Database, parameters extracted and calculations

The data used for the current study derived from a database reported in Castro-Montoya and Dickhoefer (2020)Castro-Montoya, J.M. & Dickhoefer, U. 2020. "The nutritional value of tropical legume forages fed to ruminants as affected by their growth habit and fed form: A systematic review". Animal Feed Science and Technology, 269: 1-14, ISSN: 0377-8401. https://doi.org/10.1016/j.anifeedsci.2020.114641. . For the current analysis, the studies had to report legume species, their proportion in the diet, the nutritional composition of the diet [organic matter (OM), crude protein (CP), neutral detergent fiber (NDF), acid detergent fiber (ADF) concentration], animal body weight (BW), DMI, and OM -or dry matter (DM)- total tract digestibility. If available, milk yield (kg/d) and average daily gain (ADG) were recorded. A total of 258 studies fulfilled these criteria, resulting in 1355 dietary treatments. Further parameters were estimated. The feeding level (FL), presented as a multiple of maintenance energy requirements, was estimated by dividing the metabolizable energy (ME) intake by the maintenance ME requirement for the animals. The ME requirement for each animal was calculated based on the BW reported using the estimates from the meta analysis of Salah et al. (2014)Salah, N., Sauvant, D. & Archimède, H. 2014. "Nutritional requirements of sheep, goats and cattle in warm climates: A meta-analysis". Animal, 8: 1439-1447, ISSN: 1751-732X. https://doi.org/10.1017/S1751731114001153. for tropical small ruminants and cattle (542.64 and 631.26 kJ ME/kg BW^0.75, respectively).

The majority of the studies did not provide information on the ME concentration of the diet, therefore this was estimated according to Minson (1984)Minson, D.J. 1984. "Digestibility and voluntary intake by sheep of five Digitaria species". Australian Journal of Experimental Agriculture, 24: 494-500, ISSN: 1446-5574. https://doi.org/10.1071/EA9840494. [Eq. (1) $\text{ME}\left(\frac{\text{MJ}}{\text{kg}}\right)=0.157\text{DOM}+0.059\text{CP}-1.073$ ] with CP and digestible OM as predictors, two parameters that account for differences between legumes and grasses.

Where

DOM = concentration of digestible organic matter in the diet (g/100 g DM);

CP = concentration of crude protein in the diet (g/100 g DM);

Digestible OM concentration was estimated from the concentration of OM in the diet and from the OMD (g/kg) reported in the studies. For those studies not reporting OMD but where DM digestibility was available, the former was estimated from DM digestibility (DMD (g/kg)) using regressions derived from the data used for this study [Eq. (2) $\text{OMD}\left(\frac{g}{\text{kg}}\right)=100.2+0.876\times \text{DMD};R^2=0.72;\text{for grasses only}$ , (3) $\text{OMD}\left(\frac{\text{g}}{\text{kg}}\right)=108.0+0.848\times \text{DMD};R^2=0.65;\text{for mixed legume}-\text{grasses rations}$ , (4) $\text{OMD}\left(\frac{\text{g}}{\text{kg}}\right)=44.9+0.951\times \text{DMD};R^2=0.88; \text{for legumes only}$ ].

Equations to estimate methane

For the estimation of methane production, equations considering DMI, NDF, ADF, and ME concentrations or intakes, as well as OMD were favored, parameters that capture the characteristics of legume forages, the differences within types of legume and between legumes and grasses, while also allowing for the estimation of methane from a significant amount of observations. In this regard, the predictive equations produced by Patra (2017)Patra, A.K. 2017. "Prediction of enteric methane emission from cattle using linear and non-linear statistical models in tropical production systems". Mitigation and Adaptation Strategies for Global Change, 22: 629-650, ISSN: 1573-1596. https://doi.org/10.1007/s11027-015-9691-7. and Ellis et al. (2007)Ellis, J.L., Kebreab, E., Odongo, N.E., McBride, B.W., Okine, E.K. & France, J. 2007. "Prediction of methane production from dairy and beef cattle". Journal of Dairy Science, 90: 3456-3466, ISSN: 1525-3198. https://doi.org/10.3168/jds.2006-675. were considered. From the equations produced by Patra (2017)Patra, A.K. 2017. "Prediction of enteric methane emission from cattle using linear and non-linear statistical models in tropical production systems". Mitigation and Adaptation Strategies for Global Change, 22: 629-650, ISSN: 1573-1596. https://doi.org/10.1007/s11027-015-9691-7. the Binomial 12 [Eq. (5) $C{H}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.012}_{\pm 0.709}+{0.308}_{\pm 0.249}\times \text{DMI}+{0.0404}_{\pm 0.0119}\times {\text{DMI}}^2+{2.424}_{\pm 0.415}\times \text{NDFI}-{0.290}_{\pm 0.0409}\times {\text{NDFI}}^2$ ], Binomial 13 [Eq. (6) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.490}_{\pm 0.745}+{0.418}_{\pm 0.232}\times \text{DMI}+{0.0415}_{\pm 0.0118}\times {\text{DMI}}^2+{4.311}_{\pm 0.718}\times \text{ADFI}-{0.977}_{\pm 0.138}\times {\text{ADFI}}^2$ ], Linear 17 [Eq. (7) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)={0.910}_{\pm 0.746}+{1.472}_{\pm 0.154}\times \text{DMI}-{1.388}_{\pm 0.451}\times \text{FL}-{0.669}_{\pm 0.338}\times \text{ADFI}$ ] and Linear 18 [Eq. (8) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.559}_{\pm 2.010}+{1.217}_{\pm 0.164}\times \text{DMI}-{2.418}_{\pm 0.724}\times \text{FL}+{0.00714}_{\pm 0.00316}\times \text{OMDm}$ ] were selected based on their performance and the availability of the predictor variables within this dataset.

Where

DMI = dry matter intake in kg/d;

NDFI = NDF intake in kg/d;

ADFI = ADF intake (kg/d);

FL = feeding level as multiple of the maintenance requirements.

OMDm = organic matter digestibility (g/kg) at a maintenance level of feeding (estimated based on OMD following the procedure outlined by Ramin and Huhtanen (2013)Ramin, M. & Huhtanen, P. 2013. "Development of equations for predicting methane emissions from ruminants". Journal of Dairy Science, 96: 2476-2493, ISSN: 1525-3198. https://doi.org/10.3168/jds.2012-6095. .

Where DMI = dry matter intake in kg/d; NDFI = NDF intake in kg/d; ADFI = ADF intake (kg/d); FL = feeding level as multiple of the maintenance requirements; OMDm = organic matter digestibility (g/kg) at a maintenance level of feeding (estimated based on OMD following the procedure outlined by Ramin and Huhtanen (2013)Ramin, M. & Huhtanen, P. 2013. "Development of equations for predicting methane emissions from ruminants". Journal of Dairy Science, 96: 2476-2493, ISSN: 1525-3198. https://doi.org/10.3168/jds.2012-6095. .

From the work of Ellis et al. (2007)Ellis, J.L., Kebreab, E., Odongo, N.E., McBride, B.W., Okine, E.K. & France, J. 2007. "Prediction of methane production from dairy and beef cattle". Journal of Dairy Science, 90: 3456-3466, ISSN: 1525-3198. https://doi.org/10.3168/jds.2006-675. predictive equations 6c [Eq. (9) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)={3.44}_{\pm 0.937}+{0.502}_{\pm 0.115}\times \text{DMI}+{0.506}_{\pm 0.211}\times \text{NDFI}$ ], 7c [Eq. (10) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)={3.63}_{\pm 0.921}+{0.0549}_{\pm 0.00939}\times \text{MEI}+{0.606}_{\pm 0.306}\times \text{ADFI}$ ], 8c [Eq. (11) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)={4.41}_{\pm 1.13}+{0.0224}_{\pm 0.0106}\times \text{MEI}+{0.980}_{\pm 0.241}\times \text{NDFI}$ ], and 10c [Eq. (12) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)={3.41}_{\pm 0.973}+{0.520}_{\pm 0.120}\times \text{DMI}-{0.996}_{\pm 0.447}\times \text{ADFI}+{1.15}_{\pm 0.321}\times \text{NDFI}$ ] were used to estimate methane in this dataset:

Where

DMI = dry matter intake in kg/d;

NDFI = NDF intake in kg/d;

MEI = ME intake in MJ/d;

ADFI = ADF intake in kg/d.

Subsetting of data

The dataset was divided into cattle and small ruminants, and subsets were created according to physiological stage (adult or growing), species (goat and sheep). and legume’s growth habit (herb, shrub and tree legumes). The latter classification was based on the PLANTS database (https://plants.usda.gov) of the Natural Conservation service of the United Sates as described by Castro-Montoya and Dickhoefer (2018)Castro‐Montoya, J.M. & Dickhoefer, U. 2018. "Effects of tropical legume silages on intake, digestibility and performance in large and small ruminants: A review". Grass and Forage Science, 73: 26-39, ISSN: 1365-2494. https://doi.org/10.1111/gfs.12324. . Descriptive statistics of the variables used in the study are presented in table 1. Another subset of data was created including only diets with grass forage with and without concentrate supplementation from small ruminants (table 1). This was not done for cattle because of the reduced number of observations available.

Using an energy density of 55.5 MJ/kg (Elert 2006Elert, G. 2020. The Physics Hypertextbook. Available: https://physics.info/energy-chemical/. June 4.), the estimated methane in MJ/d from Eq. (5 $C{H}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.012}_{\pm 0.709}+{0.308}_{\pm 0.249}\times \text{DMI}+{0.0404}_{\pm 0.0119}\times {\text{DMI}}^2+{2.424}_{\pm 0.415}\times \text{NDFI}-{0.290}_{\pm 0.0409}\times {\text{NDFI}}^2$ - 12) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)={3.41}_{\pm 0.973}+{0.520}_{\pm 0.120}\times \text{DMI}-{0.996}_{\pm 0.447}\times \text{ADFI}+{1.15}_{\pm 0.321}\times \text{NDFI}$ were converted into g/d. Methane estimated from Eq. (5) $C{H}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.012}_{\pm 0.709}+{0.308}_{\pm 0.249}\times \text{DMI}+{0.0404}_{\pm 0.0119}\times {\text{DMI}}^2+{2.424}_{\pm 0.415}\times \text{NDFI}-{0.290}_{\pm 0.0409}\times {\text{NDFI}}^2$ for cattle and from Eq. (8) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.559}_{\pm 2.010}+{1.217}_{\pm 0.164}\times \text{DMI}-{2.418}_{\pm 0.724}\times \text{FL}+{0.00714}_{\pm 0.00316}\times \text{OMDm}$ for small ruminants was further expressed as g/kg metabolic body weight (MBW), g/kg DMI, g/kg digestible organic matter intake (DOMI), g/kg milk, and g/kg ADG.

With the objective of comparing different forages without influence of other ingredients, a subset of data was created where small ruminants were fed exclusively on grasses, straw/stover, herb legumes, shrub legumes, or tree legumes. Boxplots were created with the forages categories for methane in g/kg DMI and in g/kg DOMI. Also, a subset was created to compare different forages when supplemented with concentrate feed in small ruminants. First, observations were selected where grasses or straw/stover were fed in proportions between 60 and 90 g/100 g DM, with the remainder of the DM coming from concentrate feeds. Second, observations were selected where legumes were fed mixed with grasses, where the legume inclusion ranged between 20 and 50 g/100 g DM, and where the total forage proportion (grass + legume) ranged between 60 and 90 g/100 g DM, with the remainder of the DM coming from concentrate feeds. Legumes were then divided into herbs, shrubs and trees. Boxplots were created with the forage categories for methane in g/kg DMI, in g/kg DOMI, and in g/kg ADG.

Statistical analysis

Outliers in the predicted methane from each equation were identified and removed based on the interquartile range. Pearson correlation was estimated within the methane predictions from all equations using the cor ( ) function of R program. The estimated methane production from Eq. (5) $C{H}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.012}_{\pm 0.709}+{0.308}_{\pm 0.249}\times \text{DMI}+{0.0404}_{\pm 0.0119}\times {\text{DMI}}^2+{2.424}_{\pm 0.415}\times \text{NDFI}-{0.290}_{\pm 0.0409}\times {\text{NDFI}}^2$ for cattle and from Eq. (8) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.559}_{\pm 2.010}+{1.217}_{\pm 0.164}\times \text{DMI}-{2.418}_{\pm 0.724}\times \text{FL}+{0.00714}_{\pm 0.00316}\times \text{OMDm}$ for small ruminants was regressed on the legume proportion in the diet using the lm ( ) function of R. Regression analyses were conducted independently for the cattle and the small ruminant datasets. Within each of these datasets, further regression analyses were conducted for subsets based on the physiological stage (adult or growing), species (goat or sheep), and growth habit of legumes (herb, shrub, tree). For the dataset including only control treatments in small ruminants, predicted methane emissions were regressed on the proportion of concentrate in the diet following the procedure described above. Significances of the slopes were declared at P < 0.05, whereas tendencies were declared at 0.05 < P < 0.10.

Predictive equations used

Equations to predict methane were selected based on the available parameters in our dataset. Therefore, equations including e.g. fat, non-structural carbohydrates, or gross energy concentration were not considered. Equations based exclusively on DMI were also disregarded, as these do not consider the differences in the nutritional composition of legumes and grasses. Hence, predictive equations considering DMI, OMD, NDF and ADF concentration or intake were purposely targeted. Recent studies have shown that, when accounting for the nutritional composition and at high inclusion levels, diets with increasing legume level tend to decrease DMI and OMD (da Silva et al. 2017da Silva, T., Pereira, O., Martins, R., Agarussi, M., da Silva, L., Rufino, L., Valadares, F. & Ribeiro, K. 2017. "Stylosanthes cv. Campo Grande silage and concentrate levels in diets for beef cattle". Animal Production Science, 58: 539-545, ISSN: 1836-5787. https://doi.org/10.1071/AN15781. and Montoya-Flores et al. 2020Montoya-Flores, M.D., Molina-Botero, I.C., Arango, J., Romano-Muñoz, J.L., Solorio-Sánchez, F.J., Aguilar-Pérez, C.F. & Ku-Vera, J.C. 2020. "Effect of dried leaves of Leucaena leucocephala on rumen fermentation, rumen microbial population, and enteric methane production in crossbred heifers". Animals, 10: 300, ISSN: 2076-2615. https://doi.org/10.3390/ani10020300. ) with clear differences between herb, shrub and tree legumes (Castro-Montoya and Dickhoefer 2018Castro‐Montoya, J.M. & Dickhoefer, U. 2018. "Effects of tropical legume silages on intake, digestibility and performance in large and small ruminants: A review". Grass and Forage Science, 73: 26-39, ISSN: 1365-2494. https://doi.org/10.1111/gfs.12324. ). Along this line, tropical legumes have a fairly high concentration of NDF and ADF, which reflects on a lower in vitro OMD than grasses, particularly for shrub legumes (Castro-Montoya and Dickhoefer 2020Castro-Montoya, J.M. & Dickhoefer, U. 2020. "The nutritional value of tropical legume forages fed to ruminants as affected by their growth habit and fed form: A systematic review". Animal Feed Science and Technology, 269: 1-14, ISSN: 0377-8401. https://doi.org/10.1016/j.anifeedsci.2020.114641. ), hence, this study aimed at capturing this variation into the estimation of methane.

For the cattle dataset, including both adult and growing animals, all the equations used yielded estimates of methane within expected values for large ruminants (table 3). For example, estimates of Eq. (5) $C{H}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.012}_{\pm 0.709}+{0.308}_{\pm 0.249}\times \text{DMI}+{0.0404}_{\pm 0.0119}\times {\text{DMI}}^2+{2.424}_{\pm 0.415}\times \text{NDFI}-{0.290}_{\pm 0.0409}\times {\text{NDFI}}^2$ to (7) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)={0.910}_{\pm 0.746}+{1.472}_{\pm 0.154}\times \text{DMI}-{1.388}_{\pm 0.451}\times \text{FL}-{0.669}_{\pm 0.338}\times \text{ADFI}$ ranged between 0.76 and 1.58 MJ/kg DMI, while Ellis et al. (2007)Ellis, J.L., Kebreab, E., Odongo, N.E., McBride, B.W., Okine, E.K. & France, J. 2007. "Prediction of methane production from dairy and beef cattle". Journal of Dairy Science, 90: 3456-3466, ISSN: 1525-3198. https://doi.org/10.3168/jds.2006-675. and Yan et al. (2009)Yan, T., Porter, M.G. & Mayne, C.S. 2009. "Prediction of methane emission from beef cattle using data measured in indirect open-circuit respiration calorimeters". Animal, 3: 1455-1462, ISSN: 1751-732X. https://doi.org/10.1017/S175173110900473X. , reported methane emissions ranging from 1.12 to 1.49 MJ/kg DMI. For the cattle dataset, there was a high correlation between all predictive equations (r ≥ 0.71; data not shown). The equations of Ellis et al. (2007)Ellis, J.L., Kebreab, E., Odongo, N.E., McBride, B.W., Okine, E.K. & France, J. 2007. "Prediction of methane production from dairy and beef cattle". Journal of Dairy Science, 90: 3456-3466, ISSN: 1525-3198. https://doi.org/10.3168/jds.2006-675. had consistently higher estimated methane than that of Patra (2017)Patra, A.K. 2017. "Prediction of enteric methane emission from cattle using linear and non-linear statistical models in tropical production systems". Mitigation and Adaptation Strategies for Global Change, 22: 629-650, ISSN: 1573-1596. https://doi.org/10.1007/s11027-015-9691-7. (table 2) likely due to the higher overall methane production of the temperate cattle used by Ellis et al. (2007)Ellis, J.L., Kebreab, E., Odongo, N.E., McBride, B.W., Okine, E.K. & France, J. 2007. "Prediction of methane production from dairy and beef cattle". Journal of Dairy Science, 90: 3456-3466, ISSN: 1525-3198. https://doi.org/10.3168/jds.2006-675. . Not only because Eq. (5) $C{H}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.012}_{\pm 0.709}+{0.308}_{\pm 0.249}\times \text{DMI}+{0.0404}_{\pm 0.0119}\times {\text{DMI}}^2+{2.424}_{\pm 0.415}\times \text{NDFI}-{0.290}_{\pm 0.0409}\times {\text{NDFI}}^2$ was developed from tropical cattle, but also because it was the equation that allowed for the maximum number of observations predicted, the estimates of Eq. (5) $C{H}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.012}_{\pm 0.709}+{0.308}_{\pm 0.249}\times \text{DMI}+{0.0404}_{\pm 0.0119}\times {\text{DMI}}^2+{2.424}_{\pm 0.415}\times \text{NDFI}-{0.290}_{\pm 0.0409}\times {\text{NDFI}}^2$ were used for the calculation of methane relative to MBW, DMI, DOMI, milk yield, and ADG, and the subsequent regression analyses.

For small ruminants Eq. (5) $C{H}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.012}_{\pm 0.709}+{0.308}_{\pm 0.249}\times \text{DMI}+{0.0404}_{\pm 0.0119}\times {\text{DMI}}^2+{2.424}_{\pm 0.415}\times \text{NDFI}-{0.290}_{\pm 0.0409}\times {\text{NDFI}}^2$ , (6) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.490}_{\pm 0.745}+{0.418}_{\pm 0.232}\times \text{DMI}+{0.0415}_{\pm 0.0118}\times {\text{DMI}}^2+{4.311}_{\pm 0.718}\times \text{ADFI}-{0.977}_{\pm 0.138}\times {\text{ADFI}}^2$ and (7) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)={0.910}_{\pm 0.746}+{1.472}_{\pm 0.154}\times \text{DMI}-{1.388}_{\pm 0.451}\times \text{FL}-{0.669}_{\pm 0.338}\times \text{ADFI}$ produced a high number of negative estimations, likely due to the quadratic nature of NDFI and ADFI in Eq. (5) $C{H}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.012}_{\pm 0.709}+{0.308}_{\pm 0.249}\times \text{DMI}+{0.0404}_{\pm 0.0119}\times {\text{DMI}}^2+{2.424}_{\pm 0.415}\times \text{NDFI}-{0.290}_{\pm 0.0409}\times {\text{NDFI}}^2$ and (6) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.490}_{\pm 0.745}+{0.418}_{\pm 0.232}\times \text{DMI}+{0.0415}_{\pm 0.0118}\times {\text{DMI}}^2+{4.311}_{\pm 0.718}\times \text{ADFI}-{0.977}_{\pm 0.138}\times {\text{ADFI}}^2$ and due to the high ADF concentrations in diets in the tropics. Equation (8) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.559}_{\pm 2.010}+{1.217}_{\pm 0.164}\times \text{DMI}-{2.418}_{\pm 0.724}\times \text{FL}+{0.00714}_{\pm 0.00316}\times \text{OMDm}$ , based on the OMD, FL and DMI, yielded methane estimates with mean, min and max (g/d) of 25.9, 2.12 and 62.7 (table 2). The estimates of Eq. (8) ${\text{CH}}_4\left(\frac{\text{MJ}}{\text{d}}\right)=-{1.559}_{\pm 2.010}+{1.217}_{\pm 0.164}\times \text{DMI}-{2.418}_{\pm 0.724}\times \text{FL}+{0.00714}_{\pm 0.00316}\times \text{OMDm}$ still appeared to be on the higher end of daily methane emissions for small ruminants (Pelchen and Peters 1998Pelchen, A. & Peters, K.J. 1998. "Methane emissions from sheep". Small Ruminant Research, 27: 137-150, ISSN: 1879-0941. https://doi.org/10.1016/S0921-4488(97)00031-X. ), but despite the possible overestimation, these estimates were used to calculate methane relative to MBW, DMI, DOMI, and ADG, and the subsequent regression analyses. The estimates from the equations of Ellis et al. (2007)Ellis, J.L., Kebreab, E., Odongo, N.E., McBride, B.W., Okine, E.K. & France, J. 2007. "Prediction of methane production from dairy and beef cattle". Journal of Dairy Science, 90: 3456-3466, ISSN: 1525-3198. https://doi.org/10.3168/jds.2006-675. averaged methane emissions between 71.4 and 88.9 g/d and were deemed not plausible for tropical small ruminants (data not shown).

In vivo methane production when feeding tropical legumes

Methane synthesis is a function of intake, therefore, robust conclusions can only be obtained when DMI has been reported in a study, thus manuscripts with grazing animals are not included in the following discussion. Under those considerations, ten studies were found reporting methane emissions and DMI from ruminants fed on tropical legumes comparing them with a control diet. Six of those in vivo studies tested the effects of leucaena, with four of them finding decreases in methane emissions (Archimède et al. 2016Archimède, H., Rira, M., Barde, D.J., Labirin, F., Marie‐Magdeleine, C., Calif, B., Périacarpin, F., Fleury, J., Rochette, Y., Morgavi, D.P. & Doreau, M. 2016. "Potential of tannin‐rich plants, Leucaena leucocephala, Glyricidia sepium and Manihot esculenta, to reduce enteric methane emissions in sheep". Journal of Animal Physiology and Animal Nutrition, 100: 1149-1158, ISSN: 1439-0396. https://doi.org/10.1111/jpn.12423. , Montoya-Flores et al. 2020Montoya-Flores, M.D., Molina-Botero, I.C., Arango, J., Romano-Muñoz, J.L., Solorio-Sánchez, F.J., Aguilar-Pérez, C.F. & Ku-Vera, J.C. 2020. "Effect of dried leaves of Leucaena leucocephala on rumen fermentation, rumen microbial population, and enteric methane production in crossbred heifers". Animals, 10: 300, ISSN: 2076-2615. https://doi.org/10.3390/ani10020300. , Pineiro-Vásquez et al. 2018Pineiro-Vázquez, A.T., Canul-Solis, J.R., Jiménez-Ferrer, G.O., Alayón-Gamboa, J.A., Chay-Canul, A.J., Ayala-Burgos, A.J., Aguilar-Pérez, C.F. & Ku-Vera, J.C. 2018. "Effect of condensed tannins from Leucaena leucocephala on rumen fermentation, methane production and population of rumen protozoa in heifers fed low-quality forage". Asian-Australasian Journal of Animal Sciences, 31: 1738-1746, ISSN: 1976-5517. https://doi.org/10.5713/ajas.17.0192. and Possenti et al. 2008Possenti, R.A., Franzolin, R., Schammas, E.A., Demarchi, J.J., Frighetto, R.T.S. & Lima, M.A.D. 2008. "Efeitos de dietas contendo Leucaena leucocephala e Saccharomyces cerevisiae sobre a fermentação ruminal e a emissão de gás metano em bovinos". Revista Brasileira de Zootecnia, 37: 1509-1516, ISSN: 1806-9290. http://dx.doi.org/10.1590/S1516-35982008000800025. ) while two other finding no effects (Delgado et al. 2013Delgado, D.C., Galindo, J., Cairo, J., Orta, I. & Dorta, N. 2013. "Supplementation with foliage of L. leucocephala. Its effect on the apparent digestibility of nutrients and methane production in sheep". Cuban Journal of Agricultural Science, 47(3): 267-271, ISSN: 2079-3480. and Moreira et al. 2013Moreira, G.D., Lima, P.D., Borges, B.O., Primavesi, O., Longo, C., McManus, C., Abdalla, A. & Louvandini, H. 2013. "Tropical tanniniferous legumes used as an option to mitigate sheep enteric methane emission". Tropical Animal Health and Production, 45: 879-882, ISSN: 1573-7438. https://doi.org/10.1007/s11250-012-0284-0. ). Importantly, three of the studies reporting a decrease in methane when feeding leucaena also found a decrease in OMD (Archimède et al. 2016Archimède, H., Rira, M., Barde, D.J., Labirin, F., Marie‐Magdeleine, C., Calif, B., Périacarpin, F., Fleury, J., Rochette, Y., Morgavi, D.P. & Doreau, M. 2016. "Potential of tannin‐rich plants, Leucaena leucocephala, Glyricidia sepium and Manihot esculenta, to reduce enteric methane emissions in sheep". Journal of Animal Physiology and Animal Nutrition, 100: 1149-1158, ISSN: 1439-0396. https://doi.org/10.1111/jpn.12423. , Pineiro-Vásquez et al. 2018Pineiro-Vázquez, A.T., Canul-Solis, J.R., Jiménez-Ferrer, G.O., Alayón-Gamboa, J.A., Chay-Canul, A.J., Ayala-Burgos, A.J., Aguilar-Pérez, C.F. & Ku-Vera, J.C. 2018. "Effect of condensed tannins from Leucaena leucocephala on rumen fermentation, methane production and population of rumen protozoa in heifers fed low-quality forage". Asian-Australasian Journal of Animal Sciences, 31: 1738-1746, ISSN: 1976-5517. https://doi.org/10.5713/ajas.17.0192. and Montoya-Flores et al. 2020Montoya-Flores, M.D., Molina-Botero, I.C., Arango, J., Romano-Muñoz, J.L., Solorio-Sánchez, F.J., Aguilar-Pérez, C.F. & Ku-Vera, J.C. 2020. "Effect of dried leaves of Leucaena leucocephala on rumen fermentation, rumen microbial population, and enteric methane production in crossbred heifers". Animals, 10: 300, ISSN: 2076-2615. https://doi.org/10.3390/ani10020300. ). When testing other legume species, Suybeng et al. (2020)Suybeng, B., Charmley, E., Gardiner, C.P., Malau-Aduli, B.S. & Malau-Aduli, A.E. 2020. "Supplementing Northern Australian beef cattle with Desmanthus tropical legume reduces in-vivo methane emissions". Animals, 10: 2097, ISSN 2076-2615. https://doi.org/10.3390/ani10112097. found decreases in methane (g/kg DMI) with increasing levels of Descanthus spp. but ADG decreased at the highest inclusion level (31 g/100 g DM). Three additional studies did not find any effect of feeding Lablab purpureus, Vigna unguiculata, Styzolobium aterrimum, Mimosa caesalpiniaefolia and Macrotyloma axiliare on methane emissions in g/kg DMI (Moreira et al., 2013Moreira, G.D., Lima, P.D., Borges, B.O., Primavesi, O., Longo, C., McManus, C., Abdalla, A. & Louvandini, H. 2013. "Tropical tanniniferous legumes used as an option to mitigate sheep enteric methane emission". Tropical Animal Health and Production, 45: 879-882, ISSN: 1573-7438. https://doi.org/10.1007/s11250-012-0284-0. ; Washaya et al., 2018Washaya, S., Mupangwa, J. & Muchenje, V. 2018. "Chemical composition of Lablab purpureus and Vigna unguiculata and their subsequent effects on methane production in Xhosa lop-eared goats". South African Journal of Animal Science, 48: 445-458, ISSN: 2221-4062. http://dx.doi.org/10.4314/sajas.v48i3.5. and Lima et al., 2020Lima, P.D., Abdalla Filho, A.L., Issakowicz, J., Ieda, E.H., Corrêa, P.S., de Mattos, W.T., Gerdes, L., McManus, C., Abdalla, A.L. & Louvandini, H. 2020. "Methane emission, ruminal fermentation parameters and fatty acid profile of meat in Santa Inês lambs fed the legume macrotiloma". Animal Production Science, 60: 665-673, ISSN: 1836-5787. https://doi.org/10.1071/AN19127. ).

Moreover, Kennedy and Charmley (2012)Kennedy, P.M. & Charmley, E. 2012. "Methane yields from Brahman cattle fed tropical grasses and legumes". Animal Production Science, 52: 225-239, ISSN: 1836-5787. http://dx.doi.org/10.1071/AN11103. quantified the methane emission from several diets containing Lablab purpureus, Leucaena leucocephala, Stylosantes hamata, and Macroptilium bracteatum, but the study was not designed to compare those emissions to those of other diets. The authors found that when increasing the legume proportion in the diet from 20 to 40 g/100 g DM, the changes in methane production where not consistent and depended on the basal grass, the legume species, and the unit in which methane was expressed (Kennedy and Charmley 2012Kennedy, P.M. & Charmley, E. 2012. "Methane yields from Brahman cattle fed tropical grasses and legumes". Animal Production Science, 52: 225-239, ISSN: 1836-5787. http://dx.doi.org/10.1071/AN11103. ). The meta-analysis of Archimède et al. (2011)Archimède, H., Eugène, M., Magdeleine, C.M., Boval, M., Martin, C., Morgavi, D. P., Lecomte, P. & Doreau, M. 2011. "Comparison of methane production between C3 and C4 grasses and legumes". Animal Feed Science and Technology, 166: 59-64, ISSN: 0377-8401. https://doi.org/10.1016/j.anifeedsci.2011.04.003. concluded that legumes from warm environments produced less methane than their grasses counterparts, but the analysis was based on only 12 observations comparing legumes versus grasses. All of these studies point towards reductions in methane production when tropical legumes are fed, but the heterogeneity in legume species, animal characteristics, and diet’s nutritional composition make it venturous to draw a conclusion on the potential of these forages to reduce methane emissions.

Overall effects of legume forages on estimated methane production

Methane decreased with increasing legume proportion in the diet when expressed relative to MBW, however, emissions could decrease simply because of a lower DMI or degradability of the feed. Indeed, recent studies showed that high proportions of tropical legumes decrease intake, as well as OM and NDF degradability of the diet (Castro-Montoya and Dickhoefer 2018Castro‐Montoya, J.M. & Dickhoefer, U. 2018. "Effects of tropical legume silages on intake, digestibility and performance in large and small ruminants: A review". Grass and Forage Science, 73: 26-39, ISSN: 1365-2494. https://doi.org/10.1111/gfs.12324. and da Silva et al. 2017da Silva, T., Pereira, O., Martins, R., Agarussi, M., da Silva, L., Rufino, L., Valadares, F. & Ribeiro, K. 2017. "Stylosanthes cv. Campo Grande silage and concentrate levels in diets for beef cattle". Animal Production Science, 58: 539-545, ISSN: 1836-5787. https://doi.org/10.1071/AN15781. ). In this regard, when expressed relative to DMI, methane still decreased with increasing legume proportion in the diet. However, when expressed relative to DOMI tropical legumes caused a less obvious response in methane, with no effects when fed to cattle and only a tendency to decrease in small ruminants.

Several mechanisms by which legumes may exert this decrease in methane production are discussed. First, microbial degradation of cellulose and hemicellulose yields H₂ that must be later removed from the rumen environment (Carroll and Hungate 1955Carroll, E.J. & Hungate, R.E. 1955. "Formate dissimilation and methane production in bovine rumen contents". Archives of Biochemistry and Biophysics, 56: 525-536, ISSN: 1096-0384. https://doi.org/10.1016/0003-9861(55)90272-1. ). A lower fiber degradability compared with grasses has been observed in tropical legumes, resulting in less substrate for methane formation. Second, diets conducive of higher propionate production sequester H₂ away from methane (Marty and Demeyer 1973Marty, R.J. & Demeyer, D.I. 1973. "The effect of inhibitors of methane production of fermentation pattern and stoichiometry in vitro using rumen contents from sheep given molasses". British Journal of Nutrition, 30: 369-376, ISSN: 1475-2662. https://doi.org/10.1079/BJN19730041. ), however, studies have shown that tropical legumes have higher acetate to propionate ratios (e.g. Castro-Montoya et al. 2020Castro-Montoya, J.M., Goetz, K. & Dickhoefer, U. 2020. "In vitro fermentation characteristics of tropical legumes and grasses of good and poor nutritional quality and the degradability of their neutral detergent fibre". Animal Production Science, 61: 645-654, ISSN: 1836-5787. https://doi.org/10.1071/AN20136. and Montoya-Flores et al. 2020Montoya-Flores, M.D., Molina-Botero, I.C., Arango, J., Romano-Muñoz, J.L., Solorio-Sánchez, F.J., Aguilar-Pérez, C.F. & Ku-Vera, J.C. 2020. "Effect of dried leaves of Leucaena leucocephala on rumen fermentation, rumen microbial population, and enteric methane production in crossbred heifers". Animals, 10: 300, ISSN: 2076-2615. https://doi.org/10.3390/ani10020300. ), thus, a higher sequestration of methane from propionate seems unlikely.

Also, it has been argued that the presence of secondary compounds in legumes may play a role on activity and number of methanogens (Makkar et al. 2007Makkar, H.P.S., Francis, G. & Becker, K. 2007. "Bioactivity of phytochemicals in some lesser-known plants and their effects and potential applications in livestock and aquaculture production systems". Animal, 1: 1371-1391, ISSN: 1751-732X. https://doi.org/10.1017/S1751731107000298. ). The in vivo studies reporting microbial activity and methane are inconclusive. The study of Montoya-Flores et al. (2020)Montoya-Flores, M.D., Molina-Botero, I.C., Arango, J., Romano-Muñoz, J.L., Solorio-Sánchez, F.J., Aguilar-Pérez, C.F. & Ku-Vera, J.C. 2020. "Effect of dried leaves of Leucaena leucocephala on rumen fermentation, rumen microbial population, and enteric methane production in crossbred heifers". Animals, 10: 300, ISSN: 2076-2615. https://doi.org/10.3390/ani10020300. with leucaena did not find any effect of the legume on methanogens number, despite the high concentration of total and condensed tannins and a decreased methane production. Conversely, Lima et al. (2020)Lima, P.D., Abdalla Filho, A.L., Issakowicz, J., Ieda, E.H., Corrêa, P.S., de Mattos, W.T., Gerdes, L., McManus, C., Abdalla, A.L. & Louvandini, H. 2020. "Methane emission, ruminal fermentation parameters and fatty acid profile of meat in Santa Inês lambs fed the legume macrotiloma". Animal Production Science, 60: 665-673, ISSN: 1836-5787. https://doi.org/10.1071/AN19127. found no decrease in methane production when macrotyloma was fed to lambs but found a decrease in the abundance of methanogens (also an increase in Ruminococcus flavefaciesn and Fibrobacter succionogens). In this study it was not possible to prove the direct effects of secondary compounds on methane, nevertheless, the indirect effects of those compounds, if any, could reflect on OMD and DMI, variables considered in these estimations. Importantly, not all tropical legumes are rich in secondary compounds neither it is proven that their activity is biologically significant to exert an effect on methanogenesis, therefore, not all the effects of legumes on methane production can be ascribed to secondary compounds.

Decreased degradability and intake can have negative consequences if their extent affects the supply of nutrients to the animal, ultimately decreasing its performance. Therefore, the best indicator of the effectiveness of a strategy to reduce methane is the reduction of intensity of emissions. In this regard, methane relative to milk yield was not affected by legume proportion (even though a negative slope was observed), whereas methane relative to ADG declined in cattle fed legumes. For methane relative to milk yield, it is important to notice that in the current dataset 95 % of the yields recorded was below 16.7 kg/d (average + 2 standard deviations), therefore the current results represent methane emissions of cattle with low to medium production level. Much higher milk yields are commonly found across the tropics. Not only are diets of high yielding cows usually well balanced in terms of nutrients, but also marginal improvements tend to decrease with higher yields. It is therefore uncertain and worth researching how tropical legumes would influence the intensity of emissions at high milk yield levels. The scenario presented for growing cattle appears to cover a larger range of growth for tropical cattle, where the majority of the ADG’s recorded were below 903 g/d.

Contrary to the cattle dataset, no effect of legumes was observed on methane in g/kg ADG for small ruminants. For goats, decreases in methane relative to MBW, DMI and DOMI were observed, hence a decrease in methane in g/kg ADG would be expected provided that ADG is maintained or increased when feeding legumes. Indeed, median ADG for sheep and goats was 48.0 and 38.7 g/d, respectively, higher than median ADG of small ruminants fed no legumes (36.2 g/d) (table 1). A previous study suggested a quadratic relationship between legume level and ADG, with ADG being maximized around 400 g legume/kg DM (Castro-Montoya and Dickhoefer 2018Castro‐Montoya, J.M. & Dickhoefer, U. 2018. "Effects of tropical legume silages on intake, digestibility and performance in large and small ruminants: A review". Grass and Forage Science, 73: 26-39, ISSN: 1365-2494. https://doi.org/10.1111/gfs.12324. ). High levels of legume inclusion were more common in small ruminants than in cattle (figure 1, Figure 3), therefore, a quadratic effect of legume proportion in the diet on methane in g/kg ADG was tested, but was found not significant (data not shown). Maybe more relevant, for small ruminants, but not for cattle, a number of observations reported a negative ADG, that is when animals lost weight throughout the experimental period. These data were not used for this analysis, as such calculation would produce a negative methane emission which is biologically impossible. Therefore, 12, 9 and 2 observations were removed from the calculations of methane in g/kg ADG of small ruminants without legume, growing sheep, and growing goat, respectively. When considering those negative observations, the median ADG of diets without legumes (25.0 g/d) was much lower than the median observed for growing sheep and goats, (48.0 and 38.1 g/d, respectively; data not shown). This means that the effect of feeding legumes to small ruminants on methane relative to ADG are indeed underestimated in this study, and that a greater reduction in methane emissions, for example over the lifespan of an animal, is likely when improving their feeding via legumes.

Effects of legumes forages according to their growth habit on estimated methane emissions

Recent literature reviews on tropical legumes have shown strong differences in the nutritional characteristics between herb, shrub and tree legumes (Castro-Montoya and Dickhoefer 2018Castro‐Montoya, J.M. & Dickhoefer, U. 2018. "Effects of tropical legume silages on intake, digestibility and performance in large and small ruminants: A review". Grass and Forage Science, 73: 26-39, ISSN: 1365-2494. https://doi.org/10.1111/gfs.12324. and Castro-Montoya and Dickhoefer 2020Castro-Montoya, J.M. & Dickhoefer, U. 2020. "The nutritional value of tropical legume forages fed to ruminants as affected by their growth habit and fed form: A systematic review". Animal Feed Science and Technology, 269: 1-14, ISSN: 0377-8401. https://doi.org/10.1016/j.anifeedsci.2020.114641. ) with shrubs having the highest NDF concentrations, and herbs having the greatest in vivo and in vitro OMD. Herb, shrub and tree legumes also differ in terms of ME, secondary compounds, and CP, to mention some (Castro-Montoya and Dickhoefer 2020Castro-Montoya, J.M. & Dickhoefer, U. 2020. "The nutritional value of tropical legume forages fed to ruminants as affected by their growth habit and fed form: A systematic review". Animal Feed Science and Technology, 269: 1-14, ISSN: 0377-8401. https://doi.org/10.1016/j.anifeedsci.2020.114641. ). All of these parameters will influence intake level, fermentation patterns, total tract digestibility, and performance. It is therefore expectable that different types of legumes will elicit different responses on methane production. Remarkably, the differential effects of herb, shrub and tree legumes were consistent across cattle and small ruminants and reflected previously outlined differences in their nutritional composition. A greater degradability would lead to higher methane emissions, and indeed, higher methane production was observed for herbs when expressed relative to MBW. When expressed relative to DMI and DOMI, herb legumes have higher methane emissions at levels of inclusion below 500 g/kg DM, but because of the steeper slope, with greater inclusion levels herbs produce less methane than shrubs and trees. Shrubs clearly show the lesser potential to decrease methane emissions, even having a tendency for higher emissions when expressed relative to DOMI, likely related to their higher NDF content directly linked to methane production.

According to the current analysis, tree legumes appear to be intermediate in their potential to decrease methane. Tree legumes have a higher lignin concentration and a lower OMD than herbs (Castro-Montoya and Dickhoefer 2020Castro-Montoya, J.M., Goetz, K. & Dickhoefer, U. 2020. "In vitro fermentation characteristics of tropical legumes and grasses of good and poor nutritional quality and the degradability of their neutral detergent fibre". Animal Production Science, 61: 645-654, ISSN: 1836-5787. https://doi.org/10.1071/AN20136. ), as well as a decreased DMI and NDF digestibility (Castro-Montoya and Dickhoefer 2018Castro‐Montoya, J.M. & Dickhoefer, U. 2018. "Effects of tropical legume silages on intake, digestibility and performance in large and small ruminants: A review". Grass and Forage Science, 73: 26-39, ISSN: 1365-2494. https://doi.org/10.1111/gfs.12324. ). The lower digestibility an intake of tree legumes may reflect on a lower animal performance, and ultimately in their potential to decrease methane. Only the foliage of tree legumes is fed to ruminants, resulting in that, compared to herbs, tree legumes have a higher CP content, but also higher secondary compounds content. The presence of secondary compounds in trees may further influence their effects on methane, but the extent and direction of such effects are uncertain, as outlined before. For example, Tiemann et al. (2008)Tiemann, T., Lascano, C., Kreuzer, M. & Hess, H. 2008. "The ruminal degradability of fibre explains part of the low nutritional value and reduced methanogenesis in highly tanniniferous tropical legumes". Journal of the Science of Food and Agriculture, 88: 1794-1803, ISSN: 1097-0010. https://doi.org/10.1002/jsfa.3282. found no correlation between the concentration of condensed tannins and NDF degradability of tropical legumes in vitro. The unclear effects of secondary compounds on ruminants’ nutrition and methane are likely related to their diversity, and must not be neglected from research, this, however, escapes the scope of this study.

The higher degradability and Metabolizable energy of herbs (Castro-Montoya and Dickhoefer 2020Castro-Montoya, J.M. & Dickhoefer, U. 2020. "The nutritional value of tropical legume forages fed to ruminants as affected by their growth habit and fed form: A systematic review". Animal Feed Science and Technology, 269: 1-14, ISSN: 0377-8401. https://doi.org/10.1016/j.anifeedsci.2020.114641. ) is likely conducive of a greater positive effect on animal performance, whereas shrubs would have the lowest. Indeed, when expressed relative to ADG and milk yield, the slopes of the regression of legume herbs on methane were consistently negative and showed a greater potential to decrease intensity of emissions than trees and shrubs (figure 2).

Estimated methane emissions from diets of small ruminants without legumes

In general, the results of the current study evidence that legumes have potential to decrease methane emissions, particularly when expressed relative to DMI and DOMI. The decreases, however, where not comparable to those achieved when concentrate is supplemented to grasses or crop residues from cereals, as reflected in the greater negative slopes in table 5 compared with those of Table 4. Interestingly, the effects of feeding concentrate were contrasting when the basal forage was a grass or a crop residue. Concentrate feeding is regarded as a strategy to decrease methane emissions as less substrate for H₂ producing is available, while steering the fermentation towards more propionate production (Carroll and Hungate, 1955Carroll, E.J. & Hungate, R.E. 1955. "Formate dissimilation and methane production in bovine rumen contents". Archives of Biochemistry and Biophysics, 56: 525-536, ISSN: 1096-0384. https://doi.org/10.1016/0003-9861(55)90272-1. ), while increasing animal productivity, thereby decreasing the intensity of emissions. In the current analysis, higher concentrate supplementation in the diet of small ruminants fed on grass constantly decreased methane emissions, including when expressed relative to ADG, an effect not always observed with the legumes. However, when concentrate was supplemented to straw, methane tended to increase when expressed relative to MBW and had no effect on methane relative to DMI. The supplementation of straw with concentrate feed certainly caused an increase in the overall diet digestibility, thereby producing more methane. Indeed, when methane was expressed relative to DOMI the supplementation of concentrate tended to decrease methane emissions. The effects of supplementing concentrate feeds to crop residues were not obvious when expressed relative to ADG. The reduction of methane when feeding a concentrate feed is remarkable considering that in most cases the portion referred as concentrate consisted of only a cereal meal, brans, cakes or byproducts from other industry, or mixtures of cereal and protein meals. This highlights, on the one hand, the low quality of basal forages found across the tropics, but on the other hand, the great potential that adding even a small proportion of readily digestible energy or protein sources may have in the utilization of that basal forage, performance, and emissions.

Importantly, the authors do not want to suggest that concentrates should be used in detriment of legume forages. Utilizing legume forages has several advantages for the agricultural production system. Moreover, enteric methane is not the only measurement of environmental impact of a given strategy, and other factors like carbon footprint, water footprint, or biodiversity should be considered for the legume forages vs. concentrates comparison.

Archimède et al. (2011)Archimède, H., Eugène, M., Magdeleine, C.M., Boval, M., Martin, C., Morgavi, D. P., Lecomte, P. & Doreau, M. 2011. "Comparison of methane production between C3 and C4 grasses and legumes". Animal Feed Science and Technology, 166: 59-64, ISSN: 0377-8401. https://doi.org/10.1016/j.anifeedsci.2011.04.003. found that sole legumes decrease methane emissions compared with sole grasses. In an attempt to do a similar evaluation boxplots were created comparing estimated methane for diets containing only grasses, only straw/stover, or only legumes (figure 6A-B). The boxplots clearly show higher methane emissions from both grasses and straw compared with only legumes for both methane in g/kg DMI and in g/kg DOMI, with herbs consistently having less methane than that of shrubs and trees. Unfortunately, not enough observations were available to compare sole forages on the basis of methane relative to ADG. Furthermore, when the grasses supplemented with concentrates were compared with mixed legume-grass rations supplemented with concentrates (figure 6C-D-E), relative to DMI and DOMI, median methane produced by grasses was lower than that of herbs and shrubs, and appeared similar to that of trees. However, when expressed relative to ADG the lowest median methane was found for diets where herbs were combined with grasses and supplemented with concentrate, slightly lower than those of only grass supplemented with concentrate. The synergism between legumes and grasses has been previously reported (Minson 1990Minson, D.J. 1990. Forage in ruminant nutrition. Academic Press Inc. CA, USA.), albeit not specific for methane emissions, but it could be related to a greater partitioning of energy away from metabolic losses.

Worth noticing is that the median estimates of methane (g/kg ADG) from shrubs seem to be higher even than those where straws as sole forage were supplemented with concentrates. It is also important to see the variation in the effects of diets with tree legumes on methane relative to ADG. This further highlights the need to evaluate the effects of legumes depending on their growth habit or other classifications that account for their different nutritional quality. It becomes obvious that not all tropical legumes share the same nutritional attributes.

The results of the current analysis must be still validated with further studies in vivo, where the main varying effect should be the inclusion of the legume. If possible, factors like the forage to concentrate ratio or the NDF, CP, and ME concentration across dietary treatments should remain constant or at least similar. For forage based diets the comparison of diets with and without legumes is still valuable, but adjusting for the nutrients composition is not possible. In this case, the use of relevant nutrients concentration as covariates in the statistical analysis might be an option to account for the changes in the nutritional value of the diet. Expressing methane relative to DMI and digestible OM or DM intake, can also help to draw stronger conclusions. More importantly, aiming at obtaining a measure of intensity of emissions should be a priority of any study. Knowing the ability of rumen microbes to adapt to different conditions, mid to long-term trials would be ideal to assess the potential of tropical legumes to decrease methane emissions when routinely included in the rations of ruminants.

The results of the current analysis clearly show the potential of legumes to decrease methane emissions in ruminants in tropical environments, even though the evidence on the reduction of intensity of methane emissions is still inconclusive. Strong differences appeared in the effects of tropical legumes on estimated methane between herb, shrub, and tree legumes, where herbs showed a clear advantage in reducing methane. While methane estimates from only grasses versus only legumes appeared to favor the latter, the reduction of methane estimates when feeding tropical legumes is less than that of supplementing concentrate feeds to grasses. In terms of intensity of methane emissions, a good potential was observed when herb legumes were fed mixed with a grass and supplemented with a concentrate, highlighting the complementarity and the potential that exists in diets containing all these feed resources.

Base de datos, parámetros extraídos y cálculos

Los datos utilizados para este estudio provienen de una base de datos informada en Castro-Montoya y Dickhoefer (2020)Castro-Montoya, J.M., Goetz, K. & Dickhoefer, U. 2020. "In vitro fermentation characteristics of tropical legumes and grasses of good and poor nutritional quality and the degradability of their neutral detergent fibre". Animal Production Science, 61: 645-654, ISSN: 1836-5787. https://doi.org/10.1071/AN20136. . Para el análisis actual, los estudios debían informar las especies de leguminosas, su proporción en la dieta, la composición nutricional de esta [materia orgánica (MO), proteína cruda (PC), fibra detergente neutro (FDN), fibra detergente ácida (FAD)], peso corporal del animal (PC), CMS y digestibilidad total de MO -o materia seca (MS)- del tracto. Si estaba disponible, se registraron el rendimiento de la leche (kg/d) y la ganancia media diaria (GMD). Un total de 258 estudios cumplieron con estos criterios, lo que resultó en 1355 tratamientos dietéticos y se estimaron otros parámetros. El nivel de alimentación (FL, siglas en inglés), presentado como un múltiplo de los requerimientos de energía de mantenimiento, se estimó dividiendo el consumo de energía metabolizable (EM) por el requerimiento de EM de mantenimiento para los animales. El requerimiento de EM para cada animal se calculó basado en el peso corporal informado, utilizando las estimaciones del metaanálisis de Salah et al. (2014)Salah, N., Sauvant, D. & Archimède, H. 2014. "Nutritional requirements of sheep, goats and cattle in warm climates: A meta-analysis". Animal, 8: 1439-1447, ISSN: 1751-732X. https://doi.org/10.1017/S1751731114001153. para pequeños rumiantes y bovinos tropicales (542.64 y 631.26 kJ EM/kg PC^0.75, respectivamente).

La mayoría de los estudios no proporcionaron información sobre la concentración de EM de la dieta, por lo que esta se estimó según Minson (1984)Minson, D.J. 1984. "Digestibility and voluntary intake by sheep of five Digitaria species". Australian Journal of Experimental Agriculture, 24: 494-500, ISSN: 1446-5574. https://doi.org/10.1071/EA9840494. [Eq. (1) $ME \left(MJ/kg\right)= 0.157 DOM + 0.059 CP - 1.073$ ] con PC y MO digerible como predictores. Estos dos parámetros explican las diferencias entre leguminosas y hebáceas.

Donde:

DOM = concentración de materia orgánica digestible en la dieta (g/100 g MS)

CP = concentración de proteína cruda en la dieta (g/100 g MS).

La concentración de MO digestible se estimó a partir de la concentración de MO en la dieta y de la DMO (g/kg) informada en los estudios. En aquellos estudios que no reportaron la DMO, pero la digestibilidad de la MS estaba disponible, la primera se estimó a partir de la digestibilidad de la MS (DMS (g/kg)), utilizando regresiones derivadas de los datos empleados para este estudio [Eq. (2) $OMD \left(g/kg\right)= 100.2 + 0.876 \times DMD;R^2=0.72; ; solo para pastos$ , (3) $OMD \left(g/kg\right)= 108.0 + 0.848 \times DMD; R^2=0.65;para leguminosas mixtas-raciones de gramíneas$ , (4) $OMD \left(\frac{g}{kg}\right)= 44.9 + 0.951 \times DMD;R^2=0.88; solo para leguminosas$ ].

Ecuaciones para estimar el metano

Para estimar la producción de metano, se favorecieron las ecuaciones que consideran las concentraciones o consumos de CMS, FND, FDA y EM, así como la DMO, parámetros que capturan las características de los forrajes de leguminosas, las diferencias entre los tipos de leguminosas y entre leguminosas y hebáceas, mientras que permitiendo también la estimación de metano a partir de una cantidad significativa de observaciones. En este sentido, se consideraron las ecuaciones predictivas de Patra (2017)Patra, A.K. 2017. "Prediction of enteric methane emission from cattle using linear and non-linear statistical models in tropical production systems". Mitigation and Adaptation Strategies for Global Change, 22: 629-650, ISSN: 1573-1596. https://doi.org/10.1007/s11027-015-9691-7. y Ellis et al. (2007)Ellis, J.L., Kebreab, E., Odongo, N.E., McBride, B.W., Okine, E.K. & France, J. 2007. "Prediction of methane production from dairy and beef cattle". Journal of Dairy Science, 90: 3456-3466, ISSN: 1525-3198. https://doi.org/10.3168/jds.2006-675. . De las ecuaciones de Patra (2017)Patra, A.K. 2017. "Prediction of enteric methane emission from cattle using linear and non-linear statistical models in tropical production systems". Mitigation and Adaptation Strategies for Global Change, 22: 629-650, ISSN: 1573-1596. https://doi.org/10.1007/s11027-015-9691-7. se seleccionaron el Binomio 12 [Eq. (5) $C{H}_4 \left(\frac{MJ}{d}\right)= -{1.012}_{\pm 0.709} + {0.308}_{\pm 0.249} \times DMI + {0.0404}_{\pm 0.0119} \times DM{I}^2 + {2.424}_{\pm 0.415} \times NDFI - {0.290}_{\pm 0.0409} \times NDF{I}^2$ ], Binomio 13 [Eq. (6) $C{H}_4 \left(\frac{MJ}{d}\right)= -{1.490}_{\pm 0.745} + {0.418}_{\pm 0.232} \times DMI + {0.0415}_{\pm 0.0118} \times DM{I}^2 + {4.311}_{\pm 0.718} \times ADFI - {0.977}_{\pm 0.138} \times ADF{I}^2$ ], Lineal 17 [Eq. (7) $C{H}_4 \left(\frac{MJ}{d}\right)= {0.910}_{\pm 0.746} + {1.472}_{\pm 0.154} \times DMI - {1.388}_{\pm 0.451} \times FL - {0.669}_{\pm 0.338} \times ADFI$ ] y Lineal 18 [Eq. (8) $C{H}_4 \left(\frac{MJ}{d}\right)= -{1.559}_{\pm 2.010} + {1.217}_{\pm 0.164} \times DMI - {2.418}_{\pm 0.724} \times FL + {0.00714}_{\pm 0.00316} \times OMDm$ ] en función de su rendimiento y la disponibilidad de las variables predictoras dentro de este conjunto de datos.

Donde:

DMI = consumo de materia seca en kg/d

NDFI = consumo de FDN en kg/d

ADFI = consumo de FDA (kg/d)

FL = nivel de alimentación como múltiplo de los requisitos de mantenimiento

DMOm = digestibilidad de la materia orgánica (g/kg) a nivel de alimentación de mantenimiento (estimado basado en la DMO según Ramin y Huhtanen (2013)Ramin, M. & Huhtanen, P. 2013. "Development of equations for predicting methane emissions from ruminants". Journal of Dairy Science, 96: 2476-2493, ISSN: 1525-3198. https://doi.org/10.3168/jds.2012-6095.

De los estudios de Ellis et al. (2007)Ellis, J.L., Kebreab, E., Odongo, N.E., McBride, B.W., Okine, E.K. & France, J. 2007. "Prediction of methane production from dairy and beef cattle". Journal of Dairy Science, 90: 3456-3466, ISSN: 1525-3198. https://doi.org/10.3168/jds.2006-675. , se utilizaron las ecuaciones predictivas 6c [Eq. (9) $C{H}_4 \left(\frac{MJ}{d}\right)= {3.44}_{\pm 0.937} + {0.502}_{\pm 0.115} \times DMI + {0.506}_{\pm 0.211} \times NDFI$ ], 7c [Eq. (10) $C{H}_4 \left(\frac{MJ}{d}\right)= {3.63}_{\pm 0.921} + {0.0549}_{\pm 0.00939} \times MEI + {0.606}_{\pm 0.306} \times ADFI$ ], 8c [Eq. (11) $C{H}_4 \left(\frac{MJ}{d}\right)= {4.41}_{\pm 1.13} + {0.0224}_{\pm 0.0106} \times MEI + {0.980}_{\pm 0.241} \times NDFI$ ], y 10c [Eq. (12) $C{H}_4 \left(\frac{MJ}{d}\right)= {3.41}_{\pm 0.973} + {0.520}_{\pm 0.120} \times DMI - {0.996}_{\pm 0.447} \times ADFI + {1.15}_{\pm 0.321} \times NDFI$ ] para estimar el metano en este conjunto de datos:

Donde:

DMI = consumo de materia seca en kg/d

NDFI = consumo de FDN en kg/d

MEI = consumo de EM en MJ/d

ADFI = consumo de FDA en kg/d

Subconjunto de datos

El conjunto de datos se dividió en bovino y pequeños rumiantes, y se crearon subconjuntos según la etapa fisiológica (adulto o en crecimiento), especies (cabras y ovejas), y el hábito de crecimiento de las leguminosas (herbáceas, arbustivas y arbóreas). Para esta última clasificación se utilizó la base de datos PLANTS (https://plants.usda.gov) del servicio de Conservación Natural de los Estados Unidos, según lo descrito por Castro-Montoya y Dickhoefer (2018)Castro‐Montoya, J.M. & Dickhoefer, U. 2018. "Effects of tropical legume silages on intake, digestibility and performance in large and small ruminants: A review". Grass and Forage Science, 73: 26-39, ISSN: 1365-2494. https://doi.org/10.1111/gfs.12324. . Las estadísticas descriptivas de las variables utilizadas en el estudio se presentan en la tabla 1. Se creó otro subconjunto de datos que incluye solo dietas con forraje de pasto con y sin suplementos concentrados para pequeños rumiantes (tabla 1). Esto no se realizó para el ganado debido al reducido número de observaciones disponibles.