ESTIMATION OF THE NUTRITIVE VALUE OF GRASS SILAGES INTRODUCTION

ESTIMATION OF THE NUTRITIVE VALUE OF GRASS SILAGES P. Huhtanen 1, M. Rinne 2 and J. Nousiainen 3 1 Department of Animal Science, Cornell University 2 Animal Production Research, MTT Agrifood Research Finland 3 Farm Services, Valio Ltd. Finland INTRODUCTION The main objective of feed evaluation including chemical laboratory analysis, in vitro digestibility with rumen bacteria or enzymes as well as adjunct methods such as in situ incubation in nylon bags and near infrared reflectance spectroscopy (NIRS) is to predict the availability of nutrients and feeding value for animal production systems. Accurate estimation of the nutritive value of forages is more important than that of concentrate feedstuffs due to the large variation in the nutritive value of forages and also because in most cases the contribution of forage to total diet DM is markedly greater than that of any of the individual concentrate ingredients. In addition to direct influence of forage quality on nutrient availability in ruminant diets, indirect effects to the total nutrient supply can be great because of the large impact of both digestibility and fermentation characteristics on silage DM intake (Rinne 2000; Huhtanen et al. 2007a). Accurate estimation of forage digestibility is a prerequisite for diet formulation, economic evaluation of forages and prediction of animal responses. Determination of in vivo digestibility is time consuming and expensive for routine use and even in nutrition research and therefore different laboratory methods have been developed to estimate digestibility of forages. Biological laboratory methods, even those based on rumen bacteria, very seldom estimate digestibility values directly. This does not mean that these methods are not useful, since they often present very close empirical relationships to in vivo digestibility, and hence empirical correction equations are required for estimating digestibility. It is also evident that these equations are often specific for different forages, environments and even laboratories (Weiss 1994, Van Soest 1994, Nousiainen 2004). The objectives of this paper are to evaluate the methods predicting digestibility of grass silages using empirical, biological and summative approaches. This paper is to a large extent based on results presented in Huhtanen et al. (2006a). DIGESTIBILITY The discussion is based on a dataset comprising of Finnish silages with known in vivo digestibility in sheep determined using total fecal collection method. The grass silages were harvested from mixed timothy - meadow fescue leys in primary growth (PG, n=33) and in regrowth (RG, n=27) and ensiled with formic acid based additives in pilot-scale tower silos or farm-scale bunker silos. The data set also included 19 legume

(red clover) and 7 whole crop silages (barley or wheat). Overview of the silages used is presented in Table 1. Table 1.Chemical composition and digestibility of silages Concentration (g/kg DM) Digestibility Ash CP NDF indf Lignin a OMS OMD NDFD pdndfd Mean 85 158 502 97 33 757 71.1 68.4 85.5 SD 15.6 43 100.3 41.2 12.8 58 5.52 9.19 5.73 Min 49 79 274 17 17 628 58.1 47.7 67.5 Max 122 301 669 211 79 878 84.0 86.9 95.9 a Analyzed as permanganate lignin Much effort has been directed into developing empirical regression equations that relate various chemical components to digestibility, although these attempts have not been very successful because of large interspecies and environmental variation (Van Soest 1994). We evaluated the empirical relationships between selected chemical parameters (CP, NDF, ADF and lignin) and in vivo organic matter digestibility (OMD) by regression analysis. The concentrations of feed chemical fractions were poorly related to in vivo OMD of silages. Although the relationships were statistically significant, prediction error using CP, NDF and ADF as independent variables was not markedly less than the standard deviation of OMD in the data set. Lignin was the best single predictor of OMD, but it explained only 43% of the variation and prediction error (4.2 %-units) is too high for practical feed evaluation and ration formulation. A large proportion of unexplained variation was related to forage type, and the use of forage specific equations decreased the error of predictions markedly. The decrease was greater for the cell wall components than for CP. The better relationship between NDF and ADF, and especially that of lignin, is due to these components being causative factors and related to the biological availability, whereas CP has no direct effect on digestibility provided that minimum N requirements of rumen microbes are met. When random study effect was included in the model RMSE decreased markedly, i.e. within forage and year, the changes in digestibility with advancing maturity can be predicted accurately from the chemical composition (Figure 1). Organic Matter Pepsin Cellulase Solubility (OMS) For digestibility determination, laboratory in vitro methods have been developed and widely exploited, based on ruminal fluid (introduced by Tilley and Terry 1963; extensively reviewed by Weiss 1994) or commercial fungal cellulases. Due to difficulties in obtaining rumen fluid in commercial laboratories and standardization of the system we have evaluated the enzymatic in vitro procedure in the determination of forage digestibility.

Enzymatic digestion procedures have been described and discussed in detail in a review by Jones and Theodorou (2000). Basically the method includes removing of cell solubles either by HCl-pepsin or neutral detergent followed by a 24 or 48 h incubation in buffered enzyme solution. The cellulase method differs from the in vivo digestion at least in two aspects: no metabolic matter is produced, i.e. the solubility reflects true rather than apparent digestibility and secondly, the capacity of commercial enzymes to degrade cell wall carbohydrates is smaller than that of rumen microbes (McQueen and Van Soest 1975, Nousiainen 2004). 6.0 5.0 RMSE (%) 4.0 3.0 2.0 s.d. A B C 1.0 0.0 CP NDF ADF Lignin Figure 1. Residual mean squared errors (RMSE) of the empirical relationships between feed components and organic matter digestibility estimated with different models (s.d. = standard deviation of the data; A = general relationship; B = forage type specific equations; C = variation from a random study effect excluded). Data contains all forage types. Our data indicates that the relationship between OMS and in vivo OMD is not uniform between the forage types (Table 2), because the prediction error within each forage type was markedly smaller than that estimated using the general correction equation. However, compared to chemical components, the prediction error was much smaller even with the general equation suggesting that OMS reflects the mechanisms behind digestibility better than components of the proximal analysis or detergent fractionation. When the forage specific equation was used, prediction error for OMD decreased from 2.45 to 1.53%. The differences in RMSE between fixed and mixed model regressions with grass silages suggest that the relationship between OMS and OMD may also be year-dependent. In addition to forage specific equations, the laboratory specific equations may be needed. Despite of serious attempts, the laboratories of Valio Ltd. and MTT Agrifood Research Finland were not able to standardize the methods (Nousiainen 2004). There was a difference in the intercept, but the slope was 1.00 and R 2 high (0.97). The intercept difference suggests particle loss during the procedures (manual filtration vs.

Tecator crucibles). The lower OMS with centrifugation method compared with Tecator crucibles (71.6 vs. 74.7%) indicates particle loss (Rinne et al. unpublished), but the material was too small to conclude whether the precision of OMS method can be improved by replacing filtration with centrifugation. It is also noteworthy that in our study with legume silages (Rinne et al. 2006), in vitro OMD determined by Tilley and Terry (1963) method significantly underestimated in vivo OMD. In their original evaluation, Tilley and Terry (1963) underlined that despite a close relationship between in vivo and in vitro digestibility, these values are not correspondent and a specific correction equation within laboratory and possibly within forage type are needed. Previous interpretation by Weiss (1994) of the between laboratory variations using ruminal in vitro system clearly suggests that corrections need to be laboratory specific. Table 2. Empirical relationships between pepsin-cellulase organic matter (OM) solubility (%) and in vivo OM digestibility determined with fixed or mixed regression analysis with random study effect (adapted from Huhtanen et al. 2006a). Silage raw material Model a Intercept Slope s.e. RMSE Adj. R 2 Primary growth grass F 10.3 0.83 0.038 1.51 0.937 Primary growth grass M 7.7 0.86 0.027 0.85 0.981 Regrowth grass F -7.0 1.01 0.136 1.93 0.676 Regrowth grass M -15.4 1.12 0.082 0.91 0.921 Legume F 0.2 0.93 0.044 1.22 0.962 Legume M 0.3 0.93 0.044 1.21 0.962 Whole-crop F 18.2 0.66 0.064 1.09 0.947 Whole-crop M 29.0 0.52 0.129 0.90 0.942 All F 6.4 0.86 0.046 2.45 0.804 All M 4.0 0.89 0.026 0.99 0.964 a Regression model: F = Fixed effect model; M = Mixed model with random study effect. Organic matter digestibility of grass silages can be predicted from OMS of ensiled herbage as precisely as from OMS of the resultant silages provided that silages are well preserved with low or moderate ensiling losses (Huhtanen et al. 2005). For practical ration formulation, sampling of herbage during silage harvesting allows obtaining more representative samples and provides a better illustration of the variation in silage digestibility than samples taken from the silos. Advance information of silage digestibility would also be useful in the ration planning. Indigestible Neutral Detergent Fiber (indf) A part of the forage cell wall is protected from all microbial and enzymatic digestion in ruminants, although the total tract residence time of fibre could be extended to ultimate time period (Allen and Mertens 1988, Van Soest 1994). This forage DM fraction can be called potentially indigestible fibre, here referred to as indigestible NDF (indf). In addition to neutral detergent solubles (NDS), indf represents by definition a uniform

feed fraction with zero true digestibility. Potentially digestible fibre (pdndf) may then be calculated as pdndf = NDF indf. Several methods may be used in dividing forage NDF to digestible and indigestible fractions, e.g. end-point measurement with long-term (up to 144 h) in vitro batch rumen fluid incubation (Traxler et al. 1998) or fitting time-dependent (0-96 h) in vitro or in situ NDF degradation data to the single digestion pool rumen model. Van Soest et al. (2005) used 216 or 240 h in vitro incubation and re-inoculated samples after each 72 h. We have used a long-term (12 d) in situ incubation to ensure complete digestion of pdndf. The small pore size (6 or 17 μm) combined with a relatively large open surface area of the nylon bag cloth used allows moderate microbial activity within the bags (Huhtanen et al. 1998) and prevents particle in- and out-flow from the bags. The incubations are carried out in cows fed diets ensuring optimal rumen environment for cell wall digestion. After in situ incubation, the residues are washed with water and treated with ND solution to remove metabolic matter. The details of the procedures have been described by Huhtanen et al. (1994) and Ahvenjärvi et al. (2000). We have also evaluated the use of NIRS in predicting grass silage indf with promising results (Nousiainen et al. 2004), but the standardization of the reference method is a vital prerequisite in developing robust multi-purpose calibrations. Because forage indf fraction is attributable to cross-linking between cell wall lignin and hemicellulose when plants mature (Van Soest 1994), several attempts to predict indf concentration from lignin concentration in DM or NDF have been made (see Traxler et al. 1998). The Cornell Net Carbohydrate and Protein system uses a factor 2.4 lignin concentration in NDF in describing indf concentration of forages (Van Soest et al. 2005). This factor is presumed to be universal across forage species and growth environments. Validation of this concept with data containing several forage species (n=21) resulted in satisfactory regression (R 2 = 0.94) between observed and predicted (2.4 lignin) indf concentrations (Van Soest et al. 2005). In contrast, our data (n=87) does not support a generally applicable relationship between lignin and indf concentrations Although the overall slope was 2.4, a general regression equation predicted indf concentration with an unsatisfactory accuracy (R 2 = 0.56; RMSE = 27.4 g/kg DM). The relationships were reasonably good within forage types, but varied greatly between the forage types. Mertens and Huhtanen (2007) concluded that indf concentration is correlated to lignin concentration, but the relationship may not be consistent. The results for grass silages (Nousiainen et al. 2003, Nousiainen 2004) suggested that indf concentration can be used in a single empirical linear regression equation to predict forage OMD relatively universally over a range of species and harvesting conditions. The intercept of this equation represents a theoretical maximum of forage OMD provided that all NDF is potentially digestible and that the rate of pdndf digestion (k d ) is the only limiting factor for availability when the forage is fed to sheep at maintenance feed intake. The slope of the regression describes the decline in OM digestibility with increasing indf concentration. When the original data comprising mainly of grass silages was extended to include legume and whole crop silages, the

relationship between indf and OMD seems not to be completely uniform (Huhtanen et al. 2006a). However, the relationship was more uniform compared with OMS both in terms of RSME (1.90 vs. 2.45%) and R 2 (0.883 vs. 0.804). The respective negative slope of indf was smaller for legume compared with grass silages, probably because of faster digestion rate of legume silages at the same indf concentration. SUMMATIVE MODELS Van Soest (1967) developed a comprehensive system of feed analysis and its application to forages. He divided the feed into NDS fraction which is essentially completely available but its digestibility is apparently incomplete because of faecal endogenous and microbial non-cell wall material. The second fraction corresponds to plant cell wall and its availability is controlled by structural features that link cellulose, hemicellulose and lignin together. Cell wall fraction is not uniform between forages. Goering and Van Soest (1970) presented a summative model to describe availability of forage DM: ddm = 0.98 NDS + NDFD NDF M, where ddm = digestible DM, NDFD = coefficient of NDF digestibility, and M = metabolic fecal output. Digestibility of NDS fraction in this equation is based on the Lucas test (see Van Soest 1994). The purpose of the Lucas test is to identify ideal nutritional entities that have uniform digestibility over a wide range of feedstuffs by plotting the digestible nutrient concentration in DM against the nutrient concentration in DM. The slope of regression estimates the true digestibility and the intercept is an estimate of the metabolic fecal matter (M) for the nutrient. Theoretically the Goering and Van Soest (1970) model is model is sound, but generally NDFD is not known. Conrad et al. (1984) developed this model by dividing feeds into NDS and potentially digestible NDF. They applied surface area law (mass raised to power 0.67) to calculate NDF that is covered by lignin, and this proportion was multiplied by lignin-free NDF to obtain an estimate of the potentially digestible NDF. Weiss et al. (1992) revised the Conrad et al. (1984) model and it was adopted by NRC (2001) to estimate total digestible nutrients (TDN). Huhtanen (2003) evaluated the NRC (2001) model using in vivo digestibility data. The predicted and observed digestible OM concentrations were relatively well correlated, but NRC (2001) clearly underestimated the digestibility and there was a considerable slope bias. Because the in vivo pdndf digestibility is markedly less variable than the total NDF digestibility, and because the fraction subjected to this variation (pdndf vs. NDF) is smaller, the accuracy of the summative systems based on three fractions (NDS, pdndf and indf) could be improved compared to systems dividing feeds only to total NDF and NS solubles. In our data set, the coefficient of variation of pdndf and NDF digestibility was 0.064 and 0.135, and respective concentrations 403 and 500 g/kg DM. In the total data set R 2 for the pdndf Lucas test was high (0.95) and intercept close to zero. However, between the forage types the intercepts varied from negative to positive, which both are biologically meaningless.

Mertens (2002) proposed a simple equation in which dndf is a linear function of NDF and lignin concentrations: dndf = a NDF (g/kg DM) + b Lignin (g/kg DM), where a and b are constants. Constant a represents the proportion of NDF that is potentially digestible and constant b is the proportion of cell walls protected by lignin. This equation has no intercept, i.e. no endogenous or metabolic excretion of NDF. We compared three methods in estimating dndf and digestible OM (dom) concentrations of the silages: NRC (2001), Mertens (2002) and the Lucas test. Digestible OM was calculated as dndf + dnds. Digestible NDS was estimated by the Lucas test for ashfree NDS (OM NDS). Because indf concentration predicted in vivo OMD better than lignin concentration, the NRC (2001) and Mertens (2002) models were also tested by replacing lignin with indf. For all models, both general equations derived from all data and forage specific equations were used for both dndf and dnds. Parameter values for the NRC (2001) and Mertens (2002) equations were estimated from the data instead of using original parameter values. The results of the evaluation of the general summative models are shown in Table 3. The more complex function of lignin adopted in the NRC (2001) model did not improve the precision of dndf prediction compared to purely empirical prediction based on NDF concentration and linear relationship between NDF digestibility and lignin concentration (MSPE = 33.9 g/kg DM). Mertens (2002) equation resulted in a similar error to that of NRC and empirical approach. Interestingly, the prediction errors of NRC and Mertens equations were strongly correlated (R 2 =0.99) indicating that the form of lignin function had no influence on dndf prediction. Prediction of dndf by the Lucas test resulted in a much lower MSPE than lignin models, and MSPE was only marginally higher than with the NRC (2001) and Mertens (2002) models using indf. RMSE was markedly reduced when indf rather than lignin was used in the NRC (2001) and Mertens (2002) models. Table 3. Prediction of the digestible NDF (dndf) and organic matter (dom) using different summative equations for all forage types; general equations were used to predict both dndf and dnds (Huhtanen et al. 2006a). Independent Distribution of MSPE Trait/method variable Intercept Slope R 2 MSPE Bias Slope Random dndf (g/kg DM) NRC Lignin 11 0.97 0.853 33.6 0.00 0.01 0.99 NRC indf 9 0.98 0.954 18.7 0.00 0.01 0.99 Mertens Lignin 9 0.98 0.849 34.0 0.00 0.00 1.00 Mertens indf 8 0.98 0.957 18.5 0.00 0.01 0.99 Lucas test pdndf 0 1.00 0.952 19.1 0.00 0.00 1.00 dom (g/kg DM) NRC Lignin -104 1.16 0.551 35.3 0.00 0.02 0.98 NRC indf -80 1.12 0.904 17.0 0.00 0.10 0.90 Mertens Lignin -121 1.19 0.542 35.7 0.00 0.03 0.97 Mertens indf -78 1.12 0.911 16.4 0.00 0.11 0.89 Lucas test dndf -152 1.24 0.908 18.3 0.00 0.26 0.74

When the forage specific equations were applied to predict both dndf and dnds (Table 4), prediction errors were markedly reduced compared with the general equations. This was especially the case with lignin models supporting the view that the ratio between lignin and indf is not constant for all forage types. Prediction errors for dndf were only 11-12 g/kg DM for the three models. The decrease in prediction error of dom was partly due to the fact that true digestibility of NDS and fecal output of metabolic and endogenous matter were not entirely constant. Prediction error of this fraction decreased from 16.2 to 8.6 g/kg DM when forage specific rather than general Lucas equations were used. Metabolic fecal output was markedly higher for the regrowth silages, but the reason for this is not known. The Mertens (2002) equation can also be formulated as dndf = a (NDF indf) + b indf. In this equation coefficient a represents digestibility of pdndf (=NDF indf) and b is a coefficient for indf allowing pdndf digestibility to vary with indf concentration. Table 4. Prediction of digestible NDF (dndf) and organic matter (dom) using different summative equations from data comprising of silages made from primary or regrowth grass and leguminous or whole crop forages; forage specific equations were used to predict both dndf and dnds (Huhtanen et al. 2006a). Independent Distribution of MSPE variable Intercept Slope R 2 MSPE Bias Slope Random dndf (g/kg DM) NRC Lignin 15 0.96 0.944 20.9 0.00 0.01 0.99 NRC indf 4 0.99 0.984 11.2 0.00 0.01 0.99 Mertens Lignin 13 0.96 0.946 20.5 0.00 0.00 1.00 Mertens indf 2 0.99 0.983 11.4 0.00 0.01 0.99 Lucas test pdndf 0 1.00 0.980 12.4 0.00 0.00 1.00 dom (g/kg DM) NRC Lignin -50 1.08 0.802 23.4 0.00 0.02 0.98 NRC indf -16 1.03 0.932 13.6 0.00 0.01 0.99 Mertens Lignin -33 1.05 0.809 23.0 0.00 0.01 0.99 Mertens indf -13 1.02 0.930 13.8 0.00 0.01 0.99 Lucas test dndf -50 1.08 0.922 15.0 0.00 0.06 0.94 Evaluation of the different approaches reveals that for accurate and precise prediction of D-value, forage-specific equations are needed irrespective of the method used (empirical vs. summative). Basically this is because the metabolic fecal output of NDF and the relationship between indf concentration and the rate of pdndf digestion are not constant among forages. An advantage of the summative systems is that they are based on physical and biochemical factors that influence the bioavailability of various feed fractions. Prediction errors of dom were similar for forage specific empirical OMS and indf equations and summative equations, but with general equations the error was much greater for OMS (22.3) compared with indf (16.9) and summative (16.4) models.

MECHANISTIC MODELS Modern feed evaluation models have mechanistic dynamic elements. Sensitivity analysis of the Nordic dairy cow model Karoline (Danfær et al. 2006) clearly indicated that forage indf concentration is the key parameter in predicting the nutrient supply of dairy cows (Huhtanen et al., 2006b). The effect of digestion rate of pdndf on nutrient supply was smaller than that of indf concentration. Variation in pdndf digestibility is related to differences in the digestion rate of pdndf associated with cell wall characteristics, since the variation in compartmental residence time in the rumen is likely to be small in sheep fed at maintenance. Considering the relatively small coefficient of variation in the pdndf digestibility of grass silages (4.1% in the present data; n=52) the potential to improve the accuracy dndf or dom predictions by mechanistic models is rather limited. Digestion rate must be predicted accurately and precisely and/or the errors in predicting fecal NDS must be reduced compared with the general Lucas test. Much better relationship between indf concentration and NDFD than that between pdndfd (reflects digestion rate) indicate that accurate determination of potentially indigestible NDF fraction is more important than accurate estimation of digestion rate (Figure 2). A prerequisite of accurate determination of indf is that incubation time is long enough to reach potential extent of digestion. It is often argued that the measurement of digestion after long fermentation times is irrelevant because average retention times of feeds in the rumen are much shorter (Mertens and Huhtanen, 2007). However, this contention is wrong for two reasons. First, selective retention of feeds in the rumen can result in long retention times in the rumen for some feed fractions. Second, and more importantly, long times of digestion are needed to measure the indigestible fraction so that the potentially digestible fraction can be determined. Only by partitioning feed components into potentially digestible and indigestible pools can we measure or assign accurate digestion kinetics. NDFD (%) 90 85 80 75 70 65 60 55 y = -0.166x + 87.6 R 2 = 0.883 50 0 50 100 150 200 indf (g/kg DM) NDFD (%) 90 85 80 75 70 65 60 55 y = 1.162x - 28.3 R 2 = 0.371 50 78 80 82 84 86 88 90 92 94 pdndfd (%) Figure 2. Effects of indf concentration and potentially digestible NDF digestibility (pdndfd) on NDF digestibility of grass silages in sheep fed at maintenance level of intake.

Errors in estimating digestion rate of pdndf can result in greater prediction errors of mechanistic models in predicting forage feeding values compared with empirical or summative models. Although ruminal in situ incubation generally ranks digestion rates correctly, the method underestimates the in vivo digestion rate (Huhtanen et al. 2006b). In vitro gas production technique (for review see Schofield 2000) is a promising tool for estimating digestion rate of NDF. When the parameter values derived from gas production kinetics of isolated NDF were used in dynamic mechanistic rumen models, in vivo NDF digestibility was predicted accurately and precisely (Huhtanen et al. 2007b). Another alternative is to estimate pdndf digestion rate indirectly from dom concentration. Digestible NDS can be estimated by the Lucas test and consequently dndf can be computed. Digestibility of pdndf can be estimated by dividing dndf with pdndf (=NDF indf). Digestion rate of pdndf is then calculated by solving the equation of Allen and Mertens (1988) for digestion rate by assuming a fixed compartmental residence time as described by Huhtanen et al. 2006b). Digestion rates estimated from isolated silage NDF with in vitro gas production technique and those calculated from the in vivo data were strongly correlated without mean or slope bias (Huhtanen et al. 2007b). A problem of in vitro gas production system in estimating digestion rate of pdndf is that the rates appear to be time-dependent, which are difficult to use in steady state rumen models. However, prediction error of in vivo NDFD was not greater when single exponential model with discrete lag model was used to fit gas production profile compared with more complicated two pool models with time-dependent digestion rates (Huhtanen et al. 2007b). We also demonstrated that a single first-order digestion rate can be estimated from more complicated models by using the kinetic parameters in twocompartment rumen model to estimate pdndf digestibility and solving digestion rate as described above. CONCLUSIONS Predicting in vivo organic matter digestibility with empirical equations using chemical parameters based either on proximate or detergent analysis gave unsatisfactory results. Pepsin-cellulase solubility predicts OMD precisely especially when forage specific equations are used. Indirect comparisons indicate that the accuracy of the method is at least similar to rumen fluid in vitro methods. The method does not require rumen cannulated animals and the activity of enzyme is most likely to be less variable than that of rumen fluid. The relationship between cellulase solubility and OMD are not uniform between forage types and therefore forage specific equations are needed to calculate OMD from solubility. Determination of indf by a long (10-12 days) in situ or in vitro incubation is a powerful tool in forage evaluation. The empirical relationship between indf and OMD was good (R2 0.88, RMSE 1.9%). The relationship between forage indf concentration and OMD is more uniform between forage types than that between cellulase solubility and OMD. Determination of indf concentration divides feed OM into three fractions (indf, pdndf and NDS), of which indf and NDS are uniform entities and pdndf is also much more uniform than NDF. This allows developing summative

equations to predict dom. Modern feed evaluation models require digestion rates of feed components in predicting nutrient supply. However, to improve the feed evaluation models by including dynamic and mechanistic elements into the models, digestion rate of pdndf must be estimated both accurately and precisely. The potential of mechanistic models is that they predict feeding level effects and interactions between dietary components better than factorial models. REFERENCES Ahvenjärvi, S., A. Vanhatalo, P. Huhtanen, and T. Varvikko. 2000. Determination of reticulo-rumen and whole-stomach digestion in lactating cows by omasal canal or duodenal sampling. Br. J. Nutr. 83:67-77. Allen, M.S., and D.R. Mertens. 1988. Evaluating constraints of fibre digestion by rumen microbes. J. Nutr. 118:261-270. Conrad, H.R., W.P. Weiss, W.O. Odgonwongon, and Shockey, W.L. 1984. Estimating net energy of lactation from components of cell solubles and cell walls. J. Dairy Sci. 67:427-436. Danfær, A., P. Huhtanen, P. Udén, J. Sveinbjörnsson, and H. Volden. 2005. Karoline - a Nordic cow model for feed evaluation - model description. In: Kebreab, E. Dikstra, J, Bannink, A, Gerrits, J., France, J. (eds) Nutrient Digestion and Utilization in Farm Animals: Modelling approaches. CAB International, pp. 407-415. Goering, H.R., and P.J. Van Soest. 1970. Forage fiber analysis. Agricultural Handbook No. 379. USDA, Washington, 22 p. Huhtanen, P. 2003. Factors influencing on voluntary intake of silage-based diets, and the responses of silage quality in milk production. In: T. Garmo (ed). Proc.Int. Symp.: Early harvested forage in milk and meat production. Kringler, Nannestad, Norway. Ås: pp. 59-82 Huhtanen, P., S. Ahvenjärvi, M.R. Weisbjerg, and P., Nørgaard. 2006b. Digestion and passage of fibre in ruminants. In: Sejrsen, K., Hvelpund, T. and Nielsen, M.O. (eds.), Ruminant Physiology. Wageningen Academic Publishers, The Netherlands, pp. 87-135. Huhtanen, P., K. Kaustell, and S. Jaakkola. 1994. The use of internal markers to predict total digestibility and duodenal flow of nutrients in cattle given six different diets. Anim. Feed Sci.Technol. 48:211-227 Huhtanen, P., J. Nousiainen, and M. Rinne. 2005. Prediction of silage composition and organic matter digestibility from herbage composition and pepsin-cellulase solubility. Agric.Food Sci. 14:154-165. Huhtanen, P., J. Nousiainen, and M. Rinne. 2006a. Recent developments in forage evaluation with special reference to practical applications. Agric. Food Sci. 3: 293-323 Huhtanen, P., A. Vanhatalo, and T. Varvikko. 1998. Enzyme activities of rumen particles and feed samples incubated in situ with differing types of cloth. Br. J. Nutr. 79:161-168.

Huhtanen, P., M. Rinne, and J. Nousiainen. 2007a. Evaluation of the factors affecting silage intake of dairy cows; a revision of the relative silage dry matter intake index. Animal 1, 758-770. Huhtanen, P., A. Seppälä, M. Ots, S. Ahvenjärvi, and M. Rinne. 2007b. Use of in vitro gas production profiles in estimating digestion rate parameters: prediction of in vivo cell wall digestibility and effective first-order digestion rate in the rumen. Manuscript. Jones, D.I.H., and M.K. Theodorou. 2000. Enzyme techniques for estimating digestibility. In: Givens, D.I., Owen, E., Axford, R.F.E., Omed, H.M., (eds.), Forage Evaluation in Ruminant Nutrition. CABI Publishing, Oxon, pp. 155-173. McQueen, R., and P.J. Van Soest. 1975. Fungal cellulase and hemicellulase in prediction of forage digestibility. J.Dairy Sci. 58:1482-1491. Mertens, D.R. 2002. Fiber: measuring, modelling and feeding. In: Proceedings Cornell Nutrition Conference for Feed Manufacturers. 23-25 Oct. 2002. East Syracuse, NY. p. 1-18 Mertens, D., and P. Huhtanen. 2007. Grass forages: Dynamics of digestion in the rumen. 2007 Proceedings of the New York Sate Ruminant Health-Nutrition Conference. Syracuse March 27, 2007. 20p. Nousiainen, J. 2004. Development of tools for the nutritional management of dairy cows on silage-based diets. Academic dissertation. University of Helsinki, Department of Animal Science Publications 72, 61 p. + 5 encl. Available at http://ethesis.helsinki.fi/julkaisut/maa/kotie/vk/nousiainen/ Nousiainen, J., S. Ahvenjärvi, M. Rinne, M. Hellämäki, and P. Huhtanen. 2004. Prediction of indigestible cell wall fraction of grass silage by near infrared reflectance spectroscopy. Anim. Feed Sci. Technol. 115:295-311. Nousiainen, J., M. Rinne, M. Hellämäki, and P. Huhtanen. 2003. Prediction of the digestibility of the primary growth and regrowth grass silages from chemical composition, pepsin-cellulase solubility and indigestible cell wall content. Anim. Feed Sci. Technol. 110:61-74. NRC. 2001. National Research Council. Nutrient Requirements of Dairy Cattle. 7 th revised Edition. National Academy Press, Washington, DC. pp. 381 Rinne, M. 2000. Influence of the timing of the harvest of primary grass growth on herbage quality and subsequent digestion and performance in the ruminant animal. Univ.Helsinki, Dep.Anim. Sci.. Publ. 54. 42 p. + 5 encl. Academic dissertation. Available at: http://ethesis.helsinki.fi/julkaisut/maa/kotie/vk/rinne. Rinne, M., A. Olt, J. Nousiainen, A. Seppälä, M. Tuori, C. Paul, M.D Fraser, and P. Huhtanen. 2006. Prediction of legume silage digestibility from various laboratory methods. Grass Forage Sci. 61: 354-362. Scofield, P. 2000. Gas production methods. In: D Mello, J.P.F. (Ed.), Farm Animal Metabolism and Nutrition. CAB International, p. 209-232. Tilley, J.M.A., and R.A. Terry. 1963. A two stage technique for the in vitro digestion of forage crops. J. Br. Grassl. Soc. 18:104-111. Traxler, M.J., D.G. Fox, P.J. Van Soest, A.N. Pell, C.E. Lascano, D.P.D. Lanna, J.E. Moore, R.P. Lana, M. Velez, and A. Flores. 1998. Predicting forage indigestible NDF from lignin concentration. J. Anim. Sci. 76:1469-1480.

Van Soest, P.J. 1967. Developments of a comprehensive system of feed analysis and its application to forages. J.Anim. Sci. 26:119-128. Van Soest, P.J. 1994. Nutritional Ecology of the Ruminant. Second Edition. Comstock Publishing Associates, Cornell University Press, Ithaca and London, 476 p. Van Soest, P.J., M.E. Van Amburgh, J.B. Robertson, and W.F. Knaus. 2005. Validation of the 2.4 times lignin factor for ultimate extent of NDF digestion, and curve peeling rate of fermentation curves into pools. In: Proc. Cornell Nutr. Conf. for Feed Manufacturers. 24-26 Oct. 2005. East Syracuse, NY. p. 139-149 Weiss, W.P. 1994. Estimation of digestibility of forages by laboratory methods. In: Fahey, G.C. Jr (Ed.), Forage Quality, Evaluation and Utilization, American Society of Agronomy, Madison, WI. pp. 644-681. Weiss, W.P., H.R. Conrad, and N.R. St.Pierre. 1992. A theoretically-based model for predicting total digestible nutrients of forages and concentrates. Anim. Feed Sci. Technol. 39:95-110.