Search

Direct measurements of methane (CH 4 ) from individual animals are difficult and expensive. Predictions based on proxies for CH 4 are a viable alternative. Most prediction models are based on multiple linear regressions (MLR) and predictor variables that are not routinely available in commercial farms, such as dry matter intake (DMI) and diet composition. The use of machine learning (ML) algorithms to predict CH 4 emissions from across-country heterogeneous data sets has not been reported. The objectives were to compare performances of ML ensemble algorithm random forest (RF) and MLR models in predicting CH 4 emissions from proxies in dairy cows, and assess effects of imputing missing data points on prediction accuracy. Data on CH 4 emissions and proxies for CH 4 from 20 herds were provided by 10 countries. The integrated data set contained 43,519 records from 3,483 cows, with 18.7% missing data points imputed using k-nearest neighbor imputation. Three data sets were created, 3k (no missing records), 21k (missing DMI imputed from milk, fat, protein, body weight), and 41k (missing DMI, milk fat, and protein records imputed). These data sets were used to test scenarios (with or without DMI, imputed vs. nonimputed DMI, milk fat, and protein), and prediction models (RF vs. MLR). Model predictive ability was evaluated within and between herds through 10-fold cross-validation. Prediction accuracy was measured as correlation between observed and predicted CH 4 , root mean squared error (RMSE) and mean normalized discounted cumulative gain (NDCG). Inclusion of DMI in the model improved within and between-herd prediction accuracy to 0.77 (RMSE = 23.3%) and 0.58 (RMSE = 31.9%) in RF and to 0.50 (RMSE = 0.327) and 0.13 (RMSE = 42.71) in MLR, respectively than when DMI was not included in the predictive model. When missing DMI records were imputed, within and between-herd accuracy increased to 0.84 (RMSE = 18.5%) and 0.63 (RMSE = 29.9%), respectively. In all scenarios, RF models out-performed MLR models. Results suggest routinely measured variables from dairy farms can be used in developing globally robust prediction models for CH 4 if coupled with state-of-the-art techniques for imputation and advanced ML algorithms for predictive modeling., (© 2022, The Authors. Published by Elsevier Inc. and Fass Inc. on behalf of the American Dairy Science Association®. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).)

Organic dairy cows in Denmark are often kept indoors during the winter and outside at least part time in the summer. Consequently, their diet changes by the season. We hypothesized that grazing might affect enteric CH 4 emissions due to changes in the nutrition, maintenance, and activity of the cows, and they might differentially respond to these factors. This study assessed the repeatability of enteric CH 4 emission measurements for Jersey cattle in a commercial organic dairy herd in Denmark. It also evaluated the effects of a gradual transition from indoor winter feeding to outdoor spring grazing. Further, it assessed the individual-level correlations between measurements during the consecutive feeding periods (phenotype × environment, P × E) as neither pedigrees nor genotypes were available to estimate a genotype by environment effect. Ninety-six mixed-parity lactating Jersey cows were monitored for 30 d before grazing and for 24 d while grazing. The cows spent 8 to 11 h grazing each day and had free access to an in-barn automatic milking system (AMS). For each visit to the AMS, milk yield was recorded and logged along with date and time. Monitoring equipment installed in the AMS feed bins continuously measured enteric CH 4 and CO 2 concentrations (ppm) using a noninvasive "sniffer" method. Raw enteric CH 4 and CO 2 concentrations and their ratio (CH 4 :CO 2 ) were derived from average concentrations measured during milking and per day for each cow. We used mixed models equations to estimate variance components and adjust for the fixed and random effects influencing the analyzed gas concentrations. Univariate models were used to precorrect the gas measurements for diurnal variation and to estimate the direct effect of grazing on the analyzed concentrations. A bivariate model was used to assess the correlation between the 2 periods (in-barn vs. grazing) for each gas concentration. Grazing had a weak P × E interaction for daily average CH 4 and CO 2 gas concentrations. Bivariate repeatability estimates for average CH 4 and CO 2 concentrations and CH 4 :CO 2 were 0.77 to 0.78, 0.73 to 0.80, and 0.26, respectively. Repeatability for CH 4 :CO 2 was low (0.26) but indicated some between-animal variation. In conclusion, grazing does not create significant shifts compared with indoor feeding in how animals rank for average CH 4 and CO 2 concentrations and CH 4 :CO 2 . We found no evidence that separate evaluation is needed to quantify enteric CH 4 and CO 2 emissions from Jersey cows during in-barn and grazing periods., (The Authors. Published by FASS Inc. and Elsevier Inc. on behalf of the American Dairy Science Association®. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).)

Highlights • CH 4 from cows can be measured noninvasively on-farm with different methods. • Laser methane detector and breath samplers (‘sniffers’) rank the cows similarly. • Due to different experimental setups: differences in measured CH 4 values. Abstract To measure methane (CH 4) emissions from cattle on-farm, a number of methods have been developed. Combining measurements made with different methods in one data set could lead to an increased power of further analyses. Before combining the measurements, their agreement must be evaluated. We analysed data obtained with a handheld laser methane detector (LMD) and the GreenFeed system (GF), as well as data obtained with LMD and Fourier Transformed Infrared (FTIR) and Non-dispersive Infrared (NDIR) breath analysers (sniffers) installed in the feed bin of automatic milking systems. These devices record short-term breath CH 4 concentrations from cows and make it possible to estimate daily CH 4 production in g/d which is used for national CH 4 emission inventories and genetic studies. The CH 4 is released by cows during eructation and breathing events, resulting in peaks of CH 4 concentrations during a measurement which represent the respiratory cycle. For LMD, the average CH 4 concentration of all peaks during the measurement (P_MEAN in ppm × meter) was compared with the average daily CH 4 production (g/d) measured by GF on 11 cows. The comparison showed a low concordance correlation coefficient (CCC; 0.02) and coefficient of individual agreement (CIA; 0.06) between the methods. The repeated measures correlation (r p) of LMD and GF, which can be seen as a proxy for the genetic correlation, was, however, relatively strong (0.66). Next, based on GF, a prediction equation for estimating CH 4 in g/d (LMD_cal) using LMD measurements was developed. LMD_cal showed an improved agreement with GF (CCC = 0.22, CIA = 0.99, r p = 0.74). This prediction equation was used to compare repeated LMD measurements (LMD_val in g/d) with CH 4 (g/d) measured with FTIR (n = 34 cows; Data Set A) or NDIR (n = 39 cows; Data Set B) sniffer. A low CCC (A: 0.28; B: 0.17), high CIA (A: 0.91; B: 0.87) and strong r p (A: 0.57; B: 0.60) indicated that there was some agreement and a minimal re-ranking of the cows between sniffer and LMD. Possible sources of disagreement were cow activity (LMD: standing idle; sniffer: eating and being milked) and the larger influence of wind speed on LMD measurement. The LMD measurement was less repeatable (0.14–0.27) than the other techniques studied (0.47–0.77). Nevertheless, GF, LMD and the sniffers ranked the cows similarly. The LMD, due to its portability and flexibility, could be used to study CH 4 emissions on herd or group level, as a validation tool, or to strengthen estimates of genetic relationships between small-scale research populations. [ABSTRACT FROM AUTHOR]