𝑗 = 𝛽0+ 𝑓 𝑡𝑖𝑗 + 𝜖𝑖𝑗;with intercept 𝛽0and spline basis function 𝑓 𝑡𝑖𝑗 •𝜖𝑖𝑗~𝑁 0, 𝜎2and σ𝑖=1 𝑛𝑦𝑖𝑗 − 𝑓(𝑡𝑖𝑗)2+ 𝜆 𝑎 𝑏𝑓′′(𝑡) 2𝑑𝑡 Figure 3 -CH4production (g/d) across Exp A-G for the 14 preset sampling schemes [i.e. sampling every x.x h starting at y.y h from morning feeding (x.x_y.y) and at four unequally spaced sampling schemes relative to morning feeding] being either significantly different (✓), decreased (↓) or increased (↑) compared to the mean of all data points. Figure 1 –Diurnal CH4emission rates recorded for 72 h for one cow with a) all data points retained and b,c) retaining data according 2 of the 14 preset sampling schemes (viz. Reference, 0.5_0.0 and 8.0_2.0). Vertical dashed lines indicate feed delivery times. Figure 2 -CH4production in Exp D computed for 14 preset sampling schemes [i.e. sampling every x.x h starting at y.y h from morning feeding (x.x_y.y) and at four unequally spaced sampling schemes relative to morning feeding] compared to the mean of all data points (Ref). Figure 5 -CH4production computed from data points sampled from fitted penalized smoothing splines and their 𝜎2(i.e. colored points in Figure 4) for 3 sampling schemes using 1000 sampling iterations (or virtual number days of CH4emission). Figure 6 -95% confidence intervals (solid lines) and means (dashed lines) relative to the number of observations/sample size for CH4productions obtained using the 0.5_0.0, 3.0_1.0 and 1.0+7.0 sampling schemes for the dietary treatments of experiment A with ത 𝑌 CH4and SDCH4computed as visualized in Figures 4 and 5. Vertical lines indicate 95% confidence band widths of CH4equal to 80 g/d. Figure 4 - Measured diurnal CH4emission rates vs. time (grey points) with penalized smoothing spline fits (black dots) and values sampled from 𝜎2of the fits (colored points) for 3 sampling schemes. Respiration chamber datasets Exp. Location Cows * Periods Measurement interval Twice daily feeding at Dietary treatments A) Wageningen 56 12-15 min 82 - 95% of ad lib Control + linseed oil B) Wageningen 59 12-15 min Ad lib Control + 3-NOP C) Wageningen 27 10-12.5 min 82 - 95% of ad lib N fertilization + regrowth period of fresh grass herbage D) Wageningen 32 10-12.5 min 82 - 95% of ad lib Grass/corn silage ratios E) Wageningen 28 10-12.5 min 82 - 95% of ad lib Corn silage maturity stages F) Wageningen 12 10-12.5 min 82 - 95% of ad lib Control + essential oils G) Zurich 32 10 min Ad lib Control + 3-NOP Evaluate CH4production (g/d) for 14 preset samplings schemes (e.g. every 0.5 h from 0.0 h from morning feeding, every 8.0 h from 2.0 h from morning feeding, 1 + 7 h from morning feeding, etc) and all data points retained (Reference) using: •𝑦𝑖𝑗𝑘𝑙𝑚𝑛 = 𝜇 + 𝐱𝑖𝑗𝑘𝑙𝑚 T𝜷 + 𝑎𝑛+ 𝜀𝑖𝑗𝑘𝑙𝑚 •𝐱𝑖𝑗𝑘𝑙𝑚𝑛 Trepresents the fixed-effects matrix, 𝜷includes parameters for sampling schemes j, diets kand chambers l, 𝑎𝑚represents cow effect with 𝑎𝑚~𝑁(0, 𝜎𝑎 2)and 𝜖𝑖𝑗𝑘𝑙𝑚~𝑁(0, 𝛿𝑗𝜎2) Statistical models 95% confidence interval for sampled CH4production (g/d) relative to sampling size: •ത 𝑌 CH4± 1.96 ×SDCH4/ 𝑛 Times to measure CH4across studies •Although increased sample sizes/number of days decrease confidence intervals, sampling scheme effects remain, with their magnitude differing across experiments •Hourly and specific 2-hourly or 3-hourly CH4emission samplings result in accurate CH4production estimates, particularly for dairy cattle fed restrictedly •These datasets allow evaluation of sampling frequency and timing (i.e. sampling schemes for sampling as few as twice a day up to twice per hour) for accurate enteric CH4estimation
