www.elsevier.com / locate / livprodsci
The prediction of grass silage intake by beef cattle receiving
barley-based supplements
a ,
*
b a,b a,bB.F. McNamee
, D.J. Kilpatrick , R.W.J. Steen
, F.J. Gordon
a
Agricultural Research Institute of Northern Ireland, Hillsborough, Co. Down, BT26 6DP, UK
b
Department of Agriculture for Northern Ireland and the Queen’s University, Belfast, UK
Received 12 July 1999; received in revised form 9 May 2000; accepted 8 June 2000
Abstract
Data on the intake by beef cattle of 11 silages either unsupplemented or supplemented with barley-based concentrates were provided from six studies carried out at the Institute. An empirical model was developed from these data which described an exponential decrease in silage intake as concentrate intake was increased, with y-axis intercepts corresponding
0.75
to unsupplemented silage intakes and a common x-axis intercept of 73.2 (S.E.52.00) g / kg W corresponding to concentrate intake when offered as a sole feed. The model curvature (r parameter) was also shown to be constant for all silages (1.025 (S.E.50.0038)). When the model was validated against external data (11 published studies), the relationship
2
between actual and predicted silage DMI gave an R of 0.89 with an intercept and slope not significantly different from 0 and 1, respectively. Seventy-six percent of silage DMIs predicted using the developed model were within 0.5 kg DM of the observed intakes in the validation data. The developed model showed an improvement over a previously published model, as indicated by a significant reduction in the mean squared prediction error. 2001 Elsevier Science B.V. All rights reserved.
Keywords: Beef cattle; Predicted grass silage ad libitum intake; Concentrate supplementation; Empirical model; Substitution rate
1. Introduction dairy systems. While some models aim to directly
predict the total intake in mixed diets (Vadiveloo and Estimates of voluntary intake form an important Holmes, 1979; Agricultural Research Council, 1980; part of feed rationing systems for ruminants and Rook and Gill, 1990), Lewis (1981) has argued that many approaches have been adopted for the develop- a more logical approach in mixed grass silage / con-ment of empirical models to predict the ad libitum centrate diets may be to adopt a two-stage prediction. dry matter intake of grass silage (SDMI) in beef and The first stage being to derive a model to predict the intake potential of silage when given as the sole feed, and the second to correct this model for the substitution effect of concentrates. This concept is *Corresponding author. Present address: Galen Ltd, Seagoe
supported by recognition that the extent to which the Industrial Estate, Craigavon, BT63 5UA, Northern Ireland, UK.
intake of grass silage is reduced by supplementary Tel.: 144-28-3833-4974; fax:144-28-3835-0206.
E-mail address: [email protected] (B.F. McNamee). concentrates depends, to a large extent, on the intake
of the forage given as a sole feed (Wilkinson, 1985; Irwin, 1970; Forbes and Jackson, 1971; Steen and
Thomas, 1987). McIlmoyle, 1982; Steen and Kilpatrick, 1999; Steen
A number of models for predicting the intake of and Kilpatrick, unpublished; Patterson et al., un-silage as a sole feed are available in the literature. published). In the studies, 11 silages (187–510 g / kg Such models have included the independent variates DM) were offered to a total of 681 heifers, bulls and of silage digestibility and dry matter concentration steers of mixed breeds (initial liveweight 250–575 (Baumgardt et al., 1976; Lewis, 1981), whole diet kg) for periods ranging from 28 to 230 days. Of the gross energy concentration and metabolisability (Ag- 11 silages examined, eight were offered as a sole ricultural Research Council, 1980), silage ammonia feed and with 2, 3 or 4 levels of concentrate nitrogen concentration (Lewis, 1981), silage neutral supplementation. The other three silages were only detergent fibre concentration (Baumgardt et al., offered to the animals supplemented with 3, 4 or 5 1976) and near infrared reflectance spectroscopy levels of concentrates. The concentrate supplements (NIRS) (Steen et al., 1998). The latter approach has offered in the studies were mostly barley based proved to be both accurate and simple and par- (600–975 g / kg barley) and in all cases were offered ticularly suitable for use within a technical support separately to the silage portion of the diet. In all role for the agricultural industry. The objective of studies the animals were housed in open pens and this paper is to develop this approach further by were offered silage through a barrier. Silage was providing a prediction model that will adjust ad refreshed twice daily.
0.75
libitum silage intake for the effect of level of The SDMIs (g / kg W ) were analysed using supplementary concentrates offered in the diet of the Genstat analysis of variance procedures (Genstat 5 beef animal. If this can be accurately achieved then Committee, 1989). Silage3concentrate means from the combined prediction of silage intake as a sole the analysis of variance were used in a regression feed and the effect of concentrate level on that intake analysis to investigate the relationship between y5 would enable more accurate feed rationing systems silage intake and x5concentrate intake on a g / kg
0.75
to be developed. While Lewis (1981) has followed a W basis. An exponential relationship of the form
x
similar approach, this has proved to have limited y5a1br was examined, where (a1b) repre-accuracy in previous independent validations (Ag- sented silage intake, when silage was offered as a ricultural and Food Research Council, 1991). sole feed. Three sets of models were fitted:
1. with separate a, b and r parameters for each of
2. Materials and methods the 11 silages;
2. as (1) but with a common r parameter;
The approach adopted in the present study was to 3. as (2) but extending to take account of the develop an empirical model from data obtained from constraint that the intercept of the equation on the six studies carried out with growing cattle at this x-axis, equating to the concentrate intake at which Institute and to validate the model against 11 pub- the silage intake was zero, should be constant for lished studies carried out outside the Institute. Be- all silages i.e. the concentrate intake when con-cause of the effect of liveweight on feed intake, all centrate as the sole feed was independent of data were expressed in the form (kg silage DM silage type. This introduced the following
con-0.75
intake / kg liveweight )31000 and (kg supplement straint and modified equation: log (e 2a /b ) / logi i e
0.75 k x
DM intake / kg liveweight )31000. (r)5k giving y5 2b (ri 2r ) where k is a constant and is equal to concentrate intake for zero silage intake and i is silage type for i51–11. 2.1. Model development
2.2. Model validation the model in establishing the relationship between silage and concentrate DMI. A strong exponential The developed model was validated against data relationship was identified, which when fitted in its
2
from 11 published studies, carried out at various unconstrained form, showed an R value of 0.99. locations outside this Institute (Leaver, 1973; Dren- Statistical tests used to compare the parameters for nan and Lawlor, 1976; Flynn, 1978; Drennan, 1979; the individual silages showed that a common r-value Aston and Tayler, 1980; Petchey and Broadbent, of 1.025 (S.E.50.0038) could be assumed for the 1980; Veira et al., 1983, 1985, 1990; Drennan and exponential relationship.
2 Keane, 1987a,b). These studies incorporated 15 grass When fitted in its constrained form, the R value silages (189–254 g / kg DM) offered to a total of 919 was 0.98, with a corresponding estimate for k of 73.2
0.75
steers, bulls and weanlings of mixed breeds and (S.E.52.00) g concentrate DM / kg W . The differ-ranging from 93 to 491 kg liveweight for periods of ence in goodness of fit between the unconstrained 84–154 days. Silages were offered either unsup- and constrained forms was non-significant (P.0.05). plemented or with barley-based supplements (.875 The SDMIs for the 11 silages used to develop the g / kg fresh barley) at 1, 2, 3 or 4 levels (0–63.3 g / kg model are presented in Fig. 1. The plotted points
0.75
W / day). represent the observed SDMIs at the range of CDMIs
In the validation exercise, predicted silage intakes for the 11 silages while the plotted curves represent (P) in supplemented diets, using the developed the predicted SDMIs for the same 11 silages in model, were compared against actual observed silage accordance with the above constrained equation. intakes (A), using the mean-square prediction error
(MSPE) defined as 3.2. Model validation
1
2
] In order to apply the developed model to the data
MSPE5
O
sA2Pd nvalidation set, it was necessary to evaluate the silage-where n is the number of pairs of values of A and P specific parameters in the model using the silage dry being compared. The MSPE can be regarded as the matter intake when silage i was offered as the sole sum of the mean bias, the line bias and the random feed (SDMImax), thus:
variation about the regression line (Rook et al., k
SDMImaxi5b (ri 21) 1990). Results are presented in terms of the
propor-tional contribution of each of these three components giving the equation:
to the MSPE. The square root of the MSPE is the k x k
y5SDMImax(r 2r ) /(r 21). mean prediction error (MPE) and is reported as a
proportion of the mean observed silage intake.
This equation was applied to calculate the pre-Finally the predicted silage intakes using the present
dicted silage intakes for the validation data using the model were compared against silage intakes
pre-SDMImax for the particular silages and the values dicted using the Lewis (1981) model. Again the
for k, the concentrate intake for zero silage intake MSPE was used as the parameter of accuracy.
estimated from the development data set, and the estimated value for r. This gives the equation (ARINI model): The wide range of SDMIs recorded for the 11 silage intakes was found to be linear with intercepts silages when offered as the sole feed (50.6 to 88.8 g non-significantly (P.0.05) different from zero and
0.75
DM / kg W ), together with the range of concen- regression coefficients non-significantly (P.0.05) trate supplementation offered to the animals (8.8 to different from unity i.e. 458lines through the origin.
0.75
Fig. 1. The observed and predicted silage DMI of 11 silages in supplemented diets with various levels of DMI when offered as a sole feed.
Fig. 2. Actual versus predicted silage DMI using the ARINI model from 11 published feeding studies carried out at various locations outside the Institute.
Table 1
a b
Example substitution rates (SR ) and corresponding marginal substitution rates (MSR ) as determined by the ARINI model (all intakes are
0.75
expressed as g / kg W )
Concentrate DMI 10 20 30 40 50 60
SDMI as
SDMI SR SDMI SR MSR SDMI SR MSR SDMI SR MSR SDMI SR MSR SDMI SR MSR
sole feed
50.0 47.3 0.27 43.8 0.31 0.04 39.4 0.35 0.04 33.6 0.41 0.06 26.2 0.48 0.07 16.8 0.55 0.07 60.0 56.8 0.32 52.6 0.37 0.05 47.2 0.43 0.06 40.3 0.49 0.06 31.5 0.57 0.08 20.1 0.66 0.09 70.0 66.2 0.38 61.3 0.43 0.05 55.1 0.50 0.07 47.1 0.57 0.07 36.7 0.67 0.10 23.5 0.78 0.11 80.0 75.7 0.43 70.1 0.49 0.06 63.0 0.57 0.08 53.8 0.66 0.09 42.0 0.76 0.10 26.8 0.89 0.13 90.0 85.1 0.49 78.9 0.56 0.07 70.8 0.64 0.08 60.5 0.74 0.10 47.2 0.86 0.12 30.2 1.00 0.14
a
kg of silage DM reduced per kg of concentrate offered.
b 0.75
The difference in SR by increasing the concentrate intake by 10 g / kg W .
which was developed from a similar standpoint. The wide range of weights were used. This large vari-validation parameters are shown in Table 2, together ation necessitates accurate correction of intakes for with the MSPE and the proportion of MSPE attribut- liveweight differences — a correction which is not able to mean bias, line bias and random error. The taken into account in the Lewis (1981) model. In mean bias and the MPE are expressed as a propor- addition, the Lewis equation is based on a series of tion of the observed silage intake. The mean pre- parallel curves, of SDMI versus concentrate DMI, dicted intake using the Lewis (1981) model was for silages with differing intakes as a sole feed considerably lower than the actual mean intake with whereas, the ARINI model uses a series of
con-0.75
a bias of 4.29 g / kg W compared to 1.01 g / kg vergent curves. It is expected that the use of con-0.75
W for the ARINI model. This large mean bias vergent curves, to describe the decreasing SDMI as was responsible for the inaccuracies encountered CDMI is increased, is biologically more appropriate with the Lewis model when validated using these as it reflects the decreasing significance of the silage data, with the mean bias accounting for 30% of the quality (and hence its intake as a sole feed) as silage calculated MSPE. The result was that the MPE for represents a decreasing proportion of the total diet. the Lewis model was 0.140 compared to 0.088 for
2
the ARINI model. Furthermore, the R of the
predicted versus actual SDMIs for the validation data 4. Conclusions
set showed a significant improvement from 0.76 to
0.89 with the ARINI as compared to the Lewis A model has been developed to predict the SDMI model indicating that the new model accounts for at a given level of concentrate supplementation more of the variance. providing the intake of the silage when consumed as The poor relative performance of the Lewis a sole feed is known or may be accurately predicted (1981) model in the present validation exercise may from silage compositional parameters. Validation of have been due to the fact that the validation data the model with independent data revealed that 76% were derived from experiments in which animals of a of SDMI predictions were within 8% of the actual
Table 2
0.75
Prediction precision of ARINI model as compared to the Lewis (1981) model (actual mean intake555.7 (S.E. 2.71; n533) g DM / kg W )
a
Model Predicted intake Proportion of MSPE
b a
Mean S.E. MSPE Bias Line Random Bias MPE
ARINI 54.7 2.88 26.0 0.026 0.141 0.833 1.01 0.091
Lewis 51.4 3.08 65.1 0.276 0.190 0.534 4.29 0.145
a 0.75
g DM / kg W .
b 0.75 2
Reference Manual. Oxford Scientific Publications, Clarendon intakes. The developed model has been designed for
Press, Oxford. use in animals offered silage and concentrates of
Leaver, J.D., 1973. Rearing of dairy cattle 4. Effect of concentrate similar composition, and in accordance with the supplementation on the liveweight gain and feed intake of feeding regimen, to that adopted in the studies from calves offered roughages ad libitum. Anim. Prod. 17, 43–52. which the model was derived, i.e. silage5187–510 Lewis, M., 1981. Equations predicting silage intake by beef and
dairy cattle. In: Proceedings of the Sixth Silage Conference, g / kg DM, concentrates5600–975 g / kg barley and
Edinburgh, pp. 35–36. silage provided ad libitum, separate to the
concen-Petchey, A.M., Broadbent, P.J., 1980. The performance of fatten-trate supplementation. ing cattle offered barley and grass silage in various proportions either as discrete feeds or as a complete diet. Anim. Prod. 31, 251–257.
Rook, A.J., Gill, M., 1990. Prediction of the voluntary intake of
References
grass silages by beef cattle 1. Linear regression analyses. Anim. Prod. 50, 425–438.
Agricultural and Food Research Council, 1991. AFRC Technical Rook, A.J., Dhanoa, M.S., Gill, M., 1990. Prediction of the Committee on Responses to Nutrients. Report no. 8. Voluntary voluntary intake of grass silages by beef cattle 3. Precision of intake of cattle. Nutr. Abs. Rev. (Series B) 61, 815–822. alternative prediction models. Anim. Prod. 50, 455–466. Agricultural Research Council, 1980. The Nutrient Requirements Steen, R.W.J., Kilpatrick, D.J., 1999. The relative effects of grass
of Ruminant Livestock. Commonwealth Agricultural Bureaux, silage:concentrate ratio, metabolisability of the diet and ad Slough. libitum versus restricted dry matter intake at equivalent energy Aston, K., Tayler, J.C., 1980. Effects of supplementing maize and intake on the performance and carcass composition of beef
grass silages with barley, and maize silage with urea or cattle. Livest. Prod. Sci., In press.
ammonia, on the intake and performance of fattening bulls. Steen, R.W.J., McIlmoyle, W.A., 1982. The effect of frequency of Anim. Prod. 31, 243–250. harvesting grass for silage and level of concentrate supple-Baumgardt, B.R., Crabill, L.F., Gobble, J.L., Wangness, J., 1976. mentation on the intake and performance of beef cattle. Anim.
In: Proceedings 1st International Symposium on Feed Com- Prod. 35, 245–252.
position, Animal Nutrient Requirements and Computerisation Steen, R.W.J., Gordon, F.J., Dawson, L.E.R., Park, R.S., Mayne, of Diets, pp. 464–469. C.S., Agnew, R.E. et al., 1998. Factors affecting the intake of Drennan, M.J., 1979. Compensatory growth in cattle 1. Influence grass silage by cattle and prediction of silage intake. Anim. Sci.
of feeding level during the first winter (9 to 14 months of age) 66, 115–127.
on subsequent performance and carcass composition. Ir. J. Thomas, C., 1987. Factors affecting substitution rates in dairy Agric. Res. 18, 131–143. cows on silage based rations. In: Harseign, W., Cole, D.J.A. Drennan, M.J., Keane, M.G., 1987a. Concentrate feeding levels (Eds.), Recent Advances in Animal Nutrition, pp. 205–218.
for unimplanted and implanted finishing steers fed silage. Ir. J. Vadiveloo, J., Holmes, W., 1979. The prediction of the voluntary Agric. Res. 26, 129–137. feed intake of dairy cows. J. Agric. Sci. Cambr. 93, 553–562. Drennan, M.J., Keane, M.G., 1987b. Responses to supplementary Veira, D.M., Butler, G., Ivan, M., Proulx, J.G., 1985. Utilisation of concentrates for finishing steers fed silage. Ir. J. Agric. Res. 26, grass silage by cattle: effect of barley and fishmeal
supple-115–127. ments. Can. J. Anim. Sci. 65, 897–903.
Drennan, M.J., Lawlor, M.J., 1976. Evaluation of pelletted dried Veira, D.M., Fortin, A., Ivan, M., Butler, G., Proulx, J.G., 1983. grass as a supplement to grass silage for fattening steers. Anim. Performance of Hereford
3Shorthorn bulls and steers raised in Prod. 22, 97–103. Northern Ontario and fed different levels of grass silage and Flynn, A.V., 1978. Supplementing early and late cut silages with high moisture barley. Can. J. Anim. Sci. 63, 861–869.
barley for store cattle. An Foras Taluntais. Anim. Prod. Res. Veira, D.M., Proulx, J.G., Seoane, J.R., 1990. Performance of beef Rep., 31–32. steers fed grass silage with or without supplements of soybean Forbes, T.J., Irwin, J.H.D., 1970. Silage for winter fattening. J. Br. meal, fish meal and barley. Can. J. Anim. Sci. 70, 313–317.
Grassld. Soc. 25, 96–103. Wilkinson, J.M., 1985. Evaluation of conserved forages. In: Forbes, T.J., Jackson, N., 1971. A study of the utilization of Whittemore, C.T., Simpson, K. (Eds.), Beef Production from silages of different dry-matter content by young beef cattle Silage and Other Conserved Forages. Longman, London, pp. with or without supplementary barley. J. Br. Grassld. Soc. 26, 67–74.
257–264.
